Nominalization in Asian Languages
Typological Studies in Language (TSL) A companion series to the journal Studies in Language. Volumes in this series are functionally and typologically oriented, covering specific topics in language by collecting together data from a wide variety of languages and language typologies. For an overview of all books published in this series, please see http://benjamins.com/catalog/tsl
Editor Spike Gildea
University of Oregon
Editorial Board Balthasar Bickel
John Haiman
Marianne Mithun
Bernard Comrie
Martin Haspelmath
Doris L. Payne
Denis Creissels
Bernd Heine
Franz Plank
William Croft
Paul J. Hopper
Anna Siewierska
Nicholas Evans
Andrej A. Kibrik
Dan I. Slobin
Carol Genetti
František Lichtenberk
Sandra A. Thompson
Leipzig
Leipzig / Santa Barbara Lyon
Albuquerque Canberra
Santa Barbara
St Paul
Leipzig Köln
Pittsburgh Moscow
Auckland
Santa Barbara Eugene, OR Konstanz
Lancaster Berkeley
Santa Barbara
Volume 96 Nominalization in Asian Languages. Diachronic and typological perspectives Edited by Foong Ha Yap, Karen Grunow-Hårsta and Janick Wrona
Nominalization in Asian Languages Diachronic and typological perspectives Edited by
Foong Ha Yap Hong Kong Polytechnic University
Karen Grunow-Hårsta Hong Kong Polytechnic University
Janick Wrona Oxford University
John Benjamins Publishing Company Amsterdamâ•›/â•›Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data Nominalization in Asian languages: diachronic and typological perpsectives / edited by Foong Ha Yap, Karen Grunow-Hårsta, Janick Wrona. p. cm. (Typological Studies in Language, issn 0167-7373 ; v. 96) Includes bibliographical references and index. 1. Asia--Languages. 2. Grammar, Comparative and general--Nominals. 3. Grammar, Comparative and general--Syntax. 4. Grammar, Comparative and general-Noun. 5. Typology (Linguistics) 6. Historical linguistics. I. Yap, Foong Ha. II. Grunow-Hårsta, Karen. III. Wrona, Janick, 1971P381.A75.N66â•… 2011 495--dc22 isbn 978 90 272 0677 0 (Hb ; alk. paper) isbn 978 90 272 8724 3 (Eb)
2010051886
© 2011 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
This volume is dedicated to Michael Noonan (14 September 1947, Auburn NY – 23 February 2009, Whitefish Bay WI) whose insightful work on nominalization has inspired many of us.
Table of contents Preface
xi
Acknowledgment to reviewers
xv
List of contributors Introduction: Nominalization strategies in Asian languages Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
xvii 1
part i. Sinitic languages From light noun to nominalizer and more: The grammaticalization of zhe and suo in Old and Middle Chinese Foong Ha Yap & Jiao Wang
61
On the polyfunctionality and grammaticalization of the morpheme kai in the Chaozhou dialect Hui Ling Xu & Stephen Matthews
109
The Cantonese ge3 Joanna Ut-Seong Sio
125
On gerundive nominalization in Mandarin and Cantonese Sze-Wing Tang
147
part ii. Tibeto-Burman languages Nominalization in Tibeto-Burman languages of the Himalayan area: A typological perspective Carol Genetti
163
Aspects of the historical development of nominalizers in the Tamangic languages Michael Noonan
195
Innovation in nominalization in Magar, a Tibeto-Burman language of Nepal Karen Grunow-Hårsta
215
Nominalization and nominalization-based constructions in Galo Mark W. Post
255
 Nominalization in Asian Languages
Nominalization in Numhpuk Singpho Stephen Morey
289
Nominalization in Nuosu Yi Hongyong Liu & Yang Gu
313
Finite structures from clausal nominalization in Tibeto-Burman Scott DeLancey
343
part iii. Iranian languages Linker, relativizer, nominalizer, tense-particle: On the Ezafe in West Iranian Geoffrey Haig
363
part iv. Korean and Japanese languages Nominalization and stance marking in Korean Seongha Rhee
393
A case of non-derived stand-alone nominalization: Evidence from Japanese Janick Wrona
423
Nominalization in Okinawan: From a diachronic and comparative perspective Rumiko Shinzato
445
Versatility of nominalizations: Where Japanese and Korean contrast Kaoru Horie
473
part v. Austronesian languages The functions of -an and =ay in Kavalan Fuhui Hsieh
499
Clausal nominalization in Budai Rukai Li-May Sung
523
Nominalization in Saisiyat Marie Mei-li Yeh
561
Rise and fall of referentiality: Articles in Philippine languages Naonori Nagaya
589
Referential and non-referential uses of nominalization constructions in Malay Foong Ha Yap
627
Expressing exclamatives in Malagasy Eric Potsdam
659
Table of contents 
Nominalizations in Toqabaqita and closely related languages Frantisek Lichtenberk
685
Exclamatives and temporal nominalizations in Austronesian Daniel Kaufman
721
part vi. Papuan languages Discourse-structuring functions of Abui demonstratives František Kratochvíl
757
Index
789
Language index
795
Preface This volume grew out of collaborative efforts to investigate why morphemes that signal nominalization and relativization constructions were also found as sentence final particles in a wide range of Asian languages. Investigations originally had an East Asian focus (e.g. Japanese, Korean and Chinese languages). However, a workshop at the conference on New Reflections on Grammaticalization (NRG3) at Santiago de Compostela in Spain in July 2005 highlighted that related nominalization phenomena exist in languages farther afield as well (e.g. Burmese, Jingpo, and the Bodic languages of Nepal, among many others). Following this, a special workshop focusing on nominalizers and copulas was held at the Chinese University of Hong Kong in January 2007. These combined research efforts gave rise to several publications, among them a series of papers on nominalizations in Asian languages in Rethinking Grammaticalization: New Perspectives, edited by Mariá José Lo´pez-Couso & Elena Seoane, in collaboration with Teresa Fanego (eds). These include papers by Michael Noonan (“Nominalizations in Bodic languages”), Andrew Simpson (“The grammaticalization of clausal nominalizers in Burmese”), Kaoru Horie (“The grammaticalization of nominalizers in Japanese and Korean: a contrastive study”), Seongha Rhee (“On the rise and fall of Korean nominalizers”), and Foong Ha Yap & Stephen Matthews (“The development of nominalizers in East Asian and Tibeto-Burman languages”). Our understanding of nominalization phenomena has benefited greatly from the extensive fieldwork and analytical insights of numerous linguists. In particular, Tibeto-Burman scholarship, as early as the 1970’s in the work of James Matisoff, recognized relationships between nominalization, relativization and genitivization. These investigations continued into the 1980’s with the work of Scott DeLancey and Michael Noonan, among others. A new generation of active and productive scholars in the 1990’s built upon this earlier research, particularly works such as Michael Noonan’s seminal paper “Versatile Nominalizations” published in (1997). The result of these investigations has been a number of insightful publications on nominalization, such as those by Bickel, Coupe, Genetti, Grunow-Hårsta, Lahaussois, LaPolla, and Watters, among others. Austronesian scholars have also worked extensively on issues related to nominalization, motivated in part by a debate within the community on issues related to the frequently observed isomorphism between nominalization constructions and finite clause structures, which gives rise to questions about the chronology (and sometimes primacy) of one construction relative to the other. The debate – often pitched
 Nominalization in Asian Languages
in terms of the nominalist vs. non-nominalist hypothesis – has greatly deepened and broadened our thinking of nominalization issues from both diachronic and typological perspectives, and we wish here to recognize the contributions of the following, among many others: the works of Stanley Starosta, Andrew Pawley and Lawrence Reid (the SPR trio), Malcolm Ross, Nikolaus Himmelmann, Daniel Kaufman and Masayoshi Shibatani for a comprehensive typological perspective on nominalization phenomena across Austronesian languages, and Elizabeth Zeitoun and the contributions of many Taiwanese linguists in her 2002 Language and Linguistics special journal issue on nominalization in Formosan languages. There has been growing recognition and appreciation of the significance of nominalization processes. Recent investigations across Asian languages, such as those seen in this volume, have both affirmed and advanced previous scholarship. For a number of languages in this volume (Chinese, Tibeto-Burman, Japanese, and Korean) it is possible to trace the diachronic development from light nouns to nominalizers. Works in this volume have confirmed that nominalizers are indeed versatile, often developing extended functions. Moreover, it has demonstrated that this extension frequently follows a trajectory from referential to non-referential functions (such as nominalization to subordination and complementation, and the marking of tense, aspect and mood, and of speaker stance). It has also become clear that nominalization constructions can be signaled not only by dedicated nominalizers. They are often also signaled by noun phrase markers – ranging from the more cross-linguistically common ones such as demonstratives and case markers, as well as possessive pronouns and genitive case (as frequently seen across the Austronesian languages), to the less commonly attested ones such as plural markers (as seen for example in Rawang and Magar from the Tibeto-Burman family) and classifiers (as evidenced in Min and Yue dialects from the Sinitic family). These findings among the Asian languages investigated within this volume are consistent with earlier typological observations that languages frequently rely on noun phrase markers to help identify not only noun phrases but also nominalized clauses – a strategy nicely referred to as ‘substantivization strategies’ (see Malchukov 2006; see also Hopper & Thompson 1984, 1985; inter alia). The papers in the present volume provide in-depth analysis of languages representing several language families that span the Asian landscape, in some cases further clarifying earlier typological insights with diachronic evidence, and in some cases with discourse corpus data. Of particular interest in terms of new directions in research is growing evidence that noun phrase markers often extend their functions along pathways parallel to those of nominalizers. It is our hope that the contributions in this volume will deepen our understanding of nominalization strategies, their diachronic development, and the significance of their ubiquitous presence on other grammatical constructions, as well as their pragmatic effects in discourse, the latter an area of study we hope many more researchers will pursue (see for example recent
Preface 
works on subjectivity and intersubjectivity among Asian scholars such as Shoichi Iwasaki, Seiko Fujii and Nina Yoshida within the discourse analysis tradition). We hope both scholars and students will greatly benefit from the data and Â�analyses in the twenty-six papers (including the introductory chapter) of this volume. A lot of people have put a lot of work into preparing this volume, and we would like to take this opportunity to express our deep appreciation to those who have made the publication of this volume possible. First, we wish to thank our contributors and our reviewers for generously sharing with us their expert knowledge of the languages of Asia, often situated within the context of a broader ongoing typological discussion on nominalizations in the languages of the world. The names and affiliations of our Â�reviewers – because the list is appreciably long – are listed immediately following this preface. We also wish to thank our research assistants for their help in preparing for the publication of this volume, among them Shirley Tsang, Jenny Tse, and Jiao Wang. We are also very grateful to Spike Gildea, chief editor of TSL, and the editing staff at John Benjamins for overseeing the publication of this volume. In particular we wish to express our warm and very special thanks to Martine van Marsbergen, Patricia Leplae and Jan-Kees van Oord for their prompt and supportive supervision over the proofcopies, and to Kees Vaes at John Benjamins for his encouragement, support and patient guidance over the long years that have brought this volume to fruition. The emotional support from our families over the years has also played a crucial role in making this volume possible. Our families have endured, with patient understanding, our many late hours at work and sometimes long-distance absences away from home as we sit down to pull together the diverse voices of our contributors and search for common themes. Finally we wish to gratefully acknowledge the generous support of the Hong Kong Polytechnic University and the Chinese University of Hong Kong through the following grants (PI: Foong Ha Yap): CUHK Direct Research Grant 2006–2007 (Development of Stance Markers in East and Southeast Asian Languages); CUHK Direct Research Grant 2007–2008 (Diachronic Syntax in East Asian Languages); CUHK Linguistics and Modern Languages Department Research Grant 2008–2009 (Nominalization Strategies in Chinese and Other Asian Languages); HKPU Faculty of Humanities Start-Up Fund 2009–2011 (Non-Referential Uses of Nominalization Constructions: Asian Perspectives). We also wish to acknowledge generous support from a grant-in-aid from the Japan Society for the Promotion of Science awarded to co-editor Janick Wrona in the form of a postdoctoral fellowship at Kyoto University (2006–2008). We hope the analysis within this volume will contribute to increased attention to in-depth typological research not only ‘on’ but also ‘beyond’ nominalization Â�phenomena, especially now that we increasingly recognize that nominalization constructions seamlessly evolve and transform themselves for new functions in new environments.
 Nominalization in Asian Languages
References Hopper, Paul & Thompson, Sandra A. 1984. Lexical categories in universal grammar. Language 60(4): 703–752. Hopper, Paul & Thompson, Sandra A. 1985. The iconicity of the universal categories ‘noun’ and ‘verbs’. In Iconicity in Syntax [Typological Studies in Language 6], John Haiman (ed.), 151–186. Amsterdam: John Benjamins. Lo´pez-Couso, Mariá José & Seoane, Elena (eds.), in collaboration with Teresa Fanego. 2008. Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], 169–187. Amsterdam: John Benjamins. Malchukov, Andrej L. 2006. Constraining nominalization: Function/form competition. Linguistics 44(5): 973–1009. Noonan, Michael, 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givo´n, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Zeitoun, Elizabeth (ed.). 2002. Nominalization in Formosan languages. Language and Linguistics 3(2), (Special Issue).
Acknowledgment to reviewers Aldridge, Edith University of Washington, Seattle Au-Yeung, Wai-hoo Ben Chinese University of Hong Kong Chang, Henry Yungli Institute of Linguistics, Academia Sinica, Taiwan Cheung, Chi-hang Candice Chinese University of Hong Kong DeLancey, Scott University of Oregon Ding, Picus University of Hong Kong Foley, William University of Sydney Gerner, Matthias City University of Hong Kong Ghomeshi, Gila University of Manitoba Grunow-Hårsta, Karen Hong Kong Polytechnic University Haig, John University of Hawaii Kaufman, Daniel Cornell University Kim, Minju Claremont McKenna College, Los Angeles Koptjevskaja-Tamm, Maria Stockholm University Kratochvíl, František Nanyang Technological University, Singapore LaPolla, Randy La Trobe University, Melbourne Larson, Richard SUNY-Stony Brook, New York Law, Paul City University of Hong Kong Lien, Chin-fa National Tsing Hua University, Taiwan Liu, Hongyong South China Normal University, Guangzhou Malchukov, Andrej Max Planck Institute of Evolutionary Anthropology, Leipzig Matisoff, James University of California, Berkeley Morey, Stephen La Trobe University, Melbourne Nagaya, Naonori Rice University Ohori, Toshio University of Tokyo Pearson, Matt Reed College, Oregon Post, Mark The Cairns Institute, James Cook University, Queensland, Australia Potsdam, Eric University of Florida Reid, Lawrence University of Hawaii Rhee, Seongha Hankuk University of Foreign Studies, Seoul Rijkhoff, Jan University of Amsterdam Ross, Malcolm Australian National University
 Nominalization in Asian Languages
Rubino, Carl Shibasaki, Reijirou Okinawa International University Shibatani, Masayoshi Rice University Shimoji, Michinori Gunma Prefectural Women’s University, Japan Shinzato, Rumiko Georgia Institute of Technology Simpson, Andrew University of Southern California Sio, Ut-seong Joanna Nanyang Technological University, Singapore Sohn, Sung-ock University of California, Los Angeles Tao, Hongyin University of California, Los Angeles Travis, Lisa McGill University, Montreal Tsunoda, Tasaku National Institute for Japanese Language and Linguistics, Tokyo Vovin, Alexander University of Hawaii Watters, David La Trobe University, Melbourne Wee, Lionel National University of Singapore Wrona, Janick University of Oxford Yap, Foong Ha Hong Kong Polytechnic University Zeitoun, Elizabeth Institute of Linguistics, Academia Sinica, Taiwan
List of contributors Delancey, Scott Genetti, Carol Grunow-Hårsta, Karen Gu, Yang Haig, Geoffrey Horie, Kaoru Hsieh, Fu-hui Kaufman, Daniel Kratochvíl, František Lichtenberk, Frantisek Liu, Hongyong Morey, Stephen Matthews, Stephen Nagaya, Naonori Noonan, Michael Post, Mark Potsdam, Eric Rhee, Seongha Shinzato, Rumiko Sio, Joanna Sung, Li-May Tang, Sze-Wing Wang, Jiao Wrona, Janick Xu, Huiling Yap, Foong Ha Yeh, Meili Marie
[email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] Introduction Nominalization strategies in Asian languages Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona† Hong Kong Polytechnic University & Oxford University†
This paper examines a wide range of nominalization strategies found in Asian languages, and identifies robust grammaticalization pathways that trace the life of versatile nominalizers, particularly those that develop from light nouns and noun phrase markers such as classifiers, plural markers, demonstratives, and case markers. It also focuses on the extended uses of nominalization constructions – from referential to non-referential functions, among them adnominal (e.g. relative clause and genitive) marking, tense-aspect-mood (TAM) marking, speaker stance marking, and subordinate adverbial marking. Examples come from Sino-Tibetan, Iranian, Korean, Japanese, Austronesian, and Papuan languages.
1.â•… Introduction Much has been written about nominalization over the past fifty years, and we are now in a better position to consolidate some robust generalizations and to chart some new areas for further investigation. In this volume on Nominalization in Asian Languages, we approach this task from an areal typological perspective, covering a number of language families that span the Asian continent and the great oceans that surround it. More specifically, we first focus on mainland Asia, covering Sino-Tibetan and Iranian languages, then we expand seaward toward the Korean, Japanese and Austronesian (inclusive of Oceanic) languages, with a representation from the Papuan family as well. Future work will clearly need to redress the lack of representation from other language families, including those from South and Central Asia. We hope papers from this volume will serve as a springboard for further investigations into other Asian languages, with a clear eye toward language universals. The typological approach adopted here is not the classic questionnaire method such as was undertaken by Koptjevskaja-Tamm (1993) or Malchukov (2004) for nominalization. Rather we opt instead for a collection of in-depth analyses of specific languages grouped around language families, in order to facilitate cross-linguistic comparison of language phyla, analyses that will bring us closer to a comprehensive

Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
understanding of nominalizers, nominalization constructions and nominalization processes. This approach is also chosen because it allows for diachronic analyses; for example for Chinese, Tibetan, Iranian, and Japanese, which have data of appreciable time-depth. This diachronic approach complements the typological approach, providing insights into how variable, language-specific semantic extensions nevertheless often conform at some macro-level to linguistically-robust grammaticalization pathways. It helps shed light on the origins of nominalizers as well as the nature and direction of the grammaticalization pathways that nominalizers undergo. This introductory chapter will revisit some of the major issues and thorny questions pertaining to nominalization. Primary and critical questions are, first and foremost, “What is nominalization?” and “What constitutes a nominalizer?” The answers to these questions are complicated by the fact that the functions and forms of nominalizers and nominalization constructions are diverse and extended. Nominalizers often extend beyond their core function of deriving nominal expressions. For example, nominalizers may go on to derive adjectives and relative clauses. This is prevalent in Tibeto-Burman and particularly the Bodic sub-group. Moreover across Asian languages nominalization constructions may also be used as adverbial clauses, as well as complement structures and main clauses, and in some languages they have also come to express tense-aspect-mood and stance. There is also considerable formal diversity among nominalizers, within and between languages. Languages may possess a single versatile nominalizer with multiple functions; or languages may possess multiple nominalizers each with specialized functions. Furthermore, there are languages in which nominalized constructions have no dedicated nominalizer; rather they utilize other categories, such as classifiers, plural markers, possessive pronouns, demonstratives, definiteness markers, and case markers to signal the nominal status of a word or construction. This raises another interesting question: “Can such noun phrase markers be considered nominalizers?” The body of this paper is organized as follows. Section 2 defines and describes types of nominalization constructions; in doing so, it helps answer the question “What is nominalization?” Section 3 identifies diverse nominalization strategies including: morphological nominalization, zero nominalization, and the use of noun phrase markers (i.e. substantivization strategies) and marked argument structure to signal the presence of nominalization constructions. Section 4 sheds light on how nominalization constructions develop from referential to non-referential uses. In particular, it looks at the relationships between nominalization and relative clause constructions, between nominalization and tense-aspect-mood (TAM) marking, between nominalization and speaker stance, and between nominalization and subordination. Where possible, it presents a diachronic perspective to changes undergone by nominalizers and nominalization constructions.
Introduction
2.â•… Nominalization types Nominalization in its core sense refers to the process by which we derive nominal expressions (e.g. Comrie & Thompson 1985/2007) – for example, from verbs (e.g. watch > watcher) or adjectives (e.g. narrow > narrowness, narrowing). Clauses may also be nominalized (e.g. awaken the public conscience > awakening (of) the public conscience). Nominalization constructions are often distinguished in terms of the following types: participant vs. event; lexical vs. clausal; embedded vs. non-embedded. These we briefly describe in the following sections, with examples drawn primarily from the Asian languages represented in the present volume.
2.1â•… Participant vs. event nominalization The term ‘participant nominalization’ (Koptjevskaja-Tamm 1993) refers to derived nominal constituents that function as arguments with referential status within a clause. More specifically, participant nominalizations refer to first order ontological entities (e.g. persons, objects, locations) and they assume semantic roles such as agents, patients, locations or instruments, as exemplified in (1) from Cantonese (Sinitic). An event nominalization, on the other hand, refers to the nominalization of an action (i.e. second order ontological entities); hence often also referred to as action nominalization, as illustrated in (2) from Toqabaqita (Oceanic), where the event of planting is nominalized with -aa.1 Cantonese (1) daai3 mou6 ge3 ho2 ji3 jap6 nei4 wear hat nmz can enter ‘the ones who wear hats may enter’ Toqabaqita
(Lichtenberk, this volume: 703)
(2) fasi-laa qoe qana baqu qena ki plant-nmz 2sg genp banana that pl ‘your planting of those bananas’
.╅ In this paper, the transcriptions appear as they are used in the papers of individual authors. In some cases, these are IPA; in others these are a traditional Romanized orthography. Glosses are defined in each of the individual papers.


Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Some constructions are ambiguous between event and participant nominalization interpretations, as seen in (3) from Numhpuk Singpho (Tibeto-Burman), and their proper interpretation is necessarily dependent on context. Numhpuk Singpho (3)
(Morey, this volume: 297)
maam1 thuu5 phaa4 waa1 muu4 n4-ŋaa5 rice pound nmz def also neg-have a. ‘There is no event of rice pounding.’ b. ‘The rice pounding machine is also not here.’
As mentioned earlier, languages differ in the number of nominalizers they employ to express participant and event nominalizations. Some rely on one highly versatile nominalizer, as for example Chantyal (Tibeto-Burman; see Noonan 1997, 2008, this volume), where the nominalizer -wa is used in the formation of agent, patient and event nominals, as seen in (4). Other languages with similarly versatile nominalizers include Cantonese ge3 (see Sio, this volume), Chaozhou kai (see Xu & Matthews, this volume), Mandarin Chinese de, Japanese no, and Korean kes (see Simpson & Wu 2001; see also Horie 1998; Rhee 2008; Yap & Matthews 2008). Chantyal (4) a.
capa ca-wa meal eat-nmz ‘the one who was eating’ ~ ‘the eater’
(Noonan 1997)
(p. 379)
b. mәnchi-sә ca-si-wa gay-ye sya person-erg eat-ant-nmz cow-gen meat ‘the beef that the person ate’
(p. 378)
bәnnu-ye nal tato ta-si-wa gun-gen barrel hot become-ant-nmz ‘The barrel of the gun had become hot!’
(p. 381)
c.
Languages may also have multiple nominalizers, each dedicated to a specific nominalization function. For example, Lhasa Tibetan (DeLancey 1986) has replaced its aging and once-versatile nominalizer -pa with a wave of new and more specialized participant nominalizers: a dedicated agent nominalizer mkhan, patient nominalizer yag (non-perfective contexts), and locative/dative nominalizer -sa.2 Magar (GrunowHårsta, this volume), Nuosu Yi (Liu & Gu, this volume), Dolakha Newar, Dongwang Tibetan, Mongsen Ao and Zhuokeji rGyalrong (Genetti, Coupe, Bartee, Hilderbrandt & Lin 2008; Genetti, this volume) provide further examples of Tibeto-Burman languages with multiple nominalizers. There is also evidence of the rise of newer and more .â•… Nominalizer -pa is still retained for non-agentive referents in perfective contexts (DeLancey 1986).
Introduction
Â� specific nominalizers in other languages. In the case of Japanese, relatively new nominalizers include mono (< ‘thing’), koto (< ‘matter’) and tokoro (< ‘place’) (see Horie 2008). Highly specialized nominalizers which still retain their lexical meaning have also been observed in Korean (Rhee, this volume) and in Galo (Tibeto-Burman) (Post, this volume); these newer (and still semantically transparent) nominalizers have been referred to as ‘secondary nominalizers’ and are often (still) of limited productivity.
2.2â•… Lexical vs. clausal nominalization Genetti (this volume: 164) has observed that “One of the reasons that nominalization is so pervasive in Tibeto-Burman is that it applies at two different levels of the grammar.” That is, nominalization functions on both lexical and clausal levels, a phenomenon also observed across the language families represented in these volumes. Lexical nominalizations have the same morphosyntactic characteristics as nonderived nouns (Comrie & Thompson 1985: 392), as for example in Chantyal (5), where the lexical nominalization (reysi thũ-wa ‘raksi drinker’) bears the genitive case-marker (-ye), or in (6) from Magar where the lexical nominalization (armBuscyo ‘fragrance’) receives plural-marking (-ko). Examples from Toqabaqita contrast lexical and clausal nominalization; in (7a) the deverbalizer (i.e the nominalizer) -a results in a lexical noun; in (7b) the nominalizer -la scopes over an entire clause. Example (7b) also demonstrates that clausal nominalization may be void of nominal features but may exhibit verbal ones such as person-number agreement. Furthermore, example (8) from Magar shows that nominalizations may additionally encode tense-aspect-mood. Chantyal
(Noonan 2008: 224)
(5) na-sә reysi thũ-wa-ye naku khway-kәy mu i-erg raksi drink-nmz-gen dog feed-prog be.n.pst ‘I’m feeding the raksi-drinker’s dog.’ Magar
(Grunow-Hårsta, field notes)
(6) ho-sa-i armBus-cyo-ko seB-cyo ale d.dem-def-foc smell-nmz-pl good-nmz cop ‘These fragrances are beautiful.’ Toqabaqita (7) a.
uusi-a buy-dvn ‘market’
b. uusi-la-na suka buy-nmz-3sg.pers sugar ‘(the act of) buying sugar’, ‘buying of sugar’
(Lichtenberk, this volume: 687)


Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Magar
(Grunow-Hårsta, this volume: 223)
(8) ŋa lasargBa-aŋ ŋa-mu-o=le-a-aŋ 1s Lasargha-loc 1pro-sit-hab.nmz=impf-pst-1pro ‘I used to live in Lasargha.’
Example (9) from Korean, in which the nominalization construction with ilk ‘read’ is marked for both tense and for case, demonstrates (as observed by Comrie & Thompson 1985/2007; Koptjevskaja-Tamm 1993, 2003; Malchukov 2006; inter alia) that nominalization constructions frequently exhibit a mix of both nominal and verbal features. Korean
(Yoon 1991; cited in Sung, this volume: 543)
(9) Chelswu-ka chayk-ul ppalli ilk-ess-um-i pwunmyengha-ta Chelswu-nom book-acc quickly read-pst-nmz-nom evident-decl ‘It is evident that Chelswu read the book quickly.’
Post (this volume: 265), in his analysis on Galo, identifies two distinct sets of nominalization constructions derived from clauses, which he terms “nominalized clauses” and “clausal nominalizations”. The former have the syntax of a noun phrase, and express event nominalizations, noun complements and relative clause constructions. The latter resemble predicative clauses in that they retain verbal features such as tenseaspect-mood marking. They frequently occur as subordinate clause constructions for framing and backgrounding functions. They also often occur as cleft and stand-alone constructions, with focus and exclamative effects respectively. An example of a nominalized clause is shown in (10a), with hocә´r àp- nám ‘stag-shooting’ accompanied by a noun phrase marker in the form of second person possessor nó-kә`. An example of a clausal nominalization is shown in (10b), with pá-tà-rә´-nà ‘going to chop/sacrifice’ retaining tense-aspect-mood features in the form of irrealis marker -rә´ and used as a nominalized clause within a cleft-like focus construction. Galo (10) a.
(Post, this volume)
nó-kә` hocә´r ôàp- nám gó mèn-bә`ә-dùu 2.sg-gen deer shoot-nzr:rls=ind say-hab-ipfv ‘(You’re) always talking about your stag-shooting (incident).’
b. ŋó dór-rĭi pá-tà-rә´-nà = ´ә 1.sg clf:high.animal-ten chop-incp-irr-nzr:sub=cop ‘(It is the case that) I am going to sacrifice ten (mithuns)!’
(p. 267)
(p. 275)
An analysis of lexical and clausal nominalization raises the issue of diachronic priority. Lexical nominalization is often presumed to be historically prior to clausal nominalization. In Tibeto-Burman, the nominalizer pa ( ‘apparently’, ‘I guess’, ‘I suppose’, etc. in sentences such as agak-nya bahaya juga ‘I guess it’s quite dangerous after all’). Below we analyze these various non-referential uses of nominalization constructions seen in our language samples.
4.1â•… Nominalization and relativization Multiple studies have shown that there is an intimate relationship between nominalization and relativization. In numerous Sino-Tibetan languages as well as in Japanese and Korean, the same morpheme used to construct nominalizations is also used to construct relative clauses (Bickel 1999; DeLancey 1999, 2005; Genetti 1992, this volume; Genetti et al. 2008; Horie 1998; Matisoff 1972; Noonan 1997, 2008; Rhee 2008; Shibatani 2009; Simpson 2008; Yap & Matthews 2008; Zeitoun 2002 and papers therein). This is not an historical accident, as we will proceed to elaborate in this section. Essentially, many nominalization constructions develop their nominalizing morpheme from erstwhile (light) nouns, which themselves were at an earlier time lexical nouns juxtaposed to modifying expressions that were reanalyzed as relative clauses (see LaPolla 1994). For this reason, these light-noun-based nominalization
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
constructions often display considerable structural parallelism with lexically-headed relative clause constructions, as schematized in Figure 1.20 In some cases the light nouns can be traced back to full lexical nouns. a.
[_i VP]i Ni = lexically-headed relative clause
b.
[_i VP]i ni = nominalization construction (or ‘light-headed’ RC); its head noun (n) often of light noun origin
Figure 1.╇ Structural parallelism between some nominalization and relative clause constructions
Yap and Wang (this volume) distinguish these two types of constructions in terms of ‘headed relative clauses’ versus ‘light-headed relative clauses’ (indicated by [VP] N vs. [VP] n). The latter forms the basis of a nominalization construction headed by a light noun or nominalizer and is often seen in Sino-Tibetan languages, as well as in Korean and Japanese. In a sense, a nominalization construction is often a relative clause with a very general or grammaticalized head noun (see also Aldridge 2008; Shibatani 2009). The light noun or nominalizer may, in turn, be further reanalyzed as a relativizer when the nominalization construction combines with a lexical noun to form a new, secondary-type relative construction (see Genetti 1991, this volume; Genetti et al. 2008; LaPolla 1994, 2008). Among our language samples, Japanese, Korean and Sino-Tibetan languages show a strong propensity for recruiting nominalizers from general nouns, hence resulting in a lexical noun (N) becoming a light noun or nominalizer (n). This is illustrated in Table 1. For example, Middle and Modern Korean nominalizer kes is derived from a general noun meaning ‘thing’ (Rhee 2008). The nominal status of kes is still semantically transparent, and [VP] kes type nominalization constructions resist juxtaposition with another lexical noun; not surprisingly, nominalizer kes has not been reanalyzed as a relativizer. Similar developments involving light nouns being reanalyzed as nominalizers are attested in Japanese, involving general nouns such as mono ‘thing’ and tokoro ‘place’ (Horie 2008), and in Lhasa Tibetan, involving general nouns such as mkhan ‘person’ and sa ‘place’ (DeLancey 1986).
.â•… We use the term ‘primary relative clause constructions’ to refer to appositive-type relative clause constructions, and we use the term ‘secondary relative clause constructions’ to refer to externally-headed relative clause constructions whose relativizers are also known to have nominalizing functions. These terms tend to implicitly suggest derivation of the latter from the former. This point is made explicit using examples of Classical Chinese zhe in the discussion that follows (see also Yap & Wang, this volume).
Introduction 
Table 1.╇ Examples of nominalization and relative clause constructions from various Asian languages (with diachronic implications) Stage 1
Stage 2
Stage 3
Lexical noun
Light noun
Nominalizer Relativizer
Old Korean kes ‘thing’
kes ‘thing, animal’
Middle & Modern Korean kes ‘thing’
VP kes attachable kes ‘thing, animal’; attachable to demonstratives and pronouns, e.g. ku-kes ‘that thing, that one, that’, na-uy kes (‘my thing, my one, mine’)
Stage 4
*[VP kes] N
Pre-modern Japanese mono ‘thing’ tokoro ‘place’
mono ‘thing, animal, person’ tokoro ‘place’
VP mono VP tokoro
Modern Japanese mono ‘thing’ tokoro ‘place’
mono ‘thing, animal, person’ tokoro ‘place’
VP mono VP tokoro
*[VP mono] N ?[VP tokoro] N Possible calque-ing from Chinese [VP de] N
mkhan ‘person’ sa ‘place’
VP mkhan VP sa
*[VP mkhan] N *[VP sa] N
zhe ‘person, thing, time, situation, etc.’
VP zhe
Pre-modern Lhasa Tibetan mkhan ‘person’ mkhan ‘person’ sa ‘place’ sa ‘place’ Modern Lhasa Tibetan mkhan ‘person sa ‘place’ Old Chinese zhe (etymology unknown) Middle Chinese zhe (lexical use no longer productive) Middle & Modern Chinese di ‘bottom’ di ‘bottom’; interrogative pronoun ‘what, how, etc.’
VP zhe
[VP zhe] N
VP di VP de
[VP ╛↜di] N [VP de] N
Note: *= ungrammatical form; ?? = variable judgment across native speakers and limited productivity
As highlighted in Stages 3 and 4 in Table 1 above, evidence from Middle Chinese shows that the [VP] n nominalization construction can sometimes develop into a secondary relative clause. This development is diachronically highlighted in Table 2 below, where we see a light noun or nominalizer (n) grammaticalizing into a relativizer (REL).
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Basically, when a nominalizer of lexical noun origin is highly grammaticalized as in the case of Middle Chinese zhe, the nominalization construction (e.g. VP n) can be apposed to another lexical noun (N). The resulting apposition, namely [VP n] N, facilitates �reanalysis of the (VP n) nominalization construction into a secondary-type relative clause construction, yielding [VP RELn N], with the appositive lexical noun reinterpreted as the relative head noun and the nominalizer as a relativer or linking element.21 Table 2.╇ Recruitment of nominalization constructions in the formation of secondary relative clause constructions in Middle Chinese (see Aldridge 2008 and Yap & Wang, this volume) Stage 1
Stage 2
Stage 3
Old Chinese
Early Middle Chinese
Late Middle Chinese
(i) DEM NUM [n╇╖] DEM NUM [zhe] Zhe as lexical noun fan ci wu zhe all these five ZHE ‘all these five items’ (Xun Zi) (ii) [VP] zhi n [VP] zhi zhe Zhe as head noun in relative clause qi wei yun zhi zhe those call fall REL ZHE ‘those so called falling things’ (Lun Heng)
(iii) [VP n╇╖] [VP zhe] Zhe as agent nominalizer [zhi yu] zhe control carriage NMZ ‘the one driving the carriage’ (Lun Yu)
(iv) [VP n╇╖] zhi N [VP zhe] zhi N Zhe-nominalization in relative clause with relativizer zhi yu [you sang zhe] zhi ce at be.in.mourning NMZ REL side ‘at the side of someone who is in mourning’ (Lun Yu)
> (v) [VP nâ•… ] N > [VP zheâ•–] N Zhe-nominalization in relative clause without zhi [ding yin zhe] _ [jiangshi] settle Yin NMZ general ‘the ones who are settled in Yin, (those) generals’ (Shi Ji)
> (vi) [VP REL N] > [VP zhe╇╛N] Reanalysis of zhe as relativizer [zuori lai zhe taishiguan] yesterday come REL teacher ‘the teacher who came yesterday’ (Yi You)
.╅ A parallel development can also be seen in the case of Modern Chinese di ~ de (see Table 1 above).
Introduction 
As seen in Table 2 above, as a light noun at Stage 1 in Old Chinese, zhe displayed nominal properties such as the ability to take noun phrase markers (e.g. demonstratives and numerals), as highlighted in (i) above. Light noun zhe at this stage could also be modified by a relative clause, either with or without relativizer zhi, as seen in (ii) and (iii) respectively. With frequent direct juxtaposition with the relative clause (especially in contexts without relativizer zhi as a linker), light noun zhe was reanalyzed as a nominalizer. Clauses nominalized by zhe could also be apposed to lexical nouns, yielding either a genitive construction or a secondary-type relative clause, initially with genitive/relativizer zhi as linker, as in (iv). With the decline of genitive/relativizer zhi from Late Old Chinese onward, however, nominalizer zhe came to be reinterpreted as a relativizer in Middle Chinese, as seen in the transition from Stage 2 to Stage 3, as seen in the reanalysis form (v) to (vi). Relativizer zhe however has not survived into Modern Chinese. See Aldridge (2008) and Yap and Wang (this volume) for a more detailed diachronic account.22 The above scenario may help resolve some of the debates in Tibeto-Burman literature surrounding the direction of grammaticalization – namely, whether semantic extension is from relativization to nominalization (e.g. LaPolla 1994) or from nominalization to relativization (DeLancey 1986, 2005; Noonan 1997 and this volume). Essentially, it supports the position as posited by Genetti (this volume; see also Genetti et al. 2008) that nominalizations may arise from, but can also give rise to, relative clauses. Note however that the grammaticalization pathways linking nominalization and relativization constructions are not necessarily bidirectional. The relativization-to-nominalization account corresponds to early developments from Stage 1 to Stage 2 in Table 2 above, while the nominalization-to-secondary-relativization account corresponds to later developments from Stage 2 to Stage 3. Many old nominalizers such as Classical Tibetan -pa and its cognates in various Tibeto-Burman languages appear to have likewise run the full course from lexical noun → nominalizer → relativizer.23 Newer nominalizers within the Tibeto-Burman language family, which are still semantically transparent, such as Lhasa Tibetan agent nominalizer mkhan and locative nominalizer sa as noted in Table 1 earlier, may still be in the early phases between Stages 1 and 2, and unlike -pa, have not (yet) extended to the secondary relativization process between
.â•… It is also possible for demonstratives to be reanalyzed as a relativizer, as attested in the case of Old Chinese demonstrative zhi (see Yue 1998; Shi & Li 2002). Reanalysis takes the form of [VP] [zhi N] > [VP zhi N]. .â•… Interestingly, Classical Tibetan showed evidence of a [VP-pa] ‘i N relative clause construction, with adnominal (i.e. genitive and relativizer) ‘i linking the [VP-pa] nominalization clause to the lexical head noun (N). This structure parallels the Old Chinese [VP zhe] zhi N relative clause construction found in Stage 1 in Table 2 above, where adnominal (i.e. genitive and relativizer) zhi also links the [VP zhe] nominalization construction to the lexical head noun (N).
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Stage 2 to Stage 3. The same can be said for Japanese nominalizers mono and tokoro, and Korean kes, among others. An analogous situation may hold for Austronesian languages as well. Li-May Sung has observed a similar development for the Formosan language Budai Rukai, where relative clause constructions emerge as secondary developments from nominalization constructions.24 Closer analysis reveals some interesting additional insights. As seen in the Budai Rukai example in (62) below, nominalization constructions such as ni-sinav-anә ‘the ones that got washed’ can be apposed to another noun phrase with more explicit reference such as kai laymay ‘these clothes’. Such apposition easily facilitates reinterpretation of the nominalization construction as a relative clause, and the appositive noun phrase as the head noun. Interestingly, in many Austronesian languages including Budai Rukai, it is not the affixal nominalizer (e.g. -anә) but rather the noun phrase marker (e.g. case marker ka/ku or demonstrative kai) which accompanies the lexical head noun that is reanalyzed as the relativizer.25 (63) below provides a similar example from Saisiyat, where case marker ka which accompanies the lexical head noun is reanalyzed as the relativizer.26 Budai Rukai
(Sung, this volume: 534)
(62) saŋu-a-әsai [ka ni-sinav-anә ka laymay] smell-fin-fragrant nom ni-wash-nmz rel clothes ‘The clothes that got washed smell good.’ Saisiyat
(Yeh, this volume: 570)
(63) [ka-hiwa:-en rim’an ka waliSan] rae’iw ila ka-cut-pv tomorrow lig boar run.away incpt ‘The boar that is going to be killed tomorrow has run away.’
In Budai Rukai, reanalysis of a case marker or demonstrative as a relativizer is found not only with pre-nominal relative clauses as seen in (62) above, but also in post-nominal relative clauses as seen in (64) below. Not surprisingly, in the post-nominal relative clause, it is the noun phrase marker preceding the nominalization construction (now reanalyzed as the relative clause) that is reanalyzed as the relativizer.27 .â•… Personal communication (April 2010). .â•… Diachronically, it is possible that case marker ka may have been recruited as a relativizer first, with demonstrative kai belonging to a later wave of noun phrase markers and thus developing a relativizer function via analogy to case marker ka. Interestingly, as noted in Ross (2006), case markers such as ka were themselves derived from old demonstratives. .â•… The term ‘ligature’ (LIG) in (63) is often used in Austronesian studies; it encompasses a wide range of functions, including relativizer and genitive. .â•… In both pre-nominal and post-nominal relative clauses, if there are multiple noun phrase markers (e.g. case marker ka/ku and demonstrative kai), one noun phrase marker is
Budai Rukai
Introduction 
(Sung, this volume: 534)
(64) ma-Tas-aku kanә [ka lacәŋә [ka ta-aga-anә stat.fin-hate-1sg.nom nfin.eat obl vegetable rel nfut-cook-nmz ki TaiTai ]] gen lailai ‘I hate to eat the vegetables that Lailai has cooked (because they did not look good the way that Lailai cooked them).’
An interesting question is whether recruitment of the noun phrase marker as relativizer was triggered by the suffixal status of nominalizer -anә. That is, as a post-predicate nominalizer (possibly originally enclitic), -anә could have been reanalyzed as a perfective/irrealis tense-aspect-mood marker, and hence was no longer available for reanalysis as a relativizer. Such a development would explain the greater tendency among Austronesian languages to recruit noun phrase markers (e.g. case marker ka/ku and demonstrative kai) as a relativizer, while Japanese, Korean and Sino-Tibetan languages are more inclined to recruit light nouns as nominalizers, and in some cases as relativizers as well.
4.2â•… Nominalization and tense-aspect-mood marking Another important development in the non-referential uses of nominalization constructions is their reanalysis as finite clauses, with the morphological nominalizer being reinterpreted as a tense-aspect-mood (TAM) marker. Data from several papers in this volume provide ample evidence that nominalizers and nominalized constructions can come to signal tense, aspect and mood. This development is frequent in Tibeto-Burman languages, as recounted in DeLancey (this volume) and GrunowHårsta (2009 and this volume; see also Grunow-Hårsta & Yap 2009, 2010 and Yap & Grunow-Hårsta 2010). In Korean, adnominals also express tense, aspect and mood – for example, adnominals -n and -(u)l, which appear to be linked to erstwhile nominalizers -n and -l respectively, have evolved into anterior/past and prospective markers (e.g. Lee 1993; Whitman 1997; Rhee 2008 and this volume; Yap & Matthews 2008). The Japanese rentaikei (non-finite) forms that are often associated with nominalization constructions have also extended into contexts where they have replaced the shushikei (‘conclusive’ or finite) forms (e.g. Iwasaki 2000; Horie 2008; Shinzato, this volume). Among Sinitic languages, it has also been noted that nominalizer de in some northern Mandarin dialects has been reinterpreted as a past tense marker (e.g. Simpson & Wu 2001). Southern Chinese dialects such as Cantonese, however, do not manifest this recruited as the relativizer, and often another is reinterepreted as the noun phrase marker for the entire relative clause construction. Other noun phrase markers such as ki are also often used to signal oblique arguments (i.e. non-focus arguments).
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
phenomenon. Interestingly, Tang (this volume) provides morphosyntactic arguments that suggest the nominalizer ge3 in Cantonese has not grammaticalized as far as Mandarin de, and in particular Cantonese ge3 has not developed the wide range of gerundive constructions attested in the case of Mandarin de. In many Austronesian languages, focus/voice markers also often convey tense-aspect-mood information. For example, in Cebuano (a Philippine language), realis/non-realis distinctions are made between actor voice markers mi-/mu-, patient voice markers gi-/-un, conveyance voice markers gi-/i-, among others (e.g. Himmelmann 2005: 126, 168). Crucially for our present discussion, Starosta, Pawley and Reid (1982) have suggested that some of these voice markers are derived from affixal nominalizers.28 In Formosan languages such as Saisiyat, prefix ka has both nominalizing and future/non-realis marking functions (Yeh, this volume).29 Below we examine more closely this fairly robust nominalizer-to-TAM marker development with examples from Magar. In this Tibeto-Burman language, all of the suffixal nominalizers have come to function within the TAM system, as illustrated in (65). Specifically, -o has developed into a past habitual aspect marker, -m% into a continuous aspect marker, -cyo into a mirative mood marker, and -ke into an imminent aspect marker and a future tense marker (Grunow-Hårsta 2009; Yap & Grunow-Hårsta 2010). Magar (65) a.
(Yap & Grunow-Hårsta 2010)
ŋa lasargBa-aŋ ŋa-mu-o=le-a-aŋ 1s Lasargha-loc 1pro-sit-hab.nmz=impf-pst-1pro ‘I used to live in Lasargha.’
b. ŋa lasargBa-aŋ ŋa-mu-m=le-a-aŋ 1s Lasargha-loc 1pro-sit-cont.nmz=impf-pst-1pro ‘I was living in Lasargha.’ c.
ram lasargBa-aŋ raB-cyo Ram Lasargha-loc come-nmz ‘Ram has come to Lasargha! (to my surprise)’
.â•… See also Kaufman (2009) for an interesting discussion defending the nominalist hypothesis. .â•… From the Iranian language family, the Bahdini Kurdish “ezafe” (derived from a demonstrative) has in modern times also developed a TAM marking function, more specifically as a present perfect marker (see Haig, this volume). This development has not been reported for other Iranian languages such as Persian. Interestingly enough, “ezafe” nominalization and relative clause constructions are attested in Bahdini Kurdish but not in Persian, prompting again the question of whether there is a link between nominalization/relativization constructions and the potential for reanalysis as a tense-aspect-mood marker. Among Papuan languages, Abui also possesses demonstratives that have developed evidential functions. Abui demonstratives can signal nominalization, but as reported in Kratochvil (this volume), these demonstratives have not been reanalyzed as nominalizers. This would suggest that there may be multiple paths to the emergence of sentence final stance/mood particles.
Introduction 
d. ram lasargBa-aŋ raB-ke=le Ram Lasargha-loc come-nmz=impf ‘Ram has yet to come to Lasargha.’
The nominalizer -ke may provide further insight into how a morpheme which marks event nominalizations (a referential element) can come to serve a non-referential function.30 Specifically, this nominalizer is now reanalyzed as a tense-aspect marker, as illustrated in (66). Magar
(Yap & Grunow-Hårsta 2010)
(66) a.
kan-ko jBurum-ke ŋeB=le 2.pl-pl assemble-nmz beg-impf ‘We begged to assemble.’
b.
kan-ko jBurum-ke=le 2.pl-pl assemble-nmz=impf ‘(It is the case that) we are yet to assemble’ ~ ‘(It is the case that) we will assemble.’
c.
kan-ko jBurum-ke 2.pl-pl assemble-nmz ‘We have yet to assemble’ ~ ‘We will assemble.’ ~ ‘Will we assemble?’31
Nominalized constructions with -ke can be supported by the imperfective copula (impf) as in (66a) and (66b), or the copula may be elided, resulting in a free-standing nominalization as in (66c). With elision, the nominalizer -ke alone signals tense and aspect. A similar development has also been observed in Mongsen Ao (Coupe 2009), with evidence of syncretism between agentive nominalizer -ә (mid tone) and present tense marker -ә (low tone), as well as between purposive complementizer -u (mid tone) and future tense marker -u (low tone).32 This development from a nominalizing to a tense marking function supports DeLancey’s position (this volume) that nominalized clause constructions (often supported by a copula) are a major source of finite main clauses in Tibeto-Burman, a position which he demonstrates with diachronic evidence from Kuki Chin, Modern and Classical Tibetan, Lolo-Burmese and Sunwar (Kiranti).
.â•… The Magar nominalizer -ke may have been derived from the dative case marker, with which it is homophonous (Grunow-Hårsta 2008). .â•… Question-answer interchanges are generally stand-alone nominalizations in Magar. .â•… There is also evidence of isomorphism between nominalizer -pa (and its allomorphs) and future or irrealis marking in a number of Tibeto-Burman languages, among them Bodish languages such as Manange (e.g. Hilderbrandt 2003; see also Genetti, this volume) and languages of Nagaland such as Yimchungrü (Coupe 2009).
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
It is possible, however, for a nominalizer to develop into a TAM marker without the mediation of a copula. This is attested even among Tibeto-Burman languages, and not surprisingly in languages without copulas as well. For example, Austronesian languages, which are known to generally lack a copula, frequently show morphosyntactic indetermination (i.e. underspecification or ambiguity) between a nominalization and a finite clause interpretation (e.g. Starosta, Pawley & Reid 1982; Ross 1995, 2002; Blust 2002; Himmelmann 2005). Disambiguation is often dependent on context. For example, as seen in Saisiyat (Yeh, this volume), in (67a) below, it is possible to construe an identity relation between hini’ ka-i:ba:-en ‘this dress’ and ma’an rim’an ka-i:ba:-en ‘what I am going to wear’, but in (67b), no identity relation is possible between hiza’ ka-ma-manra:an ‘that man’ and ka-ma-’omalop ka waliSan ‘the boar that hunts’. That is, it does not make sense to interpret (67b) as ‘the man is the boar that hunts’. A more natural interpretation for (67b) is a specificational (not identity) reading, namely, ‘the man is one that hunts boars’, which is reducible to ‘the man hunts boars’. This shift from referential to specificational use of the nominalization construction paves the way for the rise of finite clauses. Semantic changes are often accompanied by morpho-syntactic changes, and the finite clause interpretation is accompanied by reanalysis of nominalizer ka- as a non-realis mood marker (in this case, with a habitual reading). The use of accusative ka (instead of genitive ni) preceding the object complement waliSan ‘boar’ is also indicative that the nominalization construction ka-ma-’omalop ka waliSan has lost some nominal properties and is acquiring instead some verbal ones. In other words, the interpretation of (67b) clearly means ‘That man hunts boars’ and not ‘That man is (in an equative sense) the boar that hunts’ nor ‘That man is (again in an equative sense) the hunting of the boars’. Saisiyat
(Yeh, this volume: 571)
(67) a.
hini’ ka-i:ba:-en [ma’an rim’an ka-i:ba:-en] this nmz-wear-pv 1sg.gen tomorrow nmz-wear-pv ‘This dress (is) what I am going to wear tomorrow.’
b.
hiza’ ka-ma-manra:an [ka-ma-’omalop ka waliSan] that nmz-av-walk nmz-av-hunt acc boar ‘That man always hunts boars.’ (Lit. ‘That man is one who hunts boars’)
There is also suggestive evidence that many Austronesian languages similarly reanalyze suffixal nominalizers as TAM markers. As noted earlier, Starosta, Pawley and Reid (1982) have posited that certain affixes in nominalization constructions have come to be reinterpreted as focus/voice markers. These markers have long been known to also express tense, aspect and mood (e.g. Himmelmann 2005). As seen from transcriptions of Saisiyat conversations in (68) below (examples taken from Huang & Su 2005), the verbal suffix -en can be used as patient focus (PF) marker
Introduction 
as well as resultative/anterior/perfective aspect marker, and this suffix has variant phonological forms such as -oen, -on, and -in. Saisiyat (68) a.
(Huang & Su 2005)
ka boway ’aewpoe’-oen ’ara∫-en. nom fruit hold-pf take-pf ‘The fruits are taken away.’
(p. 351)
b. ka tataa’ tono’-on noka kapapama’an m-asay ila. nom chicken bump-pf gen vehicle af-die pfv ‘The chicken was bumped by a car and died.’ c.
nisia ’aehae’ kala’ sisil-in ’a∫kan-en ray kapapama’an. 3sg.gen one basket lift-pf put-pf loc vehicle ‘He lifted one basket and placed (it) on (his) bicycle.’
(p. 354)
(p. 352)
Saisiyat suffix -en may thus have evolved via the following pathway: Nominalizer -en → Focus/voice marker -en → Resultative/Anterior/Perfective -en/-oen/-on/-in. Saisiyat also has a verbal affix , which is often used as a resultative, anterior or perfective/past marker, as seen in (69). Saisiyat
(Huang & Su 2005: 348)
(69) onoka-aewha∫-a kiko m-in-potoeh ila. poss-rat-poss tail af-past-break pfv ‘The tail of the rat divided into two.’
Yeh (this volume), citing cross-linguistic parallels identified in Bybee, Pagliuca and Perkins (1994), posits that affixal perfective/past marker emerged via the following grammaticalization pathway: resultative > anterior (i.e. perfect) > perfective/past. As noted earlier in (21c) in §3.1, verbal affix is also used to derive lexical nominalizations (e.g. ka:at ‘book, letter’ (< ‘that thing that results from writing’). Yeh further posits that affixal resultative is the precursor not only of perfect(ive)/past affix but also of derivational nominalizing affix as well, the latter used for deriving patient nominals. Yeh captures these polygrammaticalization pathways as follows: Resultative → Anterior → Perfective/Past → Nominalizer used for deriving patient nominals
The grammaticalization pathways for both suffix -en/-oen/-on/-in and affix bear some strong similarities. An intriguing question of course is whether there is more than just functional relatedness between the two sets of affixes, more specifically whether there is not some formal relatedness in diachronic terms, although it is also equally possible that affix may belong to an older wave of patient focus markers that were likewise recruited to serve TAM functions. At any rate, the link between patient focus and perfective aspect is widely attested among Austronesian languages
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
(see for example Reid 1992 for similar observations in Philippine languages such as Ilokano and Tagalog).33 In sum, we have seen that in some languages nominalization constructions develop non-referential functions in which nominalizers are reinterpreted as tenseaspect-mood (TAM) markers. Reanalysis of nominalizers into focus/voice markers has received some attention in Austronesian studies through the works of Starosta, Pawley and Reid (1982), and Ross (1995, 2002), inter alia, while reanalysis of focus/ voice markers into TAM markers has been noted in Himmelmann (2005) and Yeh (this volume), among others. There is now also growing attention to this phenomenon in other language families, among them Sino-Tibetan (represented in this volume through the works of Scott DeLancey, Karen Grunow-Hårsta, and to some degree that of Michael Noonan). Languages such as Iranian (Haig, this volume) and Abui (Kratochvíl, this volume) also suggest telling relationships between demonstratives and the development of non-referential uses of nominalization constructions, including the TAM functions. More research is needed to better understand the close relationship between nominalization and TAM marking. As intimated through our discussions in this section, it would be particularly interesting to investigate if nominalizers derived from demonstratives, given their inherent deictic values, are more amenable than those derived from light nouns to be reanalyzed as tense markers.
4.3â•… Nominalization and stance marking In Tibeto-Burman languages, nominalized constructions also frequently express speaker stance, specifically mirativity (e.g. Noonan 1997; DeLancey 1986, 1997, 2001; Saxena 2000; Grunow-Hårsta 2007), which as defined by DeLancey (1986) as the expression of surprise, amazement or counter-expectation, as seen in (70) and (71). Chantyal
(Noonan 2008: 380)
(70) aay, kattay talay tha-i nә a-tha-wa tane gosh definitely cut-ant foc neg-cut-nmz affirmation ‘Gosh, it didn’t even cut, right!’
.â•… Reid (1992: 68) identifies affix in Tagalog as a completive aspect marker that previously was a derivational affix that “derived nouns that were the result of the action of the verb” (e.g. bili ‘something that was bought’; bili ‘buy’), and he identifies suffix -in with patient focus suffix -en that historically “derived nouns that would receive the action of the verb” (e.g. bilihin ‘something to be bought’). Reid’s suggestion that the two affixes and -in are historically related is partly supported by the fact that these two affixes cannot co-occur (Starosta, Pawley & Reid 1982: 162–163).
Introduction 
Magar
(Grunow-Hårsta & Yap 2009)
(71) ram-e sita-o rBa sat-cyo Ram-erg Sita-gen goat kill-mir.nmz ’Ram killed Sita’s goat!’ (to the speaker’s surprise)
(72) exemplifies the syncretism and versatility of -cyo, specifically how a nominalization construction may express either a predicate nominal or a predicate adjective (72a), as well as a mirative expression (72b), the latter as a non-embedded construction (i.e. unsupported by a copula). This array of functions suggests an extension from referential to specificational to stance usage. Magar (72) a.
(Grunow-Hårsta & Yap 2009)
ho-se seB-cyo ale d.dem-def be.beautiful–att.nmz cop ‘She is a beauty.’ ~ ‘She is beautiful.’
b. ho-se seB-cyo d.dem-def be.beautiful–att.nmz ‘She is beautiful!’
The use of a nominalizer as a mood or stance marker is not exclusive to Tibeto-Burman languages. As seen in (73), in Budai Rukai (Austronesian: Formosan), the nominalization construction signaled by suffixal nominalizer -anә in a non-embedded construction also conveys strong speaker stance. In Cantonese (Sinitic), nominalizer ge3 can be reanalyzed as a ‘sentence final particle’ to convey speaker attitude, as in (74). Similar developments have also been observed in Japanese and Korean (e.g. Horie 2008; see also Horie, this volume, and Rhee, this volume ). Budai Rukai
(Sung, this volume: 543)
(73) a. A: a-su Tumay ki saTabu? why-2sg hit obl saTabu ‘Why did you hit Salabu?’ B:
lisi-anә-li (inianә)! angry-nmz-1sg.gen (3sg.obl) ‘I am (very) angry (at him)!’ (Lit: ‘My anger (at him)!’)
b. A: ani kamani ka muni ka mu-a-cu-cubungu? how.come.3sg kamani nom muni rel go-fin-red-front ‘How come it is Muni who went on stage?’ B: θarir-anә-ini! beautiful-nmz-3sg.gen ‘She is beautiful!’ (Lit. ‘Her beauty!’)
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Cantonese
(Sio, this volume: 126)
(74) keoi5 wui5 lei4 ge3 3sg will come ge ‘(Don’t worry) He will come.’
Parallel to their development as TAM markers, as discussed earlier in §4.2, noun phrase markers, which can be used to signal nominalization constructions, can also come to express stance. This is exemplified by the third person singular enclitic =nya in Malay, as seen in (75). The demonstratives of Abui (Papuan) likewise can contribute to a stance interpretation, as in (76). Malay (75) (Begitu) cantik=nya! So (< ‘like that’) beautiful=3sg.gen ‘(He/She/It is) so beautiful!’ Abui
(Kratochvíl, this volume: 775)
(76) hai, ni kul yaar to! intj 1pl.e must go.cpl prx.ad ‘Oh, we really must go (as you can see/understand).’
It is not fully clear to us, however, if all these noun phrase markers need to first develop a nominalizing function before acquiring a stance-marking function. Kratochvíl (this volume) in particular maintains that Abui demonstratives did not grammaticalize into nominalizers, arguing instead that some noun phrase markers can directly develop into stance/mood markers. There appears to be some support for his argument from Austronesian languages such as colloquial Malay/Indonesian, as seen in (77), where the post-nominal demonstratives ’ni (< ini ‘this’) and ’tu (< itu ‘that’) serve as sentencefinal mood particles without developing a nominalizing function.34 Colloquial Malay/Indonesian (77) dia macam nak nangis ’ni (*ini)/’tu (*itu). 3sg be.like want cry prt ‘He’s about to cry, I tell you.’
In the foregoing examples, stance has been expressed by sentence-final particles, which occupy a right-periphery syntactic position that easily allows a stand-alone nominalization construction to be grounded via illocutionary force. In other words, a nominalizer or a noun phrase marker in sentence-final position is in an ideal position
.â•… Note that only the reduced form of the demonstratives are used as stance markers; the full forms ini ‘this’ and itu ‘that’ convey spatial deictic meaning and are not used to express evidentiality, epistemicity or speaker assessment/attitude.
Introduction 
to host prosodic cues that reflect the speaker’s stance, and in the process is itself often reanalyzed as a speaker mood marker.35 Stance however can also be signaled by ‘non-sentence-final’ particles. This is evident in Malagasy where demonstrative izany appears in pre-nominal (rather than post-nominal) position, as in (78) below.36 Malagasy (78)
(Potsdam, this volume: 667)
izany sotrohiny dem drink.pass.3sg Lit. ‘that amount/stuff drunk by him!’ ‘The amount/stuff he drinks!’/‘The amount/stuff he’s drinking!’
Stand-alone nominalization constructions with ‘non-sentence-final’ particles are also evident in other Austronesian languages, among them Buru, as seen in (79), where =an serves as an emphatic clitic which gives pragmatic prominence to the constituent to which it is attached (Grimes 1991; cited in Himmelmann 2005: 133). Buru
(Austronesian; Grimes 1991: 193)
(79) Tawe, yako=an naa te=keha moo. friend 1s=foc prx able=ascend neg ‘Friend, it is me here who can’t climb (the tree).’
Similarly, in Tibeto-Burman languages, nominalizers can serve also as focus markers (e.g. Bickel 1999; Grunow-Hårsta 2008) and in these instances they may also be non-final, as in mi-ja-c% in (80). Magar
(Grunow-Hårsta 2008: 188)
(80) ho-ta-i jogi-e j% men-o mi-ja-c% d.dem-foc yogi-erg emph 3-gen poss-child-foc.nmz nunB-o=le-a ta take-mir.nmz=impf-pst rep ‘They say, then, like that, indeed, the yogi took her child!’
.â•… As has often been noted in the literature, the right periphery is a natural site for hosting sentential stress or marked illocutionary force (e.g. Stowell 1981; Radford 1990; Barton & Progovac 2005; Kaufman 2005; Tang 2007). .â•… Potsdam (this volume) points out that some Malagasy exclamatives are in fact standalone relative clauses, as highlighted in the example below. The izany-type exclamatives in (78) above has an elided head noun (e.g. a general noun such as zavatra ‘thing’), but the nominalization construction (or ‘headless relative clause’) can of course still be signaled by noun phrase markers such as izany. (i)
izany zavatra [relative clause sotrohiny] ! dem thing drink.pass.3sg ‘the things that he’s drinking!’
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
In addition to stand-alone nominalization constructions signaled by nominalizers and noun phrase markers, those signaled by genitive subjects are also often recruited to express stance functions, as is often found across Austronesian languages, as in Tagalog (81), Botolan Sambal (82) and Malay (83). Note also the frequent use of nominative case markers such as ang and hay to signal the nominal status of these stance constructions. Tagalog
(Kaufman, this volume: 731)
(81) ang ganda mo! nom beauty 2s.gen ‘How beautiful you are!’ Botolan Sambal
(Kaufman, this volume: 731)
(82) hay ganda nin babayi! nom beauty gen woman ‘How beautiful the woman is!’ Malay/Indonesian (83) Agak-nya dia tak akan datang. guess-3sg.gen 3sg neg fut come ‘Apparently s/he won’t be coming.’
A similar development involving nominalization constructions with genitive subjects being recruited for stance marking is found in Old Chinese as well, as seen in (84) below. Old Chinese
(Yue 1998: gloss added)
(84) han zhi guang ye! (name of river) gen broad prt ‘How broad the Han River is!’
As seen in Cebuano (85) and Wolio (86), these stance constructions still retain traces of erstwhile nominalizers, among them prefix ka- (as opposed to prefix ma-, which has been recruited instead as a finiteness marker within the predicate domain).37 Cebuano (85) a.
(Kaufman, this volume: 728)
ka-tambok niya! cf. b. ma-tambok siya ext-fat 3s.gen adj-fat 3s.nom ‘How fat he is!’ ‘He’s fat.’
.╅ In many Formosan languages, ka and ma- are also used as stative markers and ma may also be a voice marker (personal comm., Li-May Sung, October 31, 2009 and May 21, 2010). The relationship between stativity, nominalization and voice has been investigated in Zeitoun and Huang (2000).
Introduction 
Wolio
(Anceaux 1988: 53; cited in Kaufman, this volume: 729)
(86) a. ka-luntu-na o mia sii cf. ext-lazy-3.gen det person this ‘How lazy this person is!’
b. ma-luntu o mia sii adj-lazy det person this ‘This person is lazy.’
Stand-alone nominalizations often raise the issue of whether or not such non-embedded nominalizations are the product of copula ellipsis or are unelided independent utterances. Since Matisoff (1972), it has often been contended that free-standing nominalizations in Tibeto-Burman, although often translated as cleft-constructions in English, are not necessarily derived from focus constructions supported by copulas. In Matisoff ’s words, “this [the availability of stand-alone nominalizations] does not mean that we have to assume that there is some higher verb floating around” (Matisoff 1972: 246–247). Wrona (this volume) similarly points out that in Japanese in particular, but also in Korean, Chinese, Rawang and Chantyal, there is insufficient evidence to propose that stand-alone nominalizations are necessarily derived from copula-type cleft constructions. This appears to be particularly true of mirative-type (i.e. exclamative) stand-alone nominalizations. Nevertheless, cleft-like constructions have been observed to play a role in the emergence of stand-alone stance constructions in at least some languages (see Grunow-Hårsta 2009 for Magar; Watters 2008 for Kiranti and Central Himalayish languages; Yap & Matthews 2008 for a general discussion; Shinzato, this volume, for Okinawan).38 It should also be noted that copula-less Austronesian languages can likewise form cleft-type focus constructions, as seen in the Tagalog example in (87) below. Tagalog (87) a.
(Kaufman 2005: 194)
[Si Juan] [ang lumalangoy]. p.nom Juan nom av.imp:swim ‘It’s Juan that is swimming.’
b. [Dagat] [ang nilalanguyan ni Juan]. sea nom lv.imp:swim p.gen Juan ‘It’s in the sea that Juan swims.’
.â•… The term ‘cleft-like’ is often used not only to highlight formal differences but also functional similarities in the focus constructions across languages. Essentially, the copula plays a major role in cleft constructions of copula-obligatory languages such as English, but they are optional in languages where the copula can be omitted (as is often the case among Sino-Tibetan languages), and they are simply absent in copula-less languages (as in most Austronesian languages). Korean and Japanese in general are copula-loving languages, but unlike English whose copula occupies pre-predicate position, the copulas in Korean and Japanese are post-predicate and frequently develop stance/mood marking functions (often merging with other sentence final particles). This development evinces naturally, given that copulas in these two languages occupy post-predicate position, and hence often hosts sentence-final prosody.
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
As noted in Horie (this volume), nominalization constructions often coalesce with copulas as in Japanese no-da (nominalizer + copula) and Korean kes-i-ta (nominalizer + copula + sentence ender) to form complex TAM markers or sentence-final particles (see also Horie 2008). In sum, we have seen various types of referential nominalization constructions being reinterpreted as stance constructions. These nominalization constructions may be ‘cleft-type’ focus constructions supported by copulas, where, in some languages, the copulas themselves are reinterpreted as sentence final stance particles. Nominalized stance constructions may also be ‘stand-alones’ – that is, not accompanied by copulas.39 Such constructions may be the result of copula elision (as in Japanese, Korean, Sinitic and in some Tibeto-Burman languages), or nominalized constructions may simply be independent, i.e. lacking a copula (as perhaps in the case of many Austronesian languages). Regardless of type, it is clear that there is a cross-linguistic tendency for nominalizers, and noun phrase markers that signal the nominalization constructions, to be reinterpreted as stance markers.
4.4â•… Nominalization and clausal subordination In this section, we examine another frequently attested non-referential use of nominalization constructions, namely adverbial-type subordinate clauses. While languages often make use of simple adjunction to form adverbial clauses of time and condition, as in Anything moves, we fire, they also frequently rely on subordinators, some of which are derived from nominalizing elements. Old Chinese zhe is a case in point (see Harbsmeier 1981; see also Yap & Wang, this volume). Semantically, in addition to its primary role as agent nominalizer, zhe also serves as a light noun referring to manner, time, condition, etc., often yielding an adverbial-type clausal nominalization construction.40 When this construction is used in topic position, it provides background information (i.e. it sets the stage) for the upcoming clause, as in (88).41 In this particular example, the zhe nominalizer has been reinterpreted as an adverbial subordinator in a subordinate temporal or conditional clause.
.â•… For this reason, syntacticians often refer to them as ‘syntactically incomplete’ (e.g. Stowell 1981; Radford 1990; Barton & Progovac 2005; Kaufman 2005; Tang 2007). What ‘licenses’ (or legitimizes) them as grammatical constructions is the illocutionary force expressed through the speaker’s prosody. .â•… These constructiuons are sometimes referred to as ‘headless’ relative clause Â�constructions. .â•… This topic-like ‘mental staging’ function has been extensively discussed in the literature (e.g. Haiman 1978).
Introduction 
Old Chinese (88) [nongfu duo yu tian zhe] ze guo pin ye farmer lazy loc field zhe then country poor sfp ‘(The time) when/if the peasants are lazy in the field, then the state will be poor.’ (Han Fei Zi, Warring States period, 475–221 bc)
Austronesian languages such as Budai Rukai and Kavalan (both Formosan) also show a development involving the extension of nominalization constructions to non-referential uses as adverbial subordinate clauses, as seen in (89) and (90). Budai Rukai
(Sung, this volume: 534)
(89) lu [ka ta-ki-tai-tai-anә] ŋu-a-daT anә madu maganaәT ә if/when obl nfut-get-red-taro-nmz go-fin-field have.to all ‘At the time of harvesting taros, all (of us) have to go to the field.’ (90) Kavalan a. Raya dames utuz a yau big especially earthquake lnk that
(Hsieh, this volume: 509)42
[m-suRaw=ay ci-aliung tu m-qaytun] af-fall=nmz ncm-pn obl af-vehicle
‘There was a strong earthquake (at the time) when Aliung fell over from the vehicle’
b. [wama zana iRuR=ay] zana iRuR=ay ma tamun-ta only 3.poss river= nmz 3.poss river= nmz only dish-1incl.pl-gen
[zana lazing=ay wama] zana lazing=ay tamun-ta 3.poss sea= nmz only 3.poss sea= nmz dish-1incl.pl-gen
‘If/When we have only those from the river, those from the river will be our dishes. If/When we have those from the sea only, those from the sea will be our dishes.’
The use of a nominalizer as subordinator is also found in Tibeto-Burman, as in Manange (91), where nominalizer -pa is reinterpreted as a causal subordinator. Manange (91)
[ŋ%=tse22 1=erg
(Hildebrandt 2004: 83; also cited in Genetti, this volume: 174) mwi42
phr%42 ky%=ri52 pim-p%22] money 100 2sg=loc give-nmz
nese22 ky%52 kola52 kyu-p%52 tomorrow 2sg clothes buy-nom ‘Because I gave you 100 rupees, you will buy a dress tomorrow.’
.╅ Fuhui Hsieh (personal comm., October 21, 2009) agrees that =ay in constructions such as (90) above have a nominalizing function, although she consistently glosses =ay as REL throughout her paper (Hsieh, this volume).
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Such nominalizations functioning as adverbials may be signaled by a nominalizer alone or accompanied by noun phrase markers as seen in Mongsen Ao (Tibeto-Burman) (92). Here, the nominalizer -pa combines with case marker nә to form causal subordinator -panә ‘because’. Genetti (1986, 1991, this volume; see also Genetti et al. 2008; Coupe 2007; Aikhenvald 2008) shows that nominalization constructions accompanied by certain postpositions (or case markers), particularly oblique ones not typically used in participant nominalizations, are often recruited to form subordinate clauses. Mongesen Ao
(Coupe 2007: 11; also cited in Genetti, this volume: 175)
(92) [pa tә-lәm ku tsә`-pànә] tә-lәm tſu apak-tſuk 3sg rl-head loc peck-causal rl-head dist flat-pfv.pst ‘Because [the other birds] pecked on her head, [Owl’s] head became flat.’
A parallel development is found in Japanese (93) where nominalizer no combines with case marker de to yield causal subordinator no de ‘because’, and with case marker ni to yield concessive subordinator no ni ‘although’ (see also Horie 2008: 175).43 Modern Japanese (93) a.
(Yap & Matthews 2008: 317)
undo shita no de taiju ga herakatta exercise do.pst nmz.prt weight nom reduce.npst ‘Because I exercised, I lost weight.’
b. undo shita no ni taiju ga heranakatta exercise do.pst nmz.prt weight nom reduce.neg.npst ‘Although I exercised, I did not lose weight.’
As was demonstrated in the foregoing sections, noun phrase markers which nominalize can develop analogously to nominalizers. That is to say, they can develop the same extended non-referential functions. This is true of the subordinating function as well. That the isomorphic relationship between nominalization and adverbial subordination is fairly robust crosslinguistically has also been clearly demonstrated by Haspelmath and König (1995) and papers in their volume on converbal constructions. It has also been documented for Japanese that case markers can also be reinterpreted as subordinators in the absence of overt morphological nominalizers, as in (94). 44
.â•… Korean also shows a similar development involving the combination of nominalizer kes and case markers such as ni (> kes-i-ni, nominalizer + copula + casemarker ‘cause’) and nikka (> kes-i-nikka, nominalizer + copula + casemarker ‘reason’) (see Rhee 2002). .â•… Genetti (1991) has described an analogous development of case-marker to subordinator for Newari (Tibeto-Burman).
Introduction 
Japanese
(Horie 2008: 173)
(94) Tometa-ni mo kakawarazu dete it-ta stop.pst.attr-dat also concern.neg leave-ger go-pst ‘Though I stopped, he departed.’
This phenomenon is also attested in Austronesian, as seen in the Pazeh examples in (95) below, where there is no evidence of an overt nominalizer. Worth noting is that the presence of a case marker is optional. Pazeh (95) a.
(Zeitoun & Huang 2000: 400; 406, fn 25)
[sutun-i siatu,] ini ka-akux-ay take.off-imp clothes neg stat-hot-irr ‘Take off your clothes (and) you will not be(/feel) hot.’
b. [sutun-i siatu ka,] ini (ka)-akux-ay take.off-imp clothes top neg stat-hot-irr ‘Take off your clothes (so that) you will not feel hot.’ Lit. ‘If you take off your clothes, you will not feel hot.’
Abui (a Papuan language) makes extensive use of demonstratives in adverbial clauses to express the relative time of the event described in the main clause (see Kratochvíl, this volume). As seen in (96) below, Abui makes use of proximal demonstrative do to refer to both the event of shooting and the moment of shooting that the addressee needs to pay attention to. This example provides a bridge context for the semantic extension of nominalization constructions from an event-referring to a temporalreferencing function. As Kratochvíl (this volume: 785) notes, such adverbial uses are, like their noun-modifying (i.e. relative clause) counterparts, still “referential” – that is, we are referring to the notion of time, enabling circumstance, etc. The relationship between clausal subordination and nominalization here is thus both intimate and still fairly transparent.45 Abui
(Kratochvíl, this volume: 771)
(96) [na ha-tak do] a he-roa 1sg 3ii.pat-shoot prx 2sg 3ii.loc-watch.cnt ‘When I shoot (with the bow), you watch it.’
.â•… František Kratochvil (personal comm., October 21, 2009) observes that since demonstratives are often recruited to indicate tense as well, “it is not entirely clear whether the demonstrative on the adverbial clause is more a tense-marker or more a nominalizer.” For this reason, Kratochvil (this volume) maintains that Abui adverbial clauses such as (96) above, in their role as constituents identifying the settings of an event, “are formally treated as nominals” (following De Vries 2006), but it is still not clear whether the adverbial constructions themselves are derived from nominalization constructions.
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Furthermore, in Mandarin Chinese, it can be demonstrated for the construction [VP de hua] that the subordinate clause has the structure of a relative clause, as seen in (97) below. Modern Chinese
(Yap, Choi & Cheung 2010:â•›69)
(97) [e le de hua] jiu qu chi di dongxi be.hungry asp de hua then go eat cl something ‘If you are hungry, go and get something to eat.’
In a recent corpus-based diachronic analysis, Jiang (2004) reported the following stages in the emergence of the (shuo) [VP de hua] subordinate clause (see Table 3). In Stage 1, the predicate (VP) nominalized by de, i.e. the [VP de] construction, merges with appositive light noun hua ‘speech’ in relative clause fashion. In Stage 2, nominalizer de combines with head noun hua to form a dehua complementizer. In Stage 3, the [VP dehua] complement clause can be preposed to precede the controlling verb shuo ‘say’. In stage 4, elision of the ‘say’ verb results in the modern Chinese [VP dehua] conditional construction, with dehua as a subordinator, and which can be further reinforced by another conditional subordinator such as ruguo ‘if ’ in clause-initial position. Parallel developments can be found in other Chinese dialects (e.g. Cantonese ge waa conditional constructions).46 Table 3.╇ Development of the [VP dehua] subordinate conditional clause in Chinese [VP de hua] subordinate clause in Chinese
English translation
Stage 1 shuo ‘say’ [VP de] [hua ‘speech’], main clause ‘say’ clause, main clause Stage 2 shuo ‘say’ [VP dehua], main clause ‘say’ clause, main clause Stage 3 [VP dehua] [shuo ‘say’], main clause preposed complement + ‘say’ clause, main clause Stage 4 [VP dehua], main clause conditional ‘if ’ clause, main clause
This section has demonstrated an additional development from referential to nonreferential functions. Specifically, that nominalizers and noun phrase markers that signal nominalizations can cross-linguistically develop subordinating and adverbial functions and can come to express manner, condition, temporal and causal relations.
.╅ Note however that Stage 3 is missing in some dialects such as Cantonese (*[VP ge waa] gong). This suggests the possibility that the Mandarin [VP dehua] subordinate clause could have emerged directly from Stage 2 to Stage 4, without going through Stage 3.
Introduction 
5.â•… Conclusion and future directions In this introductory chapter we have approached nominalization phenomena from a crosslinguistic perspective, often with an eye to diachronic developments. Based on data and insights from previous literature and the papers presented in this volume, we have identified a wide range of nominalization strategies among Asian languages. In addition, we have traced robust grammaticalization pathways for nominalization constructions, specifically as they move from referential to non-referential functions. Among the frequently observed developments are: (i) the use of nominalizers for nominal specification/modification, hence the frequently observed syncretism between nominalization, relativization and genitivization reported in the literature; (ii) the reanalysis of nominalized constructions as finite clauses, which accounts for frequent crosslinguistic evidence of isomorphism between nominalizers and tenseaspect-mood (TAM) markers; (iii) the reanalysis of post-predicate nominalizers as sentence final mood particles, as well as the reanalysis of stand-alone nominalization constructions as miratives and the use of embedded nominalization constructions in the formation of cleft-type focus constructions; and, finally, (iv) the reanalysis of nominalizers as clausal subordinators. Several papers in the present volume have also focused on the contributions of referentiality marking devices – that is, noun phrase markers such as case markers, demonstratives and definiteness markers, possessive pronouns, plural markers, and classifiers – in signaling nominalization constructions. Based on data and insights from these papers, we are able to also identify some strong parallels in terms of functions between nominalizers originating from light nouns and referentiality marking devices originating from deictic elements such as demonstratives. That both nominalizing and nominal-signaling elements often share similar non-referential functions such as marking relativization, genitivization, subordination, tense-aspect and speaker mood/ stance is worthy of further investigation in future research. Of special value would be diachronic studies and conversational/discourse analyses to complement typological work, particulary in identifying fine distinctions among the non-referential functions – for example, the relationship between proximal/distal deixis and the positioning of temporal reference and speaker stance. As with most typological work, the present findings identify robust patterns that will be valuable in the effort to decode the identity of semantically bleached but versatile morphemes, in this case those participating in nominalization-related phenomena, whether referential or non-referential. Further work can expand, as well as refine, the analyses presented in this introductory paper and in the 25 more detailed contributions in the present volume. In terms of organizational structure, the papers in this volume are arranged primarily along areal and genetic lines. We begin the first half of the volume with the
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona
Sinitic, Tibeto-Burman and Iranian languages – expanding westwards across the Asian continent. The second half of the volume focuses on Asian languages with close affinity to the Pacific and Indian oceans, among them the languages of Korea and Japan, followed by Austronesian languages including the Formosan-Philippine and MalayIndonesian varieties, and also Malagasy (at the south-eastern edge of Africa) and Toqabaqita (representing the Oceanic language grouping), as well as Abui (spoken within the Papuan enclave). It is intended that typological insights from these languages will spur similar research in other languages, and that through these combined efforts, we will come closer to decoding the recurrent mechanisms and principles that steer the course, and constrain the structures and evolutionary paths, of human language.
Acknowledgments We wish to gratefully acknowledge generous research funding support from the Hong Kong Polytechnic University (Faculty of Humanities Start-Up Fund 2010–2012) and from the Chinese University of Hong Kong (Faculty of Arts Direct Research Grants 2006–2007, 2007–2008; Linguistics and Modern Languages Departmental Research Grant 2008–2009) awarded to the first author, as well as NSF grant BCS-0618928 and ELDP grant FTG0104 awarded to the second author. This areal typological study (inclusive of editorial work for the remaining 25 papers in the present volume entitled Nominalization in Asian Languages: Diachronic and Typological Perspectives) was made possible through grants for the following specific projects: Development of Stance Markers in East and Southeast Asian Languages (CUHK Direct Research Grant 2006–2007); Diachronic Syntax in East Asian Languages (CUHK Direct Research Grant 2007–2008); Nominalization Strategies in Chinese and Other Asian Languages (CUHK Linguistics and Modern Languages Department Research Grant 2008–2009); Non-Referential Uses of Nominalization Constructions: Asian Perspectives (HKPU Faculty of Humanities Start-Up Fund 2009–2010). We are also deeply grateful for invaluable discussions of data with fellow contributors to this volume: Scott DeLancey, Carol Genetti, Yang Gu, Geoff Haig, Kaoru Horie, Fuhui Hsieh, Daniel Kaufman, František Kratochvíl, Frank Lichtenberk, Hongyong Liu, Stephen Matthews, Stephen Morey, Naonori Nagaya, Michael Noonan, Mark Post, Eric Potsdam, Seongha Rhee, Rumiko Shinzato, Joanna Ut-seong Sio, Li-May Sung, Sze-Wing Tang, Jiao Wang, Huiling Xu, and Marie Mei-li Yeh for prompt replies to questions on Papuan, Austronesian, Okinawan, Japanese, Korean, Sinitic, Tibeto-Burman and Iranian languages. We also wish to express special thanks to Peter Cole, Alec Coupe, Gabrielle Hermon, Shuping Huang, Haowen Jiang, You-jing Lin, Laurie Reid, Andrej Malchukov, Matt Shibatani, Michael Tanangkingsing, Stacy Teng and Elizabeth Zeitoun for valuable discussions on nominalization phenomena. This paper also owes much in terms of accuracy of
Introduction 
� interpretation of numerous languages to the invaluable feedback from the many external reviewers who have generously shared with us both their time and expertise.
References Adelaar, Alexander & Himmelmann, Nikolaus P. (eds). 2005. The Austronesian Languages of Asia and Madagascar. London: Routledge. Aikhenvald, Alexandra. 2008. Versatile cases. Journal of Linguistics 44(3): 565–603. Aldridge, Edith. 2008. Old Chinese determiner zhe. In Historical Syntax and Linguistic Theory, Paola Crisma & Giuseppe Longobardi (eds), 233–248. Oxford: OUP. Arka, I Wayan & Ross, Malcolm (eds). 2005. The Many Faces of Austronesian Voice Systems: Some New Empirical Studies. Canberra: Pacific Linguistics. Barton, Ellen & Progovac, Ljiljana. 2005. Nonsententials in Minimalism. In Ellipsis and Non-sentential Speech, Reinaldo Elugardo & Robert Stainton (eds), 71–93. Dordrecht: Kluwer. Benedict, Paul. 1972. Sino-Tibetan: A Conspectus [Princeton-Cambridge Studies in Chinese Linguistics II]. Cambridge: CUP. van den Berg, Rene. 1989. A Grammar of the Muna Language. Dordrecht: Foris. Bickel, Balthasar. 1999. Nominalization and focus in some Kiranti languages. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 271–296. Kathmandu: Royal Nepal Academy. Blust, Robert A. 1999. Notes on Pazeh phonology and morphology. Oceanic Linguistics 38: 321–365. Blust, Robert A. 2002. Notes on the history of ‘focus’ in Austronesian languages. In The History and Typology of Western Austronesian Voice Systems, Fay Wouk & Malcolm Ross (eds), 17–62. Canberra: Pacific Linguistics. Chen, Shilin. 1989. Specificity and definiteness in Nuosu Yi. Minzu Yuwen 2: 21–28. Chen, Kang & Wu, Da. 1998. Nuosu Yi Grammar. Beijing: Central University for Nationalities Press. Comrie, Bernard & Thompson, Sandra A. 1985. Lexical nominalization. In Language Typology and Syntactic Description, Vol. III: Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 349–398. Cambridge: CUP. Coupe, Alec. 2007. A Grammar of Mongsen Ao [Mouton Grammar Series 39]. Berlin: Mouton de Gruyter. Coupe, Alec R. 2009. The origins of tense marking in Ao. Paper presented at the 42nd International Conference on Sino-Tibetan Languages, Payap University, November 2–4. DeLancey, Scott. 1986a. Relativization as nominalization in Tibetan and Newari. Paper presented at the 19th International Conference on Sino-Tibetan Languages and Linguistics, Ohio State University, September 11. DeLancey, Scott. 1986b. Evidentiality and volitionality in Tibetan. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe & Johanna Nichols (eds), 203–213. Norwood NJ: Ablex. DeLancey, Scott. 1997. Mirativity: The grammatical marking of unexpected information. Linguistic Typology 1(1): 33–52.
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona DeLancey, Scott. 1999. Relativization in Tibetan. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 231–249. Kathmandu: Royal Nepal Academy. DeLancey, Scott. 2001. The mirative and evidentiality. Journal of Pragmatics 33(3): 371–384. DeLancey, Scott. 2005. Relativization and nominalization in Bodic. Tibeto-Burman Linguistics: Proceedings of the Twenty-Eighth Annual Meeting of the Berkeley Linguistics Society, 5–72. Berkeley CA: Berkeley Linguistics Society. DeLancey, Scott. This volume. Finite structures from nominalization constructions in Tibeto-Burman. In Yap, Grunow-Hårsta & Wrona (eds). Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68: 81–138. Englebretson, Robert. 2003. Searching for Structure: The Problem of Complementation in Colloquial Indonesian Conversation [Studies in Discourse and Grammar 13]. Amsterdam: John Benjamins. Genetti, Carol. 1986. The grammatical development of postpositions to subordinators in Bodic languages. In Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistics Society, Vassiliki Nikiforidou, Mary VanClay, Mary Niepokuj & Debra Feder (eds), 387–400. Berkeley CA: Berkeley Linguistic Society. Genetti, Carol. 1991. From postposition to subordinator in Newari. In Approaches to Gramamticalization, Vol. 2: Focus on Types of Grammatical Markers [Typological Studies in Language 19], Elizabeth Closs Traugott & Bernd Heine (eds), 227–255. Amsterdam: John Benjamins. Genetti, Carol 1992. Semantic and grammatical categories of relative-clause morphology in the languages of Nepal. Studies in Language 16(2): 405–427. Genetti, Carol. 2009. An introduction to Dzala. Paper presented at the 15th Himalayan Linguistics Symposium (HLS-15), University of Oregon, July 31-August 1. Genetti, Carol. This volume. Nominalization in Tibeto-Burman languages of the Himalayan area: A typological perspective. In Yap, Grunow-Hårsta & Wrona (eds). Genetti, Carol, Coupe, Alec R., Bartee, Ellen, Hilderbrandt, Kristine & Lin, You-Jing. 2008. Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himalayan area. Linguistics of the Tibeto-Burman Area 31(2): 97–143. Greenberg, Joseph H. 1966. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, Joseph H. Greenberg (ed.), 73–113. Cambridge MA: The MIT Press. Grimes, Charles E. 1991. The Buru Language of Eastern Indonesia. Ph.D. dissertation, Australian National University. Grunow-Hårsta, Karen. 2007. Evidentiality and mirativity in Magar. Linguistics of the Tibeto-Burman Area 30(2): 151–194. Grunow-Hårsta, Karen. 2008. A Descriptive Grammar of Two Dialects of Magar: Tanahu and Syangja. Volumes 1 & 2. Ph.D. dissertation, University of Wisconsin-Milwaukee. Grunow-Hårsta, Karen 2009. Plurifunctionality in the Magar nominalization system. Nepalese Linguistics 24: 37–48. Grunow-Hårsta, Karen. This volume. Innovation in nominalization in Magar, a Tibeto-Â� Burman language of Nepal. In Yap, Grunow-Hårsta & Wrona (eds). Grunow-Hårsta, Karen & Yap, Foong Ha. 2009. Referential and non-referential uses of nominalization constructions in Magar. Paper presented at the 15th Himalayan Languages Symposium (HLS-15), University of Oregon, July 31-August 1. Grunow-Hårsta, Karen & Yap, Foong Ha. 2010. From (ad)nominalizer to tense-aspect-mood marker: A case of semantic extensions from referential to non-referential uses. Paper presented at the 4th Conference on Language, Discourse and Cognition (CLDC-2010), National Taiwan University, May 1–2.
Introduction 
Haig, Geoffrey. This volume. Linker, relativizer, nominalizer, tense-particle: On the Ezafe in West Iranian. In Yap, Grunow-Hårsta & Wrona (eds). Haiman, John. 1978. Conditionals are topics. Language 54: 565–589. Harbsmeier, Christoph. 1981. Aspects of Classical Chinese Syntax [Scandinavian Institute of Asian Studies Monograph Series 45]. London: Curzon Press. Haspelmath, Martin & König, Ekkehard. 1995. Converbs in Crosslinguistic Perspective: Structure and Meaning in Adverbial Verb Forms – Adverbial Participles, Gerunds, etc. Berlin: Mouton de Gruyter. Hawkins, John A. 1994. A Performance Theory of Order and Constituency. Cambridge: CUP. Hawkins, John A. 2004. Efficiency and Complexity in Grammars. Oxford: OUP. He, Jin-song. 1988. Zhuyu weiyu zhijian zhi de ciyi ji qi yuanliu (The origin and semantics of zhi between subject and predicate). Zhong Nan Minzu Xueyuan Xuebao: Zhe She Ban (Wuhan) (Journal of the South Central University for Nationalities (Philosophy and Social Sciences Edition)) 6: 106–117. Hildebrandt, Kristine. 2004. A grammar and glossary of the Manange language. In TibetoBurman Languages of Nepal: Manange and Sherpa [Pacific Linguistics 557], Carol Genetti (ed.), 3–189. Canberra: Pacific Linguistics. Himmelmann, Nikolaus P. 2005. The Austronesian languages of Asia and Madagascar: Typological characteristics. In The Austronesian Languages of Asia and Madagascar. Alexander Adelaar & Nikolaus P. Himmelmann (eds), 110–181. London: Routledge. Horie, Kaoru. 1998a. On the polyfunctionality of the Japanese particle no: From the perspectives of ontology and grammaticalization. In Studies in Japanese Grammaticalization: Cognitive and Discourse Perspectives, Toshio Ohori (ed.), 169–192. Tokyo: Kurosio. Horie, Kaoru. 1998b. Functional duality of case-marking particles in Japanese and its implications for grammaticalization: A contrastive study with Korean. In Japanese/Korean Linguistics 8, David J. Silva (ed.), 147–159. Stanford CA: CSLI. Horie, Kaoru. 2000. Complementation in Japanese and Korean: A contrastive and cognitive linguistic approach. In Complementation: Cognitive and Functional Perspectives, Kaoru Horie (ed.), 11–32. Amsterdam: John Benjamins. Horie, Kaoru. 2008. The grammaticalization of nominalizers in Japanese and Korean: A constrastive study. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 169–187. Amsterdam: John Benjamins. Horie, Kaoru. This volume. Versatility of nominalizations: Where Japanese and Korean contrast. In In Yap, Grunow-Hårsta & Wrona (eds). Huang, Shuanfan. 2002. The pragmatics of focus in Tsou and Seediq. Language and Linguistics 3(4): 665–694. Huang, Shuping & Su, Lily I-wen. 2005. Iconicity as evidenced in Saisiyat linguistic coding of causative events. Oceanic Linguistics 44(2): 341–356. Hsieh, Fuhui. This volume. The functions of -an and =ay in Kavalan. In Yap, Grunow-Hårsta & Wrona (eds). Ivens, Walter G. 1929a. A Dictionary of the Language of Sa‘a (Mala) and Ulawa, South-east Solomon Islands. London: OUP; also Melbourne: Melbourne University Press. Iwasaki, Shoichi. 2000. Suppressed assertion and the functions of the final-attributive in prose and poetry of Heian Japanese. In Textual Parameters in Older Languages, Susan C. Herring, Pieter van Reenen & Lene Schøsler (eds), 237–272. Amsterdam: John Benjamins. Jiang, Lansheng. 2004. About the lexicalization of the cross-structural combination “de hua”. Zhongguo Yuwen 5: 387–400.
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona Kaufman, Daniel. 2005. Aspects of pragmatic focus in Tagalog. In The Many Faces of Austronesian Voice Systems: Some New Empirical Studies, I. Wayan Arka & Malcom Ross (eds), 175–196. Canberra: Pacific Linguistics. Kaufman, Daniel. 2009. Austronesian typology and the nominalist hypothesis. In Austronesian Historical Linguistics and Culture History: A Festschrift for Robert Blust, Alexander Adelaar & Andrew Pawley (eds). Canberra: Pacific Linguistics. Kaufman, Daniel. This volume. Exclamatives and temporal nominalizations in Austronesian. In Yap, Grunow-Hårsta & Wrona (eds). Konnerth, Linda. In press. The nominalizing velar prefix *gV- in Tibeto-Burman languages of Northeast India. In North East Indian Linguistics, Vol. 4, Gwendolyn Hyslop, Stephen Morey & Mark Post (eds). New Delhi: Cambridge University Press India. Koptjevskaja-Tamm, Maria. 1993. Nominalizations. Oxford: Routledge. Koptjevskaja-Tamm, Maria. 2003. Action nominal constructions in the languages of Europe. In Noun Phrase Structure in the Languages of Europe, Frans Plank (ed.), 723–761. Berlin: Mouton de Gruyter. Kratochvíl, František. This volume. Discourse-structuring functions of Abui demonstratives. In Yap, Grunow-Hårsta & Wrona (eds). Kuo, John Ching-hua. 1979. Rukai Complementation. MA thesis, Fu Jen Catholic University, Taipei. LaPolla, Randy J. 1994. Parallel grammaticalizations in Tibeto-Burman: Evidence of Sapir’s drift. Linguistics of the Tibeto-Burman Area 17(1): 61–80. LaPolla, Randy J. 2008. Nominalization in Rawang. Linguistics of the Tibeto-Burman Area 31(2): 45–65. LaPolla, Randy J. & Poa, Dory. 2001. Rawang Texts, with Grammatical Analysis and English Translation. Munich: Lincom. Lee, Hyo-Sang. 1993. The temporal system of noun-modifying (attributive) clauses in Korean from a typological perspective. Studies in Language 17(1): 75–110. Lehmann, Christian and Moravcsik, Edith. 2000. Noun. In Morphology: An International Handbook on Inflection and Word-Formation,Vol. 1, Geert Booij, Christian Lehmann & Joachim Mugdan (eds), 732–757. Berlin: Mouton de Gruyter. Li, Jen-kuei Paul. 2002. Nominalization in Pazih. Language and Linguistics 3(2): 227–239. Lichtenberk, Frank. This volume. Nominalization in Toqabaqita and closely related languages. In Yap, Grunow-Hårsta & Wrona (eds). Liu, Hongyong & Yang, Gu. This volume. Nominalization in Nuosu Yi. In Yap, Grunow-Hårsta & Wrona (eds). MacKenzie, Lachlan. 1987. Nominalization and basic constituent ordering. In The Ins and Outs of Predications, Johan van der Auwera & Louis Goossens (eds), 93–106. Dordrecht: Foris. Malchukov, Andrej. 2004. Nominalization/Verbalization: Constraining a Typology of Transcategorial Operations [Lincom Studies in Language Typology 8]. Munich: Lincom. Malchukov, Andrej L. 2006. Constraining nominalization: Function/form competition. Linguistics 44(5): 973–1009. Martin, Samuel E. 1975. A Reference Grammar of Japanese. New Haven CT: Yale University Press. Matisoff, James A.1972. Lahu nominalization, relativization, and genetivization. In Syntax and Semantics, Vol. 1, John P. Kimball (ed.), 237–257. New York: Academic Press. Matthews, Stephen & Yeung, Lousia Y.Y. 2001. Processing motivations for topicalization in Cantonese. In Cognitive-functional Linguistics in an East Asian Context, Kaoru Horie & Shigeru Sato (eds), 81–102. Tokyo: Kurosio.
Introduction 
Morey, Stephen. This volume. Nominalization in Numhpuk Singpho. In Yap, Grunow-Hårsta & Wrona (eds). Nagaya, Naonori. This volume. Rise and fall of referentiality: Articles in Philippine languages. In Yap, Grunow-Hårsta & Wrona (eds). Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Noonan, Michael. 2008. Nominalization in Bodic languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins. Noonan, Michael. This volume. Aspects of the historical development of nominalizers in the Tamangic languages. In Yap, Grunow-Hårsta & Wrona (eds). Post, Mark. This volume. Nominalization and nominalization-based constructions in Galo. In In Yap, Grunow-Hårsta & Wrona (eds). Potsdam, Eric. This volume. Expressing exclamatives in Malagasy. In Yap, Grunow-Hårsta & Wrona (eds). Radford, Andrew. 1990. Syntactic Theory and the Acquisition of English Syntax. Oxford: Basil Blackwell. Reid, Lawrence A. 1992. On the development of the aspect system in some Philippine languages. Oceanic Linguistics 31(1): 65–91. Rhee, Seongha. 2002. From silence to grammar: Grammaticalization of ellipsis in Korean. Paper presented at New Reflections on Grammaticalization 2, University of Amsterdam, April 4–6. Rhee, Seongha. 2008. On the rise and fall of Korean nominalizers. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 239–264. Amsterdam: John Benjamins. Rhee, Seongha. This volume. Nominalization and stance marking in Korean. In Yap, Grunow-Hårsta & Wrona (eds). Rijkhoff, Jan. 2002. The Noun Phrase. Oxford: OUP. (revised and extended paperback edition 2004). Ross, Malcom D. 1995. Reconstructing Proto-Austronesian verbal morphology: Evidence from Taiwan. In Austronesian Studies Relating to Taiwan [Symposium Series of the Institute of History and Philology, Academia Sinica, Number 3], Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho & Chiu-yu Tseng (eds), 727–791. Taipei: Academia Sinica. Ross, Malcolm. 2002. The history and transitivity of Western Austronesian voice and voicemarking. In The History and Typology of Western Austronesian Voice Systems, Fay Wouk & Malcolm Ross (eds), 17–62. Canberra: Pacific Linguistics. Ross, Malcolm. 2006. Reconstructing the case-marking and personal pronoun systems of Proto Austronesian. In Streams Converging into an Ocean: Festschrift in Honor of Professor Paul Jen-kuei Li on his 70th Birthday, Henry Yungli Chang, Lillian M. Huang & Dah-an Ho (eds), 521–564. Taipei: Institute of Linguistics, Academia Sinica. Ross, Malcolm. 2009. Proto Austronesian verbal morphology: A reappraisal. In Austronesian Historial Linguistics and Culture History: A Festschrift for Robert Blust, Alexander Adelaar & Andrew Pawley (eds), 295–326. Canberra: Pacific Linguistics. Ross, Malcolm & Teng, Stacy Fang-ching. 2005. Formosan languages and linguistic typology. Language and Linguistics 6: 739–781.
 Foong Ha Yap, Karen Grunow-Hårsta & Janick Wrona Saxena, Anju. 2000. Evidentiality in Kinnauri. In Evidentials: Turkic, Iranian and Neighbouring Languages, Lars Johanson & Bo Utas (eds), 471–483. Berlin: Mouton de Gruyter. Shi, Yuzhi & Li, Charles N. 2002. The establishment of the classifier system and the grammaticalization of the morphosyntactic particle de in Chinese. Language Sciences 24: 1–15. Shibatani, Masayoshi. 2009. Elements of complex structures, where recursion isn’t: The case of relativization. In Syntactic Complexity: Diachrony, Acquisition, Neuro-cognition, Evolution, Talmy Givon & Masayoshi Shibatani (eds), 163–198. Amsterdam: John Benjamins. Shibatani, Masayoshi and Khaled Awadh bin Makhashen. 2009. Nominalization in Soqotri, a South Arabian language of Yemen. In Endangered Languages: Contributions to Morphology and Morpho-Syntax, W. Leo Wetzels (ed.), 9–31. Leiden: Brill. Shinzato, Rumiko. This volume. Nominalization in Okinawan: From a diachronic and comparative perspective. In Yap, Grunow-Hårsta & Wrona (eds). Simpson, Andrew. 2003. On the reanalysis of nominalizers in Chinese, Japanese and Korean. In Functional Structures, Form and Interpretation: Perspectives from East Asian Languages. Audrey Y.H. Li & Andrew Simpson (eds), 131–160. London: Curzon Routledge. Simpson, Andrew. 2008. The grammaticalization of clausal nominalizers in Burmese. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 265–288. Amsterdam: John Benjamins. Simpson, Andrew and Wu, Xiu-Zhi Zoe. 2001. The grammaticalization of formal nouns and nominalizers in Chinese, Japanese and Korean. Language Change in East Asia, N. McCauley (ed.), 250–283. Richmond, Surrey: Curzon. Starosta, Stanley, Pawley, Andrew K. and Reid, Lawrence A. 1982. The evolution of focus in Austronesian. In Papers from the Third International Conference on Austronesian Linguistics, Vol. 2: Tracking the Travelers, Amran Halim, Lois Carrington & S.A. Wurm (eds), 145–170. Canberra: Pacific Linguistics. Stowell, Tim. 1981. Origins of Phrase Structure. Ph.D. dissertation, MIT. Sung, Li-May. This volume. Clausal nominalization in Budai Rukai. In Yap, Grunow-Hårsta & Wrona (eds). Tang, Sze-Wing. 2007. Nominalization in Mandarin and Cantonese: Variations under a unified approach. Ms, Department of Chinese and Bilingual Studies, Hong Kong Polytechnic University. Tang, Sze-Wing. This volume. On gerundive nominalization in Mandarin and Cantonese. In Yap, Grunow-Hårsta & Wrona (eds). Teng, Stacy Fang-ching, 2008. A Reference Grammar of Puyuma, an Austronesian Language of Taiwan. Canberra: Pacific Linguistics. Thurgood, Graham & LaPolla, Randy J. (eds). 2003. The Sino-Tibetan Languages. London: Routledge. Van der Auwera, Johan (in collaboration with Dónall P.Ó. Baoill). 1998. Adverbial Constructions in the Languages of Europe. Berlin: Mouton de Gruyter. Watters, David. 2008. Nominalization in the Kiranti and Central Himalayish languages of Nepal. Linguistics of the Tibeto-Burman Area 31(2): 1–43. Whitman, John. 1997. Kakarimusubi from a comparative perspective. Japanese/Korean Linguistics 6: 161–178. Stanford CA: CSLI. Wouk, Fay & Ross, Malcolm (eds). 2002. The History and Typology of Western Austronesian Voice Systems. Canberra: Pacific Linguistics.
Introduction 
Wrona, Janick. This volume. A case of non-derived stand-alone nominalization: Evidence from Japanese. In Yap, Grunow-Hårsta & Wrona (eds). Xu, Huiling & Matthews, Stephen. This volume. On the polyfunctionality and grammaticalization of the morpheme kai in the Chaozhou dialect. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha. This volume. Referential and non-referential uses of nominalization constructions in Malay. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha, Choi, Pik-ling & Cheung, Kam-siu. 2010. Delexicalizing di: How a Chinese noun has evolved into an attitudinal nominalizer. In Formal Evidence in Grammaticalization Research [Typological Studies in Language 94], An Van linden, Jean-Christophe Verstraete & Kristin Davidse (eds), 63–91. Amsterdam: John Benjamins. Yap, Foong Ha & Grunow-Hårsta, Karen. 2010. Non-referential uses of nominalization constructions: Asian perspectives. Language and Linguistics Compass 3(1): 1–21. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins. Yap, Foong Ha, Mathews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Yap, Foong Ha & Wang, Jiao. This volume. From light noun to nominalizer and more: On the grammaticalization of zhe and suo in Old and Middle Chinese. In Yap, Grunow-Hårsta & Wrona (eds). Yeh, Marie Meili. This volume. Nominalization in Saisiyat. In Yap, Grunow-Hårsta & Wrona (eds). Yoon, James Hye Suk. 1991. Korean nominalizations, lexicalism, and morphosyntactic interface. Ms, University of Illinois, Urbana-Champaign. Yue, Anne O. 1998. Zhi in Pre-Qin Chinese. T’oung Pao, LXXXIV, 239–292. Yue, Anne. 2004. Two focus markers in the Zhongshan Bronze Inscriptions. In Meaning and Form: Pre-Modern Chinese Grammar, Ken’ichi Takashima & Jiang Shaoyu (eds), 241–266. Munich: Lincom. Zeitoun, Elizabeth (ed.). 2002. Nominalization in Formosan languages. Language and Linguistics 3(2). (Special Issue). Zeitoun, Elizabeth & Huang, Lillian M. 2000. Concerning ka-, an overlooked marker of verbal derivation in Formosan languages. Oceanic Linguistics 39(2): 391–414.
part i
Sinitic languages
From light noun to nominalizer and more The grammaticalization of zhe and suo in Old and Middle Chinese† Foong Ha Yap & Jiao Wang
Hong Kong Polytechnic University/Chinese University of Hong Kong In this paper, we examine how Chinese light nouns zhe and suo evolved into agent and patient nominalizers respectively. The link between nominalization, relativization and genitivization is explored from a grammaticalization perspective; likewise the frequent relationship between nominalization and clausal subordination. Other uses derived from nominalization constructions are discussed as well, among them speaker stance marking and passive voice marking.
1.â•… Introduction This paper examines nominalization phenomena in Old and Middle Chinese, in particular the strategies deployed in the formation of nominalization constructions formed by the light nouns zhe and suo. The term ‘light noun’ is used in the sense of semantically generalized or bleached nouns (e.g. Huang 2008; Aldridge 2008). Of special interest is the different pathway taken by each of these morphemes on their way to becoming a nominalizer. Among the questions we seek to answer are the following: In what way does etymology (or lexical origin) contribute to their different grammaticalization trajectories? And despite their different grammaticalization histories, what characteristics do they have in common that contribute to their functional status as †Chinese scholars often divide Chinese texts into three major periods: Old, Middle and Modern. The Western and Eastern Han periods are typically considered as the transition period between Old Chinese and Middle Chinese. In this paper we adopt the following period classification based largely on Norman (1988). Western Zhou: 11th to 6th century BC, Spring & Autumn: 770–476 BC, Warring States: 475–221 BC, Qin: 221–206 BC, Western Han: 206 BC-9 AD, Eastern Han: 9–220 AD, Three Kingdoms: 220–265 AD, Jin: 265–420 AD, Southern & Northern: 420–589 AD, Sui: 581–618 AD, Tang: 618–907 AD, Song: 960–1279 AD, Yuan: 1271–1368 AD, Ming: 1368–1644 AD, and Qing: 1644–1912 AD.
 Foong Ha Yap & Jiao Wang
‘nominalizers’? In this paper we will also highlight how each nominalizer’s development was influenced, not only by lexical origin (semantics), but also by structural constraints (morphosyntax). This paper is organized as follows. Section 2 traces the development of zhe from a light noun of unknown origin to agent nominalizer and conditional subordinator, with additional focus on how light noun zhe was reinterpreted as an adnominal, i.e. relativizer and genitive marker, in the presence of another lexical head noun, as well as its reinterpretation as a mood particle when it occurs in sentence-final position. Section 3 then focuses on the grammaticalization of suo from a light noun of known locative origin to patient nominalizer and adverbial subordinator, as well as other extended functions in possessive and passive constructions. In this paper we highlight that whereas possessee pronominal suo emerged directly from light noun suo, the development of passive wei … suo constructions was mediated by suo nominalization constructions. Section 4 concludes with some general observations with implications for crosslinguistic studies.
2.â•… The development of agent nominalizer zhe One of the best known nominalizers in Old Chinese was agent nominalizer zhe.1 Its usage was attested as early as the ancient odes of the Zhou period (11th to 6th century BC), as seen in (1) from Shijing (The Book of Odes), where a verbal predicate zhi wo (‘know me’) followed by zhe gets reinterpreted as a nominal expression meaning ‘the ones that understand me’. The use of zhe to create agent nominals is still productive in Modern Chinese, although often restricted to the formal and classical register. In this section we will examine how zhe evolves from a light noun into a nominalizer, as well as a wide range of other functions including adnominal, possessee pronominal, marker of clausal subordination, and sentence-final mood particle.2
.╅ Zhe sometimes marks a patient nominalization, as in [suoi VP _i] zhei and [_�i V + yu NP] zhei constructions, where suo is a pronominal clitic (also often referred to as patient nominalizer) and yu can be used as a case marker for demoted agents (e.g. Pulleyblank 1995; Zhu 1983). .╅ Most of the examples of zhe constructions discussed in this paper come from Huang (2005) and Yap and Huang (2006). Many of the zhe and suo tokens in our analysis were electronically accessed from The Chinese Ancient Texts (CHANT) database, which cover excavated oracle bone and bronze inscriptions, wood/bamboo and silk scripts, and texts from the Pre-Han period to the Southern and Northern period in the 6th century AD. This database is compiled and maintained by the Institute of Chinese Studies at the Chinese University of Hong Kong.
From light noun to nominalizer and more 
(1) [知我]者, 謂我心憂
zhi wo zhe wei wo xin you know 1sg zhe say 1sg heart worried ‘Those (< the ones) that understand me will say that I am sad and worried.’ (Shi Jing: Wang Feng: Shu Li)
2.1â•… Zhe as a light noun Although the lexical origin of zhe is far from clear, we do have some evidence indicating that zhe was used as a semantically general (or ‘light’) noun in Old Chinese. As seen in (2) and (3), zhe could be modified by quantifiers, demonstratives and numerals. These examples, both from the Warring States period (475–221 BC), show that zhe could refer to both animate and inanimate referents such as ‘kinds of people’ or ‘a list of items’ previously mentioned in preceding discourse. In other words, as a light noun, zhe could be instantiated, individuated, enumerated, and identified as either definite or indefinite within the unfolding discourse.
(2) 此四者,天下之窮民而無告者
ci si zhe, tian xia zhi qiong min er wu gao zhe these four zhe sky below gen poor people conn neg appeal nomz ‘These four kinds of people, (they are) those people in the world who are poor and have no place to appeal to.’ (Meng Zi, 1B/5)
(3) 凡此五者,將莫不聞
fan ci wu zhe, jiang mo bu wen all these five zhe general cannot neg hear ‘All these five items, the generals cannot afford to ignore.’
(Xun Zi)
The use of zhe as ‘light’ noun, which we here represent with the symbol n (see also Aldridge 2008), is highlighted in (4). Lexical nouns, represented by N, could alternate with zhe, and for the most part in Old Chinese, they show complementary distribution, suggesting that zhe still retained strong nominal features.3
(4) [Quantifier Demonstrative Numeral n/N]
[fan [all
ci si/ wu zhe/ren/wu] these four/five ‘one(s)’/people/things]
The nominal status of zhe was also evident in texts from the Eastern Han period, often identified as the transition period between Old and Middle Chinese. In (5), an example from Lun Xing (97–220 AD), zhe clearly occupies the position of a head noun with
.â•… As we shall see later in §2.2, co-occurrence of light noun zhe (n) and another lexical head noun (N) in Middle Chinese was symptomatic of the reanalysis of zhe as a relativizer.
 Foong Ha Yap & Jiao Wang
relativizer zhi linking it to pre-nominal modifying expression wei yun ‘so-called falling things’. Given that genitive zhi abhors the absence of a head noun, there is good reason to treat zhe in this particular construction as a light noun, in this particular case referring to ‘falling stars’. The resulting construction has the following literal reading: ‘those so-called ‘falling thing’ kind of stuff ’, with zhe functioning as the light head noun (in this context equivalent to English general noun expression kind of stuff).
(5) 其谓霣之者,皆是星也
qi wei yun zhi zhe, jie shi xing ye those call fall rel zhe all cop star sfp ‘Those so-called falling things, in fact are all stars!’
(Lun Xing)
Light noun zhe could also be used to refer to abstract referents such as ‘reason’, as seen in (6), an example taken from a Western Han text, Nan Jing (225 BC). Worth noting is that, in Old Chinese, nouns could be modified without recourse to an overt linker such as genitive or relativizer zhi.4 Structurally, absence of an overt linker makes it easier for light noun zhe to be grammaticalized into a nominalization marker (i.e. nominalizer).
(6) 望而知之者,望見其五色
[[wang er zhi zhi] zhe] wang jian qi wu se [[watch conn know it] zhe ] watch see his five colors ‘The reason that one watches and knows it (=the sickness) is that one watches and observes his (=the patient’s) different color tones …’ (Nan Jing)
2.2â•… Zhe as a derivational nominalizer Many expressions involving nominalizer zhe have become lexicalized as a consequence of frequent usage. These expressions include sha zhe (‘killer ~ one that kills’, a [verb] + zhe construction) and zhi zhe (‘the wise ones ~ those that are wise’, an [adjective] + zhe construction). These high frequency zhe expressions constitute instances of lexical nominalization, with light noun zhe functioning as a derivational (i.e. lexical-type) nominalizer. A close link between lexical and clausal nominalization has also been observed in other languages (see for example Genetti (this volume) for crosslinguistic discussion on Tibeto-Burman languages; likewise see Liu and Gu (this volume) on Nuosu Yi; see also Sung (this volume) and Yeh (this volume) for similar discussions with respect to the Austronesian languages Budai Rukai and Saisiyat respectively; see also Shibatani and Khaled Awadh (2009) on Soqotri, a Semitic language spoken on the island of Soqotra off the southern coast of Yemen).
.╅ Note that insertion of relativizer zhi in this particular example would have been excessive in terms of homonymy, given that the relative clause already makes use of object pronoun zhi.
From light noun to nominalizer and more 
Temporal expressions such as gu zhe ‘in ancient times’, xi zhe ‘in former times’, jin zhe ‘at the present time’, and shi zhe ‘in times past, in the past’ also involve the use of zhe as a light noun. In these lexicalized expressions, zhe was (and still is) used to refer to time as an abstract entity. Similarly, light noun zhe could be modified by locative nouns to yield lexicalized expressions with both spatial and temporal orientations. Examples of [locative noun] + zhe constructions include qian zhe (lit. ‘front + zhe’) and hou zhe (lit. ‘back + zhe). The former was used spatially and temporally in the sense of both ‘in front’ and ‘before, previously’, while the latter was (and still is) used in the opposite sense of ‘at the back’ and ‘later, in the future’. Having examined evidence of zhe being used as a ‘light’ noun in Old Chinese, as well as its development into a derivational nominalizer, we now turn our attention to the use of zhe in clausal nominalizations, focusing in particular on its role in relative clause constructions.
2.3â•… From light noun to clausal nominalizer Crosslinguistically, it has been observed that light nouns often develop into nominalizers. This phenomenon is well attested among verb-final (OV) languages with pre-nominal modification (see, for example, DeLancey 1986 for Lhasa Tibetan; Horie 1998 for Japanese; Rhee 2008 for Korean). Although Modern Chinese is a non-verb final (VO) language, it nevertheless relies on pre-nominal modification. This pattern is typologically rare (Greenberg 1966; Dryer 1992), and may represent retention of verb-final (OV) characteristics from earlier prehistoric times.5 In this section, we will highlight how availability of pre-nominal modification constructions can help provide favorable structural conditions for the grammaticalization of light nouns into nominalizers within the Chinese language, and in some instances giving rise to attitudinal sentence final particles as well. In Old Chinese, light noun zhe was often used in contexts where the referent needed to be further specified, as in the expression [_ i zhi wo]i zhei (‘the onesi (whoi) [_ i know me]i’) shown in (1) earlier. Such usage facilitated the reinterpretation of light noun
.â•… By ‘prehistoric times’, we mean a period that predates Old Chinese, since both VO and OV word order were of comparable frequency in available written records from the Old Chinese period (Xu 2006). There has clearly been a shift toward dominant VO order in Middle and Modern Chinese. Whether Old Chinese represents an intermediate stage in a shift from OV to VO word order is unclear and deserves further investigation. In this regard, it is worth investigating whether embedded/subordinate clauses proved more conservative and resistant toward word order shift, given findings from previous works that subordinate clauses tend to be more conservative than main clauses in terms of language change (e.g. Bybee 2001).
 Foong Ha Yap & Jiao Wang
zhe as an agent nominalizer. That is, the zhe-headed relative clause structure [_i VP]i pre-nominal modifier [zhei] head noun readily yielded an agent nominalization construction with the meaning ‘one that VPs’. This type of agent-oriented nominalization construction was productive in Old Chinese and has survived to the present-day in formal registers of Modern Chinese.6 In (7), an example attributed to Confucius (Early Warring States period, ca. 5th century BC), the presence of demonstrative fu contributes a definiteness reading to the clause nominalized by zhe – yielding the interpretation ‘Who is the one driving the carriage?’ Worth noting is that demonstrative fu occurs before the [VP zhe] nominalization construction in the same pre-nominal position that demonstrative ci occurs before zhe as a light noun, as seen in (2) and (3) earlier, and essentially for the same reason, to provide a definiteness interpretation. In short, whether interpreted more lexically as a light noun or more grammatically as an agent nominalizer, zhe can often be accompanied by noun phrase elements, such as demonstratives.
(7) 夫執輿者為誰
fu zhi yu zhe wei shei dem control carriage zhe cop who ‘Who is the one driving the carriage?’
(Lun Yu, cited in Aldridge 2008: 1)
In (8), an example from Zhan Guo Ce, a text from the Han period (206 BC to 25 AD), light noun zhe served as the lexical head noun in an expression that referred to folks who did not know of the custom of women in Zheng and Zhou painting their faces white and black and standing along the highways and alleyways, and who as a result sometimes mistook them for non-earthly beings. Worth noting here is that light noun
.â•… In fact, Modern Chinese agent-nominalizer de developed via analogy to classical agentnominalizer zhe. In formal syntactic analyses, Whitman (2001, 2005) has posited the possibility of ‘relabelling’, via structural pruning, which gives rise to simpler, more economical structures. Consider the following example where light noun zhe was reinterpreted as a complementizer (i.e. a type of nominalizer) in Old Chinese, with Middle and Modern Chinese nominalizer de developing via analogy to complementizer/nominalizer zhe:
(i)
Old Chinese zhe
[CP [IP _ [VP zhi wo]]] [n zhe] → [CP [IP _ [VP zhi wo]] zhe] know me zhe know me zhe ‘[the ones [(that) [ know me]]]’ ‘[those who/that [know me]]’
(ii)
Middle and Modern Chinese de (via analogy to zhe)
[CP [IP _ [VP zhi wo]] de] know me de ‘[those who/that [know me]]’
From light noun to nominalizer and more 
or nominalizer zhe could be modified by a variety of clausal constructions, including coordinate ones as seen in the expression [fei zhi er jian zhi] zhe (‘the ones [that do not know of this custom and that see them (i.e. the painted women)]’).
(8) 張子曰: 彼鄭周之女, 粉白墨黑,
Zhang Zi yue: pi Zheng Zhou zhi nue, fen bai mo hei, Zhang Zi say: that Zheng Zhou gen female powder white ink black
立於衢閭, 非知而見之者, 以為神. li yu ju lu, [fei zhi er jian zhi] zhe, yi stand near highway alley [not know and see them] zhe, think wei shen as god ‘Zhang Zi said: “Those women from Zheng and Zhou, painting their faces white and black, and standing along highways and alleyways, those not knowing of this custom and seeing them will think they are gods.”’ (Zhan Guo Ce)
Like some of the examples in §2.1 earlier, nominalization constructions such as (8) above, which involve the pre-nominal modification of light noun zhe, are structurally equivalent to a subset of relative clause constructions that are sometimes referred to in the literature as ‘headless’ relative clauses.7 In terms of terminology, the ‘headless’ relative clause (involving light nouns such as zhe in Old Chinese) is often contrasted with the ‘headed’ relative clause, which necessarily requires the presence of a semantically substantive lexical head noun. Their structural difference is highlighted in (9) below. Crucially, we see that ‘headless’ relative clauses –at least the zhe-nominalization type –are not actually headless, but are in fact headed by a light noun. In a sense, then, the term ‘light-headed’ relative clause is more appropriate for zhe-type nominalization constructions, rather than the term ‘headless’ relative clause. As shown in (9a), agent nominalizer zhe is in effect a grammaticalized light noun.8
.â•… The term ‘S-pronominal’ has also been used in the literature (e.g. Horie 1998; Yap, Matthews & Horie 2004; Shinzato, this volume). This choice was motivated by a desire to highlight the transition between pronominal ‘one’ uses of light nouns to their ‘headless’ relative clause uses. The term NP-pronominal was used for pronominal ‘one’, while the term S-pronominal was coined for ‘light-headed’ relative clauses, otherwise referred to as ‘nominalization constructions’. Polinsky (2008) refers to the NP-pronominals as ‘N-pronouns’, distinguishing them from ‘D-pronouns’. In Chinese, light noun zhe would be an N-pronoun, while demonstrative/person pronoun zhi (‘this, that, him, her, it’) would be a D-pronoun. .â•… It is interesting to consider what grammatical role zhe assumes in ‘headed’ vs. ‘lightheaded’ relative clause constructions. In cognitive terms, it makes a lot of sense to posit that the role of zhe within a headed relative clause is that of ‘a nominalizer reanalyzed as a relativizer’, such reanalysis being motivated by higher efficiency in processing time. In other words,
 Foong Ha Yap & Jiao Wang
(9) a.
Agent nominalization (or ‘light-headed’ relative clause): [ _ i vo] zhe i [ _ i zhi wo] zhe i [ _ i know me] one(s) i [ _ i know me] nmz i = ‘the ones that know me’
‘Headed’ relative clause with nominalizer zhe reanalyzed as relativizer zhe: [[ _ i VO] zhe i ] i n i [[ _ i ding yin] zhe i ] i jiang shi i [[ _ i settled in Yin] ‘one(s)’ i ] i generals i [[ _ i settled in Yin] nmz i] i generals i = ‘the generals, the ones who are settled in Yin’ => [[ _ i settled in Yin] rel] generals i = ‘the generals who are settled in Yin’ b.
The example for the ‘headed’ relative clause structure in (9b) was attested in the Western Han period (206 BC to 25 AD). This example from the transition period between Old and Middle Chinese, with jiang shi ‘generals’ as head noun, is reproduced with fuller context in (10) below. As highlighted in (9b) above, the use of zhe in ‘headed’ relative clause constructions such as (10) was a secondary development based on its more basic use in nominalization constructions (for discussions of similar phenomena in Tibeto-Burman languages, see DeLancey 1986, 2002; Genetti, 1991 and this volume; Genetti et al. 2008; LaPolla 1994, 2008).9 (10) 項王怒, 將誅定殷者將史 xiang wang nu, jiang zhu [[ding yin zhe] jiang shi] personal.name king angry will kill settle place.name zhe general.officer ‘The king of Xiang became angry, (and said) he would kill those generals who (Shi Ji) are settled in Yin.’
This secondary nominalizer-to-relativizer development was made possible by a simple process of apposition, whereby the nominal interpretation of a clause nominalized by zhe was further reinforced by the additional presence of an explicit lexical noun. This appositive full (as opposed to ‘light’) noun came to be reanalyzed as the external head
the assumption here is that clausal embedding (facilitated by ‘routinization’, or frequent usage) would lead to faster processing. It would be interesting, in future research, to examine whether simple juxtapositioning of two nominal constituents – for example, positioning of a modifying predicate nominalized by zhe (or de) followed by a lexical head noun – would be slightly more costly in terms of processing time. .â•… Picus Ding (personal comm., April 2009) noted that the same sentence appears without adnominal zhe in Zi Zhi Tongjian, indicating that the use of adnominal zhe in Middle Chinese was optional.
From light noun to nominalizer and more 
noun (N) of the secondary relative clause construction, while the ertswhile nominalizer zhe came to be reinterpreted as the ‘relativizer’ (see also Aldridge 2008 for discussion of light noun zhe developing into a relativizer with the decline of zhi in Middle Chinese). Usage of zhe as a ‘nominalizer-turned-relativizer’ in a ‘headed’ relative clause construction such as (10) was rather sparse in Old Chinese, and even in Middle Chinese this development was generally seen to be a brief encroachment of nominalizer zhe into the territory of genitive and relativizer zhi, particularly during the Sui period (581–618 AD) when the use of zhi began to decline noticeably as vernacular texts became increasingly popular (Cao 1995, 1999). By Late Middle Chinese, during the Tang and Song period (618–1279 AD), the relativizing functions of zhi and zhe were gradually taken up by a new wave of relativizers, namely, di and de (Lu 1943; Jiang 1999; Feng 2000).10 A similar fate befell other adnominal uses of zhe, including the genitive-type uses, a phenomenon which we now turn to.
2.4â•… Rise and fall of adnominal uses of zhe Given its semantically bleached status, it was not surprising to find light noun zhe sometimes being accompanied by additional (initially appositive) lexical head nouns to explicitly clarify the intended referent, as in relative clauses such as (10) above and in other adnominal constructions such as (11) to (13) below. In (11), an example from the ancient odes of the Zhou period, zhe refers to The Almighty One in Heaven, i.e. God, with the help of pre-nominal modifier cang ‘enormous, expansive, limitless’ to allude to His greatness. To ensure proper identification of the intended referent, a more explicit lexical head noun is typically added, in this case another term tian ‘sky, heaven’, yielding an unambiguous idiomatic expression which distinctly identifies the ‘almighty one’ as a heavenly being, i.e. God. (11) 彼蒼者天, 殲我良人
bi cang zhe tian, qian wo liang ren that enormous zhe sky kill 1sg good person ‘God, O God, (lit. The Almighty One, Heaven), killing this fine person of mine (and taking him away from me).’ (Shi Jing: Qin Feng)
The expression bi cang zhe tian structurally appears to be like a relative clause with tian ‘Heaven’ as the head noun and zhe as the relativizer. However, it is interpreted by
.╅ Genitive zhi still survives in the formal register of contemporary Chinese, and as noted in Shi & Li (2002), di/de did not develop genitive functions till relatively late, apparently in deference to the supremacy of zhi within the genitive domain. This also highlights an interesting developmental sequence, in which the development of relativer di/de preceded that of genitive di/de, similar to the sequence seen for relativizer and genitive zhe.
 Foong Ha Yap & Jiao Wang
Chinese speakers as an idiomatic expression referring to the Almighty, and not literally to the blue sky above. In other words, zhe in (11) was not a relativizer linking the modifying expression cang ‘enormous, expansive’ to a head noun such as tian ‘sky’ to produce the meaning ‘the sky that was expansive’ or ‘the endless sky’. Rather, tian ‘sky’ was used figuratively to refer to ‘Heaven’ (and metonymically to ‘God’), and it was structurally juxtaposed with a nominalized clause bi cang zhe ‘the Almighty One’. The appositive (rather than relativizing) function of zhe observed in (11) would thus be consistent with claims by Chinese scholars that zhe did not develop into a relativizer until Middle Chinese (e.g. Aldridge 2008). In (12), pre-nominal modification of zhe with the reduplicated verbal expression huang huang ‘flourish flourish’ produced the nominal expression ‘(the) flourishing ones’, and when lexical noun hua ‘flower(s)’ is further added to huang huang zhe, we obtain the more explicit but still generic interpretation ‘(the) flourishing flowers’ rather than the more specific interpretation ‘(the) flowers that are flourishing’. (12) 皇皇者華, 于彼原濕 huang huang zhe hua, yu bi yuan shi flourish flourish zhe flower loc that hill wetland ‘Flourishing flowers, all over the hills and wetlands’
(Shi Jing: Xiao Ya)
Such use of additional (appositive) head nouns in the ancient odes of the Zhou period, however, suggests that the potential of reanalyzing nominalizer zhe as a relativizer was already structurally viable in Old Chinese. However, for several hundred years, further development along this path was not vigorously pursued, apparently in deference to the more well-established genitive and relativizer zhi (Feng 2000). As noted in §2.3 above, in Middle Chinese there was a brief flurry of zhe extending into the relative clause domain, when zhi began retreating from vernacular use during the Sui period (581–618 ad). A similar extension of zhe within the genitive domain of zhi was attested as well. Aldridge (2008) posits the following account: as genitive zhi in constructions such as (13) began to decline, zhi-less genitive constructions such as (14) began to predominate and pave the way for nominalizer zhe to be reanalyzed as a genitive marker, as seen in (15). As seen in (16), during the same period, there was also analogous extension involving nominalizer zhe being reanalyzed as a relativizer. (13)
子 食 於 有 喪 者 之 側,未 嘗 飽 也。
zi shi yu [[you sang zhe] zhi ce], wei chang bao ye master eat at be.in mourning zhe zhi side not ever sate prt ‘When the master dines at the side of someone who is in mourning, he never eats his fill.’ (Lun Yu 7, cited in Aldridge 2008: 15)
From light noun to nominalizer and more 
(14) 臣 願 得 笑 臣 者 頭。 chen yuan de [[xiao chen zhe] ___ tou] 1sg want get laugh me zhe (zhi) head ‘I want the head of the one who laughed at me.’ (Shi Ji, cited in Aldridge 2008: 15) (15) 聞 弦 者 音 wen [xian zhe yin] hear string zhe sound ‘hearing the sound of string(s)’ < ‘hearing the sound of stringed instrument(s)’ (Zhan Guo Ce, adapted from Aldridge 2008: 15) (16) 項王怒, 將誅定殷者將史 xiang wang nu, jiang zhu [[ding yin zhe] personal.name king angry will kill settle place.name zhe ___ jiang shi] (zhi) general.officer ‘The king of Xiang became angry, (and said) he would kill those generals who are settled in Yin.’ (Shi Ji)
Note that zhe in (13) is clearly still a nominalizer, with canonical genitive zhi visibly doing the work of adjoining the modifying expression you sang zhe (‘someone who is mourning’) to the lexical head noun ce (‘side’). In (14), zhe could still be used as a nominalizer (as in xiao zhen zhe ‘the one who laughed at me’), very likely followed by a slight pause in the absence of an overt or explicit genitive such as zhi before the lexical head noun tou (‘head’). The example in (15), on the other hand, is a bridging example, with zhe interpretable either as a nominalizer (as in xian zhe ‘the stringed thing(s)’ referring to stringed instruments) or as a genitive linker (as in xian zhe yin ‘the sound of strings (or stringed instruments)’). Note that the relationship here between the modifying expression and the head noun is one of ‘associativity’ (e.g. ‘the sound of strings’, meaning the sound belonging to or associated with strings). The construction with zhe in (16) yields a relative clause interpretation rather than a genitive (possessive or associative) interpretation. This is because the pre-nominal modifying expression ding yin zhe (‘those who are settled in Yin’) is co-referential with its following lexical noun jiang shi (‘generals’). In other words, the relationship is one of ‘identity’, not ‘associativity’ nor a ‘part-whole relation’ nor a ‘possessor-possessee relation’. Like (15), the example in (16) is also a bridging example, only this time zhe is interpretable either as a nominalizer (ding yin zhe ‘those who are settled in Yin’) or as a relativizer (ding yin zhe jiang shi ‘those generals who are settled in Yin’). As has often been discussed in the literature (see Givon 2009 for a recent updated summary), powerful principles of economy in cognitive and articulatory processing would tend to coerce juxtaposed constructions that are semantically related to be packaged under a single intonation unit instead of two. This cognitive principle is at work in situations
 Foong Ha Yap & Jiao Wang
such as (15) and (16) above, where nominalizer zhe becomes reanalyzed as genitive or relativizer zhe respectively. This type of extended use of nominalizer zhe as a genitive and relative clause marker was most clearly seen in Late Middle Chinese, as shown in (17) and (18) below, both examples from the Song period (960–1279 AD). (17) 父母載女來訪所謂曾秀才 fu mu zai nu lai fang [suo wei zeng xiu zai father mother take daughter come visit so.called (name of person) 者舟 不見 zhe zhou] bu jian zhe boat neg see ‘The parents took their daughter to visit the boat of the so-called scholar Zeng, but they couldn’t find it.’ (Yi Jian Zhi: Zeng Lu Gung) (18) 昨日来者太师官煞近上 [zuo ri lai zhe tai shi guan] sha jin shang yesterday come zhe emperor’s teacher very close up ‘The Emperor’s teacher who came yesterday was very close with the emperor.’ (Yi You)
As noted earlier in §2.3, both the genitive and relativizer uses of zhe began to decline toward the end of Middle Chinese. Their decline facilitated the rise of super-nominalizer di (and its phonological variant de), which assumed not only the adnominal functions of zhe but increasingly its agent nominalizer function as well.
2.5â•… Rise and fall of possessee pronominal zhe Possessee zhe constructions were also attested in Late Middle Chinese during the Tang and Song period (618–1279 AD), as seen in (19) and (20) below. These were [NPpossessor] [light noun zhepossessee] constructions, which paralleled the [NPpossessor] [NPpossessee]-type possessive constructions that were already prevalent in Old Chinese. In (19), we see light noun zhe modified by first person pronoun wo (‘I’), yielding the possessive interpretation wo zhe ‘my one’ (i.e. ‘mine’). In (20), light noun zhe is modified by wh-interrogative pronoun shei (‘who’), yielding the interrogative possessive interpretation shei zhe ‘whose one’ (i.e. ‘whose’). Essentially, the semantic lightness of zhe gives rise to its indefinite pronominal characteristic, similar to English pronominal ‘one’, and allows possessee zhe constructions to highlight the possessor NP rather than possessee zhe itself. This ‘possessor-focus’ interpretation associated with possessee zhe constructions came in especially handy in contrastive contexts such as (19) and (20) below.11 .â•… Note that this possessor-highlighting function could not be accomplished by genitive zhi since zhi requires the presence of an explicit lexical head noun.
From light noun to nominalizer and more 
(19) 麥田佔他家, 竹園皆我者 mai tian zhan ta jia zhu yuan jie [wo zhe] wheat field occupy 3sg home bamboo garden all 1sg zhe ‘Wheat fields occupy his home; the bamboo gardens all are mine (lit. ‘my ones’)’ (Poetry of Han Shan) (20) <書解> 誰者最好 ? 莫是東坡為上否?
shu jie [shei zhe] zui hao, mo shi dong bo shu (book title) who zhe most good, not true (name of person) book
wei shang fo cop best neg
‘As for Shu Jie, whose one is best? Is it not true that Dong Bo’s book is best?’ (Zhu Zi Yu Lei)
Previous studies have noted that these possessee pronominal zhe constructions emerged via analogy to possessee pronominal suo constructions (see Jiang 1999 for an excellent summary; see also §3.4). The latter was already attested in the Han period, but its eventual decline paved the way for a brief rise in possessee pronominal uses of zhe in Late Middle Chinese during the Tang and Song period. To sum up, we have thus far examined how the light noun zhe developed into an agent nominalizer in Old Chinese, and extended its function into adnominal domains such as genitivization and relativization. In addition, we have seen that in Middle Chinese light noun zhe further developed along an independent pathway into a possessee pronominal that highlighted the possessor NP rather than the possessee NP. In the next section, we will retrace our steps and return to some earlier grammatical and pragmatic functions of light noun zhe in Old Chinese that lost ground in Middle Chinese and did not survive into Modern Chinese. We turn our attention in particular to the subordinating and sentence final mood marking functions of zhe. Both these functions represent extended uses of [VP zhe] nominalization constructions into nonreferential domains.
2.6â•… Subordinating functions of zhe Crosslinguistically, numerous studies have noted a close relationship between nominalization and subordination (e.g. Harbsmeier 1981; Genetti 1991; Haspelmath & KÖnig 1995; Horie 1998b; Yap & Matthews 2008; Deutscher 2009; Kaufman this volume). This close relationship was also evident in the case of zhe, particularly in Old Chinese, as it extended its function from a light noun to a nominalizer, and further to a subordinate clause marker. We elaborate on this development below. As seen in (21), from Sun Zi, an Old Chinese text from the Spring and Autumn period (770–476 BC), zhe could be used either as an agent nominalizer, as in (i), or as a means nominalizer, as in (ii), or as a purposive complementizer, as in (iii). In other
 Foong Ha Yap & Jiao Wang
words, usage of zhe, when modified by pre-nominal expressions such as gong er bi sheng (‘attack and surely win’), could yield an agent nominalization such as ‘those who attack and win’, or a way/means nominalization construction equivalent in meaning to the English reduced relative clause the way to attack and surely win, or a purposive complementation clause equivalent in meaning to the English purposive infinitive to attack and surely win. (21) 攻而必勝者攻其所不守也 [gong er bi sheng zhe ] gong qi suo bu shou ye [attack and certainly win zhe] attack 3p all not defend sfp i. ‘Those who attack and win attack when they (= the enemy) are not prepared.’ ii. ‘The way to attack and surely win is to attack when they (= the enemy) are not prepared.’ iii. ‘To attack and surely win is to attack when they (= the enemy) are not prepared.’
In addition, clauses nominalized by zhe could appear in topic position and be used as conditional clauses. In (22), an example from the Warring States period (475–221 BC), we see zhe used as a light noun (or nominalizer) referring to ‘a hypothetical situation in which the peasants are lazy’. As pointed out in Harbsmeier (1981: 210–217), situational use of zhe such as this yielded a conditional reading. In such contexts, zhe easily came to be identified as a marker of a dependent or subordinate clause. (22) 農夫惰於田者則國貧也 [nong fu duo yu tian zhe] ze guo pin ye farmer lazy loc field zhe then country poor sfp ‘When/If the peasants are lazy in the field, then the state will be poor.’ (Han Fei Zi)
Zhe also sometimes appeared with explicit conditional subordinators such as ruo ‘if ’, as seen in (23). Such usage continued into the Song period (960–1279 AD). As an explicit subordinator, ruo could occur with or without zhe to yield a conditional interpretation. Its presence signals that the [VP zhe] construction here must not be interpreted as an agent nominalization construction within the conditional clause (i.e. *‘if the one who disobeys the order’). Rather, the presence of ruo explicitly marks a non-referential use of the [VP zhe] construction – in this case as a subordinate conditional clause. (23) 若犯令者罪死不赦 ruo fan ling zhe zui si bu she if disobey order zhe sin die neg pardon ‘If anyone disobeys, condemn them to death without pardon.’ (Guan Zi 77, 3.84–1)
From light noun to nominalizer and more 
The link between nominalizer zhe and subordinator zhe can easily be seen in bridge examples such as (24), taken from Xun Zi, another Old Chinese text from the Warring States period. Both indeterminacy of context (typical of generic conditionals) and structural ambiguity easily gave rise to two possible interpretations. The more concrete interpretation would yield an agent nominalization interpretation (‘those (< ‘the ones’) who obeyed’), while the more abstract interpretation would yield a focus on the situation, producing a subordinate adverbial clause with conditional reading (‘if anyone obeyed’). (24) 順者錯之
shun zhe cuo zhi obey zhe leave him/them i. ‘Those who obeyed, (he) left them alone.’ ii. ‘If anyone obeyed, he left them alone.’ (Xun Zi)
Essentially, then, as agent nominalizer, zhe still preserved a referential function, while as marker of clausal subordination, it served instead a topicalizing function (in the sense of Haiman 1978), i.e. it functioned as a mental staging device to prepare the hearer for new information in the upcoming clause (see also Fauconnier 1985; Rubba 1996). In this regard, the use of light noun or nominalizer zhe to represent more abstract referents such as means, manner, reason, cause, time, condition, etc. readily gives rise to adverbial-type subordinate clauses that modify the upcoming matrix clause.12 As noted earlier, extensions of light nouns (or nominalizers) into subordinating functions are robust across languages. In the case of Japanese and Korean, evidence that subordinate clauses are often derived from nominalizations can be seen in the fact that subordinators are often formed from a combination of nominalizers plus case markers (Horie 1998; Yap & Matthews 2008). This is highlighted in (25) below. (25) Japanese: no de no ni
(nominalizer + case marker > ‘because’) (nominalizer + case marker > ‘although’)
Korean: -n kes-i-ni (adnominal + light noun/nominalizer + linker + case marker > ‘because’) -ul kes-ul (adnominal + light noun/nominalizer + case marker > ‘although’)
.â•… An appositive full lexical head noun (N) is usually absent, although in Modern Chinese conditional uses of de hua show evidence that a light noun such as de (whether reanalyzed as nominalizer or relativizer) can merge with an appositive head noun such as hua ‘saying’ to yield a conditional subordinator de hua (‘say, suppose, if ’).
 Foong Ha Yap & Jiao Wang
Similar development from nominalization to subordination (with or without case markers) has long been observed in Tibeto-Burman languages (see especially Genetti 1991; see also Genetti this volume). While textual evidence indicates that both nominalizing and subordinating functions were already attested for zhe in Old Chinese, the former function was far more dominant, not only in quantitative terms, but often also as the more basic interpretation. That is, in the case of subordinate clauses, it was often still possible to reconstruct a light noun interpretation for zhe, very frequently in terms of abstract nominal concepts such as ‘way, manner, situation, time, reason, cause, condition’. These language-internal and crosslinguistic observations point to a light noun > nominalizer > subordinate clause marker development.
2.7â•… Sentence final mood particle uses of zhe Although much less visible in terms of frequency, zhe was sometimes also used in sentence-final position to express modal functions. As seen in (26) below, from the Warring States period, sentence final zhe was often functionally and structurally ambiguous between a nominalizer and a speaker mood marker. That is, zhe could be interpreted as a light noun equivalent to English pronominal one or someone, yielding a nominalization interpretation for the expression bajiushi zhe in the sense of ‘(some) one in his eighties or nineties’, or it could be interpreted as a sentence final particle, in this case indicating surprise on the part of the speaker. In this respect, sentence final zhe behaved in much the same way as numerous other sentence final particles of nominalizer origin, among them contemporary Mandarin de, Cantonese ge3, and Chaozhou kai (e.g. Xu & Matthews this volume; Yap & Matthews 2008; Yap, Choi & Cheung, in press). (26) 且年未盈五十, 而諄諄焉如八九十者 qie nian wei ying wushi er zhunzhun yan ru bajiushi zhe still age not fill fifty yet long.winded prt as(if) eighty-ninety zhe ‘(He’s) not yet fifty yet is so long-winded like one in his eighties or nineties.’ / ‘(He’s) not yet fifty yet is so long-winded as if he is in his eighties or nineties!’ (Zuo Zhuan)
Typical of sentence final particles of nominalizer origin (see for example Yap, Matthews & Horie 2004), the speaker mood interpretation tends to be highly dependent on context. Thus in interrogative contexts, often marked by interrogative particles such as an ‘where’ as seen in (27) from the Warring States period, nominalizer zhe in sentence final position could be reinterpreted as a sentence final interrogative mood particle. Such reinterpretation could arise naturally since nominalizer-derived sentence final particles such as zhe (and Mandarin de, Cantonese ge3, and Chaozhou kai, among many other cognates in other Chinese dialects for that matter) are semantically bleached ‘hat-racks’ or ‘hitching posts’ for the speaker’s sentence final
From light noun to nominalizer and more 
prosody, and thus could be associated with a wide range of speaker’s mood, including assertive mood, interrogative mood, dubitative mood, etc.13 (27) 安見方六七十如五六十而非邦也者 an jian fang liuqishi ru wuliushi er fei bang where see size six–or-seventy or five-or-sixty and neg country ye zhe cop zhe ‘Have you ever seen territories sizing up from 50 to 70 (acres of land) and not being treated as a country? (Lun Yu)
Although low in frequency, the use of sentence final particle zhe lasted for a long time, spanning a period of at least two thousand years. It was attested as early as the Warring States period as seen in (27) above, and persisted at least into the Qing period as seen in (28) below. (28) 誰為我臨期成此大節者 shui wei wo lin qi cheng ci da jie zhe who for me come period fulfill this big moral.deed zhe ‘When the time comes, who will help me fulfill this big ethical deed?’ (Mei Hua Ling Ji, Quan Zu Wang, 18th century)
To sum up, we have seen how nominalizer zhe in sentence final position could sometimes be reinterpreted as a sentence final mood particle. This comes as no big surprise given that semantically light (or bleached) particles in sentence final position are ideally situated to host speaker’s sentence final prosody. This development forms part of a natural and robust process in which versatile light nouns often shift away from referential focus toward illocutionary focus, and crosslinguistic parallels have been observed in other Chinese dialects (e.g. Xu & Matthews, this volume) and in many other languages (e.g. Noonan 1997 for Chantyal; Watters 2008 for Kiranti and the Central Himalayish languages; Rhee 2008 and this volume for Korean; Nagaya this volume for Tagalog; Yap this volume for Malay; Yap & Matthews 2008 for a summary discussion).
2.8â•… Summary on the development of zhe Figure 1 below summarizes the functions of zhe in Old, Middle and Modern Chinese. We see evidence of light noun zhe remaining true to its agent nominalizer function throughout recorded history. In Middle Chinese, zhe briefly encroached into the territory of zhi, particularly with respect to adnominal (genitive and relative clause)
.â•… The terms ‘hatrack’ and ‘hitching post’ are borrowed from various studies referring to copulas as ‘carriers of tense marking’ (e.g. Pustet 2003).
 Foong Ha Yap & Jiao Wang
functions, before being displaced by di/de. Subordinate uses of zhe were attested very early, with examples seen in Old Chinese texts from the Spring and Autumn period (770–476 BC). This suggests that the relationship between nominalization and subordination is a strong and natural one: nominalizers operate within the referential domain (essentially as referential heads for predicates that are being transformed into arguments), while subordinators operate more broadly on clausal adjuncts, often simply by exploiting the use of a light noun as a marker of abstract referents such as time, means, and hypothetical situations (or conditions). Sentence final uses of zhe were also attested very early and in fact lasted into Early Modern Chinese. Its development as a sentence final mood particle was facilitated by its semantic lightness as well as a natural tendency for speaker’s mood to be conveyed via sentence final prosody. In sum, we have seen how a versatile light noun such as zhe has grammaticalized into an agent nominalizer and how its extended uses shed light on the intimate link between nominalization and adnominalization (i.e. genitivization and relativization), as well as on adverbialization (i.e. adverbial clause subordination associated with topic marking or mental staging functions in discourse) and also on sentence final mood marking. Many of these developments are now well documented in other languages; what the present analysis of zhe does is to provide diachronic evidence of such development using data from Old, Middle and Modern Chinese covering a period of more than two thousand years. Category
Function
Nominal
Referential
Adnominal Modificational
Old Chinese Lexical NMZ
Clausal NMZ
Middle Chinese Lexical NMZ
Clausal NMZ
Associative
Relativizer
Adverbial
Event focusing
Subordinator
Subordinator
SFP
Speaker mood marking
SFP
SFP
Possessive pronoun
Modern Chinese Lexical NMZ
Clausal NMZ
Genitive
Figure 1.╇ Functions of zhe in Old, Middle and Modern Chinese
3.â•… The development of patient nominalizer suo Old Chinese also had another well-known nominalizer of light noun origin, namely patient nominalizer suo, as shown in (29) below. Unlike agent nominalizer zhe, which emerged as a grammaticalized light noun in post-predicate position, patient
From light noun to nominalizer and more 
nominalizer suo appeared as a preposed object in preverbal position. In this position, suo further developed into a pronominal clitic (Ting 2005) capable of turning the whole predicate into a nominalization construction that focuses on the theme or patient.14 (29) 是非爾所知也 shi fei [er suo zhi ] ye this neg you suo know prt ‘This is not something that you know.’
(Meng Zi)
In this section, we will focus on how suo evolved from a locative noun into a patient nominalizer. We first examine the development of locative noun suo into a semantically general (hence ‘light’ and versatile) noun in Old Chinese (§3.1), then further analyze how light noun suo (specifically in preposed object position) came to be reinterpreted as a patient nominalizer (§3.2). Special attention is given to morphosyntactic markings that reinforce the nominal interpretation associated with suo constructions. In addition, we examine the use of suo in subordinate clause constructions, in particular the suo … zhe conditional construction and the suo yi reason/purposive constructions (§3.3). We will also examine the rise of possessee pronominal and genitive suo (§3.4), as well as the use of suo in passive constructions (§3.5). In sum, we will examine the diachronic links of these constructions to the core referential uses of suo nominalization constructions.
3.1â•… Suo as locative noun The etymology of suo (所) can be traced to a lexical noun meaning ‘place’, as attested in examples from Old Chinese texts such as (30) from Lun Yu (Early Warring States period, ca. 5th century BC). This locative noun usage persisted into Middle Chinese, as seen in (31), where suo was used in the sense of ‘land’ or ‘territory’ in a text from the Tang period (618–907 AD). (30) 居其所而眾星共之 ju qi suo er zhong xing gong zhi stay its place and many star surround it ‘(The Polaris) stays in its place and other stars surround it.’
(Lun Yu)
(31) 盤之阻,誰爭子所? pan zhi zu shui zheng zi suo mountain.road zhi block who contest your place ‘The mountain roads being blocked, who would contest your territory?’ (Song Li Yuan Gui Pan Gu Xu: Han Yu)
.╅ See also Ting (2003).
 Foong Ha Yap & Jiao Wang
Locative noun suo was also extended to more abstract uses, from the spatial domain to the social domain to express relational concepts such as ‘position’ or ‘rank’, as in (32), an example from Zuo Zhuan from the Warring States period (475–221 BC). (32) 為人子者,患不孝, wei ren zi zhe huan bu xiao cop people son zhe worry.about neg filial, 不患無所。 bu huan wu suo neg worry.about neg.have position Lit. ‘To be a good son, worry about not being filial, and not worry about not having a high rank’ ‘What a son should be worried about is being filial, not his position.’ (Zuo Zhuan)
Locative noun suo also appeared in compound locative expressions such as wang suo ‘palace’ (< ‘king’s place’), as in (33) from Mencius (Warring States period, 475–221 BC), and si suo ‘place of death’, as in (34) from Shi Ji (Book of History, Western Han period, 206 BC to 25 AD). In contemporary Chinese, the locative sense of suo is still productive and is preserved in locative expressions such as ce suo ‘toilet’, zhen suo ‘clinic’, shi wu suo ‘office’, pai chu suo ‘police station’, etc. (33) 子謂薛居州善士也使之居於王所。
zi wei xue ju zhou shan shi ye shi zhi ju yu wang suo you say (name of person) good man prt let him live at king place ‘You say that Xue Juzhou is a good man and that makes it possible for him to live in the palace.’ (Meng Zi)
(34) 遂入,至紂死所。
sui ru, zhi zhou si suo so enter to (name of person) die place Lit. ‘So (he) came to Zhou’s place of death. ‘So (he) came to the place where Zhou died.’
(Shi Ji)
As a semantically ‘light’ noun, suo also came to be used as a classifier, often for buildings. Such usage was attested in texts from the Eastern Han period, as in (35).15 (35) 離宮別館三十六所。 li gong bie guan san shi liu suo away.from palace other inn three ten six cl ‘Outside the palace were thirty-six inns/guesthouses.’
(Xi Du Fu: Ban Gu)
.╅ This example is from Hanyu Da Cidian (Dictionary of Chinese Words).
From light noun to nominalizer and more 
However, long before we see evidence of suo established as a classifier in Middle Chinese, there was already ample evidence from Shi Jing (Book of Odes) of highly productive uses of suo as a nominalizer dating back at least to the Western Zhou period (1046 -771 BC).16 Evidence from the grammaticalization of light noun suo thus suggests that classifiers and nominalizers could share a common etymology but each could develop independently of the other. We will now focus on the development of suo into a versatile nominalizer in §3.2 below.
3.2â•… From light noun to patient nominalizer In Old Chinese, there was clear evidence that locative noun suo (< ‘place’) had already evolved into a highly versatile and semantically bleached light noun that could be used as a nominalizer to refer not only to locations, as seen in (36) and (37), but also to objects, people, and abstract goals, as shown in (38) to (40). More specifically, in (36), an example from Lao Zi, a text of the Western Zhou period (around 728 BC), we see a bridging context (Heine 2003; Evans & Wilkins 2000), where locative noun suo could be interpreted in either the locative sense of ‘place for an action to happen’ or the more abstract sense of ‘potential or possibility for an action to happen’. In other words, we see a shift from ‘locus of action’ to ‘potential for action’. More specifically, with respect to (36), we obtain the literal reading of ‘no place where the rhino could put its horn to good use’, often yielding the ambiguous sense of ‘nowhere/no way for the rhino to put its horn to good use’; likewise the literal reading of ‘no place where the tiger could flex its claws’ could be reinterpreted as ‘nowhere/no way for the tiger to flex its claws’; similarly, the literal reading ‘no place where the weapon could use its blade’ could be reinterpreted as ‘nowhere/no way for the weapon to be used’.
.╅ Xu and Matthews (this volume) reports on the polysemy of Chaozhou kai, which is used for a wide range of functions including classifier, nominalizer, relativizer, genitive, sentence-final particle, copula, and focus particle. Classifier kai may have been derived from a lexical noun used as a unit of measure in ancient times. In the absence of diachronic evidence, whether nominalizer kai evolved from the classifier usage or more directly from its lexical source is unclear. Classifiers are generally believed to have emerged as a strong paradigm in Early Middle Chinese (Shi & Li 2002); this was about the same time that early uses of classifier suo were attested (e.g. Eastern Han period). As noted above, nominalizer uses of suo were already highly productive long before this period, thus clearly indicating independent grammaticalization pathways for the development of nominalizer suo and classifier suo. It may be possible that classifier and nominalizer uses of kai in Chaozhou (a southern Min dialect) also developed independently as different light noun manifestations of a lexical noun frequently used as a measure unit, though more diachronic dialectal research is needed to resolve this question.
 Foong Ha Yap & Jiao Wang
(36) 兕無所投其角, si wu [suo tou qi jiao] rhino not.have suo put its horn 虎無所措其爪, hu wu [suo cuo qi zhua] tiger not.have suo put its claw 兵無所容其刃 bing wu [suo rong qi ren] weapon not.have suo use its blade ‘The rhino has no place to rest its horn; the tiger has no place to land its claws; the weapon has no place to use its blade.’ Intended meaning: ‘The rhino cannot use its horn; the tiger cannot use its claws; the weapon cannot use its blade.’ (Lao Zi)
Bridge contexts such as (36), where locative noun suo was extended to more abstract uses, from ‘place where an action happens’ to ‘place for an action to happen’ to ‘potential for an action to happen’, paved the way for the emergence of suo as a light noun capable of grammaticalizing into a versatile and highly productive locative/means nominalizer. Below we further examine various nominalizer uses of suo, with a focus on the relationship between nominalization and relativization. In (37), an example from Meng Zi, better known as Mencius in English (Warring States period, 475–221 BC), suo was still used in the locative or goal sense, in this particular instance referring to ‘the place to which (the Master) was going’. From a structural perspective, we see light noun suo, not in postverbal locative or goal NP position as in the examples in §3.1 above, but in preverbal position, yielding a [suoi V_ i] i construction with a locative or goal focus–see (37′). The prominence given to preposed locative/goal argument suo readily gave rise to a ‘locative/goal type’ nominalization construction, with light noun suo conveying a meaning similar to English ‘the place or destination where (someone) VP’. (37) 他日君出, 則必 命有司所之 ta ri jun chu, ze bi ming you si [suo zhi] other day you go.out, then certainly instruct officials suo go ‘On previous occasions when you went out, you always gave instructions to your officials as to where you were going.’ (Meng Zi, 1B/16) (37′) Non-agent nominalizer suo referring to location or goal: [suo i V _ i ] i [suo i zhi_ i ] i [‘place’i go _ i ] i = ‘the place (where) (you) were going’
From light noun to nominalizer and more 
In (38), also from the Warring States period, we see evidence of the use of light noun suo further extending from locative to theme/patient referents such as min zhi suo shi ‘what people eat’. (38) 民之所食,大抵豆飯藿羹 [min zhi suo shi] da di dou fan huo geng people gen suo eat basically beans cuisine beans soup ‘What people eat is basically cuisine and soup made of beans.’ (Zhan Guo Ce) (38′)
Patient nominalizer suo referring to thing or concrete object: [suo i V _ i ] i [suo i shi_ i ] i [‘thing’ i eat _ i ] i = ‘the thing (= what) (people) eat’
As seen in (38) above and also in (39) below, the subject NP was sometimes followed by genitive case marker zhi. In (39), for example, the expression ren zhi suo yuan ‘what people want’ is co-referential with the object of desire in topic NP position, namely ‘being rich and noble’. The presence of genitive zhi marked the subject NP ren ‘people’ as a genitive subject rather than a nominative one, and signaled that the [NP zhi suoi VÂ�_ i] i construction was available for use as either a subordinate clause or a nominal (i.e. entityreferring) expression (see also He 1988).17 In (39), the [NP zhi suoi VÂ�_ i] i construction yielded a theme/patient nominalization interpretation. Similar use of genitive subject to signal that a given construction is to be construed as a nominalized expression is robust in other languages as well (see Kaufman (this volume) and Sung (this volume) for related discussion on Austronesian languages).18 (39) 富與貴,是人之所欲也 fu yu gui, shi [ren zhi suo yu] ye rich and noble this people gen suo want prt ‘Being rich and noble, this is what people want.’
(Lun Yu)
To recap, we thus see two strategies being used to reinforce a nominal interpretation for [suo V] constructions. One strategy is via object preposing; this yields a predicate that is reanalyzed as a nominal expression that highlights the patient or theme NP. The other strategy is via genitivization; it renders the subject NP as ‘possessor-like’ and hence part of a nominal expression, rather than as the agent in a finite transitive clause construction. In other words, the presence of genitive zhi served to overtly reinforce the nominal status of the suo construction. .â•… He (1988) made no reference to genitive case, but noted that the presence of zhi in [NP zhi VP] constructions yielded either gerundive constructions or subordinate clauses. .â•… Whether these nominalization constructions with genitive subjects constitute a type of internally headed relative clause also deserves further investigation.
 Foong Ha Yap & Jiao Wang
The nominalization interpretation often accorded to [suo V] constructions was sometimes further reinforced by apposition to a lexical head noun, as in (40) and (41), or to a light head noun, namely zhe, as in (42) to (44). The former yielded a ‘headed’ relative clause construction, such as ‘the person that I am talking about’ and ‘the Han army that Ji alone has killed’. The latter yielded a nominalization (or ‘light-headed’ relative) clause construction, with light noun zhe being much less explicit about the referent, and giving rise to interpretations such as ‘the thing that one generates from the accumulation of righteousness’, ‘something that heaven gives me’, and ‘the thing that the Di tribe wants’. Not surprisingly, [suo V] zhe nominalization constructions were often used for generic reference (or ‘type/kind’ referring functions). (40) 所謂伊人,在水一方 [suo i wei] i [yi ren] i zai shui yi fang suo say this person at water one side ‘The person that I am talking about is on the other side of the water.’ (Shi Jing: Guo Feng) (41) 獨籍所殺漢軍數百人 [du ji suo i sha] i [han jun] i shu bai ren alone Ji suo kill Han army several hundred people ‘The Han army that Ji alone has killed numbered several hundred people.’ (Shi Ji) (42) 是集義所生者
shi [[ji yi suo i sheng] i zhe i] this accumulate righteousness suo generate ‘thing’ ‘This is the thing one generates from accumulating righteousness.’ = ‘This is what one generates from accumulating righteousness.’
(Meng Zi)
(43) 此天之所與我者
ci [[tian zhi suo i yu wo] i zhe i] this heaven gen suo give me ‘thing’ ‘This is something that heaven gives me.’ = ‘This is what heaven gives me.’
(Meng Zi)
(44) 狄人之所欲者,吾土地也。 [[Di ren zhi suo i yu] i zhe i] wu tu di ye Di people gen suo want ‘thing’ our land land prt ‘What the Di tribe wants is our land.’
(Meng Zi)
Evidence that the [suo V] zhe nominalization construction could be construed as a ‘light-headed’ relative clause can be seen in examples such as (45), from Mencius, where relativizer zhi was recruited as an overt linker between the [NP suo V] patient nominalization construction and the light head noun zhe.
From light noun to nominalizer and more 
(45) 其所取之者義乎 [[qi suoi qu __ i] i zhi zhe i] yi hu he suo acquire rel ‘thing’ righteous prt ‘Is the thing that he acquired righteous?’
(Meng Zi)
Parallel evidence that [suo V] N nominalization constructions were construable as ‘headed’ relative clauses can be seen in examples such as (46), also from Mencius. In this case, relativizer zhi was recruited to link the [suo NEG V] construction to a full lexical head noun chen ‘courtier’. (46) 故將大有為之君,必有所不召之臣 gu jiang da you wei zhi jun bi you [suo i bu zhao thus will great have do zhi ruler surely have suo neg summon zhi chen i] rel courtier ‘Thus a ruler who is to do great things will surely have courtiers whom he does not summon.’ (Meng Zi)
To sum up this section, we have seen how locative noun suo meaning ‘place’ evolved into a light noun capable of referring to non-locative referents as well. As a pronominal clitic (grammaticalized from a light noun) used in preposed object position, suo played a major role in the formation of theme/patient nominalization constructions. It was possible, in Old Chinese, for either a full lexical noun (N) or a light noun (zhe) to sometimes be attached as an external head noun to a [suo V] nominalization construction. The former yielded a ‘headed’ relative clause; the latter yielded a ‘light-headed’ relative clause – also referrable to as a [[suo V] zhe] double nominalization construction. This intimate relationship between nominalization and relativization phenomena is captured in structural terms in Figure 2 below.
Type of [suo V] construction
Structural form
[suo V] nominalization
[(NP) suo V] [(NP zhi) suo V] [(NP) suo V] N [(NP zhi) suo V] N
[suo V] N relativization
Old Chinese
Middle Chinese
Modern Chinese
[suo V] zhe double nominalization (or ‘light-headed’ relativization) [(NP) suo V] zhe [(NP zhi) suo V] zhe [suo V] zhi N relativization [(NP) suo V] zhi N [(NP zhi) suo V] zhi N (not attested in our corpus) [suo V] zhi zhe relativization [(NP) suo V] zhi zhe [(NP zhi) suo V] zhi zhe
Figure 2.╇ Relationship between nominalization and relativization in suo constructions in Old, Middle and Modern Chinese
 Foong Ha Yap & Jiao Wang
3.3â•… Subordinating functions of suo Earlier we noted that there is often an intimate relationship between nominalization and clausal subordination. This close relationship was observed for nominalizer zhe in Old Chinese (recall §2.6; see also Harbsemeier 1981: 210–217), and also for nominalizers (often accompanied by case markers) in other languages such as Japanese and Korean (Horie 1998; Yap & Matthews 2008), as well as a number of Tibeto-Burman languages (see Genetti 1991 and this volume). Below we examine this phenomenon with respect to suo constructions. In particular, we focus on the extended uses of suo constructions with temporal and conditional interpretations, and suo yi constructions with manner/means/reason interpretations, as well as those with purposive complementizer readings.
3.3.1â•… Temporal and conditional subordinator suo Previous studies have already noted the use of suo in adverbial subordinate constructions in Old Chinese (e.g. Harbsmeier 1981: 217–219). As seen in (47), from Zhuang Zi (Warring States period, 475–221 BC), suo nominalization constructions sometimes yielded a subordinate temporal (‘when’) clause interpretation. This was possible because pronominal clitic suo, originally derived from a locative noun, had grammaticalized and extended its semantic range such that it could be used to also refer to abstract notions such as ‘time’. (47) could thus be interpreted as ‘At the time when the Supreme One emerges, there exists its being, but it has no form.’. (47) 一之所起,有一而未形。 [yi zhi suo qi] you yi er wei xing One gen suo rise have One but neg form ‘When the One comes into existence, there is the One, but it has no form.’ (Zhuang Zi 12.38, cited in Harbsmeier 1981: 217)
As seen in (48), from The Analects (Early Warring States period), [suo V] zhe double nominalization (or ‘light-headed’ relative clause) constructions could also be used as subordinate conditional clauses. (48) 予所否者,天厭之 [yu suo pi zhe] tian yan zhi I suo do.wrong zhe heaven reject me (1sg < 3sg) ‘If I have done anything improper, may Heaven’s curse be on me.’ (Lun Yu 6.28, cited in Harbsmeier 1981: 217)
The importance of context in disambiguating between a nominalizer or subordinator use of suo is highlighted in (49) below, where subtle calls for judgment are strongly influenced by contextual cues.
From light noun to nominalizer and more 
(49) a.
君之所為百姓之所從也,
[jun zhi suo wei] [bai xing zhi suo cong] ye ruler gen suo do people gen suo follow prt ‘What the ruler does is what the people follow.’
b. [君所不為]百姓何從? [jun suo bu wei] bai xing he cong ruler suo neg do people what follow ‘When/If the ruler does not perform his job, what can the people follow?’ (Da Dai Li Ji, Warring States)
3.3.2â•… Manner, means and reason marker suo yi In Old Chinese, suo constructions were also used to express ‘manner’, ‘means’ or ‘reason’ for a given situation. This usage is associated with suo yi constructions. As seen in (50), from Mencius (Warring States period), when light noun suo was preposed before lexical verb yi (‘use’), a suo yi nominalization construction obtained that focused on the way/manner in which something was used. (50)
敢問其所以異
gang wen [qi [suo yi] yi] dare ask they suo use differ ‘May I ask in what way/manner they differ?’
(Meng Zi)
In (51), also from Mencius, the suo yi construction could also focus on ways of use in the sense of means. (51) 公事畢,然後敢治私事, gong shi bi ran hou gan zhi si shi public affair finish this after dare deal private affair 所以別野人也. [suo yi bie ye ren] ye suo use distinguish common people sfp ‘Completing their public duty and then dealing with their own business is the means of distinguishing the common people (from those in authority).’ (Meng Zi)
In (52), we see evidence from Mencius that the suo yi construction could also refer to more abstract concepts focusing on cause or reason for a given situation. This more abstract usage was often reinforced by light noun or nominalizer zhe, with both suo yi and zhe co-referring to ‘the reason (why the ancient sages surpassed others)’.
 Foong Ha Yap & Jiao Wang
(52) 古之人所以大過人者,無他焉, [gu zhi ren suo yi da guo ren zhe ], wu ta yan ancient gen people suo yi great than others zhe not.have other prt 善推其所為而已矣。 shan tui qi suo wei er yi yi good.at extend they suo do sfp (mitigative mood) ‘The reason why the ancient sages surpassed others–and there’s no reason other than this–was simply that they were good at extending what they did.’ (Meng Zi)
The nominal status of the suo yi reason construction can often be reinforced by the presence of the genitive subject marker zhi –as seen in (53) and (54). The presence of light noun zhe as external head noun is optional –absent in (53) while present in (54). Both the genitive subject marker zhi and nominalizer zhe thus signal that the suo yi reason construction is in fact a type of nominalization construction, or at least was functionally derived from it. (53) 斯民也, si min ye this.kind common.people prt 三代之所以直道而行也 [san dai zhi suo yi zhi dao er xing] ye three dynasty gen suo yi straight way lnk walk prt ‘The common people were all like that, (and this is) the reason for the three dynasties’ following the Straight Way.’ (Lun Yu) (54) 人之所以異於禽獸者幾希 [ren zhi suo yi yi yu qin shou zhe] ji xi human gen suo yi differ from bird animal zhe very little ‘The reasons for man’s being different from other creatures are quite limited.’ (Meng Zi)
3.3.3â•… Purposive complementizer suo yi In (55), we see evidence of suo yi used as a purposive (‘so that’) complementizer. Note that unlike the suo yi … (zhe) reason constructions in (53) and (54) above, which typically involves the presence of either a genitive subject marker or light noun zhe, the latter as an external head noun, or ‘additional nominalizer’, the suo yi purposive construction could entertain neither genitive subject marker zhi nor nominalizer zhe. A natural consequence of this difference is that the suo yi purposive construction, which is still used in contemporary Chinese, is visibly ‘less nouny’ than its suo yi reason counterpart, as highlighted in the contrast between (55) and (55′) below. Indeed,
From light noun to nominalizer and more 
its more clausal rather than nominal morphosyntactic characteristic renders the suo yi purposive construction in (55) more amenable for use as a purposive complement construction, with suo yi readily reanalyzable as a purposive complementizer. Within Malchukov’s (2006) typological framework on nominalization strategies, the suo yi purposive construction also ranks lower than its suo yi reason counterpart in terms of ‘substantivization cues’ since the former lacks genitive subject marker zhi as a noun phrase marker. The exclusion of the light noun zhe (or de in Modern Chinese) also indicates that the suo yi purposive marker ranks lower in terms of nominal characteristics. (55) 言飽乎仁義也, yi bao hu ren yi ye say full of benevolence righteousness top
所以不願人之膏粱之味也 [suo yi bu yuan ren zhi gao liang zhi wei (*zhe)] ye suo yi neg relish other.people gen meat rice gen taste sfp ‘That is to say, one’s soul is saturated with benevolence and righteousness so that one does not relish the taste of other people’s choice foods.’ (Meng Zi) (55′) 所以不願人之膏粱之味 *(者) [suo yi bu yuan ren zhi gao liang zhi wei *(zhe)] suo yi neg relish other.people gen meat rice gen taste nmz 飽乎仁義也 bao hu ren yi ye full of benevolence righteousness sfp ‘The reason that/why one does not relish the taste of other people’s choice foods is that one’s soul is saturated with benevolence and righteousness.’ (constructed example)
It is also worth noting that the suo yi … (*zhe) purposive construction, as seen in (55), assumes a non-core argument position. That is, unlike the suo yi … (zhe) reason construction which typically occupies subject position, as in (55′), the suo yi … (*zhe) purposive construction occupies instead a purposive complement position. Moreover, in place of light noun zhe as an overt external head noun, the suo yi purposive construction co-occurs instead with sentence-final particle ye. We thus see differences in structure and function between the two suo yi constructions (‘purposive’ vs. ‘reason’) being reinforced by morphosyntactic cues that suggest a more clausal vs. more nominal interpretation.
3.4â•… Possessee pronominal and genitive suo As seen in (56) and (57), suo was also used in possessee pronominal and genitive constructions, the former emerging in the transition period between Old and Middle Chinese (i.e. the Han period) and the latter in Middle Chinese.
 Foong Ha Yap & Jiao Wang
(56) 心持非我所是我所 xin chi fei [wo suo] shi [wo suo] heart hold neg 1sg ‘one’ cop 1sg ‘one’ Lit. ‘(My) heart holds that which isn’t something that I have to be something that I have.’ Intended meaning: ‘The heart treats that which is not mine as mine.’ (Fo Shuo Yi Ri Ni Mo Bao Jing, Han period) (57) 誰能救濟我所壽命 shui neng jiu ji wo suo shou ming who can save 1sg gen life ‘Who can save my life?’ (Zhuan Ji Bai Yuan Jing, Three Kingdom period, 220–265 AD)
There are several competing accounts on the development of locative noun suo into a possessee pronominal and genitive marker. On one account, Liang (1994) claims that possessee pronominal suo constructions were derived from nominalization suo constructions, more specifically via VP-elision. According to Liang’s analysis, a possessive construction such as wo suo (‘my stuff, mine’) in (58b) below was derived from a more extended theme/patient nominalization construction such as wo suo you (‘that which I have or possess’) in (58a) via elision of the verb you ‘have, possess’. This process of VP-elision was argued to have forced a reanalysis of suo from nominalizer to possessee pronominal.19 (58) a.
wo suo you → b. wo suo 1sg nmz have 1sg poss.pron ‘that which I have’ ‘that which is mine; mine’
We suggest, however, that the emergence of possessee pronominal suo may have been facilitated instead by the semantic extension of suo from a locative noun to a general (or ‘light’) noun from a different direction. Note that in Old Chinese the notion of possession was often expressed via simple juxtaposition of the possessor and possessee NPs, such as wang suo (‘palace, or more literally ‘king’s place’) and si suo (‘place of death’, or more literally ‘death’s place’), as seen in (33) and (34) earlier. That is, we see an [NP-possessor] [NP-possessee] word order, where the notion of ‘possessor’ applies not only to possessive relationships but to associative ones as well. The possessive construction could of course have a definite interpretation and be realized structurally as
.╅ Crosslinguistic studies often show an intimate link between nominalization and genitivization (e.g. Matisoff 1972 for Lahu; Horie 1998 for Japanese; Yap & Matthews 2008 for a summary on a wide range of Asian languages). Although related, the two processes could follow underlyingly independent pathways, with many languages opting for different markers for each function (see Rhee 2008 for Korean; Morey this volume for Numhpuk Singpho).
From light noun to nominalizer and more 
[DP-possessor] [NP-possessee]. As seen in (59a) and (59b) below, the possessee NP (namely, suo) could still retain its locative or spatial noun reading, such as ‘place’ and ‘territory’. However, as seen in (59c), the use of suo could also extend to non-locative referents, i.e. it could be used as a general noun in the sense of ‘thing, stuff, etc.’, often equivalent to English indefinite pronoun ‘one’, and hence yield a possessee pronominal interpretation, equivalent in meaning to English ‘my one’ or ‘mine’. (59) a. qi suo ‘its place’ (referring to the position of the star Polaris) b. zi suo ‘your territory’ c. wo suo ‘my place’ > ‘my thing’, ‘my stuff ’ > ‘my one’ > ‘mine’
Reanalysis of locative and spatial nouns into possessee pronominals was fairly productive in Old Chinese and Middle Chinese, with locative nouns xu ‘meeting place’ and di ‘bottom’ following in the footsteps of suo (e.g. Jiang 1994, 1999). In this respect, possessee pronominal zhe (whose lexical origin was very likely non-locative in nature, given its predominantly agentive orientation as discussed in §2.5 earlier) was a relative late-comer, emerging in Late Middle Chinese when the trend for light nouns to extend into possessive contexts was already fairly well-entrenched. Genitive uses of suo emerged when possessive suo constructions (e.g. wo suo ‘my one’) were appositively reinforced by explicit lexical nouns (e.g. shou ming ‘life’), as highlighted in (60) below. Such apposition facilitated the reanalysis of light noun or possessee pronominal suo as a genitive marker. This type of reanalysis involving genitive suo, as in the case of genitive zhe, coincided with the decline of genitive zhi beginning in Late Old Chinese and extending through much of Middle Chinese. (60) a. [wo suo] → b. [wo suo] [shou ming] 1sg poss.pron 1sg gen life ‘my one; mine’ ‘my one, life’ > ‘my life’
In sum, diachronic evidence shows that possessee pronominal uses of suo preceded genitive ones, the former attested in Late Old Chinese and the latter appearing later in Middle Chinese. This pathway, as highlighted in (61) below, began with a locative noun suo (‘room, place’) extending into a light noun with locative and non-locative interpretation (e.g. ‘place’ and ‘thing’), and in possessive constructions, its use as a possessee pronominal of ‘light noun’ origin yielded a strategy which allowed the interlocutors to subtly focus on the possessor NP rather than the possessee NP. This came about naturally, given the semantic lightness of suo in the [NP suo] possessive constructions. A rough approximation in English would be the possessive construction involving examples such as Peter’s one, where the focus of the discourse is on the possessor NP (Peter) rather than the possessee pronominal (one). This is in contrast to possessive constructions with an explicit possessee NP such as Peter’s hat, where the focus could be either on the possessor NP (Peter) or the possessee NP (hat), depending on intonational
 Foong Ha Yap & Jiao Wang
stress. In Middle Chinese, the [NP suo] possessive constructions could be followed by an appositive lexical noun that was co-referential with pronominal possessee suo. This facilitated the reanalysis of suo as a genitive marker. (61) Grammaticalization of suo in possessive contexts
Old Chinese locative noun suo > ju qi suo stay its place ‘stay in its place’
Late Old Chinese possessee pronominal suo > wo suo 1sg place/possession ‘my one; mine’
Middle Chinese genitive suo wo suo shou ming 1sg gen life ‘my life’
There are a number of parallels between the development of possessee pronominal suo into a genitive marker and the development of nominalizer suo into a relative clause marker. Both possessee pronominal suo and nominalizer suo develop as ‘light noun’ extensions from locative noun suo, and both develop a secondary function when followed by an appositive lexical noun. Both these secondary functions–i.e. genitive marking and relative clause marking respectively–are adnominal in nature, and signal the presence of phrases or clauses with specificational content. For both pathways, then, we see an extension from referential uses of suo to non-referential, specificational uses. Because of these strong parallels, some scholars (e.g. Shibatani 2009) have suggested that possessee pronominal constructions such as wo suo are a type of nominalization construction.
3.5â•… Suo in passive constructions Suo nominalization constructions are also found in passive constructions, initially with copula wei. Previous studies have shown that [wei (NP) V] passive constructions were common in Late Old Chinese, but although these constructions persisted into the Han period, they lost ground in Middle Chinese. In their place, [wei (NP) suo V] passive constructions, which were rare in Old Chinese, began to gain increasing acceptance in the Han period, in large part due to the strong association between suo and patient-focus interpretation, and eventually becoming the most common passive form in Early Middle Chinese before being replaced by [bei (NP) suo V] passive constructions in Late Middle Chinese (see Ting 2008 for a summary; see also Tang & Zhou 1985; Peyraube 1989; Pulleyblank 1995). In this section we consider the role of suo in passive environments and examine possible implications for the relationship between nominalization and passivization. We first look at wei-passives without suo, then at wei-passives with suo.
3.5.1â•… From copula wei to case-marker and passive marker wei Evidence from Old Chinese texts from the Warring States period indicates that wei was used as a copula verb meaning ‘be’ or ‘become’, as seen in (62) and (63), as well as an agent marker, as seen in (64). The latter reveals possible reanalysis of copula wei into
From light noun to nominalizer and more 
a case marker for defocused agents in passive constructions (see Shibatani 1985 for an illuminating discussion of passivization as essentially an ‘agent defocusing’ strategy). (62) 子為誰? zi wei shui you cop who ‘Who are you?’
(Lun Yu)
(63) 微子去之,箕子為之奴 wei zi qu zhi, ji zi wei [zhi nu] Weizi fled.from him Jizi become his slave ‘Weizi fled from him and Jizi became his slave.’
(Lun Yu)
(64) a.
在上為烏鳶食,在下為螻蟻食
zai shang wei [wu yuan] shi, zai xia wei [lou yi] shi loc above be crow hawk eat loc below be mole-cricket ant eat ‘Above, (it) is the crows and hawks that will eat (me); below (it) is the mole-crickets ants that will eat (me).’
b.
在上為烏鳶食,在下為螻蟻食
zai shang [wei wu yuan] shi, zai xia [wei lou yi] shi loc above by crow hawk eat loc below by mole-cricket ant eat ‘Above I’ll be eaten by the crows and hawks, below I’ll be eaten by the mole-crickets and ants.’ (Zhuang Zi)
Examples such as (64) above provide a bridging context, with two possible interpretations – namely, (a) ‘above it’s the crows and hawks that will eat (me); below it’s the mole crickets and ants that will eat (me)’, or alternatively (b) ‘above (I’ll) be eaten by the crows and hawks; below (I’ll) be eaten by the mole crickets and ants’. Thus, whereas copula wei provides contrastive focus for the agent(s), namely, wu yuan ‘the crows and hawks’ and lou yi ‘the mole-crickets and ants’, case marker wei on the other hand defocuses these same agent NPs, and gives rise instead to oblique argument interpretations ‘by the crows and hawks’ and ‘by the mole-crickets and ants’. This reanalysis of copula wei from contrastive focus marker to case-marker of a defocused agent, which was already attested in Old Chinese, is highlighted in (64′) below. (64′) Reanalysis of wei from a copula with a contrastive focus function to a case-marker with an agent-defocusing function: → b. [wei NP] V
a.
wei [NP] V
wei [wu yuan] shi [wei wu yuan] shi be [crow hawk] eat [by crow hawk] eat
wei [lou yi] shi be [mole-cricket ant] eat
[wei lou yi] shi [by mole-cricket ant] eat
 Foong Ha Yap & Jiao Wang
In short passives (where the agent NP is elided), as seen in (65), agent marker wei was reinterpreted as a passive marker by virtue of its implicit association with a defocused agent NP. Worth noting in this particular example from the Warring States period is that the short passive expression wei lu (‘was executed’) parallels jian yi (‘was suspected’), providing supporting evidence that wei, like jian, could be used as a passive marker in Old Chinese. (65) 厚者為戮,薄者見疑 Hou zhe [wei lu ], bo zhe [jian yi ] thick one pass execute thin one pass suspect ‘In the worst case the man was executed, and even in the lesser case he was suspected.’ (Han Fei Zi, cited in Pulleyblank 1995: 38)
Bridging evidence of a copula link for case marker wei can be seen in (66), where the presence of genitive zhi in (66a) imposes a nominal interpretation for the expression tianxia zhi da lu (‘a big shame in the world’) and hence a copula interpretation for wei (‘be’ or ‘become’), while the absence of genitive zhi in (66b), on the other hand, leaves open two possible interpretations for wei tianxia da lu – namely, either a nominal reading with wei still interpreted as a copula (‘it was/he became a disgrace to the world’), or a passive reading with wei reanalyzed as a case marker introducing a defocused agent (‘(he) was laughed at by the world’), arguably via an intermediate interpretation such as ‘(he) became for the world a big disgrace’. (66) a.
身死國亡,為天下之大僇
b.
然後身死國亡,為天下大僇
shen si guo wang, wei [tianxia zhi da lu ] body die country die wei world gen big shame ‘When he died and his state was lost, it was/he became a disgrace to the world.’ (Xun Zi, Qiangguo, Warring States; Ting 2008) ranhou shen si guo wang, wei tianxia da lu then body die country die wei world big shame ‘Then when he died and his state was lost,
i. (it) was/(he) became a disgrace to the world’ ii. (it was/he became) for the world a disgrace’ iii. (he) was laughed at by the world’ (Xun Zi, Feixiang, Warring States; see also Ting 2008)
Note that the passive reading emerges as a consequence of various agent defocusing strategies. As highlighted in (67a), copula wei (‘be’ or ‘become’) introduces a nominal in which the agent tianxia (‘the world’) is expressed as a possessor NP or genitive subject. In (67b), the defocused status of the agent is brought about by reanalyzing copula wei as a case marker for a defocused agent. In other words, wei is reinterpreted as a
From light noun to nominalizer and more 
Â� non-nominative agent marker. This renders wei tianxia as the equivalent of English ‘for the world’ or ‘by the world’. As seen in (67c), the agent can be further defocused via elision, paving the way for copula wei to be reanalyzed as a passive marker. (67) a.
wei [tianxia zhi da lu] ‘be/become [a disgrace to the world]’
Lit. ‘be/become [the world’s disgrace]’
b. wei tianxia [ da lu]
‘be/become [laughed at] by the world’ Lit. ‘be/become for the world [to laugh at]’
c.
wei _ [ da lu]
‘be/become [laughed at] (by someone)’
3.5.2â•… Passive-wei constructions with suo There are other ways to defocus an agent. This includes highlighting the patient at the expense of the agent. This is where the suo nominalization construction fits in perfectly. Recall that [(NP zhi) suo VP] constructions highlight the patient NP (represented by pronominal clitic suo in preverbal position); at the same time the agent NP is defocused, often by casemarking the agent np with genitive subject marker zhi or by eliding the agent NP altogether.20 This very effective strategy explains why [suo VP] nominalization constructions – sometimes also referred to as ‘suo relatives’ (e.g. Ting 2005) – became increasingly popular in wei passive constructions in Middle Chinese, extending later into bei passive constructions in Modern Chinese. Examples of [suo VP] in wei passive constructions are shown in (68) to (70) below. Note that the [suo VP] construction could be found in syntactic environments where wei could be interpreted either as copula, case marker, or passive marker. In (68), with wei as copula, the [NP suo VP] nominalization construction that serves as the verbal complement – i.e. Bineng su suo guixin ‘the one Bineng always obeys’ – clearly has a strong patient focus by virtue of pronominal clitic suo in preverbal position.21 Lacking in voice morphology, the verb guixin (‘obey’) is
.â•… Examples of [NP zhi suo VP] constructions (i.e. with genitive subject) are seen in (39), (43), and (44) in §3.2, and in (45) and (46) in §3.3.1. .â•… From a formal syntactic perspective, it is worth noting that this position of suo follows (or ‘is lower than’) both subject NP Bineng and frequency adverb su (‘always’). This lends support to Ting’s (2008) argument that preverbal suo is a pronominal clitic, which we suggest occupies a syntactic position very different from external light head noun or nominalizer zhe, as discussed earlier in §2.3. This also explains why [suo VP] zhe constructions are possible in Chinese. That is, since pronominal clitic suo occupies a clause-internal ‘head’ position, while light noun or nominalizer zhe occupies a clause-external (and initially appositive) ‘head’ position, the apparently redundant use of ‘double suo…zhe nominalization’ is permissible (recall examples (42) to (44) in §3.2). Recall also from examples (40) and (41) in §3.2 that [suo…N] constructions were also permissible in Old Chinese, where N was clearly used as the external
 Foong Ha Yap & Jiao Wang
morphologically indeterminate with respect to either an active voice or a passive voice interpretation, as seen in the acceptability of either (i) or (ii) below. (68) 上谷太守阎志….为比能素所归信 [shang gu tai shou] i [yan zhi] i wei [bi neng su Shanggu county-magistrate Yanzhi cop Bineng always suo i gui xin _ i] i suo obey ‘The county-magistrate of Shanggu, Yanzhi, … i. is [the one Bineng always obeys _ i]’ ii. is [the one (who is) always obeyed by Bineng]’ (San Guo Zhi, Jin period)
As seen in (69), from The Book of Han (Eastern Han period), depending on whether copula wei is interpreted in the sense of ‘be’ or ‘become’, there is a corresponding shift from referential focus (‘was the one Jiang Chong defeated’) to result state focus (‘became the one defeated by Jiang Chong’), as highlighted in (i) and (ii) respectively. Depending on context, the [wei NP suo V] construction could be associated with a passive ‘result state’ interpretation with wei interpreted less as a copula and more as a case marker, equivalent to English agent-defocusing preposition by, as seen in (iii). The presence of pronominal clitic suo, which highlights a patient-oriented focus, concurrently defocuses the agent NP Jiang Chong, further facilitating the reinterpretation of wei as case marker for the defocused agent. (69) 衛太子為江充所敗。 [Wei tai zi] i wei [[ jiang chong] k suoi bai ]i Wei prince wei Jiang Chong suo defeat i. ‘Prince Wei was/became the one Jiang Chong defeated.’ ii. ‘Prince Wei was/became the one defeated by Jiang Chong. iii. ‘Prince Wei was defeated by Jiang Chong.’ (Han Shu)
As seen in (70), agent defocusing via elision gives rise to short passive constructions. Wei in these short passive constructions could still retain a copula interpretation, as in (i), but more often it is interpreted instead as a passive marker, as in (ii). In other words, not only is the agent defocused (via elision) but even the focused patient (namely, preposed object pronominal suo) could lose its explicit referential and anaphoric force, a drift toward non-referential use as [suo V] nominalization constructions come to
head noun in suo relative clause constructions. These two constructions have been replaced in Modern Chinese with ‘double suo…de nominalization’ constructions and [suo…de N] relative clause constructions.
From light noun to nominalizer and more 
focus not so much on the patient argument but more on the result of a change-of-state situation. Not surprisingly then, [ruo shu jie] i qie wei _ k suo lu is often interpreted not as ‘you fellows will all be the ones captured (by him)’ but rather as ‘you fellows will all be captured (by him)’. (70) 不者若屬皆且為所虜 bu zhe [ruo shu ] i jie qie wei _ k suo lu neg zhe you fellow all just by _ k suo capture i. ‘If not, you fellows will all be the ones captured (by him).’ ii. ‘If not, you fellows will all be captured (by him).’
(Shi Ji, Western Han period, the latter cited in Pulleyblank 1995: 38)
As noted earlier, wei could already be used as a passive marker in pre-Qin texts (e.g. Han Fei Zi from the Warring States period; see (65) above). The wei suo passive was a later development, found sporadically in pre-Qin texts and gaining popularity beginning from the Han period. What is important for the purposes of the present discussion is that the development of the wei … suo passive hinged on the availability of the [(NP zhi) suo V] patient nominalization construction, and in this sense we see how nominalization and passivization can sometimes be linked to each other (see also Ting 2008 for a formal syntactic discussion that unifies suo passives and suo relatives, where her term ‘suo relatives’ is interchangeable with our term ‘suo patient nominalization construction’).22
3.6â•… Summary To sum up, we have seen that locative noun suo evolved into a light noun that was used in locative nominalization as well as theme/patient nominalization constructions (§3.2), then later extended to more abstract nominalization uses where it could be used as a marker of subordinate temporal and conditional clauses (§3.3.1), this extension being part of a natural drift from referential to non-referential uses of nominalization constructions over time. In combination with light verb yi (‘use’), light noun suo also developed into reason marker and purposive complementizer suo yi (see §3.3.2 and §3.3.3 respectively). In Middle Chinese, light noun suo further developed a possessee pronominal function which, given its semantically bleached reference to the
.╅ An interesting question is why suo was not reanalyzed as a passive marker independently of wei. We suggest that overwhelming competition from [suo V] patient nominalization constructions curbed further reduction of [wei suo V] short passives into [suo V] short passives. In other words, there was no advantage gained for Middle Chinese to adopt suo alone as a passive marker when it was far more useful as a patient nominalizer.
 Foong Ha Yap & Jiao Wang
possessee NP (e.g. wo suo ‘my one’ > ‘mine’), came in very handy for the purpose of highlighting the possessor NP (§3.4). When apposed to a co-referential lexical noun (N), an [NP suo][N] construction emerged that was reanalyzed as an [NP suogen N] construction, with suo reinterpreted as a genitive marker. Beginning from the Han period, suo was also often found in passive constructions, either with wei NP (i.e. case marker + defocused agent NP) preceding it, or with only wei alone preceding it due to agent NP elision (§3.5). We thus see a wide range of grammaticalization pathways for light noun suo. In particular, we see evidence of a strong relationship between suo patient nominalization (sometimes referred to as suo relatives) and clause subordination (particularly clauses with topic-like conditional ‘if ’ interpretations in the sense of Haiman 1978). We also see evidence of a strong relationship between suo patient nominalization and the emergence of wei … suo passive constructions. On the other hand, while we do see a common light noun origin for pronominal clitic (i.e. nominalizer) suo and possessee pronominal and genitive suo, the latter two were parallel developments with, rather than derived from, the former. That is, whereas nominalizer suo emerged via introduction of a pronominal clitic that yielded a focus on the patient argument, possessee pronominal suo emerged via apposition of NP arguments in possessive relations (i.e. possessor NP + light noun suo as possessee NP), with examples such as wo suo ‘my one’. Genitive suo involved an additional (i.e. secondary) process of NP apposition, yielding examples such as wo suo shou ming ‘my one, life’ > ‘my life’. As noted earlier in §3.4, this secondary extension from possessee pronominal suo to genitive suo is not unlike the secondary extension from nominalizer suo to relativizer suo; nevertheless, these parallel developments suggest that it is still possible to delink genitivization from nominalization and relativization. Figure 3 summarizes the functions of suo in Old, Middle and Modern Chinese for ease of comparison with the functions of zhe discussed earlier in §2. Note that the preverbal syntactic position of suo in [suo V] nominalization constructions readily accounts for the absence of suo as a sentence-final mood particle, contra the postpredicate position of zhe in [VP]-zhe nominalization constructions (and later de in [VP]-de nominalization constructions) which accounts for the availability of zhe (and de) as a sentence final mood particle (§2.7). Note that suo, like zhe discussed earlier, was most versatile in Middle Chinese, assuming some of the functions of a declining demonstrative and adnominal (i.e. genitive and relative clause marker) zhi. Eventually, however, it was the super-adnominal di/de that replaced all three suo/zhe/zhi genitives and relativizers in Early Modern Chinese (see Shi & Li 2002). Classifier uses of suo first attested in Middle Chinese have survived to some extent in Modern Chinese, but this was a function not seen with zhe. As noted earlier, possessee pronominal and genitive uses of suo did not survive (§3.4), nor did those of zhe (§2.5).
From light noun to nominalizer and more 
While suo yi preserves its reason marker and purposive complementizer functions (§3.3), these are functions never associated with zhe. The [suo V] construction has also thrived within the passive construction domain, though greatly dependent on case marker wei or passive marker wei to disambiguate between a patient nominalization vs. patient-focusing/agent-defocusing (i.e. passive) interpretation. Later in Modern Chinese, the [suo V] passive construction came to increasingly rely on passive marker bei instead. Old Chinese
Middle Chinese
Modern Chinese
Locative noun suo
Locative noun suo
Locative noun suo
Light noun suo NMZ suo
Light noun suo NMZ suo
Light noun suo CLF suo
POSS Pron.suo
NMZ suo
CLF suo
Locative NMZ suo
Locative NMZ suo
Locative NMZ suo
Patient NMZ suo
Patient NMZ suo
Patient NMZ suo Genitive suo
Conditional subordinator suo V (zhe)
Conditional subordinator suo V (zhe)
Conditional subordinator suo V (zhe)
Means suo yi
Means suo yi
Means suo yi
Reason suo yi
Reason suo yi
Reason suo yi
Purposive suo yi
Purposive suo yi
Purposive suo yi
Long passive wei NP suo V Short passive wei suo V
Long passive wei NP suo V Short passive wei suo V
Figure 3.╇ Functions of suo in Old, Middle and Modern Chinese
4.â•… Some general observations In this paper, we have seen how light nouns could develop into nominalizers. In particular, we have seen that certain nominalizers could be construed as light head
 Foong Ha Yap & Jiao Wang
nouns. In the case of zhe, the nominalizer is an external head noun, albeit a semantically light one, often in complementary distribution with full lexical head nouns, while in the case of suo, the nominalizer is perhaps an internal head noun, also semantically light, and possibly a pronominal clitic as argued in Ting (2005; see also Ting 2003). This structural difference (i.e. external vs. internal head position) could be traced to the lexical origin and hence preferred (or frequent) semantic role associated with each nominalizer.23 Zhe, as an agent nominalizer, assumes the role of an external argument acting on a situation, while suo, as a patient and other non-agent nominalizer, serves instead as an internal argument, crucially in preverbal position that highlights a non-agent argument, which nicely explains why suo nominalization constructions align themselves so readily with passive contexts, yielding the wei/bei … suo passive constructions. At the same time, the external vs. internal head noun distinction also nicely explains why post-predicate zhe could develop a sentence-final mood marking function while preverbal suo could not. Another structural difference that we see between zhe and suo as external vs. internal head nouns is that agentive zhe nominalization constructions involve agent gapping, with the gapped agent subject co-indexing with external head noun zhe, as in [ _ i VP] zhei, whereas non-agentive suo nominalization constructions could express gapped non-agent arguments, as in [(NP zhi) suoi V _ i] i (zhei /Ni), with suo referring to the non-agent. The subject NP is optional and sometimes case-marked by genitive zhi, indicating its status as a genitive subject and hence a defocused agent. The use of this defocusing strategy is consistent with the overall goal of highlighting the preverbal light noun or pronominal clitic suo to form a patient (or other non-agent) nominal construction. Moreover, assuming that light noun or pronominal clitic suo in patient (and other non-agent) nominalization constructions is an internal head, it would be possible to additionally express an external head noun, either in the form of light noun zhe or full noun N, and in fact this is what we see in Old Chinese (especially during the Warring States, as seen in text such as Mencius). Crucially, in the [(NP zhi) suoi V _ i] i zhei construction, external head noun zhe co-indexes with non-agentive suo rather than the genitive subject. We thus have evidence from Old Chinese that it is possible for a language to adopt a ‘double nominalization’ strategy using two
.╅ Zhu (1983) indicated that zhe can be used for both zizhi (self-designation, i.e. event nominalization) and zhuanzhi (other-designation, i.e. argument nominalization) whereas suo can only be used for zhuanzhi (other-designation). That suo cannot be used for zizhi (selfdesignation) suggests that light noun or pronominal clitic suo is in fact a clause-internal head noun, and hence cannot scope over the entire proposition, whereas zhe could.
From light noun to nominalizer and more 
Â� different light nouns – one internal to the clause (with suo), and another external to it (with zhe).24 Comparison with internally-headed relative clauses (IHRCs) in other languages is also revealing. Much discussed in the literature is the Japanese IHRC, with the structure [NP (ga) NPi (o) V] noi /*Ni, with no gapping of object NPs. Note that whereas light head noun no is felicitous, the use of a full lexical head noun (N) is infelicitous. One obvious reason for this asymmetry appears to be related to cognitive processing. When the internal argument (NPi) is itself a full lexical noun, there is little need to rely on another full lexical noun (Ni) as an external head noun, whereas a light head noun (no) would serve as an ideal resumptive pronoun that could further serve as a marker of clausal nominalization, a process that could transform an independent clause into a nominal complement, or it could serve as a marker of clausal subordination, a process that could transform an independent clause into a dependent one, often for the purpose of specifying circumstantial information such as time, condition, means, manner, reason, purpose, etc., related to the the upcoming matrix clause. By way of comparison, in the case of Old Chinese suo nominalization constructions, we see no constraint on an external full head noun (N). That is, we could have [(NP zhi) suoi VP _ i] i zhei /Ni. This is because the internal argument (suoi) is a light noun or pronominal clitic, thus legitimizing either light noun zhe or full lexical noun N as the external head noun. We thus see different types of IHRCs depending on the degree of semantic lightness of the internal arguments within a nominalization construction. The above discussion also blurs the line between nominalization and relativization. Structurally, many researchers have argued that a nominalized clause is a subtype of the relative clause construction. In particular, many scholars have often referred to nominalization constructions as ‘headless’ relative clauses. In recent work, Shibatani and colleagues (e.g. Shibatani 2009; see also Shibatani & Khaled Awadh 2009) argue against this practice, pointing out that the use of the term ‘headless’ relative clause is misguided since nominalization constructions are not derived from relative clauses via deletion of an external head noun, but on the contrary, crosslinguistic evidence points instead to relative clauses being secondary developments in which an external head noun is added to pre-existing nominalization constructions. In this paper, we show that availability of [VP] zhi zhe constructions such as (5) in §2.1, where zhi functioned as a relativizer linking a pre-nominal modifying expression and the head noun, suggests that zhe nominalization constructions in Old Chinese could be construed as ‘light-headed’ relative clauses.
.╅ The use of zhe as an external (albeit light) head noun in [suo V] zhe patient nominalization constructions in Old Chinese indicates that zhe had become a general nominalizer, and was not restricted only to agent nominalization.
 Foong Ha Yap & Jiao Wang
In this paper we have also seen that the relationship between nominalization, relativization and subordination is a very intimate one. From a functional perspective, the term ‘nominalization’ is used essentially with an emphasis on referentiality, while the terms ‘relativization’ and ‘subordination’ focus on specificational (or modification) functions, the former to specify nominal arguments and the latter to specify the temporal and other characteristics associated with clausal propositions. It is worth noting here that in Old Chinese the link between nominalization and clausal subordination has had a longer history, while secondary-type relativization constructions derived from light-nounbased nominalization constructions became productive much later in Middle Chinese. This brings us to another interesting observation, more specifically about the development of the adnominal (i.e. genitive and relativizer) system in Chinese. As is well known, Old Chinese initially made extensive use of genitive zhi (and to some extent relativizer zhi), but with its gradual decline in Late Old Chinese and throughout Middle Chinese, new adnominals began to emerge, rather briefly in the form of genitive suo (§3) and later more visibly in the form of both genitive and relativizer zhe (§2). This was followed by another transition period in which di and de emerged as the adnominal par excellence in Modern Chinese. Several stages of development and their distinctive characteristics can be seen here. Firstly, at Stage 1, the Old Chinese adnominal system was based on zhi, which was derived from a demonstrative meaning ‘this, that’ (see Yue 1998). Then at Stage 2, the Middle Chinese adnominal system began shifting to a new base derived from light nouns, particularly zhe (as amply discussed in this paper). This light-noun-based adnominal system was further consolidated at Stage 3 with the rise of light noun di/de in Modern Chinese (see Yap, Choi & Cheung 2010). In other words, we see a shift from demonstrative-based adnominal zhi to lightnoun-based adnominals such as zhe and di/de.25 This development explains the complementary distribution of zhi-type ‘headed’ relative clauses and zhe-type ‘light-headed’ relative clauses in Old Chinese–the latter often referred to as agent nominalization constructions. It also explains the blurring in functional differentiation between zhi and zhe constructions in Middle Chinese as the latter (zhe) extended from ‘light-headed’ relative clause environments into ‘headed’ ones. This extension to environments supported by the additional presence of a full lexical head noun (initially arguably in appositive contexts) was accompanied by a reanalysis of light head noun (or nominalizer) zhe into an adnominal that signals a
.╅ DeLancey (2005) and Noonan (this volume) posit a similar development within the Tibetan and Tamangic languages respectively, with genitive -i still accompanying pa- adnominal nominalizations in some languages, but losing ground in many other languages. Also noteworthy from a diachronic perspective, genitive -i is not found with adnominal nominalization constructions formed with newer nominalizers.
From light noun to nominalizer and more 
dependent (modifying or specifying) relationship between the preceding clause and the lexical head noun (N). This trend toward adnominalization has been both strong and pervasive, with a new wave involving nominalizer di/de recruited as super-adnominal par excellence in Modern Chinese, thus relieving the pressure on light noun zhe so it could re-focus on preserving its major role as agent nominalizer. This strong trend toward adnominalization also explains why native speakers (and scholars) of contemporary Chinese tend to identify ‘headless’ (or what we prefer to call ‘light-headed’) di/ de relative clauses with an implicit (or ‘empty’) head noun interpretation. In this paper we also see that light nouns can also evolve into a variety of other grammatical functions other than nominalizer and adnominalizer (genitive and relativizer), among them classifier and possessee pronominal. We have also seen how nominalization constructions can further develop into specialized uses, as in the case of the suo yi cause/reason and purposive markers extending from suo yi manner/ means nominalization constructions, as well as in the case of the wei … suo passive marker in contexts involving defocused agent NPs introduced by non-nominative case marker wei. It is not yet fully clear whether such specialized extensions respond largely to language-specific conditions (both structural and semantic) or whether they might not in fact be motivated by fairly robust grammaticalization tendencies across languages. Further research is needed to address such issues. To conclude, we have seen how versatile light nouns such as zhe and suo have grammaticalized in the history of the Chinese language. Preserved records of their versatility have provided us with a diachronic window into the mechanisms and conditions that support their semantic extensions, and in the process we have come to better understand how certain types of nominalizers emerge, and how these nominalizers often go on to develop other even more abstract grammatical functions.
Acknowledgments We gratefully acknowledge generous research funding from the Chinese University of Hong Kong (Faculty of Arts Direct Research Grants 2006–2007 and 2007–2008) and the Hong Kong Polytechnic University of Hong Kong (Faculty of Humanities Start-Up Research Fund 2010–2012) to the first author. We also thank Edith Aldridge, Benjamin Au-Yeung, Marco Caboara, Picus Ding, Haihua Pan, Yafei Li, Masayoshi Shibatani, Sze-wing Tang, and Hongyin Tao for constructive comments. We are also deeply grateful to Pik-ling Choi, Joan Huang, Polly Chan, Kam-siu Cheung, and Ming-kay Poon for assistance with data collection and data analysis in the preliminary stages of this project. Earlier versions of this paper were presented at the Annual Research Forum of the Linguistic Society of Hong Kong (2006, 2008) and the 11th Symposium on Contemporary Chinese Linguistics in Tianjin (2008).
 Foong Ha Yap & Jiao Wang
Abbreviations 1p 1sg 3p asp cl conn cop dem gen lnk
First person First person singular Third person pronoun Aspect Classifier Connective Copula Demonstrative Genitive Linker
loc neg nmz pass poss pron prt rel sfp top
Locative Negator Nominalizer Passive Possessive Pronoun Particle Relativizer Sentence final particle Topic marker
References Aldridge, Edith. 2008. Old Chinese determiner zhe as a determiner. In Historical Syntax and Linguistic Theory, Paola Crisma & Giuseppe Longobardi (eds), 233–248. Oxford: OUP. Bybee, Joan. 2001. Main clauses are innovative, subordinate clauses are conservative: Consequences for the nature of constructions. In Complex Sentences in Grammar and Discourse: Essays in Honor of Sandra A. Thompson, Joan Bybee & Michael Noonan (eds), 1–17. Amsterdam: John Benjamins. Cao, Guang-shun. 1995. Particles in Modern Chinese. Beijing: Language Publishers. Cao, Guang-shun. 1999. Fo Bun Xing Ji Jing zhong de xu he zhe (Xu and zhe in Fo Bun Xing Ji Jing). Zhongguo Yuwen 6: 440–444. DeLancey, Scott. 1986. Relativization as nominalization in Tibetan and Newari. Paper presented at the 19th International Conference on Sino-Tibetan Languages and Linguistics, Ohio State University, Columbus OH, September 1986. DeLancey, Scott. 2002. Relativization and nominalization in Bodic. In Proceedings of the 28th Annual Meeting of the Berkeley Linguistics Society, Parasession on Tibeto-Burman and Southeast Asian Linguistics, Patrick Chew (ed.), 55–72. Berkeley CA: Berkeley Linguistics Society. DeLancey, Scott. 2005. Relativization and nominalization in Bodic. Tibeto-Burman Linguistics: Proceedings of the 28th Annual Meeting of the Berkeley Linguistics Society. Berkeley CA: Berkeley Linguistics Society. Deutscher, Guy. 2009. Nominalization and the origin of subordination. In Syntactic Complexity: Diachrony, Acquisition, Neuro-cognition, Evolution [Typological Studies in Language 85], Talmy Givon & Masayoshi Shibatani (eds), 199–214. Amsterdam: John Benjamins. Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68: 81–138. Evans, Nicholas & Wilkins, David. 2000. In the mind’s ear: The semantic extensions of perception verbs in Australian languages. Language 76(3): 546–592. Fauconnier, Gilles. 1985. Mental spaces: Aspects of Meaning Construction in Natural Language. Cambridge MA: The MIT Press. Feng, Chun-Tian. 2000. Jin Dai Han Yu Yu Fa Yan Jiu (A Study on the Grammar of Modern Chinese). Jinan: Shandong Education Publishing Company.
From light noun to nominalizer and more 
Genetti, Carol. 1991. From postposition to subordinator in Newari. In Approaches to Gramamticalization, Vol. 2: Focus on Types of Grammatical Markers [Typological Studies in Language 19], Elizabeth Closs Traugott & Bernd Heine (eds), 227–255. Amsterdam: John Benjamins. Genetti, Carol. This volume. Nominalization in Tibeto-Burman languages of the Himalayan area: A typological perspective. In Yap, Grunow-Hårsta & Wrona (eds). Genetti, Carol, Coupe, Alec R., Bartee, Ellen, Hildebrandt, Kristine and Li, You-Jing. 2008. Syntactic aspects of nominalization in five Tibeto-Burman languages of the Hinalayan area. Linguistics of the Tibeto-Burman Area 31(2): 97–144. Givon, T. 2009. The Genesis of Syntactic Complexity: Diachrony, Ontogeny, Neuro-Cognition, Evolution. Amsterdam: John Benjamins. Greenberg, Joseph H. 1966. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, Joseph H. Greenberg (ed.), 73–113. London: MIT Press. Haiman, John. 1978. Conditionals are topics. Language 54: 564–589. Harbsmeier, Christoph. 1981. Aspects of Classical Chinese Syntax [Scandinavian Institute of Asian Studies Monograph Series 45]. London: Curzon Press. Haspelmath, Martin & König, Ekkehard (eds). 1995. Converbs in Cross-Linguistic Perspective: Structure and Meaning of Adverbial Verb Forms–Adverbial Participles, Gerunds. Berlin: Mouton de Gruyter. He, Jin-song. 1988. The origin and semantics of zhi between subject and predicate. Zhong Nan Minzu Xueyuan Xuebao: Zhe She Ban (Wuhan) 6: 106–117. Heine, Bernd. 2003. On the role of context in grammaticalization. In New Reflections on Grammaticalization [Typological Studies in Language 49], Ilse Wischer & Gabriele Diewald (eds), 83–101. Amsterdam: John Benjamins. Horie, Kaoru. 1998a. On the polyfunctionality of the Japanese particle no: From the perspectives of ontology and grammaticalization. In Studies in Japanese Grammaticalization: Cognitive and Discourse Perspectives, Toshio Ohori (ed.), 169–192. Tokyo: Kurosio. Horie, Kaoru. 1998b. Functional duality of case-marking particles in Japanese and its implications for grammaticalization: A contrastive study with Korean. In Japanese/Korean Linguistics, Vol. 8, David J. Silva (ed.), 147–159. Stanford CA: CSLI. Huang, C.-T. James. 2008. On lexical decomposition and silent categories. Plenary paper presented at the 16th Annual Conference of the International Association of Chinese Linguistics (IACL-16), Beijing, May 29-June 2. Huang, Joan. 2005. Semantic Extensions of Zhe. MA research project, Chinese University of Hong Kong. Jiang, Lansheng. 1999. The genitive use of locative words and the origin of the structural particle di3. Zhongguo Yuwen 2: 83–93. Kaufman, Daniel. This volume. Exclamatives and temporal nominalizations in Austronesian. In Yap, Grunow-Hårsta & Wrona (eds). LaPolla, Randy J. 1994. Parallel grammaticalizations in Tibeto-Burman: Evidence of Sapir’s ‘Drift’. Linguistics of the Tibeto-Burman Area 17(1): 61–80. LaPolla, Randy J. 2008. Nominalization in Rawang. Linguistics of the Tibeto-Burman Area 30(2): 45–65. Liang, Xiaohong. 1994. Fo Jiao Ci Yu De Gou Zao Yu Han Yu Ci Hui De Fa Zhan (Word Construction in Buddihist Texts and the Development of the Chinese Lexicon). Beijing : Bei Jing Yu Yan Xue Yuan Press. Liu, Hongyong & Gu, Yang. This volume. Nominalization in Nuosu Yi. In Yap, Grunow-Hårsta & Wrona (eds).
 Foong Ha Yap & Jiao Wang Lü, Shuxiang. 1943. A discussion on the controversy of di3 and di4 and the origin of di3. Jingling Qilu Hanxi Zhongguo Wenhua Huikan 3. Beijing: Commercial Press. Lü, Shuxiang. 1947. Zhong Guo Wen Fa Yao Lue (A Brief Introduction to Chinese Grammar). Beijing: Commercial Press. Malchukov, Andrej. 2006. Constraining nominalization: Function/form competition. Linguistics 44(5): 973–1009. Matisoff, James A. 1972. Lahu nominalization, relativization, and genitivization. In Syntax and Semantics, Vol. 1, John Kimball (ed), 237–258. New York NY: Seminar Press. Morey, Stephen. This volume. Nominalization in Numhpuk Singpho. In Yap, Grunow-Hårsta & Wrona (eds). Nagaya, Naonori. This volume. Rise and fall of referentiality: Articles in Philippine languages. In Yap, Grunow-Hårsta & Wrona (eds). Noonan, Michael. 1997. Versatile Nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givon, Joan Bybee, John Haiman & Sandra Thompson (eds), 373–394. Amsterdam: John Benjamins. Noonan, Michael. This volume. Aspects of the historical development of nominalizers in the Tamangic languages. In Yap, Grunow-Hårsta & Wrona (eds). Norman, Jerry. 1988. Chinese. Cambridge: CUP. Peyraube, Alain. 1989. History of passive construction in Chinese until the 10th Century. Journal of Chinese Linguistics 17: 335–372. Polinsky, Maria. 2008. Looking for traces. Keynote paper at the 2nd International Conference on East Asian Linguistics, Simon Fraser University, Vancouver, November 7–9. Pulleyblank, Edwin G. 1995. Outline of Classical Chinese Grammar. Vancouver: UBC Press. Pustet, Regina. 2003. Copulas: Universals in the Categorization of the Lexicon. Oxford: OUP. Rubba, Jo. 1996. Alternate grounds in the interpretation of deictic expressions. In Spaces, Worlds, and Grammar [Cognitive Theory of Language and Culture Series], Gilles Fauconnier & Eve Sweetser (eds), 227–261. Chicago IL: Chicago University Press. Rhee, Seongha. 2002. From silence to grammar: Grammaticalization of ellipsis in Korean. Paper presented at Second International Conference of New Reflections on Grammaticalization (NRG-2), University of Amsterdam, April 3–6. Rhee, Seongha. 2008. On the rise and fall of Korean nominalizers. In Rethinking Grammaticalization: New Perspectives, Maria Jose Lopez-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego [Typological Studies in Language 76], 239–264. Amsterdam: John Benjamins. Rhee, Seongha. This volume. Nominalization and stance marking in Korean. In Yap, GrunowHårsta & Wrona (eds). Shi, Yuzhi & Li, Charles N. 2002. The establishment of the classifier system and the grammaticalization of the morphosyntactic particle de in Chinese. Language Sciences 24: 1–15. Shibatani, Masayoshi. 1985. Passives and related constructions: A prototype analysis. Language 61(4): 821–848. Shibatani, Masayoshi. 2009. Elements of complex structures, where recursion isn’t: The case of relativization. In Syntactic complexity: Diachrony, Acquisition, Neuro-cognition, Evolution [Typological Studies in Language 85], Talmy Givón & Shibatani Masayoshi (eds), 163–198. Amsterdam: John Benjamins. Shibatani, Masayoshi & Khaled Awadh bin Makhashen. 2009. Nominalization in Soqotri, a South Arabian language of Yemen. In Endangered Languages: Contributions to Morphology and Morpho-Syntax, W. Leo Wetzels (ed.), 9–31. Leiden: Brill.
From light noun to nominalizer and more 
Shinzato, Rumiko. This volume. Nominalization in Okinawan: From a diachronic and comparative perspective. In Yap, Grunow-Hårsta & Wrona (eds). Sung, Li-May. This volume. Lexical and clausal nominalization in Budai Rukai. In Yap, GrunowHårsta & Wrona (eds). Tang, Yuming & Xifu, Zhou. 1985. Lun xian-Qin Hanyu beidong shi de fazhan. [On the development of the passive construction in pre-Qin period]. Zhongguo Yuwen 4: 281–284. Ting, Jen. 2003. The nature of the particle suo in Mandarin Chinese. Journal of East Asian Linguistics 12: 121–139. Ting, Jen. 2005. On the syntax of the suo construction in Classical Chinese. Journal of Chinese Linguistics 33 (2): 233–267. Ting, Jen. 2006. NegP and the particle suo in Mandarin Chinese. Concentric: Studies in Linguistics 32(2): 71–92. Ting, Jen. 2008. The nature of the particle suo in the passive constructions in Classical Chinese. Journal of Chinese Linguistics 36 (1): 30–72. Wang, Li. 1958. Han Yu Shi Gao (History of the Chinese Language). Beijing: Zhonghua Shuzu. Watters, David E. 2008. Nominalization in the Kiranti and Central Himalayish languages of Nepal. Languages of the Tibeto-Burman Area 31(2): 1–43. Whitman, John. 2001. Relabelling. In Diachronic Syntax: Models and Mechanisms, Susan Pintzuk, George Tsoulas & Andrew Warner (eds), 220–239. Oxford: OUP. Whitman, John & Paul, Waltraud. 2005. Reanalysis and conservancy of structure in Chinese. In Grammaticalization and Parametric Change, Montserrat Batllori, Maria-Lluïsa Hernanz, Carme Picallo & Francesc Roca (eds), 82–94. Oxford: OUP. Xu, Huiling & Matthews, Stephen. This volume. On the polyfunctionality and grammaticalization of the morpheme KAI in the Chaozhou dialect. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha. This volume. Referential and non-referential uses of nominalization constructions in Malay. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha, Choi, Pik-ling & Cheung, Kam-siu. 2010. Delexicalizing di: How a Chinese noun has evolved into an attitudinal nominalizer. In Formal Evidence in Grammaticalization Research [Typological Studies in Language 94], An Van linden, Jean-Christophe Verstraete & Kristin Davidse (eds), 63–91. Amsterdam: John Benjamins. Yap, Foong Ha & Huang, Joan. 2006. The grammaticalization of zhe in Classical Chinese. Paper presented at the 2006. Annual Research Forum of the Linguistics Society of Hong Kong, Hong Kong Polytechnic University, December 2–3. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], Maria Jose Lopez-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins. Yap, Foong Ha, Matthews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Yeh, Meili Marie. This volume. Nominalization in Saisiyat. In Yap, Grunow-Hårsta & Wrona (eds). Yue, Anne O. 1998. Zhi in pre-Qin Chinese. T’oung Pao LXXXIV: 239–292. Zhu, Dexi. 1983. Zhizhi he zhuanzhi (Self-referring and other-referring expressions). Fang Yan 1: 16–31.
On the polyfunctionality and grammaticalization of the morpheme kai in the Chaozhou dialect Hui Ling Xu & Stephen Matthews
Macquarie University/University of Hong Kong The morpheme kai is among the most frequent and polyfunctional in Chaozhou, a Min dialect of Chinese. The present paper aims to present a descriptive and analytical account of the various functions of kai from Chaozhou in order to shed more light on nominalization and the diachronic development of nominalizers. We discuss the use of the morpheme kai in classifier, genitive, possessive, pronominal, relative clause marker, nominalizer and sentence-final stance-marking functions. Our data provide further support for some pathways of grammaticalization conjectured or attested in various languages, including the links between adnominal markers and nominalization and between nominalization and stance marking.
1.â•… Introduction One of the first missionaries to describe the Chaozhou dialect of the Swatow (Shantou) variety described the morpheme kai in Chaozhou as “that ubiquitous man of all work in Chinese syntax” (Ashmore 1884: 90). This can be clearly demonstrated by the fact that even in Modern Chaozhou, kai can appear in a wide range of syntactic positions including the following: (1) tsi53-35 no35-21 kai55-11 naŋ55 this two cl person ‘these two persons’ (2) tsi53 kai55-11 this pron ‘this one’ (3) ua53 kai55-11 p'eŋ55-11iu53 1sg gen friend ‘my friends’
 Hui Ling Xu & Stephen Matthews
(4) koŋ33 meŋ55 kai55-11 kuaŋ55-11li35 citizen assoc right ‘rights of citizens’ (5) ua53 siaŋ35-21 hi53-35huã33 kai55-11 tsY33 1sg most like rel book ‘the books which I like most’ (6) ia35-21 hoô 2 tsoi11 im33am213 kai55-11 also very many dark nomz ‘…there are also many dark aspects (of society)’ (7) liaô 5 kai55-11 ŋeŋ35-21hoũ53–35 tioô 5–2 p’uaŋ213-53 si53-35 heŋ55 kai55-11 catch nomz definitely must judge dead sentence stance ‘Those who got caught will definitely be executed (I can assure you about this fact.) (8) ua53 m35-21 tsai33 ma11, m35-21 hĩ35–21 ua53 boi53-35 kai55-11 1sg not know prt, not cop 1sg buy stance ‘I didn’t know (about this). It was not me who bought it.’ (or: I didn’t buy it, I am telling you.)
A preliminary examination of the above examples shows that the syntactic positions where kai can appear instantiate a wide range of functions of this morpheme. These include that of a nominal classifier, adnominal relations (genitive, associative and relative clause), pronominal, nominalization, as well as stance marking. It is not uncommon cross-linguistically to find morphological convergence of syntactic functions such as what has been exhibited above. For example, it is a well known fact that in many Sino-Tibetan languages, the marker for adnominals is the same as that for nominalization. On the other hand, the development from a nominalizer to stance marker is not only a pan-Sinitic phenomenon but is also attested in various East Asian languages such as Japanese and Korean (Yap, Matthews and Horie, 2004). However, for the morpheme kai to perform all of the above is certainly unique and makes it an interesting and worthwhile case to study. The aim of the present paper, therefore, is to present a descriptive and analytical account of the various functions of kai from the Chaozhou dialect’s perspective1 in order to shed more light on nominalization and related issues such as the diachronic development of nominalizers. We organize the paper as follows: Section 2 provides some background information
.╅ The data used in this paper are taken from naturally occurring conversations collected by the first author of the paper.
On the polyfunctionality and grammaticalization of the morpheme kai 
about the typological features of this language.2 Section 3, which is the bulk of this paper, will be devoted to the description and discussion of the range of functions of kai. Section 4 presents some possible grammaticalization pathways by which the multiple functions, if not all, may have evolved. Section 5 closes the paper with some conclusions.
2.â•… Typological features of the language The Chaozhou dialect, spoken in the eastern part of Guangdong province in China, is one of the most conservative dialects in the Sinitic family, retaining various archaic features in its phonology, morphology and syntax. In this paper, it is represented by the Jieyang variety, one of nine varieties in the Chaozhou dialect group (see Xu 2007 for a comprehensive description of the Chaozhou grammar). It is a tonal language with eight basic tones and complex tone sandhi (TS). Two basic types of TS can be observed: (a) anterior TS whereby in spoken discourse, the syllables surface in sandhi tone except the rightmost one in a group of two syllables or more (eg. ang55 ‘red’, but ang55-11hue33 ‘flower’ = ‘red flower’), and (b) posterior TS which is a process of weakening or lowering the rightmost syllable of a compound, a phrase or a clause.3 As TS can distinguish meanings, it plays an important role in the grammar of the Jieyang dialect (however, there has been little in-depth research done so far to map out the interplay of TS and syntactic structure). The basic word order of clauses includes SV, AVO and topic-comment. In terms of morphology, there is limited affixation while compounding and reduplication processes are common. These derivational processes can serve as word-class changing strategies (such as adjectivization, but not so much nominalization). As in other Sinitic varieties, there is no marking for case, gender or number, although there are two fused genitive forms in modifying kinship terms:
(9) uã33 ‘my’ e.g. uã33pa55 ‘my father’
nia33 ‘your’ e.g. nia33pa55 ‘your father’
.╅ We consider this essential as the Chaozhou dialect, although quite an important variety of the Sinitic family, is relatively unknown. For a comprehensive synchronic description of the Chaozhou grammar, see Xu (2007). .╅ The numbers represent tone values on a conventional scale from 1 (low) to 5 (high). The citation tone is given first followed by the sandhi form. The two numerical values are separated by a hyphen.
 Hui Ling Xu & Stephen Matthews
Noun phrase structures display pre-nominal modification although there are some compound nouns, for example, some animal terms which have post-nominal modification. The linear order of noun phrase elements is shown in Figure 1: (PRO) - (DEM) - (NUM) - CL - (REL) / (ATTR) – (LW) – (HN) Figure 1.╇ Structure of the noun phrase in Chaozhou (pro = pronoun, dem=demonstrative, cl = classifier; attr= attributive, rel = relative clause, lw= linking word, hn = head noun)
As can be seen, the minimum element of a noun phrase can be just the classifier, which is normally interpreted as an indefinite singular entity with the head noun being elided. For example, uã53 zioô 5-2tsoi11 tsĩ55 (‘clbowl + how much + money’) means ‘how much is a bowl of…?’). This feature is quite unique to the Chaozhou dialect, and can be attributed to the fact that the numeral ‘one’ is often left out and that the head noun can always be unspecified if it is clear from the context or has been mentioned earlier. In numeral phrases, mensural and sortal classifiers are not only used for classification purposes but can be used to mark definiteness as well as indefiniteness of the head noun in the spoken language, the distinction of which generally has to do with the position of the classifier phrase in the sentence (further discussion in 3.1). There is no grammatical tense marking, as time reference is encoded lexically or in context, but the dialect has a rich aspectual system (see Xu 2006). Like other Sinitic varieties, the Chaozhou dialect extensively uses adjectives or verbs in post-verbal position to encode such meanings as result, extent, manner of an action and potential of an event to take place. Function words grammaticalized from lexical verbs are common. For example, the verb ‘to say’ has developed into a complementizer introducing declarative clausal complements, similar to ‘that + clause’ in English, and the verb ‘to see’ into a complementizer introducing an interrogative clausal complement, similar to ‘whether/if + clause’ in English (Xu & Matthews 2007). Other grammaticalized words include the verb ‘to give’ as a passive and permissive as well as an unaccusative marker (See Matthews, Xu & Yip 2005), and the verb ‘to cross’ as a comparative marker (Xu 2004).
3.â•… Functions of the morpheme kai 3.1â•… kai as a classifier Kai is first and foremost a general-purpose numeral classifier in the Chaozhou dialect, and is a cognate of the modern Mandarin general classifier ge, which has its origin in a Classical Chinese specific classifier for ‘bamboo’ (see Lisa L.- S, Cheng & Rint
On the polyfunctionality and grammaticalization of the morpheme kai 
Sybesma 2005, among others). Synchronically, in a demonstrative or numeral phrase, the classifier kai appears before the head noun, as schematized earlier in Figure 1 and illustrated below in (10): (10) tsi53-35 no35-21 kai55-11 naŋ55 this two cl person ‘these two persons’
sã33 kai55-11 p'eŋ55-11kue53 three cl apple ‘three apples’
The function of kai above is to classify and individuate entities. However, it is worth pointing out that numeral classifiers also exhibit referential properties in the Chaozhou dialect. Formally, definiteness of a noun entity is marked by demonstratives, pronouns or other referential nominal modifiers. However, in the spoken language, a construction of the form [CL + NP] can render different referential interpretations, depending on its position in a sentence. In pre-verbal (i.e. predicate) position, it gives the referent a definite interpretation, as shown in (11). In other words, when [CL + NP] is thematic, it is definite. But in post-verbal position, it can be definite or specific, as shown in (12) and (13) respectively, the distinction depending largely on context. (11) kai55-11 siÕ33 k’aô 2–3 soi213 cl box too small ‘The/(this or that) box is too small.’ (12) i33 kiau33 kai55-11 nou55-11kiã53 k‘Y213-53 iũ213-53zi55-11hŋ55 3sg take cl child go kindergarten ‘She took the (or: her) child to the kindergarten.’ (13) ua53 k’ioô 5–2 kai55-11 siÕ33 k'eô 5–2 lY53-35 paŋ213 1sg take cl box give 2sg put ‘I will find a box for you to put (your stuff in).’
As we will see later, the referential property of kai in marking definiteness is pivotal in its development of other uses, notably, in adnominal constructions (Section 3.2).
3.2â•… kai in adnominal functions As shown in Figure 1, adnominals, that is, modifying constituents in a noun phrase which include genitive, associative, adjectival, nominal and relative clauses, are, in general, pre-nominal. In the formal register, the pronominal modifying elements and the head noun are linked by the morpheme 的 ti, the transliteration and functional counterpart of de in Mandarin, but in the dialect’s vernacular, it is kai. Their distribution is highlighted below and in examples (14) and (15): Formal register: [modifier – ti – NP/Ø] (14) taŋ53 ti33 nia53-35 tau35 party asso leadership ‘the party’s leadership’
 Hui Ling Xu & Stephen Matthews
Colloquial usage: [modifier – kai – NP/Ø] (15) ua53 kai55-11 hiã33ti35 1sg gen /poss brothers ‘my brothers’
Below, we show more adnominal uses of kai such as in possessive, associative, adjectival attributive (16, 17), and relative clause constructions (18). (16) lY53 mo53-35 tsiam213-53 lY53-35 kai55-11 suŋ33, ua53-35 2sg cannot take.side 2sg poss grandson 1sg kai55-11 suŋ33, ts'au213 kai55-11 suŋ33, sẽ213-53keô 2–3 mÕ53 poss grandson bad lw grandson character not.good ‘You shouldn’t side with your grandson; my grandson,my badly-behaved (lit. smelly) grandson has bad temperament.’ (17) naŋ53 tsio213-53sẽ33 tso213 si35-21 ts’oô 5-2tseŋ213, u35-21 tsio213-53sẽ33 1plincl this.way do cop promote have this.way kai55-11 ts’eŋ55-11k’uaŋ213, siau33tY55-11 tsio213-53sẽ33 kai55-11 heŋ55-11siaŋ35, asso situation eliminate this.way asso phenomena tsu35-21 pĩ213–53 tso213-53 ho53 kai55-11 lau213 then change as good nomz crs ‘What I am doing is a positive approach. (When) there are such phenomena, to eliminate such phenomena, means that (my action) has turned (things) into positive (ones).’ (18) ua53 hi53-35huã33 i33 li53-35loô 5 kai55-11 ts'ai213 1sg like 3sg cook rel dish ‘I like the dishes he cooks.’
Although the use of a nominal classifier to link a modifier and its head noun is a common phenomenon in Chaozhou as well as in some Southern dialects such as Cantonese (parallels are also found in various Southeast Asian languages), this function is not found in Mandarin Chinese. For example, (19b), which is the equivalent of the Jieyang example in (19a), is ill-formed: (19) a. Jieyang: ua53 kai55-11 tiaŋ35-21si35 1sg cl TV ‘my TV’ b. Mandarin: * wo ge dianshi 1sg cl TV
Possessive constructions such as (19a) can be interpreted as an elliptical construction where an underlying demonstrative (‘this’ or ‘that’) is omitted. Thus, it is possible that [1sg + cl + tv] (‘my TV’) comes from [1sg + dem + cl + TV] (‘this/that TV of mine = my
On the polyfunctionality and grammaticalization of the morpheme kai 
TV’). This reanalysis of [cl + n] as [dem + n] may also help to account for the relative use of kai. The following three examples demonstrate the possible development of kai being reanalyzed as a relativizer after the demonstrative is elided, as shown in (21) and in a headless sentence in (22) where kai fuses two functions: as a relativizer and a pronominal. (20) [[ua53 tsau33zek5 boi53] hY53-35 kai55-11 tiaŋ35-21si35] 1sg yesterday buy that cl TV ‘that TV which I bought yesterday’ (21) [[ua53 tsau33zek5 boi53] kai55-11 tiaŋ35-21si35] 1sg yesterday buy rel TV ‘the TV which I bought yesterday’ (22) [ua53 tsau33zek5 boi53] kai55-11 Ø 1sg yesterday buy rel/pron Ø ‘the one (=TV) which I bought yesterday’
It is worth noting that due to tone sandhi, another reading of example (21) above is also possible, illustrated in (23). This is because when the verb boi ‘buy’ changes from its citation tone to a TS tone, it will be understood as forming a constituent with the following NP and together they form a VO construction (note that as the numeral ‘one’ can always be left out in a [numeral + cl + np] phrase when it appears in post-verbal position, kai in (23) should be understood as a classifier; the TS tone of boi also determines the function of kai. This possibility of two readings is important as it supports the reanalysis of the classifier kai as a relativizer as shown in (21) and (22). (23) [ua53 tsau33zek5 [boi53-35 kai55-11 tiaŋ35-21si35]] 1sg yesterday buy cl TV Reading Two of (21): ‘I bought a TV yesterday’
In what follows, we discuss the pronominal function of kai, which is an important step in its development into a nominalizer.
3.3â•… kai as a pronominal In most of the adnominal examples shown above, the head noun being modified is explicitly indicated. However, when the context is clear or when the entity presented by the head noun has been mentioned or is easily retrievable from the discourse context, the head noun does not need to be specified, thus giving rise to a ‘headless’ construction, as shown in example (22) above and (24B) below. In this [Dem + kai + Ø] construction, kai can be taken as a pronominal, similar to the English ‘one’ as in ‘this one’.
 Hui Ling Xu & Stephen Matthews
(24) A: lY53 aĩ213–53 ti11 kai55-11 kue33 2sg want which cl melon ‘Which melon do you wantô ’ B: ua53 aĩ213–53 tsi11 kai55-11 Ø 1sg want this cl Ø ‘I want this one.’
The following are further examples of pronominal uses of kai. In (25), it is a possessive pronominal, referring to something possessed as well as the possessive relationship. (25) A: tsio53-35 mueô 5 ti11tiaŋ55-11 kai55 this thing who poss ‘Whose is this thing?’ B: ua53-35 kai55 1sg poss/pron ‘It is mine.’
Examples (26) to (28) are headless relative clauses where the head noun is omitted, because it has either been mentioned or can be retrieved from the discourse context: (26) A: lY53 u35-21 p’eŋ55-11â•›iu53 ho53-35 kai213-53siau35 k'eô 2 i33 bo55-11? 2sg have friend can introduce give 3sg not.have ‘Do you have friends who you can introduce to her (as a potential partner)?’ B: [ua53 pak2 kai55-11 Ø] loŋ53-35 kaô 2-3 huŋ33 lau213 1sg know rel Ø all marry perf/crs ‘Those (friends) who I know are all married.’ (27) liau53 tsu35-21 koi33 t’au55-11 haŋ11 bue53 u35-21 naŋ55-11 prt then street head lane tail have people ts’iaŋ213-53-k’eô 2, ts’iaŋ213-53 [hia53-35 tiaŋ213-53si35 hш53-35-toĩ53 sing-song sing that.kind.of television there-suf to53-35 ts’iaŋ213 kai55-11 Ø] prog sing rel Ø ‘(Then,) there were people singing songs on streets and lanes. They sang those which were sung on TV.’ (28) liau53 tsu35-21 lai55-11 k’Y213-53 siÕ33-paŋ55, tsek5-2 kai55-11 prt then come go wing-room one cl [bo55-11 naŋ55-11 tso35 kai55-11 Ø] not.have people sit rel Ø
‘Then we went4 to the wing-rooms, to one where there was no one (sitting).’
.â•… This feature of using ‘come’ and ‘go’ to express ‘go’ is quite unique to Chaozhou dialects.
On the polyfunctionality and grammaticalization of the morpheme kai 
In what we have shown above, it is clear that like the multifunctional de in Mandarin (e.g. Yap et al. 2004), the pronominal use of kai in post-modifier environment enables it to be interpreted as having a nominalization reading, which eventually gives rise to its development as a full-fledged nominalizer, as we shall see in the next section.
3.4â•… kai in nominalization Cross-linguistically, nominalization is a process by which non-nominal elements can be converted into nominals by morphological or syntactic means. In Mandarin Chinese, for example, a common strategy of nominalization is achieved by suffixing the particle de to non-nominal elements, which can be a verb, a verb phrase, a sentence, or a portion of a sentence including the verb (Li & Thompson 1981: 575). The Chaozhou dialect employs the same strategy and the general nominalizer used is kai. As well, morphological means such as affixation is another means of deriving nominals from non-nominal elements. For example, there are a few bound morphemes which can be suffixed to verbs to derive compound nouns. It is worth noting, however, that event nominalization is usually zero marked. Before we proceed with our discussion on nominalization with kai, we look at a few examples of the nominalizing derivatives.
3.4.1â•… Nominalizing derivatives Most nominalizing derivatives in Chaozhou are cognate with those of Mandarin. These include -ki (-machine), used with verbs to derive instrumental nominals, or functional entities such as those in (29), and -uaŋ (-person) and -ke (-family), which are used with verbs to form agentive nominals, typically those which denote names of occupations, that is, ‘the one who V’, as shown in (30) and (31). (29) soi53-35-sã33- ki33 hoŋ33-kaŋ33-ki33 lok5-2-im33-ki33 wash-cloth-nomz bake-dry-nomz record-sound-nomz washing machine dryer recorder (30) pua213-53im33- uaŋ55 eŋ55-11-uaŋ55 ke213-53sai53-35-uaŋ55 broadcast-nomz perform-nomz drive-nomz broadcaster actor/actress driver (31) ue11-ke33 ko33ts’iaŋ213-53-ke33 zaô 2-5-ke33 paint-nomz sing-nomz create-nomz artist singer writer
The nominals created in this way are more or less fixed expressions. For example, while -uaŋ and -ke are both used to create agentive nominals, they are not interchangeable. There are also a number of nominalizers which are from the Classical Chinese stock, such as so for forming patient nominals and tsia in deriving predicate nominals with the meaning of ‘the one who V’. However, they belong to the formal register and are not productively used in the spoken language (although they can be heard
 Hui Ling Xu & Stephen Matthews
from speakers of the older generation). We nevertheless show some examples in (32) and (33): (32)
so53-35-tã213 nomz-say ‘(someone’s) words’ ‘what (someone) says’
so53-35-tso213 nomz-do ‘(someone’s) actions’ ‘what (someone) does’
(33) u53-35koŋ213-53hiaŋ213tsia53-35 have contribution nomz ‘contributor’
3.4.2â•… Nominalization with kai Nominalization with kai is quite productive. Elements which can be converted to nominals include adjectives, verbs, verb phrases, demonstratives which include interrogative pronouns. Below, we look at some examples of nominalization constructions formed with kai. It is clear from the examples that nominalized forms with kai can serve various functions, which are quite similar to those found in other Chinese and Tibeto-Burman languages (see Noonan 1997; Sio, this volume).
i. with adjectives
(34) tsek5-2 ts’oô 2–3 [mo53 kai55-11], meŋ55-11meŋ55 [mo53 kai55-11], one clpl not.good nomz obviously not.good nomz
tseŋ35-21lak5 u35-21piaŋ213-53 k’Y213-53 koi53-35kuaô 2 i33-11 try.utmost able go resolve 3sg ‘(As for) the bad conduct, especially obvious cases of misconduct, (we should) try our best to deal with them.’
In this example, mo53kai55-11 (‘not.good-nomz’) serves as a nominal attributive with the head noun ‘behavior’ or ‘conduct’ implied in the discourse context.
ii. with verbs and interrogative pronouns
(35) [tsiaô 5 kai55-11], [meô 2–3 kai55-11], huã33 si35-21 boi35-21 tã213–53 eat nomz what nomz still cop not.can comp su33, [k’ia35-21k’i53 kai55-11], ia35-21 si35-21 k’o53-35ĩ53 lose ,5 live nomz also cop can ‘Food and whatever (things) are not bad, and the living (conditions) are also okay.’
As can be seen, the three nominalized constructions in the above example can all serve as sentence topics. .â•… This is a colloquial way of saying ‘bad’ in a comparative construction. See Xu (2004) for a description of the comparative constructions in the Jieyang dialect.
On the polyfunctionality and grammaticalization of the morpheme kai 
iii. with verb phrases
(36) [gua11ti11 lai55-11 tsio53-35 p’aô 2kaŋ33 kai55-11] loŋ53-35 si35-21 outside come here work nomz all cop gua11 seŋ53-35- kiã53 outside province- suf ‘The laborers from outside (of Chaozhou) are from other provinces.’
Here in (36), kai derives an agent nominal ‘laborer’ which in turn serves as the sentence topic, and in (37) and (38) below, we observe kai nominalized constructions serving as copula complements. (37) na55-11tau55 tsu35-21 si35-21 tsia53-35k’i53-35 [u35-21 ts’i213 kai55-11] aloe vera then cop this.kind.of have thorn nomz ‘Aloe vera is this kind of (plant) which has thorns.’ (38) naŋ53 tsio213-53sẽ33 tã213–53, tsi53-35 ts’oô 2–3 hua55-11zeŋ55 lai55-11 1pl this.way say this clpl overseas.Chinese come tsi53-35ko213-53, [tsoi11 kai55-11] si35-21 [u35-21 buŋ55-11hue213 kai55-11] here many cl cop have culture nomz
soô 5-2i33 [i55-11meŋ55 kai55-11], to33 mi35-21 ke33 belong [migrate nomz prt not more ‘Let me put it this way: these overseas Chinese who came here, the majority (of people) are (people) who are educated. (Those who) came as migrants are not many.’
In (39), the nominalized clause, po53-35hu35-21 koŋ33 meŋ55kai55-11 (‘protect citizennomz’), is a complement to an abstract head noun (‘rights’). (39) hia53 uaŋ53 hio213-53-paĩ55 sai33-hŋ33 koô 2-3-ke33 [po53-35hu35-21 prt 1plexcl that-suf west-side country protect
koŋ33meŋ55 kai55-11] k’uaŋ55-11iaô 2 tso213-53 kau213-53 kue213-53 ho53 citizen asso right do ext over good ‘Well, in our Western countries, they do a good job in protecting the rights of the citizens.’
As mentioned earlier, nominalization with kai cannot be used to derive an event or action. That is why tsiaô 5kai55-11 (‘eat + nomz’) in (35) cannot be taken to mean ‘eating’ while p’aô 2kaŋ33 kai55-11 (‘work + nomz’) in (36) does not mean ‘working’. In other words, there are no event/action nominals derived from kai as such in the Chaozhou dialect. Event or action nominals are generally zero marked, as shown below: (40) tsiaô 5–2 te55 tui213-53 seŋ33t’i53 ho213 eat tea towards body good ‘Drinking tea is good for health.’
 Hui Ling Xu & Stephen Matthews
(41) k’Y213-53 gua11 ti11 p’aô 2kaŋ33 siaŋ35-21 ho53-35 t’aŋ213-53 tsĩ55 go outside place work most easy earn money ‘Going away to work is a good way to make money.’
3.5â•… kai as a stance marker Stance marking denotes strategies which serve to signal speaker prospective and attitudes. In some East Asian languages such as Japanese, Korean, as well as Chinese, stance marking has been shown to be quite common and the marker of speaker stance seems to evolve from a nominalizer (see Yap & Matthews 2008, among others). For example, pronominal and nominalizer ‘de’ in Mandarin Chinese can occur in sentence-final position to reflect the speaker’s mood or stance. This phenomenon is also observed in the Chaozhou dialect where there is an extremely high frequency of kai in sentence final position in spoken discourse to mark speaker affirmation of the truth of the proposition. For example, in (42), the speaker first affirms that he thinks himself better at judgment than the listener. And to reaffirm his view, he also uttered ‘Really!’ in the second clause, using a stand alone phrase consisting of the epistemic adverb ‘really’ together with kai to express such mood as ‘What I told you is nothing but the truth’: (42) ua53 kai55-11 t’oĩ53-35huap2-3 1sg poss view
kai55 boi35-21 su55 lY53-35 kai55-11, cop not lose 2sg stance
tseŋ33sek5 kai55-11 really stance ‘I can see things better than you, you know. It is true, really!’
In (43), the speaker was retelling a story about how a friend’s wife got electrocuted while hanging out her washing. He affirmed his view that her death was not caused by her touching a broken electric wire nearby but that the electric current was conducted through a wet floor. His strong assertive-matter-of-fact attitude is reinforced by another add-on sentence at the end with stance kai, quite similar to the previous sentence. (43) i33 tsu35-21 k’ia35-21 to11 zi35-21 lau55, e35 hou35 ts’ẽ55, 3sg prt live at second floor prt rain clear e35 i33 seŋ33pu35 ts’uô 2 lai55-11 iaŋ55-11t’ai55 ne55-11 prt 3sg daughter.in.law out come balcony hang sã33 k’ou213, tsi53-35- e35- mueô 5 k’eô 2- i35 to35- lok5-2- k’Y213-53, shirt pant this- one- thing pass- 3sg fall- down- go boi55-11 k’Y213-53 tiã33- tioô 5–2 kai55-11, boi35 kai55-11 not.can go touch- rvc stance not.can stance ‘She lived on the second floor. The rain had just stopped. Then her daughter-inlaw went to the balcony to put some washing out. Only just this short while, she suddenly fell down. She didn’t touch (the broken wire) at all (ie. It was not that she touched the wire). It was not the case at all.’
On the polyfunctionality and grammaticalization of the morpheme kai 
In (44), we can also see that by using the sentence-final kai, the speaker is committed to the truth value of what he had to say: (44) a33- ma53 tsek5-2 ts’uô 2-3- meŋ11 i33 tã213–53 bo55-11 pref- grandma as.soon.as out- face 3sg say not.have no35-21 ku213 kai55-11 two sentence stance ‘As soon as grandma interfered (showed up), he said nothing (he dared not say anything, I can assure you of this).’
In Mandarin, sentence final stance marking de also co-occurs with the copula shi to form a construction commonly known as the shi….de construction. According to Li and Thompson (1981: 589), this construction serves to “characterize or explain a situation by affirming or denying some supposition” (as opposed to simply reporting an event), stance marker kai also co-occurs with the copula si, as shown in (45a), in forming similar cleft-like constructions. It is worth noting that the homonym and nonverbal copula kai6 can replace the copula si in this construction, as shown in (45b). What is more interesting is that copula si and the non-verbal kai can be used together to co-occur with stance marker kai, shown in (45c). We believe that in such a case, the non-verbal copula kai serves as an emphatic marker or focus marker, a function which is beyond the scope of the current paper. (45) a.
ua53 si35-21 tio55-11ziu33 lai55-11 kai55-11 1sg cop Chaozhou come stance ‘I come from Chaozhou’ (i.e, ‘It is Chaozhou that I come from’).
b. ua53 kai55 tio55-11ziu33 lai55-11 kai55-11 1sg cop Chaozhou come stance ‘I come from Chaozhou’ (i.e, ‘It is Chaozhou that I come from’). c.
ua53 kai55 si35-21 tio55-11 ziu33 lai55-11 kai55-11 1sg cop cop Chaozhou come stance ‘I come from Chaozhou’ (i.e. ‘It is Chaozhou that I come from’).
4.â•… Grammaticalization of kai On the basis of the Chaozhou data, we hypothesize the following relationships among the various uses of kai: Classifier kai
Adnominal > Pronominal > nominalization > Stance Marking
Figure 2.╇ Relationships of the uses of kai
.╅ The use of kai as a non-verbal copula was discussed in Matthews & Xu 2002.
 Hui Ling Xu & Stephen Matthews
As the Chaozhou dialect is essentially an oral language and therefore lacks historical textual evidence, we cannot conclusively make the claim that these relationships represent a grammaticalization chain. Nevertheless, the adnominal pathway is quite widely attested among Sinitic languages, such as Cantonese ge and Mandarin de. A similar development is also found cross-linguistically, in particular, in Tibeto-Burman languages and East Asian languages such as Japanese and Korean (see Yap et al. 2004; Yap & Matthews 2008; Noonan 1997; Bickel 1999; among others). In Japanese, for example, various studies have recognized a pivotal role for the pronominal uses of the genitive marker no in the development of nominalizer no. In Korean, the pronominal morpheme kes (a dummy noun) has extended nominalizer functions via headless relative clause constructions (see Yap & Matthews 2008). It is plausible, therefore, to argue that the various functions of kai have also gone through a similar path of grammaticalization. Along this pathway, the classifier kai develops adnominal use in headed constructions, as shown in Section 3.2, and pronominal use in headless construction, as shown in Section 3.3. Like the pronominal de in Mandarin, kai in such an environment enables it to be interpreted as a nominalizer. However, up to this point, [modifier + kai +ø] is still referentially definite. The next step sees kai fully grammaticalized as a nominalizer, able to derive nominals denoting a definite or generic entity. We summarize these developments in the following diagram (the stance function is not illustrated due to space constraints; examples are given in (43–45)): CL>
Adnominal>
Pronominal>
eg.1 tsek kai p’eŋkue one CL apple ‘one apple’
kai aŋ kai p’eŋkue CL red LW apple ‘the red apple’ or ‘a red apple’
kai aŋ kai Ø aŋ kai CL red PRON Ø red NOMZ ‘the red (one) ‘red ones’ /‘the red one’ / ‘ a red (one)’ ‘any N that is red’
eg.2 ua tsi kai p’eŋkue 1sg this CL apple ‘this apple of mine’
ua kai p’eŋkue 1sg POSS apple ‘my apple(s)’
ua kai Ø 1sg POSS Ø ‘my + (N)’
eg.3 ua [tsiaô kai p’eŋkue] ua tsiaô kai [ p’eŋkue] 1sg eat REL apple 1sg eat CL apple ‘I eat one apple.’ ‘the/this/that/these/those apple(s) that I eat’ (this relative clause reading is based on TS) (+headed, +definite) (+headed, +definite)
Nominalization>Stance
ua kai 1sg NOMZ ‘mine’
tsiaô kai ua tsiaô kai Ø 1sg eat CL/REL Ø eat NOMZ I eat one (N) ‘food’ (things to eat) or (based on TS): ‘the (one/ones) I eat’ (–headed, +definite)
(–headed, +definite, +generic)
Figure 3.╇ Adnominal pathway of kai
5.â•… Conclusion The Chaozhou data presented in this paper shows that the morpheme kai is highly versatile: it is used in classifier, genitive, possessive, pronominal, relative clause marker,
On the polyfunctionality and grammaticalization of the morpheme kai 
nominalizer and sentence final stance functions. Many of these functions are shared with similar polyfunctional morphemes in other dialects, such as Mandarin de and Cantonese ge as discussed in Yap and Matthews (2008). Our data were able to further support some pathways of grammaticalization conjectured or attested in various languages, such as the robust link between adnominal markers and nominalization and between nominalization and stance marking.
Abbreviations adj asso attr cl comp cop crs dem dis emp foc gen hn lw nomz
adjective associative attributive classifier complementizer copula current relevant state demonstrative disjunctive emphatic marker focus marker genitive head noun linking word nominalizer
np num pass perf pref pro pron prt rel rvc qprt quan suf stance
noun phrase numeral passive perfective aspect prefix pronoun pronominal particle relativizer resultative complement question particle quantifier suffix stance marker
References Ashmore, William. 1884. Primary Lessons in Swatow Grammar. Shanghai: Presbyterian Mission Press. Bickel, Balthasar. 1999. Nominalization and focus constructions in some Kiranti languages. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren G. Glover (eds), 271–296. Kathmandu: Royal Nepal Academy. Cheng, Lisa L-S. & Sybesma, Rint. 2005. Classifiers in four varieties of Chinese. In The Oxford Handbook of Comparative Syntax, Guglielmo Cinque & Richard S. Kayne (eds), 259–292. Oxford: OUP. Li, Charles N. & Thompson, Sandra A. 1981. Mandarin Chinese: A functional Reference Grammar. Berkeley CA: University of California Press. Matthews, Stephen & Xu, Hui Ling. 2002. The pronominal copula in Chaozhou. Paper presented at the 11th Annual Conference of the International Association of Chinese Linguistics, Nagoya, Japan. Matthews, Stephen, Xu, Hui Ling & Yip, Virginia. 2005. Passive and unaccusative in Jieyang dialect of Chaozhou, Journal of East Asian Linguistics 14(4): 267–298.
 Hui Ling Xu & Stephen Matthews Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra Thompson (eds), 373–394. Amsterdam: John Benjamins. Xu, Hui Ling. 2004. Comparative constructions in Jieyang. Paper presented at the Research Centre of Linguistic Typology workshop on the comparative constructions. La Trobe University. Xu, Hui Ling. 2006. The aspect marking of the Jieyang dialect. Journal of Chinese Language Teaching 3(1): 45–81. Xu, Hui Ling. 2007. Aspects of Chaoshan Grammar: A Synchronic Description of the Jieyang Dialect [Journal of Chinese Linguistics Monograph Series 22]. Berkeley CA: University of California, Berkeley. Xu, Hui Ling & Matthews, Stephen. 2007. Cong dongci dao ziju jiegou biaoji: Chazhou fangyan he Taiwan Minnan hua dongci shuo he kan de xuhua guocheng (The grammaticalization of the verbs ‘say’ and ‘see’ in the Chaozhou dialect and Taiwanese Southern Min). Zhongguo Yuwen Yanjiu (Studies in Chinese Linguistics). 23: 61–72. T.T. Ng Chinese Language Research Centre, Institute of Chinese Studies, Chinese University of Hong Kong. Yap, Foong Ha, Matthews, Stephen & Horie, Kaoru 2004. From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline: The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Yap, Foong Ha and Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds), in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins.
The Cantonese ge3* Joanna Ut-Seong Sio
Hong Kong Baptist University This paper presents a descriptive overview of the interpretative and distributive properties of the grammatical marker ge3 in Cantonese. It provides a Cantonese perspective on the nature of a marker with both adnominal and sentence-final mood marking functions which prevails in numerous East and Southeast Asian languages.
1.â•… Introduction In a lot of East and Southeast Asian languages, there exists a grammatical marker that is used both adnominally and sentence-finally, for examples, the Mandarin de, the Japanese no, the Malay punya, etc. (Yap et al. 2004). The variant of such marker in Cantonese is ge3. A lot of studies have been done on the Mandarin de. In comparison, studies on the Cantonese ge3 are scanty.1 This paper explores the nature of the marker from a Cantonese perspective. This paper aims to (i) provide a descriptive account on the
*The research study reported here was carried out within the context of Sze-Wing Tang’s GRF project ‘A comparative study of definiteness in Chinese dialects’ (PolyU5476/06H). I am grateful to the reviewers for their helpful comments. .â•… The Mandarin de and the Cantonese ge3 do not have identical distributions. For instance, the Cantonese counterparts of the following Mandarin sentences are not grammatical: (1)
wo zuotian jin de cheng 1sg yesterday enter de city ‘I entered the city yesterday’
(2)
bie sheng wo de qi do-not get 1sg de angry ‘(Please) Don’t be angry with me.’
The above sentences are taken from Lü (2004: 162). In fact, the Cantonese counterparts of all the Mandarin examples on pp. 162 under the heading ‘other usages’ (Section 7) are not grammatical.
 Joanna Ut-Seong Sio
various constructions ge3 can appear in; (ii) discuss the similarities among some of these constructions; (iii) discuss the properties of ge3 in light of other similar elements in Cantonese. One of the difficulties in the study of ge3 lies in its appearances in various seemingly unrelated constructions. With respect to distribution, one can identify two environments where ge3 appears in a sentence. It either appears adnominally or sentence-finally. As for the adnominal appearances of ge3, it can be part of (i) an adnominal modifier, as in (1); (ii) a conditional, as in (2); (iii) a contrastive topic clause, as in (3). As for the sentence-final appearances of ge3, it either functions as a sentence-final-particle (SFP) that gives rise to an assertive reading, as in (4), or it forms part of the predicate in class-membership type of copular sentences, as in (5). Adnominal: (1) hou2 sau3 ge3 zyu1 very thin ge pig ‘very thin pig(s)’ (2) jyu4gwo2 gam1-jat6 lok6jyu5 ge3 (waa2/6), …. if today rain ge saying ngo1 zau6 m4 ceot1 gaai1 1sg then neg go-out street ‘If it rains today, then I am not going out.’ (3) gong2-dou3 go1sau2 ge3 (waa2/6), …. talk-about singer ge saying ngo5 zui3 zung1ji3 Nina Simone 1sg most like Nina Simone ‘When it comes to singers, I like Nina Simone best.’ Sentence-final: (4) keoi5 wui5 lei4 (ge3) 3sg will come ge ‘(Don’t worry) He will come.’ (5) bun2 syu1 hai6 hung4-sik1 *(ge3) cl book be red ge ‘The book is red.’
In (1), (2) and (3), ge3 is followed by a nominal element. Different from (1), in (2) and (3) (a conditional and a topic clause respectively), the post-ge3 nominal waa2/6 is optional and cannot be substituted by other nominals. In (2) and (3), the post-ge3 noun, if present, has to be waa2/6. In (4) and (5), ge3 occupies a sentence-final position. However, only in (4) is ge3 optional with respect to grammaticality. In the absence of ge3, (5) would be ungrammatical.
The Cantonese ge3 
If we view the data from a different angle and consider the nature of the pre-ge3 element, we can classify all the constructions containing ge3 into two groups. In the conditional, topical and assertive use of ge3, i.e. examples (2), (3) and (4), the pre-ge3 element is a clause denoting a proposition. In (1) and (5), the pre-ge3 element is a modifier denoting a property. (6) a. proposition-ge3 b. property-ge3
Constructions with the pre-ge3 element denoting a proposition share one property. It has been pointed out by Haiman (1978) that conditionals and contrastive topics are similar in that both constructions have the effect of making the propositions/entities they denote into the shared presuppositions of all conversation participants in the discourse context. For instance, the underlined parts in (7) and (8) are interpreted as given by all conversation participants, based on which, the subsequent utterances can be evaluated against.
(7) If it is raining, then I will stay in. (8) As for singers, I like Nina Simone best.
(conditional) (contrastive topic)
Similarly for assertion, Stalnaker (2000) suggests that assertion has the effect of changing the shared set of presuppositions of the conversation participants by adding the content of what is asserted by the speaker to what is presupposed by all. In other words, conditionals, contrastive topics and assertions are similar in that in all these cases, the constructions make the contents they denote into the shared presuppositions of all the conversation participants. In view of the fact that all these three constructions involve the use of ge3, it suggests that ge3 is related to the discourse function of making propositions into shared presuppositions in these three constructions. In the nominal domain, ge3 is attached to the end of a nominal modifier. In the verbal domain, there is a similar element, gam2, which is attached to the end of a preverbal manner adverbial. Unlike ge3, gam2 is not as semantically bleached. It can still be shown that it has a pointing function, similar to demonstratives. In this paper, I provide evidence to show the similarities between ge3, gam2 and demonstratives, arguing that all of them are indexicals, sharing the features [+matching] and [+contrastive]. This paper is organized in the following fashion. Section 2 discusses the adnominal and sentence-final usage of ge3. Section 3 discusses the similarities between ge3 in conditionals, contrastive topic clauses and assertive sentences, arguing that all these constructions have the effect of making their contents into shared presupposition of all conversation participants. Section 4 discusses the similarities between ge3, gam2, and demonstratives, arguing that all of them share similar features. I conclude the paper in Section 5.
 Joanna Ut-Seong Sio
2.â•… An overview of the various usages of ge3 2.1â•… Adnominal-ge3 2.1.1â•… [modifier-ge3] -noun Ge3 can appear at the end of an adnominal modifier, preceding the noun. Example (1) is repeated here as (9b).
(9) a.
[modifier-ge3]-noun
b. hou2 sau3 ge3 zyu1 very thin ge pig ‘very thin pig(s)’
I am treating ge3 as part of the modifier rather than part of the noun because in cases with multiple modifiers, as schematized in (10), each modifier-ge3 sequence is a prosodic unit in the sense that there is a natural pause after each ge3. Ge3 is always ‘grouped’ with the preceding modifier. To the extent that prosodic pauses can be taken as evidence for constituent-boundaries, this suggests that [modifier-ge3] is a constituent to the exclusion of the noun. (10) [modifier-ge3] pause [modifier-ge3] pause [modifier-ge3] noun
Both the term modifier and the term noun in (9a) need some qualification here. The term modifier here is used in the broadest sense possible, referring to syntactic categories including adjectives, possessors, relative clauses and even nominal complements etc.2 All types of modifying elements can be followed by ge3. The variety of pre-ge3 modifiers is illustrated with the following examples:
.╅ An anonymous reviewer points out that it is also possible to have ge3 following quantifying expressions, as shown in the following examples: (3)
so2-jau5 (ge3) hok6saang1 all-have ge student ‘all students’
(4) ?hou2 do1 (ge3) hok6saang1 very many ge student ‘a lot of students’ Unlike other modifier classes mentioned in the main text, some of these phrases sound marginal. For some quantifiers, combination with ge3 renders the phrase ungrammatical: (5)
*mou5mat1 ge3 hok6saang1 not-much ge student Intended reading: ‘not many students’
The Cantonese ge3 
(11) fei4 ge3 neoi5jan2 fat ge woman ‘fat woman/women’
(adjective)
(12) ngo5 ge3 ce1 1sg ge car ‘my car/cars’
(possessor)
(13) a.
ngo5 maai5 ge3 ce1 1sg buy ge car ‘car/cars that I bought’
b. Peter coeng3-go1 ge3 sing1jam1 Peter sing-song ge sound ‘The sound of Peter’s singing’ c.
keoi5 zi6-sat3 ge3 siu1sik1 3sg self-kill ge news ‘The news that she/he killed herself/himself ’
(relative clause)
(aboutness relative clause)
(nominal complement)
It seems that in Cantonese adnominal modification, the grammar treats modifiers of different syntactic categories identically by sticking a ge3 at the end of them before they can serve as proper adnominal modifiers.3 The term noun in (9a) refers to both NP and DP. By NP, I refer to the maximal projection of N which can be embedded under functional elements like classifiers, numerals and demonstratives; by DP, I refer to a noun phrase headed by a demonstrative. Thus, (9a) can in fact be re-written into two sub-cases:4 (14) a. [(dem)-(nume)-(classifier)-modifier-*(ge3)-N] b. modifier-(ge3)-[dem-(nume)-classifier-N]
The addition of the modifier-ge3 sequence does not affect the referential interpretation of the modified noun phrase. In cases where the modifier-ge3 sequence precedes a
Since the acceptability of the [quantifier – ge3] combinations vary, I do not include these examples in this paper. .â•… The same applies to the Mandarin de. Partly due to this close relationship with modification and its apparent indifference to the syntactic categories of the preceding modifiers, some scholars have opted for a purely semantic account for the Mandarin de (cf. Cheng & Sybesma 2005).The Mandarin de is sometimes treated as the overt realization of a semantic operator that is related to predication. .â•… It is also possible for [modifier-ge3] to precede a [Nume-Cl-N] phrase. In such a case, the reading is obligatorily specific (Huang 1982; Sio 2006; Zhang 2006). I assume in these cases, the D layer is also present. These cases are not discussed in this paper, but see Sio (2006) for discussion.
 Joanna Ut-Seong Sio
demonstrative (i.e. the case in 14b), ge3 is optional with respect to grammaticality: deleting it would not render the phrase ungrammatical. When the modifier-ge3 sequence precedes an np (i.e. the case of 14a), ge3, in most cases, cannot be omitted. There are exceptional cases, however, as will be discussed next. For ease of exposition, let’s call modifiers without ge3 ‘bare modifiers’ and modifiers with ge3 ‘ge3-modifiers’. There are in fact only two situations in which bare modifiers can combine with a noun directly. The first case is when the modifiee is a bare noun. In this case, the possibility of having bare modifiers depends on their types. Vaguely speaking, the only legitimate pre-bare-noun bare modifiers are adjectives relating to size, color, shape, origin or modifiers whose descriptive contents pertain to some inherent properties of the noun (see Paul 2005 and references therein).5 Some examples are given below: (15) a.
size: color: shape: daai6 toi2 b. hung4-sik1 syu1 c. jyun4 toi2 big table red-color book round table
d. origin: e. Inherent properties: zung1gwok fo3 (go3) lei5 zi6 China goods cl lei word ‘the character Lei’ f.
dik1si2 si1gei1 taxi driver ‘taxi driver’
Another case where a modifier can combine with a noun without ge3 is when the modifiee is definite and contains a classifier (Cheung 1989; Sio 2006, 2008). This includes cases where the modifier appears before a demonstrative, as in (16) and also cases where the modifier appears immediately before a definite [classifier-noun] phrase, as in (17).6 (16) daai3 mou2 go2 go3 naam4jan2 wear hat that cl man ‘the man who wears a hat’ (17) daai3 mou2 go3 naam4jan2 wear hat cl man ‘the man who wears a hat.’
Note, for reasons yet unknown, ge3 can never appear immediately preceding a classifier. The phrase below is horrendously bad: .╅ The precise constraints as to what are legitimate pre-NP bare modifiers are not clear. However, there is evidence to show that bare modifier-NP sequences are not compounds, see Paul (2005) and Sio (2006) for discussion. .╅ In Mandarin, [Cl-N] phrases cannot be interpreted as definite.
The Cantonese ge3 
(18) *daai3 mou2 ge3 go3 naam4jan2 wear hat ge cl man Intended reading: ‘the man who wears a hat’
What the above intends to show is that even though ge3 appears in nominal modification contexts, it is not the case that ge3 is always needed for modification. Bare modifiers are also allowed in certain environments. In situations where both bare and ge3-modifiers are allowed, it is well-documented that they give rise to a difference in interpretation: the latter being obligatorily contrastive but the former not (Zhu 1984). Consider the two examples below: (19) a.
dung3 pizza cold pizza
b. dung3 ge3 pizza cold ge pizza
Both (19a) and (19b) refer to cold pizzas. However, they differ in the context within which they can be used. (19a) can be used in a situation where there are only cold pizzas around. It can also be used in a situation where both cold and hot pizzas are available in the surrounding. For (19b), however, it can only be used in a situation where there are also hot pizzas (or other contrasting types) around. By using a ge3-modifier, the speaker is picking out one type of pizzas out of a list of different types of pizzas in the context. In other words, ge3-modifiers can only be used in a contrastive environment. Sproat and Shih (1990) make the following observation for Mandarin, which also holds for Cantonese. Bare adjectives obey ordering restrictions while adjectives with markers do not. Adjectives with the marker have free orderings. The following two Cantonese examples illustrate the contrast. (20) a.
daai6 jyun4 toi2 big round table
b. *jyun4 daai6 toi2 round big table (21) a.
daai6 ge3 jyun4 ge3 toi2 big ge round ge table
b. jyun4 ge3 daai6 ge3 toi2 round ge big ge table
In (20), the two pre-nominal adjectives are bare and flipping their order leads to ungrammaticality. In (21), the adjectives come with ge3. Flipping their order does not lead to ungrammaticality.7
.╅ It is, however, not true that all ge3-modifiers are not subject to any ordering restrictions. For instance, it has been observed that individual-level ge3-modifiers have to be placed closer
 Joanna Ut-Seong Sio
When both bare modifiers and ge3-modifiers are present, the ge3-modifiers have to precede the bare modifiers: (22) a.
laam4sik1 ge3 daai6 toi2 blue ge big table ‘A big table that is blue’
b. *daai6 laam4sik1 ge3 toi2 big blue ge table
The above observations paint the following picture of the properties of ge3-modifiers: (23) a. Ge3-modifiers are contrastive. b. For ge3-modifiers, ordering restrictions do not apply. c. Ge3-modifiers always precede bare modifiers.
2.1.2â•… [modifier–ge3]- Ø (nominal) The Mandarin de (as well as markers in other East Asian languages) is often treated as a nominalizer (Li & Thompson 1981) as [modifier-de] phrases can function as noun phrases in the absence of a noun. In Cantonese, [modifier-ge3] phrases can also function as noun phrases, both in subject and object positions, as shown in the following examples: (24) hung5sik6 ge3 hai6 peng4gwo2 red ge be apple ‘The red ones are apples.’
to the noun than stage-level ge3-modifiers (Del Gobbo 2005), see (6a) and (6b). Complement clauses containing ge3 also have to be placed closer to the noun than non-complement clauses (Tsao 1997; Simpson 2002), see (7a) and (7b) below. (6) a. [ngo5 kam4jat6 gin3dou2 ge3]A, 1sg yesterday see ge
[fei4 ge3]b go2 go3 naam4zai2 le1? fat ge that cl boy qp
‘Where is the fat boy that I saw yesterday?’
b. ?/*[…]B […]A go2 go3 naam4zai2 le1? that cl boy qp (7) a.
[haak3 jan4 ge3]A, scare people ge
[keoi5 zi6-sat3 ge3]B siu1sik1 3sg self-kill ge news
‘The scary news that she/he killed herself/himself ’
b. ?/* […]B […]A siu1sik1 news
The Cantonese ge3 
(25) jau5-mou5 sik6-dak1 ge3 le1? have-neg-have eat-possible ge qp ‘Do you have edible ones?’
It should be noted that in these cases, even though there are no referential elements like the demonstrative or the classifier, the ge3-nominals in (24) and (25) do not have a kind reading. The interpretations of the two ge3-nominals are referential. The interpretation of hung5-sik6 ge3 in (24) is ‘the red ones’, referring to a familiar set in the discourse out of which the red ones are picked. Similarly, the interpretation of sik6-dak1 ge3 in (25) is ‘the edible ones’, meaning the edible ones out of the familiar set of objects.8 Note that it is also possible to have more than one [modifier-ge3] phrase in an argument position: (26) daai3 mou2 ge3, daai6zek3 ge3 hai6 laam4-kau4 jyun4 wear hat ge well-built ge be basket-ball player ‘The ones who wear hats and who are well-built are basketball players.’
In view of the nominal interpretation of such phrases and the possibility of multiplying them, I assume that such noun-less [modifier-ge3] phrases are like regular adnominal modifiers. The only difference is that there is an empty noun following such phrases, representing a familiar set of entities. (27) [modifier-ge3]-Ø(nominal)
2.1.3â•… Conditionals and contrastive topic clauses Ge3 can appear in conditionals and contrastive topic clauses, consider the following examples: (28) jyu4gwo2 lok6-jyu5 ge3 (waa2/6), ngo5 zau6 m4 ceot1 gaai1 if fall-rain ge saying, 1sg then neg go-out street ‘If it rains, then I won’t go out.’ (29) gong2-dou3 go1sau2 ge3 (waa2/6), ngo5 zui3 zung1ji3 Nina Simone talk-about singer ge saying, 1sg most like Nina Simone ‘When it comes to singers, I like Nina Simone best.’
8.â•… To have a kind reading, a head noun representing a general object like je5 ‘thing’ can be added. For instance: (8) a. hung5sik1 ge3 je5 = red things b. sik6-dak1 ge3 je5 = edible things red ge things ear-possible ge thing
 Joanna Ut-Seong Sio
The usage of ge3 in conditionals and contrastive topic clauses are subsumed under the adnominal usage because in both cases, ge3 is followed by the noun waa2/6, which literally means ‘saying’. A small detour in explaining why waa2/6 is a noun is in order here. In other environments, waa6 (with low level tone) can be used as either a verb (i.e. say) or a noun (i.e. saying), as illustrated in the following examples: (30) Peter waa6 nei5 fei4 aa3 Peter say you fat sfp ‘Peter said you are fat.’ (31) mui5 jat1 geoi3 waa6 every one cl saying ‘every saying/utterance’
(verb)
(noun)
However, waa2 (with a rising tone) can only be a noun. It cannot function as a verb. (32) a.
gwong2dung1-waa2 Canton-saying ‘Cantonese’
b. *Peter waa2 nei5 fei4 aa3 Peter say you fat sfp Intended reading: ‘Peter said you are fat.’
(noun)
(verb)
In other words, while waa6 can be a noun or a verb, waa2 can only be a noun. In a position where both waa2 and waa6 are possible, it should be a position for a noun. Thus, I conclude that the post-ge3 waa2/6 in conditionals and contrastive topic clauses is a noun. Ge3 in these cases is also adnominal. Even though ge3 in conditionals and contrastive topic clauses is an adnominalge3, it differs from the other adnominal usages in several ways. Firstly, the element that precedes ge3 can be a full clause denoting a proposition; secondly, waa2/6 is the only noun that can follow ge3, though it can be optional in both cases.9
9.â•… For conditionals and contrastive topic clauses in Mandarin, the equivalent of the Cantonese post-ge3 waa2/6, i.e. hua, cannot be absent.
(9)
(10)
ruguo xia-yu de *(hua), …. if fall-rain de saying shuo-dao gexing de *(hua), ….. talk-about singer de saying
(conditional) (contrastive topic clause)
The Cantonese ge3 
2.2â•… Sentence-final ge3 2.2.1â•… Assertive-ge3 When ge3 appears at the end of a well-formed sentence, it gives rise to an assertive reading (Law 1990; Fung 2000). It is often referred to as a sentence-final-particle (henceforth SFP). Consider the following sentences: (33) Peter zung1ji3 sik6 lau4lin4 Peter like eat durian ‘Peter likes to eat durians.’ (34) Peter zung1ji3 sik6 lau41lin4 ge3 Peter like eat durian ge ‘Peter likes to eat durians.’
(33) is a simple statement saying that Peter likes durians. It can be uttered out of the blue, like in the beginning of a story about Peter and the durians. (34) has something extra. It carries the speaker’s assertion that the statement is true. It cannot be uttered out of the blue. It carries the implication that the statement is relevant to the context: an actuality reading. Sybesma (2004) suggests that ge3 is probably best characterized as an ‘actuality marker’, asserting that the statement is highly relevant to the current conversation. I assume that the assertive-ge3 is used to express the speaker’s assertion that the truth of the statement is relevant to the current conversation. Assertive-ge3 can alternate with other SFPs. For instance the ge3 in (35) can be replaced by any of the following g-particles (SFPs with the onset /g/). (35) keoi5 sik1 zung1man2 ge3/ge2/gaa2/gaa3/gaa4/gaa5/gaak3 3sg know Chinese ‘S/He knows Chinese.’ (with different epistemic nuances depending on the choice of g-particles)
Li (2006) assumes that all g-particles have the semantics of assertion. Furthermore, she suggests that ge3 seems to express the core meaning of assertion, with /g/ giving rise to assertion. The rime /e/ and tone 3 (mid-level) are the default rime and tone. They do not carry extra semantics. All the other g-particles are derived from ge3 composing with other semantically loaded vowels, codas and tones. All g-particles contain ge3 underlyingly, a view that I would adopt. 10
.╅ The composition of SFP of the g-family can be schematized as follows, with numbers indicating tones (taken from Li 2006, pp. 113): ge2 = gaa3 =
ge3 ge3 + 1 ge3 + aa
 Joanna Ut-Seong Sio
2.2.2â•… predicative-ge3 Ge3 can appear after the predicate in class-membership type of copular sentences as shown in (36) and (37): (36) gin6 saam1 hai6 hung4sik1 *(ge3) cl shirt be red ge ‘The shirt is red.’ (37) gin6 saam1 hai6 kam4jat6 maai5 *(ge3) cl shirt be yesterday buy ge ‘The shirt was bought yesterday.’
I assume that the function of the copula in this type of copular sentences is to set up a relationship in which the subject is assigned a property denoted by the predicate. Unlike assertive-ge3, if ge3 is absent in the above two cases, the sentence would be ungrammatical. At least descriptively, it can be said that the presence of ge3 makes the post-copular element a legitimate post-copular predicate. For this reason, I call it the predicative-ge3. 11
2.2.3â•… Interaction between assertive-ge3 and predicative-ge3 In this section, I present some observations on the interaction between the assertivege3 and the predicative-ge3. Even though assertive-ge3 and predicative-ge3 both occupy a sentence-final position, there are several ways to tell them apart. Predicative-ge3 appears with classmembership type of copular sentences and deleting it would render the sentence ungrammatical. Assertive-ge3, on the other hand, appears with a fully grammatical preceding sentence and is optional with respect to grammaticality. Secondly, at least to my ear, the assertive-ge3 is longer than the predicative-ge3. The longer ge3 is indicated by underlining below.
gaak3 gaa2 gaa4 gaa5
= = = =
ge3 + aa + k ge3 + aa+ 1 ge3 + aa + 4 ge3 + aa + 5 + 1
.╅ Note that it is not always the case that the post-copular predicate would require the presence of ge3. When the predicate is nominal in nature, ge3 is not only not needed, but not allowed. (11)
Peter hai6 jat1 go3 lou5si1 (*ge3) Peter be one cl teacher ‘Peter is a teacher.’
(38) gin6 saam1 hai6 hung4sik1 ge3 cl shirt be red ge ‘The shirt is red.’ (39) Peter zung1ji3 sik6 lau41lin4 ge3 Peter like eat durian ge ‘Peter likes to eat durians.’ (with assertion)
The Cantonese ge3 
(predicative-ge3)
(assertive-ge3)
Assertive-ge3 gives rise to a reading that the pre-ge3 proposition is true and it is relevant to the current conversation. Predicative-ge3 is needed for giving rise to a legitimate post-copular predicate. In other words, assertive-ge3 and predicative-ge3 have different properties. The prediction is that it should be possible then for the two ge3 to co-occur in copular sentences, as schematized below: (40) a. *[subject copula [predicate-ge3] ge3] b. *gin6 saam1 hai6 hung4sik1 ge3 ge3 cl shirt be red ge ge Intended reading: ‘The shirt is red.’ (with assertion)
(40b) is predicted to have the interpretation that the speaker is assuring the other conversation participant(s) that the shirt is red. However, (40b) is in fact ungrammatical. To convey the intended reading, only one ge3 is required in the sentence-final position, but ge3 is lengthened, as in (41).12 (41) gin6 saam1 hai6 hung4sik1 ge3 cl shirt be red ge ‘The shirt is red.’ (with assertion)13
In view of the above, I assume that there are indeed two ge3 underlyingly in (41). But the two are collapsed, i.e. haplology. (42) ge3 ge3 Æ ge3 predicative assertive predicative + assertive
Predicative-ge3 may alternate with other SFPs of the g-family. When it alters, it also brings an assertive reading, with different epistemic nuances depending on the choice of SFPs.
.╅ In fact, whenever a predicative-ge3 is lengthened in a class-membership type of copular sentences, it inevitably gives rise to an assertive reading. .╅ Suppose the hearer is worried about whether the color of the shirt she is about to receive will match her red handbag, the speaker can assure her by uttering (41).
 Joanna Ut-Seong Sio
(43) gin6 saam1 hai6 hung4sik1 ge3/ge2/gaa2/gaa3/gaa4/gaa5/gaak314 cl shirt be red ge ‘The shirt is red.’
This indicates that once the predicative-ge3 alters, the sentence in fact also contains an assertive-ge3.
3.â•… S imilarities between conditionals, contrastive topic clauses and assertive sentences If we consider the nature of the pre-ge3 elements, the usages of ge3 can be classified into two groups: (44) a.
proposition-ge3 (conditionals, contrastive topic clauses, assertive sentences) b. property-ge3 (part of the adnominal modifier, part of the predicate in class-membership type of copular sentences)
In this section, I argue for the similarity in the function of ge3 when it is preceded by a proposition-denoting element, i.e. in conditionals, in contrastive topic clauses and in assertive sentences. Let’s start with conditionals and contrastive topic clauses. Recall in both conditionals and contrastive topic clauses, the post-ge3 noun is waa2/6. Unlike other adnominal use of ge3, waa2/6 is not alternatable with other nouns. This suggests some similarities between conditionals and contrastive topic clauses. Haiman (1978) points out that conditionals and contrastive topics are marked identically in a number of unrelated languages (e.g. Hua, Turkish, Tagalog). Assuming that formal similarities indicate similarities in meaning, some hidden similarities in meaning between the two must exist. He reasons, conditionals are like polar questions with implied positive answers (Jespersen 1940):15
.╅ A rising prosody for ge2 indicative of interrogative nuances is also possible. Note that the interrogative interpretation is not available to the other extended /g/ particles. (12)
gin6 saam1 hai6 hung4sik1 ge2 cl shirt be red ge ‘The shirt is red?’
(with rising prosody)
.â•… Haiman (1978) also presents morphological evidence to support his claim He points out that in English, polar interrogative whether maybe replaced by the conditional if in indirect questions. In French, indirect polar questions are introduced by si ‘if ’.
The Cantonese ge3 
(45) a. Is he coming? (Yes.) Well then, I will stay. b. If he is coming, then I will stay.
The conditional in (45b) (if he is coming) establishes the proposition it denotes (he is coming) as given, i.e. making it presuppositional. Conditionals are established as given facts of a (maybe) hypothetical situation, based on which the apodosis can be evaluated against. They are formal devices whereby the speaker seeks the agreement of his interlocutor(s) as to their validity. In this sense, conditional clauses are like contrastive topics, which is also a device to bring in the existence of an entity (the topic) as given, i.e. presupposed, which functions as the basis for subsequent discussion. (46) As for the play, John saw it yesterday.
(Chafe 1976)
The topic sets a framework, which limits the applicability of the main predication (Chafe 1976).16 In Cantonese, the similarity between conditionals and contrastive topics is more striking in that the constituent that marks the contrastive topic is clausal and it is possible to add the word jyu4gwo2 ‘if ’ in the front, without affecting the interpretation. (47) jyu4gwo2 lok6-jyu5 ge3 (waa2/6), ngo5 zau6 m4 ceot1 gaai1 if fall-rain ge saying, 1sg then neg go-out street ‘If it rains, then I won’t go out.’ (48) (jyu4gwo2) gong2-dou3 go1sau2 ge3 (waa2/6), if talk-about singer ge saying, ngo5 zui3 zung1ji3 Nina Simone 1sg most like Nina Simone
‘(If we) talk about singers, I like Nina Simone best.’
The possibility of adding jyu4gwo2 ‘if ’ indicates that contrastive topic clauses in Cantonese can be viewed as conditionals. Thus, adopting Haiman’s (1978) view, in Cantonese, conditionals, including contrastive topic clauses, have the following function: They bring in a presupposition that is to be shared in a (maybe hypothetical) situation which sets a spatial and temporal framework that functions as the basis for further discussion (the apodosis or the comment). Waa2/6, I believe, is the nominal lexicalization of the ‘situation’. With respect to the sentence-final usage of ge3 as an assertive marker, what exactly is the effect of assertion? How is it similar to conditionals (including contrastive topic clauses)? Stalnaker (2000) suggests that in the setting of a conversation, different .â•… Haiman (1978) treats the presupposition of a proposition identically as the presupposition of an entity. For the former, it is the presupposition of a state of affair the proposition describes; for the latter, it is the presupposition of existence of the entity the topic describes.
 Joanna Ut-Seong Sio
participants can have different sets of presuppositions, some of the presuppositions are shared among all the participants and they constitute the common ground. Assertion has the effect of changing the shared set of presuppositions of the conversation participants by adding the content of what is asserted by the speaker to shared set, i.e. the common ground. In other words, conditionals, contrastive topic clauses and assertions are similar in that in all three cases, the constructions make the contents they denote into the shared presuppositions of all the conversation participants.17 In view of the fact that all these three constructions involve the use of ge3, I assume that ge3 is related to the function of making the speaker’s proposition into shared presuppositions in these three constructions. In conditionals (contrastive topic clauses included), ge3 makes the prege3 proposition presuppositional to all the interlocutors. Similarly for assertive-ge3, it brings the proposition denoted by the pre- ge3 sentence, which is only the speaker’s presupposition at the point of utterance, into the common ground. For this reason, sentence-final ge3 yields an intersubjective ‘I tell you’ or ‘believe me’ effect.
4.â•… Ge3, gam2 and demonstratives In this section, I discuss the similarities between ge3, gam3 (a marker that is employed in manner modification of events) and demonstratives.
4.1â•… Ge3 and gam2 Gam2, when used pre-verbally, is an indexical that refers to the manner property of an event. It can be used deictically, anaphorically and establishingly. When used deictically, it refers to a manner that is in the immediate context by way of demonstration. When used anaphorically, it refers to a manner that is made explicit in the previous discourse. By ‘establishingly’, I mean that the reference of gam2 is the content of the preceding description.18 This is illustrated in (49a), (49b) and (49c) respectively.
.â•… There is, however, one difference between conditionals (including contrastive topic clauses) on the one hand and assertive sentences on the other. For the former, the construction makes a proposition into a shared presupposition of a situation which can be hypothetical. For the latter, the construction makes a proposition into a shared presupposition in the actual situation. .â•… The term ‘establishing’ is borrowed from Hawkins (1978, pp. 131). He uses the term ‘referent-establishing relative clauses’ to refer to relative clauses like the woman he went out with last night, where a definite referent is established at the point the noun phrase is uttered.
The Cantonese ge3 
(49) a.
gam2 sik6 gam eat ‘eat this/that way’ (with demonstration)
(deictic)
b. gam2 sik6 gam eat ‘eat in a certain way’ (a manner that is previously mentioned in the discourse) c.
mou5 seng1 gam2 sik6 not-have noise gam eat ‘eat in a way that does not produce any noise’
(anaphoric)
(establishingly)
The property of the ‘establishing’ use of gam2 is similar to the adnominal use of ge3. Firstly, both are markers used in modification: the former is placed at the end of a manner adverbial when modifying an event; the latter is placed at the end of a modifier modifying a nominal. Secondly, not all manner adverbials need gam2. For adverbials that are inherently manner, gam2 is not needed. In such cases, the adverbial with gam2 gives rise to an obligatorily contrastive reading, similar to ge3-modification. Consider the following example: (50) a.
ceot1-seng1 gong2je5 produce-noise speak
b. # ceot1-seng1 gam2 gong2je5 produce-noise gam speak
(50a) sounds natural while (50b) sounds strange. The contrast can be explained in the following way. Under normal circumstances, the action of talking involves producing noise. Thus, “producing noise” is not a manner of talking, but rather, an inherent property of the action of talking. “Producing noise” is not contrastive with respect to other manners of ‘talking’. The contrast between (50a) and (50b) suggests that when gam2 is present, the adverbial is obligatorily contrastive. Since a contrastive reading is impossible in (50b), the sentence becomes strange. This is similar to ge3 in adnominal modification. Example (19) is repeated here as (51): (51) a.
dung3 pizza cold pizza
b. dung3 ge3 pizza cold ge pizza
(51b) is only legitimate in a situation where there are also hot pizzas. In other words, when ge3 and gam2 are present, the property denoted by the preceding description is obligatorily contrastive.
 Joanna Ut-Seong Sio
Thirdly, gam2-adverbials are also stackable with free ordering of the adverbials: (52) a.
maan6-maan6 gam2, jat1-daam6-daam6 gam2 sik6 slow-slow gam one-mouthful-mouthful gam eat ‘eat slowly, mouthful-by-mouthful’
b. jat1-daam6-daam6 gam2, maan6-maan6 gam2 sik6 one-mouthful-mouthful gam, slow-slow gam eat ‘eat mouthful-by-mouthful, slowly’
Moreover, given an appropriate noun, the pre-adnominal-ge3 description can be interpreted as the complement of the noun. Gam2, when combined with certain complement-taking verbs, can be interpreted as referring to the complement of the verb instead of manner. In (53a), the pre-ge3 description is, semantically, the complement of the noun siu1sik1 ‘news’. Gam2 in (53b) is ambiguous between being interpreted as referring to manner and referring to the complement of the verb. (53) a.
nei5 lei4-hoi1 ge3 siu1sik1 2sg leave ge news ‘the news that you left’
b. nei5 m4-hou2 gam2 gong2-je5 2sg neg-good gam speak-thing ‘Don’t speak in a certain way.’/‘Don’t say things like that.’
4.2â•… Gam2 and demonstratives Sio and Tang (2007) treat gam2 as a demonstrative-like element. Similar to gam2, demonstratives can also be used deictically, anaphorically and establishingly in Cantonese. (54) a.
go2 go3 naam4jan2 that cl man ‘that man (over there)’ ‘that man (we talked about earlier on)’
b. daai3 ngaan5geng2 go2 go3 naam4jan2 wear glasses that cl man ‘that/the man with glasses’
(deictic) (anaphoric) (establishing)
Furthermore, as is well-known, demonstratives are also contrastive in the sense that they cannot be used with unique objects like the sun: (55) # go2 go3 tai3joeng4 that cl sun
In view of the similarities, Sio and Tang (2007) conclude that demonstratives and gam2 are both indexicals. In order to point to a referent, they need a certain input. The
The Cantonese ge3 
requirement for an input can be represented by a [+matching] feature on demonstratives and gam2, which needs to be satisfied. The [+matching] feature can be satisfied in three different ways: (56) i. a demonstration (the deictic usage) ii. an antecedent in the discourse (the anaphoric use) iii. a preceding description (the establishing use)
In addition to a [+matching] feature, another shared property of demonstratives and gam2 is that they only appear in a contrastive environment. This property can be represented by a [+contrast] feature, which forces them to appear in a contrastive environment. In other words, both gam2 and demonstratives have the following feature make-up: (57) [+matching (deictic, anaphoric, establishing)], [+contrast]
4.3â•… Ge3 as a defective indexical Considering the aforementioned similarities between gam2 and ge3 on the one hand and gam2 and demonstratives on the other, one can also view ge3 as an indexical, but of a special type. Since ge3 can never be used without a preceding element, ge3 can be treated as a defective indexical that can only point establishingly, i.e. the [+matching] feature on ge3 can only be satisfied by a preceding description. It can neither be used deictically nor anaphorically. In other words, ge3 has a [+matching (establishing)] feature. In its adnominal usage, ge3 points (being an indexical) to the property denoted by the modifier, the property is then used to modify the noun. As for gam2, it points to the property denoted by the manner adverbial, which is then used to modify the verb. In conditionals, contrastive topic clauses and assertive sentences, the function of ge3 is similar: it points to the preceding proposition and put that into the shared presuppositions. With respect to the [+contrast] feature, as noted earlier on, just like demonstratives and gam2, the use of ge3 in regular adnominal modification cases is contrastive (see example (19) and (51)), i.e. the implication of the existence of alternatives. It can, in fact, also be argued that in conditionals, contrastive topic clauses and assertive sentences, the use of ge3 is also contrastive. In contrastive topic clauses, the topic that is being chosen to be talked about is contrasted with all the other potential topics that could be talked about;19 in conditionals, the speaker picks out a situation that contrasts with all the other possible situations as the protasis; in assertive sentences,
.╅ According to Haiman (1978), contrastive topics are selected by the speaker apropos of thoughts which he has not yet communicated to his listener.
 Joanna Ut-Seong Sio
one proposition (among all the other possible presuppositions of the speaker’s belief) is chosen to be placed in the shared presuppositions.20
4.4â•… Ge3 and demonstratives Treating ge3 as an indexical provides an explanation as to why in adnominal modification, when the modifiee contains a demonstrative, ge3 can be omitted. Both the demonstrative and ge3 have a [+match (establishing)] feature in that both can refer back to the preceding description denoting a property. Furthermore, the demonstrative is also [+contrastive] in nature (as shown earlier on). In other words, both the demonstrative and ge3 have the same features except that the demonstrative can also refer deictically and anaphorically. When the demonstrative is present, ge3 is thus not needed. (58) modifier-(ge3)-Dem-cl-N
5.â•… Conclusions This paper gives a descriptive overview of the usages of the grammatical marker ge3 in Cantonese. It organizes the diverse data in two ways: (i) according to whether there is a nominal element after ge3, and (ii) according to the nature of the pre-ge3 element, whether it denotes a proposition or a property. This paper has made some new observations: (a) the usages of ge3 in conditionals and contrastive topic clauses are also adnominal usages as the optional post-ge3 waa2/6 is arguably a noun; (b) the predicative-ge3 and the assertive-ge3 differ in phonological length; (c) even though the predicative-ge3 and the assertive-ge3 have different properties, the two cannot co-exist, indicating the possibility of haplology; (d) in conditionals, contrastive topic clauses and assertive sentences, the pre-ge3 element is propositional and in each of these cases, the construction has the effect of bringing the proposition it denotes into the set of presuppositions shared by all conversation participants, suggesting that ge3 is related to such discourse function; (e) it can be shown that ge3 is a defective indexical that has the features of [+matching/establishing] and [+contrast] by comparing it with similar elements in Cantonese. In view of the above, the function of ge3 seems to be the following: it points to either a property or a proposition, and brings that (if it is a property) to modify a noun or (if it is a proposition) into the shared set of presuppositions of the conversation participants. The only construction that seems to fall outside this idea is the usage of
.╅ It is unclear to me whether the [+contrast] feature is also applicable to the predicative-ge3.
The Cantonese ge3 
the predicative-ge3. Even though a unified account for all the possible usages of ge3 is not yet available, it is hoped that these new observations will provide a new perspective and act as a stepping stone in the understanding of such grammatical marker in Chinese as well as in other languages.
Abbreviations 1/2/3sg cl neg qp sfp
first/second/third person singular Classifier Negation Question particle Sentence final particle
References Chafe, Wallace. 1976. Givenness, contrastiveness, definiteness, subjects, topics and point of view. In Subject and Topic, Charles Li (ed.), 25–55. New York NY: Academic Press. Chao, Yuen-Ren. 1959. Yuyan Wenti (Questions in Linguistics). Beijing: Commercial Press. Cheng, Lisa & Sybesma, Rint. 2005. A Chinese relative. In Organizing grammar: Linguistics Studies in Honor of Henk van Riemsdijk, Hans Broekhuis, Norbert Corver, Riny Huybregts, Ursula Kleinhenz & Jan Koster (eds), 69–76. Berlin: Mouton. Cheung, Sam. 1989. Yueyu liangci yongfa de yan jiu (A study of the use of classifiers in Cantonese). Di’er Jie Gouji Hanxue Huiyi Lunwen: Yuyan Ji Wenzizu (Xi Ce). Taipei: Academia Sinica. Del Gobbo, Francesca. 2005. Chinese relative clauses: restrictive, descriptive or appositive? In Contributions to the XXX Incontro di Grammatica Generativa, Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter Schweikert & Giuseppina Turano (eds), 287–305.Venezia: Cafoscarina. Fung, Roxana. 2000. Final Particles in Standard Cantonese: Semantic Extension and Pragmatic Inference. Ph.D. dissertation, Ohio State University. Hawkins, John. 1978. Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction. London: Croom Helm. Haiman, John. 1978. Conditionals are topics. Language 54(3): 564–589. Huang, James. 1982. Logical Relations in Chinese and the Theory of Grammar. Ph.D. dissertation, MIT. Huang, Shi-Zhe. 2006. Property theory, adjectives, and modification in Chinese. Journal of Asian Linguistic 15: 343–369. Jespersen, Otto. 1940. A Modern English Grammar on Historical Principles, V: Syntax. London: George Allen and Unwin. Kitagawa, Chisato and Ross, Claudia. 1982. Prenominal modification in Chinese and Japanese. Linguistic Analysis, 9: 119–153.
 Joanna Ut-Seong Sio Law, Sam-Po. 1990. The Syntax and Phonology of Cantonese Sentence-final Particles. Ph.D. dissertation, Boston University. Li. Bo-Ya. 2006. Chinese Final Particles and the Syntax of the Periphery. Leiden: LOT Publication. Li, Charles & Thompson, Sandra. 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley CA: University of California Press. Lü, Shu-Xiang. 2004. Xiandai Hanyu Babai Ci. Beijing: Commercial Press. Paul, Waltraud. 2005. Adjectival modification in Mandarin Chinese and related issues. Linguistics 43(4): 757–793. Simpson, Andrew. 2002. On the status of modifying de and the structure of the Chinese DP. In On the Formal Way to Chinese Languages, Sze Wing Tang & Chen Sheng Liu (eds), 74–101. Stanford CA: CSLI. Sio, Joanna Ut-Seong. 2006. Modification and Reference in the Chinese Nominal. Leiden: LOT Publication. Sio, Joanna Ut-Seong. 2008. The encoding of referential properties in the Chinese nominal. Language and Linguistics 9(1): 101–126. Sio, Joanna Ut-Seong & Tang, Sze-Wing. (2007). The indexical expressions gam2 and gam3 in Cantonese. Studies in Cantonese Linguistics 2. Hong Kong: LSHK publication. Sproat, Richard & Shih, Chilin. 1991. The cross-linguistic distribution of adjectival ordering restrictions. In Interdisciplinary Approaches to Language: Essays in Honor of S.-Y. Kuroda, Carol Georgopoulos & Roberta Ishihara (eds), 565–693. Dordrecht: Kluwer. Stalnaker, Robert. 2002. Assertion. In Formal Semantics: The Essential Readings, Paul Portner & Barbara Partee (ed), 150–161. Oxford: Blackwell. Sybesma. Rint. 2004. Exploring Cantonese tense. Linguistics in the Netherlands 21(1): 169–180. Tsao, Feng-Fu. 1997. Complement and adjunct distribution and the two-place nominal in Chinese. Paper presented in the 9th North American conference on Chinese Linguistics, Victoria, Canada. Yap, Foong-Ha, Matthews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implication for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline: The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Zhang, Niina. 2006. Representing specificity by the internal order of indefinites. Linguistics 44(1): 1–21. Zhu, De-Xi. 1984. Dingyu he zhuangyu de qufen yu tici he weici de duili (On the demarcation between attributes and adverbials with reference to the opposition between substantives and verbs). In Yuyanxu Luncong, The Chinese Department of Beijing University (eds), 5–14. Beijing: Shangwu.
On gerundive nominalization in Mandarin and Cantonese* Sze-Wing Tang
The Chinese University of Hong Kong This paper examines a number of variations between Mandarin and Cantonese nominalizations, including genitive agent nominals, possessive objects, relativization of idiomatic expressions, verbless de expressions, and internal de expressions. Differences between these two languages boil down to the availability of some gerundive nominalization in these two dialects.
1.â•… Introduction There is a structural particle in nominal modification in Chinese,1 a grammatical morpheme that is used adnominally. Examples (1) and (2) show the adnominal
*The earlier version of this paper was presented at the International Workshop on Nominalization and Copulas in East Asian and Neighboring Languages (Chinese University of Hong Kong, January 7–11, 2007), the 15th Annual Meeting of the International Association of Chinese Linguistics, and the 19th Annual North American Conference on Chinese Linguistics (Columbia University, May 25–27, 2007), the Symposium on Chinese Syntax and Semantics (City University of Hong Kong, August 18–20, 2007), the 16th Annual Meeting of the International Association of Chinese Linguistics (Peking University, May 30-June 2, 2008), and invited lecture series at the Beijing University of Aeronautics and Astronautics (October 2008), the Beijing Language and Culture University (October 2008), and the Shanghai International Studies University (November 2008). The research reported here is conducted within the context of the project ‘A Comparative Study of Definiteness in Chinese Dialects’ (CUHK5476/06H), funded by the General Research Fund, University Grants Committee, HKSAR. I benefited from discussions with Lisa Cheng, Yang Gu, Jim Huang, Thomas Lee, Audrey Li, Danqing Liu, Huijuan Liu, Jianming Lu, Jiaxuan Shen, Fuzhen Si, Joanna Sio, Dylan Tsai, Ting-Chi Wei, Heyou Zhang, and Qingwen Zhang. I am particularly indebted to Foong-ha Yap for her wonderful suggestions, friendly encouragement, and endless patience through the revising process. I would also like to thank an anonymous reviewer of this paper. All errors remain my own. .â•… The term ‘Chinese’ used in this paper is a collective term referring to Chinese dialects. ‘[M]’ and ‘[C]’ are used to mark Mandarin and Cantonese sentences, respectively. The following
 Sze-Wing Tang
usages of Mandarin Chinese (hereafter called ‘Mandarin’) de. In (1), the adjective youyong ‘useful’ functions as an attributive of the head noun shu ‘book’ and the pre-de constituent wo mai ‘I bought’ in (2) is a relative clause. In these two examples, de is used to link the attributive to the head noun. (1) youyong de shu useful de book ‘useful books’
[M]
(2) wo mai de shu I buy de book ‘the books that I bought’
The counterpart of Mandarin de in Cantonese Chinese (hereafter called ‘Cantonese’) is ge.2 Examples in (3) and (4) show that the Cantonese adnominal grammatical morpheme ge shares the same properties as those of Mandarin de listed above. (3) jaujung ge syu useful ge book ‘useful books’
[C]
(4) ngo maai ge syu I buy ge book ‘the books that I bought’
Though Mandarin de and Cantonese ge share many grammatical similarities, there are a few exceptions. The goal of this paper is to examine the dialectal variations in Mandarin and Cantonese with respect to the sentences with the adnominal grammatical morpheme. This paper is organized as follows. In Section 2, I attribute the lack of the genitive agent nominals in non-object positions in Cantonese (contra Mandarin) to the absence of gerundive nominalization. In the rest of the sections, I extend the consequences of my analysis to other apparently unrelated phenomena in the two dialects. Attempts are made to better understand why gerundive nominalization is restricted to the object position in Cantonese.
abbreviations are used in giving glosses for Chinese examples: Des: descriptive complement marker, Cl: classifier, Exp: experiential aspect marker, Perf: perfective aspect marker, and SFP: sentence final particle. The third person singular pronoun in Chinese is glossed as ‘he’ in English. The Cantonese Romanization follows the Linguistic Society of Hong Kong Cantonese Romanization Scheme. For the sake of simplicity, tone markers are omitted. .â•… The default tone of ge is mid-level (tone 3). For the grammatical properties of Cantonese ge, see also Sio (this volume).
On gerundive nominalization in Mandarin and Cantonese 
2.â•… Genitive agent nominals Let us begin with some initial observations concerning the distribution of the adnominal usage of de. Consider example (5), where Mandarin has the adnominal usage of de in genitive agent nominals. The pronoun ta ‘he’ marked with de is interpreted as the agent, not as the possessor, hence the term ‘genitive agent’. Huang (1997) identifies the genitive agent nominal in (5) with the bracketed constituent in (6), in such a way that the bracketed constituent is a gerundive nominal which expresses the action of serving. In other words, ta de laoshi ‘his (serving as) a teacher (lit.: his teacher)’ in (5) is a reduced form of ta de dang laoshi ‘his serving as a teacher’. In the subject position in (5), it is interpreted in topic-like fashion as ‘Speaking of his serving as a teacher, he serves well’, while in the object position in (6), it is interpreted as ‘He does his job of serving as a teacher.’ In these cases, there is an abstract usage involving nominalization of events (second order ontological entities) rather than modification of objects (first order ontological entities). In this regard, the bracketed constituent is a reduced form of the ‘NP de VP’ gerundive nominal construction in Modern Chinese, traceable to an older lineage of the ‘NP zhi VP’ gerundive nominal construction in Old Chinese.3 (5) [Ta de laoshi] dang-de hao. he de teacher serve-Des well ‘He serves well as a teacher.’
[M]
(6) Ta dang [ta de laoshi]. he serve he de teacher ‘He does his job as a teacher.’
Huang (2008) further notes that idiomatic expressions can also occur in genitive agent nominal constructions, for example, chui-niu ‘bluff ’ (lit.: ‘blow-cow’). The bracketed constituent in (7) is arguably a reduced form of ni de chui-niu ‘your blowing cow’ and (7) literally means ‘Your cow has been overblown’. As seen in (8), Huang shows that the classifier can only be the general classifier ge, not the classifier tou ‘head’ that is specific to animals like cows. Consequently, what the classifier selects is an event, not an object in this structure, which implies that the bracketed constituent in (8) is a gerundive nominal containing an elided verb that literally means ‘blow’. (7) [Ni de niu] chui-de tai guohuo le! you de cow blow-Des too over sfp ‘You have bluffed your way too much.’
[M]
.╅ See Yap and Matthews (2008) and Yap, Choi, and Cheung (2010) for a discussion of grammaticalization pathways of Mandarin de.
 Sze-Wing Tang
(8) [Zhe ge/ *tou niu] chui-de tai guohuo le! this cl head cow blow-Des too over sfp ‘The bluff is too much.’
Let us compare (5) and (7) in Mandarin and their counterparts in Cantonese, namely (9) and (10), respectively. Such genitive agent nominals are, however, not very good in Cantonese.4 (9) *[Keoi ge lousi] zou-dak hou. ╛╛↜渀屮he ge teacher do-Des well ‘He serves well as a teacher.’
[C]
(10) *[Nei ge ngau] ceoi-dak taai gwofo laa! ╛╛↜渀屮you ge cow blow-Des too over sfp ‘You have bluffed your way too much.’
It is worth noticing that gerundive and derived nominals are different in nature (Chomsky 1970) and readers should not be confused by these two types of nominalizations in our discussion. Cantonese verbs like paiping ‘criticize’ may form acceptable derived nominals, even in the subject position, on a par with their Mandarin counterparts, as is evident by the following examples in (11) and (12). Though (12) remains somewhat marked and may sound a bit formal to the ears of native speakers, they are far better than those in (9) and (10).5 (11) [Laoshi de piping] guwu-le women. teacher de criticism inspire-Perf we ‘Teacher’s criticism has inspired us.’
[M]
(12) [lousi ge paiping] gwumou-zo ngodei. teacher ge criticism inspire-Perf we ‘Teacher’s criticism has inspired us.’
[C]
When we compare (9) and (10) with (13) and (14), however, we observe a sharp contrast in grammaticality. Cantonese genitive agent nominals are acceptable in the object position. Notice that the first conjunct of (13) is the Cantonese counterpart of (6). The gerundive nominal keoi ge lousi ‘his (serving as) a teacher’ in (13) is the object of the
.â•… The idiomatic expression ceoi-ngau ‘bluff ’ (lit.: ‘blow-cow’) exists in colloquial Cantonese. Notice that (9) is acceptable under a different interpretation: ‘His teacher did well’, where the pronoun keoi ‘he’ is the possessor, not the agent. .â•… Cantonese derived nominals are not productive and many of them seem to be borrowed from Mandarin/standard written Chinese. Native Cantonese verbs like naau ‘criticize, accuse’ cannot form a derived nominal *lousi ge naau ‘teacher’s criticism’. Productivity of Cantonese derived nominals is reminiscent of English derived nominals (Chomsky 1970).
On gerundive nominalization in Mandarin and Cantonese 
verb zou ‘perform’. Data from Cantonese show that there is an asymmetry in the distribution of gerundive nominals in the subject and the object positions. (13) Keoi zou [keoi ge lousi], ngo zou [ngo ge hoksaang]. he perform he ge teacher I perform I ge student ‘He does his job as a teacher, and I do my job as a student.’
[C]
(14) Nei ceoi [nei ge ngau], ngo se [ngo ge man]. you blow you ge cow I write I ge paper ‘You bluff your way, and I do my paper writing.’
Given the grammaticality of Mandarin examples like (5) and (7) and the ungrammaticality of Cantonese examples like (9) and (10), it appears that de in Mandarin is sufficiently grammaticalized that it could form a gerundive nominal with the verb elided, while ge in Cantonese needs to retain the verb. This would help explain why Cantonese does not have verbless gerundive nominalization in non-object positions like (9) and (10). With this in mind, consider (13) and (14) again, which contrast with the gerundives (9) and (10) in the subject. To account for the grammaticality of (13) and (14) in Cantonese, we could posit the gerundive nominal structure in (15), following Huang (1997), in which ‘NP’ is the genitive agent, and the verbs zou ‘perform, serve’ in (13) and ceoi ‘blow’ in (14) are copies of the verbs embedded within the bracketed nominalized constituent. The verbs can be elided in the object position, represented as ‘Ø’, if their contents can be recovered from the copies in the root clause, i.e. the verb V in the main clause. Their status as copies is, however, far from transparent in non-object positions, and verbal ellipsis is difficult unless the verbless genitive agent nominal construction is more grammaticalized in the language, as is the case for Mandarin but not for Cantonese. As will be shown in Section 3, reanalysis of the genitive subject in the non-object positions in Mandarin as a possessor NP qualifies it as a noun phrase marker. That is the genitive agent serves as a referentiality marking device to signal that the verbless constituent is to be construed as a gerundive nominal. Hence, the asymmetry between the object and non-object positions with respect to the distribution of the gerundive nominals is attributable to the different degree of grammaticalization of the grammatical morphemes de in Mandarin and ge in Cantonese.6 (15) Subject V [np de/ge Ø np]
.╅ See Tang (2008, 2009) for a formal analysis of the parametric variation of gerundive nominalization in Chinese dialects.
 Sze-Wing Tang
3.â•… Possessive objects Objects in some Mandarin sentences are in possessive form, for example, wo de in (16), which literally means ‘my’ but is understood as ‘me’, and as such are often referred to as ‘possessive objects’ (Chao 1968; Huang 1997). Cantonese has a similar way to say (17). (16) Ta chi wo de cu. he eat I de vinegar ‘He is jealous of me.’
[M]
(17) Keoi haap ngo ge cou. he sip I ge vinegar ‘He is jealous of me.’
[C]
The Mandarin verb chi ‘eat’ in (16) and Cantonese verb haap ‘sip’ in (17) are verbs in the root clause that can recover the contents of the elided verbs within the embedded gerundive nominal in the object position, just like the structure in (15).
4.â•… Relativation of idiomatic expressions Aoun and Li (2003: 138) point out that the object part of some Mandarin VO idioms can undergo relativization, for example, cu ‘vinegar’ of the VO idiom chi cu ‘be jealous’ (lit.: ‘eat vinegar’) in (18) can undergo relativization. (18) literally means ‘The vinegar that he eats is greater than anyone else’s’.7 (19) is interesting. Youmo ‘humor’ is a loanword borrowed from English humor and can be reanalyzed as a VO compound in Mandarin. The object part mo ‘-mor’ undergoes relativization and becomes the relative head. (18) [Ta chi de cu] bi shei dou da. ╛╛↜渀屮he eat de vinegar compare who all big ‘His jealousy is greater than anyone else’s.’
[M]
(19) Wo ting-bu-dong [ta you de mo]. I listen-not-understand ╛╛↜渀屮he hu- de -mor ‘I do not understand his humor.’
The relative clause does not just modify a noun; instead, it modifies a gerundive nominal. The noun following de is actually a gerundive in disguise with an elided verbal head, which roughly means ‘eating vinegar’ in (18) and ‘having a humor’ in (19). .â•… The metaphorical usage of chi cu ‘be jealous’ comes from the idea that those who are jealous often are highly worked up and suffer from excessive stomach acids spilling over, causing harm to one’s health, biologically and psychologically.
On gerundive nominalization in Mandarin and Cantonese 
The idiomatic uses in these two examples involve nominalization of events rather than modification of objects, in which the relativizing use of de is reinterpreted as the gerundive use of de.8 Aoun and Li (2003: 152) observe that dao ‘knife’ of the VO idiom kai dao ‘(medical) operation’ (lit.: ‘open-knife’) can be relativized in (20). After relativization, the gerundive nominal can be preceded by the demonstrative na ‘that’ and the classifier ge. Note that the Mandarin classifier ge, which is regarded as a ‘general classifier’ and which can be used with most nominals, cannot be replaced with the Mandarin classifier ba in (21), which normally goes with objects in a long shape like knives. The contrast between (20) and (21) implies that the classifier in such idiomatic relatives actually selects an event, i.e. a gerundive nominal, and what is relativized is a gerundive nominal, not just the noun dao ‘knife’. (20) [Ta kai de na ge dao] hen chenggong. he open de that cl knife very successful ‘That medical operation he performed was successful.’
[M]
(21) *[Ta kai de na ba dao] hen chenggong. ╛╛↜渀屮he open de that cl knife very successful ‘That medical operation he performed was successful.’
Notice, however, that Cantonese does not have verbless gerundive nominalization in non-object positions. Consequently, the examples in (22) and (23), the Cantonese counterparts of (18) and (19), respectively, should not exist in Cantonese. In spite of the fact that the same idiomatic chunks, haap cou ‘jealous’ (lit.: ‘sip-vinegar’) and jaumak ‘humor’, can be used in colloquial Cantonese, (22) and (23) are all unacceptable. The contrast between the Mandarin examples and their counterparts in Cantonese thus falls along the same lines as we have observed. (22) *[Keoi haap ge cou] daaigwo sojau jan. ╛╛↜渀屮he sip ge vinegar bigger all person ‘His jealousy is greater than anyone else’s.’
[C]
.â•… In their original analysis, Aoun and Li (2003) claim that the relative head is the object of the VO idiom, like cu ‘vinegar’ in (18) and mo ‘-mor’ in (19). Foong-ha Yap (personal communication) suggests that given that de is highly grammaticalized and used not only as a relativizer but also as a genitive, it seems to be easier for Mandarin to extend the relativizer function of de to a genitivizing and hence gerundive one. This may also have been facilitated by the availability of genitive zhi as a gerundive in Classical Chinese. In the case of Cantonese, however, ge is less grammaticalized and verbless gerundive ge is acceptable only in the object position. Perhaps Cantonese ge is more restricted in its gerundive use because spoken Cantonese may have been more vernacular, with fewer links to the older gerundive zhi uses found in Classical Chinese.
 Sze-Wing Tang
(23) *Ngo teng-m-ming [keoi jau ge mak]. I listen-not-understand ╛╛↜渀屮he hu- ge -mor ‘I do not understand his humor.’
5.â•… Verbless de expressions Examples (24) and (25) in Mandarin, taken from Si (2002: 35), are sentences without a verb on the surface, which I call ‘verbless de expressions’.9 The meaning of the elided verb in these examples is recoverable if provided with rich contextual information and is normally interpreted as an action of playing a certain role in a particular context, for instance, in a committee or in a drama. The nominal followed by de will be interpreted as the agent, not the possessor. But their Cantonese counterparts, as illustrated in (26) and (27), respectively, are ungrammatical.10 (24) Shei de zhuxi? who de chairperson ‘Who served/will serve as the chairperson?’
[M]
(25) Wo de Dachun, ni de Xi’er. I de Dachun you de Xi’er. ‘I play the role of “Dachun” and you “Xi’er”.’ (26) *Bingo ge zyuzik? who ge chairperson ‘Who served/will serve as the chairperson?’
[C]
(27) *Ngo ge Daaiceon, nei ge Heiji. I ge Daaiceon nei ge Heiji ‘I play the role of “Daaiceon” and you play the role of “Heiji”.’
.â•… The existence of the verbless de expressions in Mandarin may suggest that there could be development of new finite structures from clausal nominalization constructions. That is, in (24), shei de zhuxi may have been reinterpreted as ‘Who was/will be the chairperson’, where de apparently seems to function as a copula. See Delancey (this volume) and Grunow-Hårsta (this volume) for a typological study along these lines and Yap and Matthews (2008) and references cited therein for the relationship between nominalization and copula usage. .â•… (26) and (27) are acceptable under a different interpretation where the nominal followed by ge is understood to be the possessor, rather than the agent. Such genitive interpretation is also available in Mandarin, and is arguably the more basic function of Cantonese ge and Mandarin de. As in the case of zhi in Classical Chinese, the genitive function appears to be the more basic function, and the gerundive function appears to be secondary or derived (Yue 1998).
On gerundive nominalization in Mandarin and Cantonese 
Although gerundive nominals are generally embedded as arguments of a verb, as the examples we have seen so far, they can function as complete and independent utterances in Mandarin, which could be regarded as a kind of ‘stand-alone’ nominalization in the sense of Noonan (1997), Watters (2008), Grunow-Hårsta (this volume), and Wrona (this volume). The verbless de expressions in Mandarin are clauses formed by gerundive nominals,11 having a configuration like the bracketed gerundive nominal structure in (15) where the verb within the embedded gerundive nominal is elided.12 In Cantonese, we cannot have (26) and (27). These are precisely the consequences that follow from the generalization in this paper that Cantonese does not have verbless gerundive nominalization in non-object positions.
6.â•… Internal de expressions Mandarin has the adnominal usage of de that is preceded by the verb and followed by the object, apparently in the pre-object position, such as (28). Sentences like (28) in Mandarin are called the ‘internal de expressions’ for ease of presentation. Example (29) is the Cantonese counterpart of (28) and is ungrammatical (Lee & Yiu 1998).13
.â•… A significant difference between (25) and the internal de expression in (i) lies in their temporal interpretation. (25) may refer to either a past event or a non-past event (as used as an imperative). However, it is felicitous in (i) only if it refers to a past event. Thanks to Fuzhen Si and Jim Huang (personal communication) for useful discussion. Foong-ha Yap (personal communication) suggests that gerundive nominals are ‘ungrounded’ (or ‘unanchored’) in terms of temporal reference, being referentials rather than event descriptions. Being underspecified with respect to temporal reference, it is thus readily compatible with various temporal references (past and non-past). The fact that yan de ‘play DE’ can only have a past interpretation indicates that de may have also developed into a past tense marker (see Grunow-Hårsta, this volume, on Magar; see also Wu 2004 and Yap & Grunow-Hårsta 2010). (i)
Ni yan de Xi’er. you play de Xi’er ‘You played the role of “Xi’er”.’
[M]
.â•… In Si’s (2002) analysis, (24) and (25) are clauses containing an empty verb. De is C (for ‘complementizer’) and the agent nominal raises from the subject of the clause IP, i.e. the complement of C, to the specifier of CP. .â•… (29) is unacceptable under the intended clausal reading, but it is acceptable if it has the nominal reading ‘the telegram that he sent yesterday’.
 Sze-Wing Tang
Chao (1968: 297) points out that the internal de expressions are a ‘specially Northern usage’.14 (28) Ta zuotian da de dianbao. he yesterday send de telegram ‘He sent the telegram yesterday.’
[M]
(29) *Keoi camjat daa ge dinbou. he yesterday send ge telegram ‘He sent the telegram yesterday.’
[C]
According to Ono (2001), in sentences like (28), as repeated in (30), dianbao ‘telegram’ is the head of a predicate nominal, which is modified by the attributive zuotian da de ‘sent yesterday’.15 Scrutinizing Ono’s (2001) analysis, Long (2007) treats the attributive as a relative clause. Under this analysis, (30) would literally mean ‘He was the telegram that was sent yesterday’. Though this reading is incomprehensible in English, sentences like (31) are, as a matter of fact, fairly natural in colloquial Mandarin if provided with an appropriate context.16 (30) Ta zuotian da de dianbao. he yesterday send de telegram ‘It was yesterday that he sent the telegram.’
[M]
(31) Ta dianbao. he telegram ‘Lit.: He was the telegram.’
Ono’s (2001) analysis faces problems when it is extended to Cantonese. (32) is the counterpart of (31) and can be uttered in colloquial Cantonese with rich contextual
.â•… To derive (29) in Cantonese, Lee and Yiu (1998) propose that a cleft like (i) is its underlying structure, which is not possible in Cantonese because bare nouns cannot be definite in Cantonese (Li 1997; Au Yeung 1997, 2005, Cheng & Sybesma 1999, among others), and subjects in Chinese in general tend to be definite (Chao 1968; Teng 1975; Zhu 1982). A noticeable problem of their analysis is that there is not any known principle to exclude (i) from the underlying structure, particularly given that the underlying structure itself is generally not regarded as an ‘acceptable’ sentence. (i) ??Dinbou hai camjat daa ge. telegram be yesterday send ge ‘It is yesterday that the telegram was sent.’ .â•… My understanding of Ono’s (2001) analysis is based on the discussion in Kimura (2003). .â•… (31) is analyzed as an empty verb sentence, in which the two nominals are the subject and the object of an empty verb (Tang 2001).
On gerundive nominalization in Mandarin and Cantonese 
information. The acceptability of (32), however, does not imply the acceptability of (29), reproduced as (33). (32) Keoi dinbou. he telegram ‘Lit.: He is the telegram.’
[C]
(33) *Keoi camjat daa ge dinbou. he yesterday send ge telegram ‘He sent the telegram yesterday.’
Sugimura (1999) observes an interesting fact about the internal de expressions, namely, that they are anaphoric, and therefore tend to pick out an action in the preceding discourse as their antecedent. In (34), which is taken from Sugimura (1999: 50), dong de shou ‘take action’ refers to dong-le shou ‘have taken action’ in the preceding sentence in the following context.17 (34) Houlai tamen jiu dong-le shou. … Shei xian dong de shou? [M] later they then move-Perf hand who first move de hand ‘Later on, they took action. … Who did it?’
Sugimura’s (1999) claim is subsequently echoed by Kimura (2003) who further proposes that the function of the de-phrase in the internal de expressions is to modify an action by defining its property. Using our previous example in (30), we may say that the internal de expression refers to an event of sending the telegram presupposed in the discourse. The attributive is to restrict the domain of this presupposed event by expressing its temporal property, for example, the time the action took place. Having reviewed the major properties of the internal de expressions, I am now in a position to propose a finer-grained account of their anaphoric nature and the dialectal variation between Mandarin and Cantonese. I assume with Sugimura (1999) and Long (2007) that the de-phrase in the internal de expressions is an attributive relative. The de-phrase does not modify a noun only; instead, it modifies a gerundive nominal that contains an elided verbal head ‘Ø’, from which the anaphoric eventive reading of the internal de expressions is derived, and its NP object, as schematized in (35). (35) Subject Copula [ Relative de Ø np]
In the internal de expressions, the bracketed gerundive nominal is only a complement of the Mandarin copula shi ‘be’, not an object of a verb that is copied from the embedded gerundive nominal. Strictly speaking, there is no formal relation between the copula in the root clause and the elided verbal head in the embedded gerundive nominal,
.â•… Mandarin dong shou ‘take action’ is a VO compound which literally means ‘move hands’.
 Sze-Wing Tang
unlike the configuration in (15), and the contents of the elided verbal head cannot be recovered from the copula. Mandarin de is, however, sufficiently grammaticalized that it could form a gerundive nominal with the verb elided, while Cantonese ge is not grammaticalized enough to make verb ellipsis in the embedded gerundive nominal legitimate and the gerundive nominals are possible in restricted contexts only, i.e. the object position in contrastive constructions. Native speakers tend to associate Mandarin de primarily with adnominal functions, and thus the strong empty verbal head hypothesis for Mandarin. Such a dialectal difference between Mandarin and Cantonese suggests that verb ellipsis in embedded gerundive nominal structures is difficult unless the adnominal grammatical morpheme is fully grammaticalized, as is the case for de in Mandarin but not for ge in Cantonese.
7.â•… Concluding remarks In this paper, I provided an analysis of the (non-)existence of gerundive nominals in Mandarin and Cantonese. It was proposed that such a variation boils down to the availability of gerundive nominalization in these two dialects, which may eventually be attributed to differences in degree of grammaticalization of the adnominal morpheme de in Mandarin and the adnominal morpheme ge in Cantonese. We know that nominalization in Mandarin has been thoroughly and extensively discussed in the field of Chinese linguistics and it is not easy to make a significant breakthrough in such ‘age-old’ issues, but having said that, the parametric approach to the study of Chinese dialectal grammar presented in this paper hopefully gives us some new thoughts on the facts we thought we understood and takes us a considerable way towards explaining the reasons behind them.
References Aoun, Joseph & Li, Yen-hui Audrey. 2003. Essays on the Representational and Derivational Nature of Grammar. Cambridge MA: The MIT Press. Au Yeung, Wai Hoo. 1997. A parametric analysis of Cantonese and Mandarin nominals. In Proceedings of Postgraduate Research Forum on Language and Linguistics 97, 186–197. Hong Kong: City University of Hong Kong. Au Yeung, Wai Hoo. 2005. An Interface Program for Parameterization of Classifiers in Chinese. Ph.D. dissertation, Hong Kong University of Science and Technology. Chao, Yuen-Ren. 1968. A Grammar of Spoken Chinese. Berkeley CA: University of California Press. Cheng, Lisa Lai-Shen & Sybesma, Rint. 1999. Bare and not-so-bare nouns and the structure of NP. Linguistic Inquiry 30: 509–542. Chomsky, Noam. 1970. Remarks on nominalization. In Readings in English Transformational Grammar, Roderick A. Jacobs & Peter S. Rosenbaum (eds), 184–221. Waltham MA: Ginn and Company.
On gerundive nominalization in Mandarin and Cantonese 
DeLancey, Scott. This volume. Finite structures from clausal nominalization in Tibeto-Burman. In Yap, Grunow-Hårsta & Wrona (eds). Grunow-Hårsta, Karen. This volume. Innovation in nominalization in Magar, a Tibeto-Burman language of Nepal. In Yap, Grunow-Hårsta & Wrona (eds). Huang, C.-T. James. 1997. On lexical structure and syntactic projection. In Chinese Languages and Linguistics 3: 45–89. Reprinted in C.-T. James Huang. 2010. Between Syntax and Semantics, 347–376. London: Routledge. Huang, C.-T. James. 2008. Cong ‘Ta de laoshi dang-de hao’ tanqi (On Ta de laoshi dangde hao and related problems). Yuyan Kexue (Linguistic Sciences) 3: 225–241. Kimura, Hideki. 2003. ‘De’ zi ju de jushi yuyi ji ‘de’ zi de gongneng kuozhan (The semantics of the de construction and the extended function of de). Zhongguo Yuwen 4: 303–314. Lee, Thomas Hun-tak & Yiu, Carine. 1998. Final ‘de’ and ‘ge3’: A nominalization analysis for cleft sentences in Mandarin and Cantonese. Paper presented at the LSHK Annual Research Forum, The Hong Kong Polytechnic University. Li, Yen-hui Audrey. 1997. Structures and interpretations of nominal expressions. Ms, University of Southern California. Long, Haiping. 2007. Yiranyi ‘shi … de’ lei jushi de duojiaodu kaocha (A Multi-angle Approach to the Perfective shi … de Construction). Ph.D. dissertation, Central China Normal University. Ono, Hideki. 2001. The mechanism of ‘singularization’ of de (in Japanese). Contemporary Research in Modern Chinese 3: 146–158. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givon, Joan Bybee, John Haiman & Sandra Thompson (eds), 373–394. Amsterdam: John Benjamins. Si, Fuzhen. 2002. Hanyu de biaojuci ‘de’ ji xiangguan de jufa wenti (The Chinese complementizer de and the related syntactic problems). Yuyan Jiaoxue yu Yanjiu (Language Teaching and Research) 2: 35–40. Sio, Joanna Ut-Seong. This volume. The Cantonese ge3. In Yap, Grunow-Hårsta & Wrona (eds). Sugimura, Hirofumi. 1999. ‘De’ zi jiegou, chengzhi yu fenlei (De constructions, anaphoric forms, and classification). In Hanyu Xianzhuang yu Lishi de Yanjiu (Studies of the present situation and the history of Chinese), Lansheng Jiang & Jingyi Hou, (eds), 47–66. Beijing: Chinese Social Sciences Press. Tang, Sze-Wing. 2001. The (non-)existence of gapping in Chinese and its implications for the theory of gapping. Journal of East Asian Linguistics 10: 201–224. Tang, Sze-Wing. 2008. ‘Xingyi cuopei’ yu mingwuhua de canshu fenxi (On ‘syntax-semantics mismatches’ and a parametric analysis of nominalization). Hanyu Xuebao (Chinese Linguistics) 24(4): 72–79. Tang, Sze-Wing. 2009. ‘Ta-de laoshi dang-de hao’ ji Hanyu fangyan de mingwuhua (Ta-de laoshi dang-de hao and nominalization in Chinese dialects). Yuyan Kexue (Linguistic Sciences) 8(3): 239–247. Teng, Shou-hsin. 1975. A Semantic Study of Transitivity Relations in Chinese. Taipei: Student Book Co. Watters, David. 2008. Nominalization in the Kiranti and Central Himalayish languages of Nepal. Linguistics of the Tibeto-Burman Area 31(2): 1–43. Wrona, Janick. This volume. A case of non-derived stand-alone nominalization: Evidence from Japanese. In Yap, Grunow-Hårsta & Wrona (eds). Wu, Xiu-Zhi Zoe. 2004. Grammaticalization and Language Change in Chinese: A Formal View. London: Routledge Curzon.
 Sze-Wing Tang Yap, Foong Ha, Choi, Pik-ling & Cheung, Kam-siu. 2010. Delexicalizing di: How a Chinese noun has evolved into an attitudinal nominalizer. In Formal Evidence in Grammaticalization Research [Typological Studies in Language 94], An Van linden, Jean-Christophe Verstraete & Kristin Davidse (eds), 63–92. Amsterdam: John Benjamins. Yap, Foong Ha & Grunow-Hårsta, Karen. 2010. Non-referential uses of nominalization constructions: Asian perspective. Language and Linguistics Compass 3(1): 1–21. Yap, Foong-ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], Maria Jose Lopez-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins. Yue, Anne O. 1998. Zhi in pre-Qin Chinese. T’oung Pao,LXXXIV: 239–292. Zhu, Dexi. 1978. ‘De’ zi jiegou he panduanju (The de-construction and assertives). Zhongguo Yuwen 1 & 2. Zhu, Dexi. 1982. Yufa Jiangyi (Lectures on grammar). Beijing: Commercial Press.
part ii
Tibeto-Burman languages
Nominalization in Tibeto-Burman languages of the Himalayan area A typological perspective* Carol Genetti
University of California Santa Barbara, Research Centre for Linguistic Typology, La Trobe University Tibeto-Burman languages exhibit an interesting composite of behavior in nominalized structures. In particular, these languages make extensive use of clausal nominalization (found in attributive phrases, complementation, relative clauses, and independent utterances), as well as derivational nominalization. This paper describes the syntactic and functional dimensions of nominalization in five distantly-related Tibeto-Burman languages of the Himalayan area. It demonstrates that the same affixes are used to derive lexical nouns and lexical adjectives and that this bifunctionality is a natural consequence of clausal nominalization. While clausal and derivational nominalizations are distinct structures synchronically, each can give rise to the other. The center of these historical developments is relativization, which instantiates clausal syntax but is ultimately a referential device.
1.â•… Introduction One of the most frequently noted syntactic characteristics of Tibeto-Burman languages is the presence of nominalized clauses which are used in a range of syntactic structures. These include attributive phrases, nominal-complement constructions, relative clauses, verbal- complement clauses, adverbial clauses, and free-standing independent clauses. In some languages, nominalized verbs derive lexical nouns or lexical adjectives and sometimes they function within a larger system of tense-aspect marking.
*This paper was written when I was in residence at the Research Centre for Linguistic Typology. I would like to thank Professors R.M.W. Dixon and Alexandra Aikhenvald for their comments and support. I am also highly indebted to Ellen Bartee, Kristine Hildebrandt, YouJing Lin, and especially Alec Coupe for their data and insightful analyses. I would also like to thank Rebekka Siemens for assistance with various aspects of this study. All errors are my own.
 Carol Genetti
In addition, some Tibeto-Burman languages (such as Lahu; Matisoff 1972) use the same morpheme for nominalization and the marking of genitive relationships, while in others (such as some dialects of Tibetan; DeLancey 1999) nominalizations must or can co-occur with the genitive casemarker in some constructions. This particular congruence of syntactic functions has been referred to as ‘Standard Sino-Tibetan Nominalization’ (Bickel 1999: 271) and has been reported on in a number of languages (e.g. Bickel 1995, 1999; Chalise 2005; DeLancey 1986, 1999, 2002; Ebert 1994; Genetti 1992; Herring 1991; Kölver 1977; Lahaussois 2002, 2003; La Polla 2006; Matisoff 1972; Noonan 1997, 2008; O’Rourke 2000; Regmi 2005; Watters 2008). One of the reasons that nominalization is so pervasive in Tibeto-Burman is that it applies at two different levels of the grammar. Derivational nominalization, which takes as its domain the verb root or predicate, works at the morphological level to derive lexical nouns. It also, perhaps paradoxically, derives lexical adjectives. Clausal nominalization, which takes as its domain the clause or clause combination, works at the syntactic level and allows a grammatical clause to be treated as a noun phrase within a broader syntactic context. The two levels of nominalization differ both in terms of the domains to which they apply (lexical root versus clause), and in terms of the syntactic category of the resultant structures (single word versus noun phrase). These differences are illustrated in Table 1: Table 1.╇ Two types of nominalization
Derivational Clausal
Applies to
Results in
Structure
Verb or predicate Verb Clause
Lexical noun Lexical adjective Noun phrase
[V-nmz]NOUN [V-nmz]ADJ [(NP)…V-nmz]NP
Example (1) illustrates derivational nominalization in Mongsen Ao, one of the languages of this study. Here the verb root kh6lem ‘worship’ undergoes derivational nominalization by the combination of the prefix t6- and the suffix -paÌ•ô, deriving a lexical noun referring to the manner of worship. The noun is used to head a noun phrase; it co-occurs in a possessive phrase with a dependent genitive noun.
(1) Mongsen Ao lexical nominalization
(Coupe 2007: 237)
[muÌ•ŋs6n-6p t6-kh6lem-paÌ•ô]NP Mongsen-gen1 nmz-worship-nmz ‘Mongsen person’s manner of worship’
.â•… Coupe (2007: 237) actually glosses the morpheme -6p anom for “agentive nominalizer”. This morpheme is also used “for marking a genitive relationship between a human possessor and some semantic classes of noun” (2007: 253).
Nominalization in Tibeto-Burman languages of the Himalayan area 
This can be constrasted with the clausal nominalization in example (2), taken from the same language; here -paÌ•ô is functioning to nominalize the entire clause:
(2) Mongsen Ao nominalized clause
(Coupe 2007: 237)
[ts6`hŋi ku hwaŋ-6k6 m6n-pàô i]NP apu-6`p-ùô sun loc roast-sim sit-nmz prox be.good-pres-dec ‘This sitting [and] bathing in the sun is good.’
Here an entire clause combination “sitting and bathing in the sun” is nominalized with the suffix -pàô and functions as the subject of the following intransitive verb. There is no evidence that either the verb m6n-pàô ‘sit’ or the combination of the verbs hwaŋ-6k6 m6n-pàô ‘sit and roast’ function as a noun heading a noun phrase. Instead, the nominalization has scope over the whole unit; note that the initial locative noun phrase remains internal to it. A third type of nominalization, action nominalization, can be distinguished from both the derivational and clausal nominalization types. Like derivational nominalization, action nominalizations derive a noun from a verb which functions as the head a noun phrase. Like clausal nominalization, action nominalization can target a whole clause and take clausal arguments and adjuncts under its scope. Action nominalization differs from clausal nominalization, then, in the status of the head. In action nominalizations, the head is a derived noun; in clausal nominalizations, the head remains a verb. These differences are summarized in Table 2:2 Table 2.╇ Action Nominal Constructions and Clausal Nominalizations Contrasted
Head Internal syntax External syntax
Action Nominal Construction
Clausal Nominalization
derived noun from NP-like to clause-like functions as an NP
verb clause-like functions as an NP
Consider the English example [Her mother’s constant nagging]NP annoyed her to no end. This contains an action nominalization, Her mother’s constant nagging, derived from an underlying clause Her mother constantly nagged. The scope is clearly clausal, incorporating both the subject argument and the modifier. The fact that nagging is a
.â•… Koptjevskaja-Tamm warns that one needs to be careful in distinguishing clause-like action nominal constructions from clausal nominalizations. She cites cases in the literature where some constructions described as clausal nominalizations are, by her criteria, action nominal constructions. She writes: “there is probably no sharp boundary between clausal nominalizations and [action nominal constructions]. Some languages have clausal nominalizations, some have both types, and finally, some do not distinguish between the two types” (1993: 52).
 Carol Genetti
derived noun can be seen in its cooccurence with an adjectival (as opposed to adverbial) modifier. Also, the subject argument of the verb is put into a genitive phrase as a nominal dependent. Cross-linguistically, arguments of the nominalized verb are represented in different ways by different languages (Comrie & Thompson 1985; Koptjevskaja-Tamm 1993, 2005). They may “be signaled in the same way as verbargument relations in finite clauses (‘sentential marking’) or as head-dependent relations in NPs” (Koptjevskaja-Tamm 2005: 255). Action nominalizations are not attested in the five languages of the present study although they are found in other Tibeto-Burman languages of the area. Post (2007: 750) describes such a structure when he states: “Event nominalizations standing as argument noun-phrase heads have their underlying subjects marked in the Genitive”. He provides the following example:
(3) Galo action nominalization
(Post 2007: 751)
[nó-k6` r6¢-nam=66m] nó allîî=b6¢ r6¢-máa-boolo… 2sg-gen live-nmz=acc 2sg well-sbrd live/exist-neg-cond ‘If you don’t live your life properly…’
This structure can be contrasted with the following nominalized clause from Dolakha Newar. Since the verb is the head of a clause, it takes clausal arguments. Here, the notional argument of the nominalized predicate balyet- ‘stalk’ is the grammatical subject and takes ergative case in accordance with the transitivity of the verb:
(4) Dolakha Newar full clausal nominalization
(Genetti 2007: 410)
[dhũ=n ānthi ānthi balyeŋ-an coŋ-gu]NP tiger=erg that.manner that.manner stalk-part stay-nmz ju-en con-a be-part stay-3sg.pst ‘It turns out a tiger was stalking me in that manner.’
Although action nominalizations are unattested in the languages of this sample, an interesting structure of participant nominalization, that shares some properties with action nominalization, will be discussed below (Section 4.2). The current paper contributes to our typology of nominalization by illustrating that languages with clausal nominalizations may also have derivational nominalization and that the two structures are related in interesting ways. It will illustrate that the same derivational affixes can be used to derive both lexical nouns and lexical adjectives, and that this bifunctionality is a natural consequence of clausal nominalization. Finally, it will illustrate that the clausal and derivational types of nominalization, while clearly distinct synchronically, feed each other diachronically in what appears to be a cyclic process. At the center of this dynamic system is relativization, which instantiates clausal syntax but which ultimately is a referential device.
Nominalization in Tibeto-Burman languages of the Himalayan area 
The data for this paper are taken from five Tibeto-Burman languages of the Himalayan region. Listed from east to west, they are Zhuokeji rGyalrong, Dongwang Tibetan, Mongsen Ao, Dolakha Newar, and Manange. Data for the study comes from published and unpublished work by five linguists who are actively involved in fieldwork on these languages: Kristine Hildebrandt for Manange, Carol Genetti for Dolakha Newar, Alec Coupe for Mongsen Ao, Ellen Bartee for Dongwang Tibetan, and You-Jing Lin for Zhuokeji rGyalrong.3 Geographically, the languages for this study range from central Nepal to Northeastern India to Yunnan and Sichuan provinces in China, as shown in Figure 1. Gansu
Qinghai China Tibet Nepal
Sichuan
Manange Kathmandu
rGyalrong Chengdu
Lhasa
Dongwang Tibetan
Thimphu
Dolakha Newar
Bhutan
Mongsen Ao Guizhou
Northeastern India
Yunnan
Bangladesh
Kunming
Dhaka Calcutta
Vietnam
Mandalay
Hanoi Myanmar
Laos
Figure 1.╇ Locations of the five languages of the study
The genetic relationships are more difficult to state, as the sub-grouping of TibetoBurman languages is highly debated and the basic groundwork that would establish familial relationships has not been done. Recent proposals would place these languages into one (Noonan 2008, although see his Footnote 2), three (Matisoff 2003), or four (Bradley 1997) distinct sub-branches. However, everyone would agree that the languages are not closely related and that any common ancestor would have existed thousands of years in the past. The two most closely related languages are Manange .╅ Genetti et al. (2008) discusses the same phenomena but is directed at an audience of specialists in Tibeto-Burman. That paper aims to bring a deeper level of syntactic explicitness to the discussions of Tibeto-Burman nominalization and to illustrate the advantages of such explicitness for discussions of historical development.
 Carol Genetti
and Dongwang Tibetan; however they are widely separated geographically, have had no shared contact, and are significantly different in structural terms. Thus the five languages appear to be broadly representative of the Tibeto-Burman family within the broader Himalayan region.
2.â•… Typological Overview Typologically, the five languages share the following morphosyntactic features: a. b. c. d. e.
Verb-final constituent order, with flexibility in positioning of arguments Dependent clauses before matrix clauses or matrix verbs Distinction between finite and non-finite verbs Head-final or head-medial NP structures Case and number markers as clitics on the noun phrase, not nominal suffixes
Beginning with constituent order, all five languages are quite strongly verb-final. The positioning of the A and O arguments before the verb is pragmatically governed, although AOV can be said to be unmarked. Occasionally arguments are postposed after sentence-final verbs. Dependent clauses in these languages typically precede the matrix, or final, clause. Thus dependent clauses are typically non-final and matrix clauses are typically final. Dependent clause types include adverbial, complement, and converbal clauses. Relative clauses and nominal-complement clauses are also dependent, but are embedded into the noun phrase. Complex sentences are usually formed by one or more non-final clauses with nonfinite morphology followed by a final clause with finite morphology (or, in the case of complement clauses, by the finite verb). All five languages also have sentence-final particles which can occur following the finite verb. Relative clauses, which modify nouns, are most commonly positioned before the nominal head; Mongsen Ao, Dongwang Tibetan, and Zhuokeji rGyalrong also have internally-headed relatives, although in each language these are statistically rare. For this paper, I will use the term “finite” to refer to verbs which have marking typical of that of a sentence-final declarative clause in an unmarked context. Finite verbs in the five languages under discussion are characterized either by distinct morphology that marks tense-aspect-modality or person and number, or by their ability to be directly followed by auxiliary verbs or particles that indicate aspect, evidentiality, or related semantic dimensions. “Non-finite” verbs will refer to those found in sentencemedial contexts; they typically carry suffixes that indicate the nature of the syntactic and/or interpropositional relationship between clauses. These usually include nominalizers, converbs, infinitives, or markers of adverbial clauses, such as purposives. Note that in all five of the languages the nominalizer replaces the tense marker; hence
Nominalization in Tibeto-Burman languages of the Himalayan area 
all nominalizations are non-finite. Following Malchukov (2006), this can be seen as a type of deverbalization, rendering the clause more subordinate. Noun-phrase elements in the five languages include numerals (sometimes with classifiers), demonstratives, quantifiers, adjectives, and relative clauses. Headless noun phrases are possible in each of the languages studied. The relative ordering of nounphrase elements varies across the languages. An important point is that in all five languages, case and number markers are not suffixes, bound to a lexical noun, but are clitics, phonologically bound to the last element of the noun phrase (see Genetti 2007: 103–104; Bartee 2007: 257ff; Coupe 2007: 211; Hildebrandt 2004: 98; 66–67). If the noun phrase is headless, the final element may be a member of any number of lexical classes, e.g. a demonstrative, a numeral, or an adjective. This morphosyntactic characteristic of the language means that we cannot use case or number morphology to argue that a given word is a lexical nominal. It is important to note that one significant way in which the languages differ is in the number of nominalizers, and for languages with more than one, the different functions that are assumed by each. Table 3 lists the nominalizers found in each language: Table 3.╇ Nominalizers in the five languages of this study Language
Forms
Comments
Manange
1. -p
Used in all environments
Dolakha
1. -ku~gu~u 2. -a~e 3. -iuri
The nominalizer -iuri is used with irrealis predicates in limited environments; the other two have complicated distributions related to transitivity (Genetti 2007: 403–407)
Zhuokeji
1. ka~k!2. k63. sa- ~ s!-
The morpheme sa-, s!- functions only in the formation of place and instrument nominalizations as well as locational and temporal relatives; ka~k!- is used in the derivation of patientive nouns, as well as the formation of infinitive, citation and converb forms.
Mongsen
1. t62. t6-…-pàô 3. -pàô 4. -6p. 5. -tſ6n 6. -puÌ• 7. -tſhaŋ
All forms are plurifunctional. Functions will be discussed in this paper where relevant.
Dongwang
1. 〈po〉 2. 〈ba〉 3. 〈mo〉 4. 〈pa〉 5. 〈ma〉 6. 〈sa〉 7. 〈myi〉 8. -mi-n6
Forms in angled brackets represent Written Tibetan orthography; phomemic forms vary considerably. The nominalizers vary in their productivity and range of functions.
 Carol Genetti
Manange is parsimonious with only one nominalizer. Dolakha Newar has three; the irrealis nominalizer is quite restricted in distribution and will not be further discussed. rGyalrong also has three, with the prefix sa- being the most restricted in its distribution. Mongsen Ao has four distinct nominalizing morphemes, two of which can combine to make a fifth nominalizing strategy. Dongwang Tibetan has eight nominalizers, however, five have quite limited productivity.
3.â•… Clausal nominalization The essence of clausal nominalization is [clausen]NP: the placement of a clause, or a combination of clauses (represented by the superscript n), into a broader syntactic structure where it functions as a noun phrase. These broader syntactic structures include clauses where the nominalized element is an argument of a verb (verbal complement clauses; see Section 3.1), sentences where the nominalized clause is an adverbial element (adverbial and medial or ‘converbal’ clauses; Section 3.2), and noun phrases where the nominalized clause modifies a noun, in a structural configuration similar to the coding of possession (nominal complement and relative clauses; Section 3.3). In each of these syntactic environments simple noun phrases (lacking nominalized clauses) are also found. There is one syntactic environment, “non-embedded nominalization” (Matisoff 1972; also variously referred to as “main-clause” or “stand-alone” nominalization) where it is unclear whether the nominalized constituents are, indeed, noun phrases. An example from Dolakha Newar is given in (5):
(5) Dolakha Newar non-embedded nominalization
[Mansu Lal=na syeŋ-gu] ka Mansu Lal=erg teach-nmz assertion ‘Mansu Lal taught us (that song)’
Within the broader context of the discourse in which this sentence was uttered, it is clear that the function of this sentence is to emphasize who it was who taught the speaker the song, which is the known information which formed the background for this utterance. The sentence could be translated as a cleft in English (‘It was Mansu Lal that taught us that song’), however, this implies an equational structure that is lacking in the Dolakha Newar; the ergative marking on the subject clearly indicates that it is an argument of the transitive verb syen- ‘teach’. It is possible that this construction is a reduction of a complementation structure where the copula functions as an intransitive verb meaning ‘it is the case that’, and the nominalized clause is syntactically the intransitive subject (cf. Watters 2008). In such a case this “non-embedded” nominalization would still be analyzable as a noun phrase. The lack of the copula would need
Nominalization in Tibeto-Burman languages of the Himalayan area 
to be accounted for in such a case and the analysis would necessarily be abstract.4 We leave such examples for future, more detailed studies, and will here only emphasize that the nominalized element is indeed a clause or clause combination; the only way in which it differs from a non-nominalized clause is in the lack of the finite suffixes, which are necessarily replaced by the nominalizer. This structure is further discussed in Section 3.4. The fact that the domain of “clausal” nominalizations can include clause combinations is illustrated in the following two examples. The first, taken from Mongsen Ao, illustrates a complement clause that contains within it another complement clause: (6) Mongsen Ao double-complementation (Coupe 2007: 238) [a-nu-tſaŋ n6 [waÌ•zaÌ•ô-saÌ•ô t6-s6´n-ŭô ]NP.O t6 nrl-child-male agt bird-meat nmz-be.sour-dec thus sa-paÌ•ô]NP.VLESS.CL.TOP sitak say-nmz correct ‘The boy’s saying that the bird-meat is sour is correct.’
Here the complex sentence ‘the boy said that the bird-meat is sour’, which contains a quotative complement ‘the bird-meat is sour’ is nominalized with the suffix -paÌ•ô and functions as the topic of the verbless clause ‘it is correct’. Thus we find a quotative object complement structurally embedded into another complement which is a verbless clause topic. Another example of a nominalized complement with internal complexity is the following from Dolakha Newar:
(7) Dolakha Newar clause chain inside a complement
(Genetti 2007: 411–412)
… [kho dupān swat.t.a morlu-en ye-ene dudu … river inside expr bathe-part come-part milk toŋ-a]NP-O khon-ju drink-nmz see-3sg.pst
‘(They) saw (the calf) …bathe inside the river, come out, and drink milk.’
Here a sequence of three clauses is combined into a clause chain with the participial construction. This is then embedded as the grammatical object complement of khon- ‘see’. Thus the basic structure [clause]NP may be expanded into a multi-clausal constituent, represented as [clausen]NP. This structure can be found in any syntactic environment where nominalized clauses are used; thus clausal nominalization is a
.â•… Note that the final particle ka is a particle and not a verb, so it cannot be taken to be a complement-taking predicate. It typically follows a single finite clause; it is used to mark facts that the speaker is asserting to be true. For more information, see Genetti (2007: 255–256).
 Carol Genetti
highly flexible and powerful syntactic mechanism that allows for an infinite number of embedded structures. In some cases there is evidence that grammaticalization and syntactic reanalysis has moved a nominalized structure beyond that of a clause functioning as a noun phrase. For example, it is clear that in some cases nominalization is clearly deriving lexical adjectives, not lexical nouns, and that reanalysis has indeed occurred (Section 4.1). Another example is the development of true converbal clauses, with converbal suffixes, from old nominalized clauses plus casemarkers (Section 3.2; Genetti 1986, 1991). We also find nominalized structures drafted into the system of tense-aspect marking, typically co-occurring with a copula to signal imperfective aspect (as in Zhuokeji rGyalrong) or even functioning independently as a future marker (Manange). Such developments are not surprising given the pervasiveness of nominalized structures and the natural tendencies for grammaticalization and reanalysis. The following discussion will outline briefly the primary types of clausal nominalization attested in the languages of this sample: complement clauses, converbal clauses, and relative and nominal-complement clauses. The paper will then move on to derivational structures, and conclude with a discussion of the relationship between them, highlighting the role of the relative clause as the fulcrum around which these structures develop.
3.1â•… Complementation All five of the languages of this study use nominalized clauses in some constructions where they function as noun-phrase complements of a matrix verb. The languages differ in which complement-taking predicates select nominalized complements and in the argument role of the complement. In some cases, the nominalized clause functions as a complement in toto, with all of the arguments, adverbials, or other elements being realized, each with the same morphosyntactic trappings that they would have in an independent clause. This is true of the following example, in which a nominalized complement clause functions as the argument of the transitive verb thuÌ…³-⁵³ ‘see’:
(8) Dongwang Tibetan complement of thũ353 ‘see’
[khui55 kh655 dõ353-sa]NP.O ŋe13 thũ353 soÌ… 3sg.erg 3sg.abs hit-nmz 1sg.erg see ego ‘I saw him hit him.’
Here, we can see that both of the arguments of the nominalized complement clause retain the casemarking that they would have in an independent finite clause. Thus, the complement clause is being placed directly into the NP slot without any adjustments to its structure outside of the nominalizing suffix. The structure is simply [NP NP V-nmz]NP, and it serves as the syntactic object of the matrix verb.
Nominalization in Tibeto-Burman languages of the Himalayan area 
Not surprisingly, some complement clauses necessarily have adjustments to their internal syntactic structure when they are embedded as arguments of a matrix verb. This occurs when there is co-reference between the matrix subject and the complement subject. For example, in complement constructions involving the Manange desiderative predicate saŋ52 kh%22 ‘want’, the subject of the matrix verb and the subject of the complement verb are necessarily co-referential, and the subject of the complement verb is obligatorily unexpressed:
(9) Manange complement of saŋ52 kh%22 ‘want’
(Hildebrandt 2004:113)5
ŋ%22 [ya52 sh%22 ts%-p%22]NP.STIM saŋ52 kh%22 mo22 1sg [yak flesh eat-nmz] desire come cop ‘I want to eat yak meat.’
The fact that it is the complement subject that is unexpressed can be seen in the casemarking of the surface argument, the first-person singular pronoun ŋ%22; it is absolutive, in accordance with the argument structure of the matrix predicate, which requires an absolutive subject. If the pronoun were an argument of the complement clause, it would necessarily be in ergative case (carrying the clitic =tse), as the verb ts%-p%22 ‘eat’ of the complement clause is transitive. Thus we can see that the nominalized complement clause undergoes an obligatory grammatical adjustment, omitting the subject argument when it is coreferential with the subject of the complement clause. Another case where the structure of the nominalized clause may be altered in the process of embedding as an argument of matrix verb is the case of raising. This happens in Dolakha Newar, as illustrated in the following example: (10) Dolakha Newar complementation with raising
(Genetti 2007: 409)
cilā=n ninpatti āmta [kho-en coŋ-gu]NP.O khon-ai goat=erg daily 3sg.dat cry-part stay-nmz see-3sg.pres ‘The goat sees her crying everyday.’
Here the referent of ‘her’ plays a double role; she is both the one seen by the goat and the one who cries. Speakers of the language can choose whether to represent such arguments as the subject of the complement clause or as the object of the matrix clause, with their choice being reflected in the casemarking (Genetti 2007: 311–312). In this example, the dative pronoun āmta reflects the object orientation; it could be felicitously replaced with the ergative pronoun āmun, marking it as subject. The obligatory non-expression of the complement-clause subject in Manange and the variability in the casemarking of the complement-clause subject in Dolakha Newar
.â•… Note that this is a noun-verb compound meaning ‘want’ which takes an absolutive argument indicating the experiencer and a second noun phrase indicating the stimulus.
 Carol Genetti
both illustrate that the structure of a nominalized complement clause may undergo adjustments in the process of embedding, and may not surface with all elements having the same forms as in independent clauses. Thus while [clause]NP is the basic schema underlying clausal nominalization, it is important to keep in mind that the details of the internal structure of the clause may vary based on the external context into which it is embedded.
3.2â•… Adverbial and medial clauses Nominalized clauses are sometimes used as syntactically dependent, adverbial or medial clauses (sometimes referred to as “converbal clauses” (Genetti 2005; Coupe 2007)). In some of the languages, the nominalized clause functions as an adverbial or medial clause in the absence of any additional morphology. This is found with the causal adverbial clause in Manange, the sequential converbal clause in Mongsen Ao, and the medial (general converbal) clause in Zhuokeji rGyalrong: (11) Manange nominalized causal adverbial clause
(Hildebrandt 2004: 83)
[ŋ%=tse22 mwi42 phr%42 ky%=ri52 pim-p%22]ADV.CL nese22 ky%52 1=erg money 100 2sg=loc give-nmz tomorrow 2sg kola52 kyu-p%52 clothes buy-nmz ‘Because I gave you 100 rupees, you will buy a dress tomorrow.’ (12) Mongsen Ao sequential converb clause with -6p suffix
(Coupe 2007: 423)
[t6`-tſhaÌ•-6p tſhuwa-6p]CNV.CL a-ki thuÌ•ŋ thus-do-seq emerge-seq nrl-house reach.pst ‘And then, [he] came out [of the jungle] and reached home.’ (13) Zhuokeji rGyalrong clause chain with nominalized verbs6 … [ſ6kphû spo-j pt‰er6 k6-gô]CL.CHAIN pt‰er6 … tree bottom-loc then nmz-go.eastwards then [w6-spo-s stô=m6naŋor6 ſamdu k6-l!¥t]CL.CHAIN pt‰e 3sg.pos-bottom-abl upward=top gun nmz-put1 then w6t6psôk ſo t!-rw!k k6-pa na-ŋôs that.way usually n-hunting nmz-do ipfv.pst-cop2 ‘He used to go under the tree then shoot upward from the bottom; in that way he usually hunted.’
.â•… The Zhuokeji rGyalrong tone system shows a binary contrast between falling tone and zero. Lexically, a word is either prelinked with a falling tone (to the final syllable in most cases) or toneless. In this paper, toneless words in Zhuokeji rGyalrong are left unmarked, while falling tone is marked with a circumflex (¥) on the falling-toned syllable.
Nominalization in Tibeto-Burman languages of the Himalayan area 
In other cases, the nominalized clause formally carries a nominal casemarker, which specifies the interpropositional relationship held between the adverbial clause and the following clause. This can be seen in the Dolakha Newar causative construction, which requires the benefactive casemarker lāgin, which cliticizes to the nominalized clause and signals a causal relationship. Note that both a benefactive argument and a causal clause denote the reason for the action of a predicate. (14) Dolakha Newar causal adverbial clause [dukha bi-e-lāgin]ADV.CL āpen=ri jaŋgal oŋ-an con-hin hã trouble give-nmz-benefactive 3p=ind jungle go-part stay-3ppst evid ‘Because she gave them trouble, they went and stayed in the jungle.’
Such constructions are quite usual in Himalayan languages and casemarkers frequently grammaticalize into adverbial clause morphology (Genetti 1986, 1991; Horie 1998 offers similar observations for Korean and Japanese; Aikhenvald 2008 provides a typological study of this phenomenon). This process is seen in the derivation of converbal affixes in Mongsen Ao. As noted by Coupe (2007: 435–446 passim), the Mongsen general nominalizer â•‚pàô is etymologically part of three converbal suffixes: the causal -pàn6 (from the nominalizer plus the instrumental casemarker n6); the conditional -pàla (from the nominalizer plus the topic marker la); and the concessive â•‚pàkukà (from the nominalizer plus the locative casemarker ku plus the final syllable of anukaô ‘also, even, yet’). An example follows: (15) Mongsen Ao causal converbal clause
(Coupe 2007: 11)
[pa t6-l6m ku ts6`-pàn6]CNV.CL t6-l6m tſu apak-tſuk 3sg rl-head loc peck-causal rl-head dist flat-pfv.pst ‘Because [the other birds] pecked on her head, [Owl’s] head became flat.’
The basic morphosyntactic structure which allows for the affixation of clauses by nominal case or topic markers is [clause]NP, the simple structure underlying clausal nominalization. Since casemarkers (and topic markers) are NP-level enclitics, they directly follow the noun phrase. If the noun phrase consists of a clause, then the clitic will be bound to the final element, which in these languages is consistently the verb in dependent clauses, allowing for affixation and grammaticalization. In cases where a casemarker is absent, as in the Manange causal clause exemplified in (9) above, the same [clause]NP structure is found, but without the clitic. In such cases, the clausal noun phrase is placed into the sentence structure directly, in the same position as an unmarked adverbial noun phrase that isn’t clausal; i.e. one with a head noun, as in the following Dolakha Newar example: (16) Dolakha Newar adverbial noun phrase without casemarker [ām khunu]NP bud.hi=ta cheÌ…=ku almal ju-en that day old.woman=dat house=loc busy be-part ‘That day, the old woman was busy in the house…’
 Carol Genetti
Thus again we see that nominalized clauses can take the same positions as noun phrases. In adverbial function they can either function independently or can be casemarked, just like other noun phrases.
3.3â•… Relative clauses and nominal complement clauses One of the most important features of nominalized clauses in Tibeto-Burman languages is their ability to be embedded into noun phrases as modifiers of nouns. As with the other environments for clausal nominalization, we find that clausal nominaliÂ� zations take the same syntactic positions as simplex noun phrases: they may occur as dependents of noun phrases, with the relationship either explicitly marked via a possessive construction or unmarked. Let’s begin with possessive constructions. In the five languages of this study, nominal possession by a referent represented by a (non-pronominal) noun phrase is accomplished by three types of structures. The first, and most common, is a genitive construction where the possessive relationship is marked on the dependent. This is exemplified in the Dongwang Tibetan example in (17). The noun phrase has the formal structure: [ [np]=gen n ]np. (17) Dongwang Tibetan possessive construction with dependent marked by genitive mi11 mu53=ji tsho55wa53 people=gen life ‘people’s lives’
The second type of possessive construction is found in Zhuokeji rGyalrong, which has a stronger head-marking typology. In this language, the possessor is cross-referenced by a prefix on the head (possessed) noun, as exemplified in (18). The formal structure is: [ [np] poss-n ]np. (18) Zhuokeji rGyalrong possessive NP with head marked by possessive prefix ta-pu w!-p!¥ npr-child 3sg.pos-father ‘the child’s father’
The third type of possession is found in some languages in restricted semantic contexts. In that structure the formal marking of the dependent by a genitive or possessive affix is not found. Instead a noun phrase is simply positioned inside of another noun phrase in a modifying relationship, with the simple structure [[NP] N]NP (although in this example the N is expanded to two conjoined nouns). This is exemplified in (19): (19) Mongsen Ao complex noun phrase with unmarked possessive np [[lázáti-la]NP t6-paô kh6 t6-ja n6t n6]NP unmarried. woman-f rl-father conj rl-mother two agt ‘The unmarried woman’s father and mother’
Nominalization in Tibeto-Burman languages of the Himalayan area 
All three of the noun-phrase structures are found with relative clauses, which strongly suggests that relative clauses are, at their core, [clause]NP structures positioned within noun phrases as dependents of nouns. Examples (20)–(22) illustrate each of these three structures: (20) Dongwang relative clause with genitive marking on relative [[shihui s6353-sa]REL =ji dong¹³ ]NP limestone burn-nmz =gen hole ‘The hole where limestone is burned’ (21) Zhuokeji rGyalrong relative clause with possessive marking on head [[khabõi k6-pa]REL w6-mi]NP song nmz-do 3sg.pos-woman7 ‘The woman who sings’ (22) Dolakha Newar noun phrase with quantifier and relative clause [dokhunuŋ [āmu cilā=n bi-e]REL cij=pen]NP all that goat=erg give-nmz2 thing=pl ‘All the things that the goat gave.’
Note that in Dongwang Tibetan, the genitive marking of relative clauses is optional. Speakers vary in their frequency of use of the genitive marker and in their opinions about which relatives are better with or without it. No functional distinction in the presence or absence of the marker has been found. While the parallelism between the syntax of possessive phrases and the syntax of relative clauses clearly argues that relative clauses are clauses holding noun-phrase positions within a larger noun phrase, it is important to note that relative clause structures have additional complexities. The criterial feature of a relative clause is that one argument of the relative clause is coreferential with the head noun. In pre-head and post-head relative clauses in these languages, the co-referential argument in the relative clause is necessarily unexpressed, leaving a “gap” in the relative clause.8 There is then an obligatory control relationship that holds between the head noun and that gap.
.╅ Note that, unlike many other Tibeto-Burman languages, the noun mi in Zhuokeji rGyalrong indicates a female, woman or girl. The generic noun for person is rmi. .╅ Internally-headed relative clauses are found in low frequency in Mongsen Ao, Dongwang Tibetan, and Zhuokeji rGyalrong. For such constructions, the unexpressed element is external to the relative clause; see Genetti et al. (2008).
 Carol Genetti
Thus the relative clause constructions in (20)–(22) may be structurally represented as in (23)–(25) respectively: (23) Dongwang:
[ [(NP) ... Ø V-]REL= N ]NP control
(24) Zhuokeji:
[ [(NP) ... Ø V-]REL=–N ]NP control
(25) Dolakha Newar:
[ [(NP) ... Ø V-]REL
N ]NP
control
Nominal complement clauses differ from relative clauses in that there is no coreferentiality between the head noun and an argument of the relative clause. Typically the head noun is abstract, and refers to something that is spoken or understood, as in (26): (26) Dolakha Newar nominal complement clause [ām jāl ta-en ta-e]NOM.COMP khã]NP that net put-part put-nmz talk ‘The talk of that net having been put (laid)’
Nominal complement clauses are not attested in Dongwang, so the presence of genitive marking on such structures is not found. In Dolakha Newar, nominal complement clauses, like relative clauses, are used in a noun phrase without any formal marking of dependency, as in (26). Since nominal complement clauses lack the coreferentiality with the head noun found in relative clauses, their syntactic structure is simply [[clause]NP N]NP; a clause embedded as a noun phrase as dependent to a noun.
3.4â•… Non-embedded nominalizations One of the most notable structures discussed in the literature on Tibeto-Burman nominalization is the so-called “non-embedded nominalization”: a nominalized clause which is not syntactically or functionally an element of any higher “matrix” clause, but which functions as a complete and independent utterance (Bickel 1999; Hargreaves 1991: 35–40; Matisoff 1972; Noonan 1997). Matisoff (1972) was the first person to describe this structure as integral to Tibeto-Burman nominalization
Nominalization in Tibeto-Burman languages of the Himalayan area 
patterns. He succinctly summarized the functional and structural properties of such structures as follows: The verbal event is being objectified, reified, viewed as an independent fact, endowed with a reality like that inhering in physical objects – in short, nominalized. It is standing on its own, and is not a constituent of any sentence higher than the one to which it belongs itself. (1972: 246)
In the languages of this study, non-embedded nominalizations are attested in two of the five languages: Dolakha Newar and Manange.9 In languages in which they are attested, there are clearly visible differences in the frequency of such constructions. In Dolakha Newar they are relatively rare, whereas they occur frequently in Manange. Beginning with Dolakha Newar, Genetti (2007: 400–403) describes these as focus constructions, which profile new information against a backdrop of presupposed information. An example is given in (27): (27)
Dolakha Newar non-embedded nominalization phā«si bi-e le hanging give-nmz prt ‘It was by hanging (that they were executed).’
This sentence was produced during a conversation regarding the execution of the assassins of Indira Gandhi. The fact that the men had been executed had already been established in the previous discourse. The new information is the method of execution, which is focused by the use of this construction. The speaker could have chosen to use a regular finite clause to convey this information, i.e. phā«si bir-ju [hanging give-3sg.pst] ‘(they) hung (them)’, but this would have lacked the subtle invocation of the previous context performed by the nominalization. Given that this construction is used with focusing functions, it is not surprising that it occurs readily in questions, as in (28): (28) pus=na phoŋ-a rā? Pus=abl ask.for-nmz q ‘Was it in (the month of) Pus that he asked?’
Hargreaves (2005:19) describes the use of non-embedded nominalizations in interrogatives in Kathmandu Newar as “being less ‘interrogatory’ (hence polite or merely phatic) than questions in finite verb forms”.
.╅ The available rGyalrong data is all in the narrative genre. In this data simple nominalized clauses appear to be chained clauses, as discussed in section3.2. The extent to which such structures occur in conversation independent of a chaining construction is unknown at this time. They are not attested in either Mongsen Ao or Dongwang Tibetan.
 Carol Genetti
Note that while examples (27) and (28) translate felicitously into English with cleft constructions, these are not structural clefts but cohesive clauses that differ from independent finite clauses only in their non-finite, nominalized morphology. This can be seen in example (5) above, repeated here as (29) for convenience. Note the ergative marking on the agentive noun marks it as being the grammatical subject of the nominalized verb, a clear indicator of the integrated nature of the clausal elements. (29) [Mansu Lal=na syeŋ-gu] ka Mansu Lal=erg teach-nmz assertion ‘Mansu Lal taught us (that song)’
In all three of these examples, the clause is syntactically independent. There is no higher structure in the surrounding context that they are related to syntactically. Each one also is followed by a sentence-final particle, indicating that they are syntactically complete. And each one constituted a complete turn in the conversations in which they were produced. In Manange, non-embedded nominalizations are more common and differ functionally from those in Dolakha Newar. If they occur in the absence of evidential morphology, nominalized clauses are interpreted as being in future tense: (30) Manange non-embedded nominalization khi2
(Hildebrandt 2004: 83)
nu-p%42
3sg sleep-nmz ‘He will sleep’
With evidential morphology, the future interpretation is possible but not required, as shown by the following example, where the evidential particle confirms that the truth of the proposition has been checked by the speaker: (31) Manange non-embedded nominalization
(Hildebrandt 2004: 83)
ya52 k%tti55 mo-p%22 ko yak many cop-nmz evid ‘There were many yaks’
The fact that non-embedded nominalizations in Manange are interpreted as future tense in the absence of evidential morphology suggests a path of development whereby such clauses are interpreted as finite and the nominalizing suffixes are reinterpreted as markers of tense or aspect. A pattern of development from nominalizer to marker of imperfective has been noted for a number of languages of the Himalayas (Watters 2008; see also Delancey, this volume). If nominalized clauses have the syntactic structure [clause]NPâ•›, the question arises whether the non-embedded use of such structures are still structurally noun phrases, or whether they have been syntactically reanalyzed as simply [clause]. It is probably
Nominalization in Tibeto-Burman languages of the Himalayan area 
not possible to make a general statement about this across multiple languages with different functional interpretations of such structures. However, the fact that in some languages the nominalizing morphology has come to be strongly associated with particular tense-aspect values does suggest that such a reinterpretation has occurred. Thus there is a grammatical path of development from nominalizer to marker of tense-aspect, and correspondingly from [clause]NP to [clause].
3.5â•… Summary of clausal nominalization structures We have seen that the essence of clausal nominalization is [clause]NP. Such noun phrases can be used: (1) as arguments of verbs, in which case they are complement clauses; (2) as adverbial elements in a sentence, in which case they are adverbial or medial clauses; (3) as modifiers inside a noun phrase, in which case they are nominal-complement or relative clauses; (4) as independent utterances. In the latter case, it appears that [clause]NP could be reinterpreted as simply [clause]. Additional modifications of the basic clausal structures occur in some types of complements and obligatorily in relative clauses, which require the gap.
4.â•… Derivational nominalization We now turn to derivational nominalization, the derivation of lexical adjectives and lexical nouns. We will demonstrate that derivational nominalizations appear to arise from, but can also give rise to, relative clauses.
4.1â•… The derivation of lexical adjectives In all five languages of the study, verbs with adjectival meanings can be nominalized to denote attributes of nouns. Genetti and Hildebrandt (2004) and Genetti (2007: 206–207) have demonstrated that nominalized adjectival verbs in Manange and Dolakha Newar respectively have unique morphosyntactic behavior which is clearly distinct from that of nouns, partially shared with that of verbs, and partially shared with that of non-derived adjectives. In addition, in Dongwang Tibetan (Bartee 2007: 150), some adjectives derive historically from nominalized verbs, although the connection to the old lexical verb is lost in the modern language. This piece of evidence, with others, suggests that nominalized verbs that function as adjectives in these languages can be lexicalized as such, in other words, that nominalization can derive adjectives. It is the unique morphosyntactic and semantic properties of this class of terms that make this possible. Adjectival verbs are syntactically intransitive. The single argument represents the entity that is asserted to be in (or be entering into) the state denoted by the verb.
 Carol Genetti
Adjectival verbs can be used attributively through relative-clause structures. In the majority of cases, the adjectival verb will be the only element in the relative clause. The result is a one-word phrase which expresses an attribute of a following noun, a structure easily reinterpretable as a lexical adjective. This structure arises naturally from the intransitivity of the verbs and the requirement that the relative argument coreferential with the head noun be unexpressed. We can see this in the following Dolakha Newar example: (32) Dolakha Newar relative clause with adjectival verb heÌ…gā-u sona red-nmz flower ‘red flower; ‘flower that is red’
This example has two structural interpretations, represented by the two translations. It can both mean ‘flower that is red’, representing a head noun modified by a relative clause, and ‘red flower’, representing a head noun modified by an adjective. Both structures have a head noun, and a dependent modifier realized as a nominalized verb. Genetti (2007:206–207) demonstrates that heÌ…gā-u and other nominalized adjectival verbs have unique morphosyntactic behavior when compared to non-adjectival verbs and so constitute a unique lexical class (or sub-class). However, the connection with the verb is not entirely lost, as it is still possible to inflect the verb with other verbal affixes. For example, heÌ…gār-a, with the third-person singular past suffix, can mean ‘it reddened’, e.g. sona hẽgar-a [flower red-3sg.pst] ‘the flower reddened; became red’. However, the more common way of expressing this meaning is to use the nominalized form of the verb heÌ…gā-u as a derived lexical adjective. In this construction it functions as the copula complement and has the same syntactic status as a non-derived adjective. It is then obligatory followed by the copula jur- ‘become’, e.g. sona hẽga-u jur-a [flower red-nmz become-3sg.pst] ‘the flower became red’. In the other four languages of the sample, lexical adjectives and relative clauses are more distinct syntactically; nevertheless, each language derives lexical adjectives from verbs with nominalizing affixes. In Manange, there are several syntactic features which distinguish attributive adjectives from relative clauses. One of these is positioning with respect to the head noun. Relative clauses obligatorily precede the head noun. This is true even if the relative clause contains a single descriptive verb: (33) Manange pre-verbal relative with non-adjectival verb (Hildebrandt 2004: 60) ŋwo-p%52 sh%22 fry-nmz meat ‘fried meat’
Nominalization in Tibeto-Burman languages of the Himalayan area 
Adjectives, whether derived from verbs or un-derived “simple” adjectives, obligatorily follow the noun: (34) Manange post-nominal simple adjective khye42
t%rkya=ri22
ŋ%22
por52
y%22
(Genetti & Hildebrandt 2004: 54)
mo22
road white=loc 1sg take go cop ‘I take (the prayer scarf) on the white road (to heaven).’ (35) Manange post-nominal verb-like adjective
(Hildebrandt 2004: 60)
kyu31 thy%-p%=ri22 thẽ22 t%-tsi22 water big-nmz=loc throw become-perf ‘(The ashes) were thrown in big water (like a river).’
This fact, together with others (Genetti & Hildebrandt 2004), illustrates the distinct syntactic behavior of relative clauses and derived adjectives. Derived adjectives behave syntactically like simple adjectives in Manange and not like verbs. In Mongsen Ao, adjectives are derived from verbs with the nominalizing prefix t6â•‚. Examples are given in (36); an example of a derived adjective in predicative function is given in (37): (36) Mongsen attributive adjectives derived with t6 t6-a-miô t(6)-apu (nrl-person nmz-be.good) ‘good man’ a-s6 t6-m6p6m (nrl-shawl nmz-be.red) ‘red shawl’ (37) Mongsen predicative adjective derived with t6 a-tſa tſu t6-tſha-ù nrl-cooked.rice dist nmz-be.hot-dec ‘The rice is hot.’
As with Manange, there is clear evidence that in Mongsen it is possible to differentiate between a noun modified by a derived adjective and a noun modified by a relative clause formed with a descriptive verb. Coupe (2005: 232–233) provides the following four arguments: (1) The two structures have different morphology: to derive adjectives from descriptive verbs, the t6- prefix is used, as illustrated in (36); to form a relative clause with a descriptive verb, speakers use the general nominalizing suffix, -pàô. (2) Relative clauses may either precede or follow the head noun: (38) Mongsen relative clauses with descriptive verbs a.
maŋu niŋ-pàô tſu banana be.ripe-nmz dist ‘that ripe banana’/‘the banana that is/was ripe’
b. niŋ-pàô maŋu tſu be.ripe-nmz banana dist ‘that ripe banana’/‘the banana that is/was ripe’
(Coupe 2007: 208)
 Carol Genetti
By contrast, deverbal adjectives may only follow the head (as in (36)). (3) The relative clause may contribute a restrictive meaning to the head noun; however deverbal adjectives are not restrictive, only descriptive (Coupe 2007: 209). (4) Only deverbal adjectives can function as complements of verbless clauses; an example of this structure is given in (39): (39) Mongsen deverbal adjective as verbless clause complement (Coupe 2007: 209) [maŋu i]VCT [t6-hniŋ]VCC banana prox nzp-be.ripe ‘This banana is ripe.’
The analogous sentence with a relativized verb (t6)-hniŋ-pàô is ungrammatical (Coupe 2007: 209). Thus in Mongsen the two constructions have different morphology and different syntactic constraints. In Dongwang Tibetan there is a lexical class of adjectives and a distinct class of “descriptive verbs”. About one-quarter of the lexical adjectives can be historically traced to old verbs suffixed by old nominalizers. While some of these adjectives are based on recognizable synchronic verb roots (e.g. kã 53 ‘to dry’), many are not (e.g. di11m655 ‘peaceful; smooth’). Descriptive verbs in Dongwang may be productively nominalized to derive adjectival forms. As with Manange and Mongsen, these are put in the post-nominal position common to lexical adjectives: (40) Dongwang nominalized descriptive verb as attributive adjective (Bartee 2007: 155) t655 kue¹³ ja¹³-wa ta¹³ khuaâ•›53=jæ ‰æ «¹³ t.o53 re¹³ s that clothes good-nmz def 3sg.dat=dat wear caus cop.disj qtv ‘Put those good clothes on him.’
As with Manange and Mongsen, it can be argued that the nominalizers are truly deriving lexical adjectives; for example, derived adjectives may be modified by adverbs. Relative clauses, which are positioned pre-verbally, cannot be modified by adverbs. In Zhuokeji rGyalrong, adjectives are derived from verbs with one of the nominalizing prefixes used to derive relative clauses. However, nouns modified by adjectives differ from those modified by relative clauses both in the positioning of the modifier (post-nominal for adjectives and pre-nominal for relatives) and in the absence of possessive marking on the head noun. Thus (41) can be compared to (21) above: (41) Zhuokeji rGyalrong attributive adjective (Lin, personal communication) rg6nb!ˆ k6-wagze ti j6-ka-mdu temple nmz-be.splendid top:loc pfv-nmz-arrive ‘When (he) arrived at a splendid temple…’
Once again we see that relative-clause constructions and the derivation of adjectives utilize the same morphology but are syntactically distinct.
Nominalization in Tibeto-Burman languages of the Himalayan area 
In summary, all five languages of the study use nominalizers to derive adjectives, defined as a set of terms which denote property concepts, can be used attributively and predicatively, and which have distinct behavior from other attributive verbs, specifically from relative clauses. The Dolakha Newar data, which shows strong structural and functional parallels between relative clauses and derived adjectives, suggests that relative-clause structures place nominalized adjectival verbs in one-word attributive phrases expressing property concepts of the accompanying noun, and that this then opens the door to their reanalysis and distinct syntactic treatment. Thus relative-clause structures can give rise to derived lexical adjectives.
4.2â•… Derivation of nouns The five languages of this study differ noticeably in the extent to which nominalization is used to derive lexical nouns which can function as heads of noun phrases. In Manange, derivational nominalization is confined to only a few highly lexicalized examples, which seem to be unanalyzable for many speakers. An example is kyo-p%1 [mourn-nmz] ‘lungs’ (place from which mourning/melancholy comes). Mongsen Ao has multiple nominalizers that are involved in the derivation of nouns. In Mongsen Ao, abstract nouns are derived by the prefix t6-; however, the process is relatively unproductive. Some examples are provided in (42): (42) Mongsen deverbal nouns with t6- prefix
s6 tſàsi th6m ts6pha m6pak
‘die’ ‘be distressed’ ‘finish’ ‘fear’ ‘lie’
t6-s6 tſ-tſasi t6-th6m t6-ts6pha t6-m6pak
(Coupe 2007: 256)
‘death’ ‘anger, distress’ ‘end’ ‘fear’ ‘lie’
A number of nominalizers are involved in participant nominalizations; they denote an entity (including place and instrument) involved in the action or state of the verb. Examples of the four nominalizers are given in (43): (43) Mongsen participant nominalizations Instrument: n6`mphaŋ ‘covered’ t6-n6mpháŋ-pàô Locative: jip ‘sleep’ jip-tſ6n Agentive: ni-si ‘lead’ t6-ni-si-6p Patientive: tſàô ‘eat’ t6-m6`-tſàô-i
(Coupe 2007: 265) ‘cover’ (n.) ‘bed’ ‘leader’ ‘that which is not to be eaten’
There are a number of arguments which show that the words derived by these four nominalizers are indeed lexical nouns (for full argumentation, see Coupe 2007: 256–259; 263–265): they function as heads of noun phrases; they cannot co-occur with another (head) noun, they cannot co-occur with arguments or adjuncts of the verb (i.e. the
 Carol Genetti
nominalizer has scope over the verb root only and cannot also include object NPs); they function referentially, not as descriptive modifiers; and they are often highly lexicalized, e.g. jip-tſ6n ‘bed’. However, it is possible, although rare, in Mongsen for both locative nominalizers and agentive nominalizers to be used in the formation of relative clauses. The normal nominalizer that is found in relative clauses is the general nominalizer -paÌ•ô. Coupe has found, in both natural occurring discourse and elicitation data, that it is also possible to construct relative clauses with the agentive nominalizer -6p and the locative nominalizer -tſ6n. That the resulting expressions are relative clauses can be seen by the presence of a head noun and noun phrases representing other verbal arguments: (44) Mongsen Ao agentive nominalization as a relative clause
(Coupe 2007: 230)
[[m6`-tſh6n-6p] luŋ s6 n6]NP neg-flee-anom group anaph agt ‘The group that didn’t flee …’ (45) Mongsen Ao locative nominalization as a relative clause
(Coupe 2007: 231)
[[a-miô s6-tſ6n] ki n6]NP wa-6`p-uÌ•ô nrl-person die-lnom house all go-pres-dec ‘[We’re] going to the house where the person died.’
It is clear that the Mongsen agentive and patient nominalizers are used first and foremost for deriving lexical nouns. This function is overwhelmingly predominant numerically. The extension of the nominalizers to relative-clause constructions appears to be a secondary development, based on a natural fit between participant nominalizations and relative clauses. This is our first intimation of a close relationship between these two structures. A similar state of affairs is found in Zhuokeji rGyalrong. That language has three nominalizing prefixes that derive lexical nouns: the prefix s!- derives nouns which are patients or instruments; the prefix k6- derives agentive nouns; and the prefix kaderives patient nouns (and is also used in citation forms of verbs): (46) Zhuokeji rGyalrong place/instrument nominalization with s! ka-rtſi ‘to wash’ s!-rtſi ‘somewhere to wash’; ‘something used to wash’ ka-za ‘to eat’ s!-z!¥ ‘somewhere to eat’; ‘something used to eat’ ka-mot ‘to drink’ sa-môt ‘somewhere to drink’; ‘something used to drink’ (47) Zhuokeji rGyalrong agent nominalization with k6-; patient nominalization with ka k6-rKûk k6dêk
‘the runner’ ‘the chaser’
ka-dêk ‘the chased’
Nominalization in Tibeto-Burman languages of the Himalayan area 
The fact that these are lexical nouns can be seen in their ability to take nominal possessive prefixes, as in (48): (48)
Zhuokeji locative nominalization with possessive prefix indicating subject ŋ6-s!-rm! 1sg.pos-nmz-sleep1 ‘The place where I sleep; my sleeping place’
Interestingly, these prefixes are used to reference arguments of the verb not represented by the nominalizer. Thus in (49), which is an agentive nominalization, the prefix references the patient, while in (50), a patient nominalization, the prefix references the agent: (49) Zhuokeji rGyalrong agentive nominalization with attendant nominal argument rKalpô w6-k6-n6mſi king 3sg.pos-nmz-know ‘the one who knows the king’ (lit. ‘he who is knowing of the king’) (50) Zhuokeji rGyalrong patientive nominalization with attendant nominal argument rKalpô w6-ka-n6mſi king 3sg.pos-nmz-know ‘the one whom the king knows’ (lit. ‘he who the king is knowing of ’)
This structure is distinct from a relative-clause structure in two ways: (1) there is no head noun, instead the derived noun functions as a head; and (2) the nominalized element is a derived noun, as can be seen from its ability to take the possessive prefix, which references dependent entities. This construction is interesting typologically because it is reminiscent of an action nominal construction. It has both a derived noun as a head and the representation of arguments as dependents in a noun phrase. However, it differs from an action nominal construction semantically; it does not refer to the action of the verb, but to a verbal participant. Just as Mongsen Ao allows the agentive and locative derivational nominalizers to be used to form relative clauses, so does Zhuokeji rGyalrong. As in Mongsen, the use of such constructions appears to be rare. Example (51) illustrates a derived instrument nominal carrying a possessive prefix indexing the semantic patient: (51) Zhuokeji rGyalrong derived nominal with possessive prefix referencing patient [ji-w! w6-s!-rtſî]NOM 1pposs-clothes 3sg.poss-nmz-wash ‘Soap of/for our clothes’/‘what we wash our clothes with’ (lit. ‘the thing for washing of our clothes’)
 Carol Genetti
In (52) this construction is set into a relative clause, modifying a head noun féizào ‘soap’, a Mandarin borrowing. (52) Zhuokeji rGyalrong relative clause formed with derived nominal [ji-w! w6-s!-rtſî]REL w6-fejtsaw=t6 1pposs-clothes 3sg.poss-nmz-wash 3sg.poss-soap=top ‘The soap we use to wash our clothes’
The fact that it is a participant nominalization naturally means there will be an unexpressed argument within the nominalized clause (here representing the instrument) which then can be coreferential with the head noun. Note that, as with other relative clauses, the head noun also carries the possessive prefix, as illustrated above. As in Mongsen, the numerical preponderance of this construction for deriving nouns suggests that the derivational function is basic, and the use in relative clauses an extension which speakers can make, based on the strong structural and functional similarities between the two constructions. We also find overlap between a derivational nominalization and relative-clause marker in Dongwang Tibetan. There, the suffix -n6 is used for agentive nominalizations to derive occupational terms, as in the following examples: 10 (53) Dongwang agentive nominalizations (Bartee 2007: 104)
s櫹³ zu³5³ th655pæ5³ t6o5³ \"¹³ kho5³ ‰iÌ…55 m6« 5³
‘to cook food’ ‘to take a photo’ ‘to catch fish’ ‘to plough a field’
s櫹³zu²²n6¹¹ t.B655pæ5³t6o³³n6¹¹ \"¹³ kho²²n6¹¹ ‰ĩ55 m6« 5³n6¹¹
‘cook; chef ’ ‘photographer’ ‘fisherman’ ‘ploughman’
Here the domain of nominalization is the predicate and the result of the nominalization a two-word phrase, with the patient being compounded with the nominalized verb. The fact that these nominalizations are indeed lexical, creating nominal heads of noun phrases, can be seen by their ability to co-occur with noun-phrase elements, such as numerals, which follow the noun. In addition, these terms cannot occur with a head noun and still denote a person holding the denoted occupation. For example, s櫹³zu²²n6¹¹ (=ji) mi¹¹, with the optional genitive clitic and the noun 〈myi〉 ‘person’, means ‘person who is cooking’ and cannot mean ‘cook; chef ’. This example illustrates that the nominalizer -n6 is also used in the construction of relative clauses. These can be differentiated syntactically from derivational nominalizations in that relative clauses: (1) may co-occur with a head noun, (2) do not function as noun-phrase heads, (3) take an entire clause as the domain, so may include
.â•… Note that derivational nominalization applies to simple NV sequences as opposed to just V. For example \"¹³kho²²n6¹¹ ‘fisherman’ nominalizes the sequence \"¹³ ‘fish’ and kho²² ‘catch’.
Nominalization in Tibeto-Burman languages of the Himalayan area 
other arguments of the relative verb and adverbials; and (4) may be marked with the genitive clitic. All of these attributes may be found in example (54): (54) Dongwang relative clause
(Bartee 2007: 459)
[ŋe13 t655pha53 za13-ra-n6]REL =ji t611ki55 m655mæ53 t611 1sg.erg over.there hang-ra-nmz =gen coat red that ‘That red jacket I hung up over there’
An interesting point is that the nominalizer-n6, which derives agentive nouns when functioning derivationally, is used for agents, patients, benefactives, and occasionally locations when forming relative clauses. It also can be used to modify inanimate head nouns. Both of these facts are evident in example (54), where the head noun t611ki55 ‘coat’ is the inanimate patient of the relative clause verb zæ13 ‘hang’. The broader semantic range of the -n6 nominalizer in relative clauses appears to reflect semantic bleaching typical of grammaticalization. The occurrence of the same form in the two morphosyntactically distinct, if related, constructions allows for two logically possible paths of development. In one path, the morpheme started as an agentive nominalizer, which was then extended to relative clauses, as in Mongsen Ao. Once used in this syntactic environment, it underwent semantic bleaching and reanalysis as a general marker of relativization, hence its extension to inanimate head nouns and those that hold other semantic relations with the verbs. The other logically possible path of development would be from a general marker of relative clauses to agentive nominalizer. This seems less likely, as there is no reason that a general marker of relativization should specialize to deriving agentive nominals. This type of semantic specialization is not reflective of the general processes of grammaticalization. The Mongsen Ao and Zhuokeji rGyalrong cases appear to represent the first step in the grammatical extension of derivational nominalizations to relative clauses. What allow this step to be taken are the structural and functional similarities between participant nominalizations and relative clauses. Functionally, participant nominalizations refer to an entity by its involvement with the state or activity denoted by the nominalized verb, while relative clauses restrict or describe a referent by its involvement with the state or activity denoted by the nominalized verb. Structurally, participant nominalizations, in naming the participant that has a role in the activity of the verb, necessarily do not express that participant as a clausal argument. Relative clauses (unless internally-headed) also do not express the participant within them, but require a gap in the clause for the argument coreferential with the head noun. Thus the extension of a derivational nominalization to a relative clause is a short step for languages of this typological make-up. The Dongwang Tibetan case appears to represent a later stage of development, where the relative clause structure formed with the nominalizer has become firmly
 Carol Genetti
established, allowing the nominalizer to lose its semantic restriction and be extended to a greater number of environments.
5.â•… Summary and conclusions Nominalization in Tibeto-Burman languages differs from standard conceptions of nominalization in that the derivation of nouns plays a relatively minor role in the complex of nominalized structures that these languages exhibit. More significant in these languages is the extensive use of clausal nominalization for complement and adverbial or medial structures, the use of nominalized clauses in non-embedded contexts for some languages, the consistent use of nominalized clauses in relativization, and the pervasive use of nominalized clauses to derive lexical adjectives. Although clausal nominalization and derivational nominalization are clearly distinct structures, each one can give rise to the other historically. The central structure around which this occurs is the relative clause. The use of relative clauses with verbs of adjectival meanings results in nominalized one-word descriptive terms that can then be reanalyzed as adjectives. In addition, headless relative clauses appear to underlie the lexicalization of occupational terms in Dolakha Newar, showing that relativization can feed a process of derivational nominalization. On the other hand, we have seen that derivational nominalizations are sometimes placed into relative structures, modifying a noun, and expanded to include clausal arguments. A final point is that at least some relative clauses appear to give rise to complement clauses (cf. Lehmann 1980; Heine & Kuteva 2002: 254). We can see this in Dongwang Tibetan, where the nominalizer -sa, used for locational relative clauses, is the nominalizer used on complements of verbs of visual perception (example (8) above). The nominalizer has been extended from denoting the place of the action (e.g. ‘I saw (the place) where he hit him’ to denoting the action itself (‘I saw him hit him’). Once again, relativization is the key structure around which the nominalizers get extended and reinterpreted.
Abbreviations 1 1p 1sg 2 2sg 3sg abs
first person first person plural first person singular second person second person singular third person singular absolutive
abl adj adv agt caus.part cl cnv
ablative adjective adverb agentive case participial form of causative classifier converb
cont cop dat dec dist dual ego erg evid expr foc gen loc m n neg nmz nom.obl np
Nominalization in Tibeto-Burman languages of the Himalayan area 
continuative copula dative declarative distal demonstrative dual egodeictic ergative evidential expressive vocabulary focus particle genitive locative masculine gender noun negative nominalizer nominalizer.oblique noun phrase
nrl npr part pfv.pst pl pos pr pst purp pfv rel rl seq top tr v v2 vbzr vis.pfv
non-relational prefix noun prefix participle perfective past plural possessive present past purpose perfective relativizer relational prefix sequential converb topicalizer transitive verb secondary verb verbalizer perfective visual evidential
References Aikhenvald, Alexandra. 2008. Versatile cases. Journal of Linguistics 44(3): 565–603. Bartee, Ellen. 2007. A Grammar of Dongwang Tibetan. Ph.D. dissertation, University of California, Santa Barbara. Bickel, Balthasar. 1995. Relatives à antécédent interne, nominalisation et focalisation: Entre syntaxe et morphologie en Bélharien. Bulletin de La Société de Linguistique de Paris XC (1): 391–427. Bickel, Balthasar. 1999. Nominalization and focus in some Kiranti languages. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 271–296. Kathmandu: Royal Nepal Academy. Bradley, David. 1997. Tibeto-Burman languages and classification. In Papers in Southeast Asian Linguistics No. 14: Tibeto-Burman Languages of the Himalayas, David Bradley (ed.), 1–72. Canberra: Pacific Linguistics Research School of Pacific and Asian Studies, Australian National University. Chalise, Krishna Prasad. 2005. Nominalized verb forms in Tamang. In Contemporary Issues in Nepalese Linguistics, Yogendra Yadava, Govinda Bhattarai, Ram Raj Lohani, Balaram Prasain & Krishna Parajuli (eds), 479–484. Kathmandu: Linguistic Society of Nepal. Comrie, Bernard & Thompson, Sandra A. 1985. Lexical nominalization. In Language Typology and Syntactic Description, Vol. III: Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 349–398. Cambridge: CUP.
 Carol Genetti Coupe, Alec. 2007. A Grammar of Mongsen Ao [Mouton Grammar Series 39]. Berlin: Mouton de Gruyter. DeLancey, Scott. 1986a. Relativization as nominalization in Tibetan and Newari. Paper presented at the 19th International Conference on Sino-Tibetan Languages and Linguistics, Ohio State University, September 11. DeLancey, Scott. 1999. Relativization in Tibetan. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 231–249. Kathmandu: Royal Nepal Academy. DeLancey, Scott. 2002. Relativization and nominalization in Bodic. In Proceedings of the TwentyEighth Annual Meeting of the Berkeley Linguistics Society, Parasession on Tibeto-Burman and Southeast Asian Linguistics, Patrick Chew (ed.), 55–72. Berkeley CA: Berkeley Linguistic Society. DeLancey, Scott. This volume. Finite structures from nominalization constructions in TibetoBurman. In Yap, Grunow-Hårsta & Wrona (eds). Ebert, Karen. 1994. The Structure of Kiranti Languages. Zürich: ASAS-Verlag. Genetti, Carol. 1986. The grammatical development of postpositions to subordinators in Bodic languages. In Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistics Society, Vassiliki Nikiforidou, Mary VanClay, Mary Niepokuj & Debra Feder (eds), 387–400. Berkeley CA: Berkeley Linguistic Society. Genetti, Carol. 1991. From postposition to subordinator in Newari. In Approaches to Grammaticalization, Vol. 2: Focus on Types of Grammatical Markers [Typological Studies in Language 19], Elizabeth Traugott & Bernd Heine (eds), 227–256. Amsterdam: John Benjamins. Genetti, Carol. 1992. Semantic and grammatical categories of relative-clause morphology in the languages of Nepal. Studies in Language 16(2): 405–427. Genetti, Carol. 2005. The participial construction in Dolakha Newar: Syntactic implications of an Asian converb. Studies in Language 29(1): 35–87. Genetti, Carol. 2007. A Grammar of Dolakha Newar [Mouton Grammar Library 40]. Berlin: Mouton de Gruyter. Genetti, Carol & Hildebrandt, Kristine. 2004. The two adjective classes in Manange. In Adjective Classes: A Cross-Linguistic Typology, Robert Malcolm Ward Dixon & Alexandra Aikhenvald (eds), 74–97. Oxford: OUP. Genetti, Carol, Bartee, Ellen, Coupe, Alec, Hildebrandt, Kristine & Lin, You-Jing. 2008. Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himalayan area. Linguistics of the Tibeto-Burman Area 31(2): 97–143. Hargreaves, David. 1991. The Concept of Intentional Action in the Grammar of Kathmandu Newari (Nepal). Ph.D. dissertation, University of Oregon. Hargreaves, David. 2005. Agency and intentional action in Kathmandu Newar. Himalayan Linguistics Journal 5: 1–48. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Herring, Susan C. 1991. Nominalization, relativization, and attribution in Lotha, Angami, and Burmese. Linguistics of the Tibeto-Burman Area 14(1): 55–72. Hildebrandt, Kristine. 2004. A grammar and glossary of the Manange language. In TibetoBurman Languages of Nepal: Manange and Sherpa [Pacific Linguistics 557], Carol Genetti (ed.), 3–189. Canberra: Pacific Linguistics. Hildebrandt, Kristine. In preparation. A grammar of Manang-Gurung. Horie, Kaoru. 1998. Functional duality of case-marking particles in Japanese and its implications for grammaticalization: A contrastive study with Korean. In Japanese/Korean Linguistics 8, David J. Silva (ed.), 147–159. Stanford CA: CSLI.
Nominalization in Tibeto-Burman languages of the Himalayan area 
Kölver, Ulrike. 1977. Nominalization and Lexicalization in Modern Newari. Köln: Arbeiten des Kölner Universalien-Projekts. Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. Koptjevskaja-Tamm, Maria. 2005. Action nominal constructions. In World Atlas of Language Structures, Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds), 254–257. Oxford: OUP. Lahoussois, Aimée. 2002. Nominalization, relativization, and genitivization in Thulung Rai. In Proceedings of the Twenty-Eighth Annual Meeting of the Berkeley Linguistics Society, Parasession on Tibeto-Burman and Southeast Asian Linguistics, Patrick Chew (ed.), 87–98. Berkeley CA: Berkeley Linguistic Society. Lahoussois, Aimée. 2003. Nominalization its various uses in Thulung Rai. Linguistics of the Tibeto-Burman Area 26(1): 33–57. LaPolla, Randy J. 2006. Nominalization in Rawang. Paper presented at the 39th International Conference on Sino-Tibetan Languages and Linguistics. Lehmann, Winfred P. 1980. The reconstruction of non-simple sentences in PIE. In Linguistic Reconstruction and Indo-European Syntax [Current Issues in Linguistic Theory 19], Paolo Ramat (ed.), 113–144. Amsterdam: John Benjamins. Malchukov, Andrej L. 2006. Constraining nominalization: Function/form competition. Linguistics 44(5): 973–1009. Matthews, Peter H. 1997. The Concise Oxford Dictionary of Linguistics [Oxford Paperback Reference]. Oxford: OUP. Matisoff, James A.1972. Lahu nominalization, relativization, and genetivization. In Syntax and Semantics, Vol. 1, John P. Kimball (ed.), 237–257. New York NY: Academic Press. Matisoff, James A. 2003. Handbook of Proto-Tibeto-Burman: System and Philosophy of SinoTibetan Reconstruction. Berkeley CA: University of California Press. Noonan, Michael. 1997. Versatile nominalization. In Essays on Language Function and Language Type, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Noonan, Michael. 2008. Nominalization in Bodic languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins. O’Rourke, Mary Jo. 2000. Relatively Nominal: Relativisation in Kathmandu Nepal Bhasa (Newari). MA thesis, La Trobe University. Post, Mark. 2007. A Grammar of Galo. Ph.D. dissertation, Research Centre for Linguistic Typology, La Trobe University. Regmi, Dan Raj. 2005. Morphosyntax of relativization in the Bhujel language: Typological perspective. In Contemporary Issues in Nepalese Linguistics, Yogendra Yadava, Govinda Bhattarai, Ram Raj Lohani, Balaram Prasain & Krishna Parajuli (eds), 541–546. Kathmandu: Linguistic Society of Nepal. Watters, David. 2008. Nominalization in the East and Central Himalayish Languages of Nepal. Linguistics of the Tibeto-Burman Area 31(2): 1–43. Yap, Foong Ha & Wang, Jiao. This volume. From light noun to nominalizer and more: On the grammaticalization of zhe and suo in Old and Middle Chinese. In Yap, Grunow-Hårsta & Wrona (eds).
Aspects of the historical development of nominalizers in the Tamangic languages Michael Noonan
University of Wisconsin-Milwaukee This paper focuses on the form and function of nominalizations within Tamangic, a sub-group of Bodic (Tibeto-Burman). Nominalizations have played an important role in the grammar of Tamangic languages from the Proto-Tamangic stage until the present. The paper posits historical developments for nominalizers and addresses issues related to this development. It concludes that the basic features of the use of nominalizations have remained constant; however there have been significant innovations, largely induced by contact with the lingua franca, Nepali. Among these innovations are: the rise of new nominalizers, the loss of the genitive with nominalizations used adnominally, as well as the development in some Tamangic languages, Chantyal in particular, of nominalizations exhibiting tense distinctions, and the development of periphrastic constructions involving nominalized verbals in the verb complex.1
1.â•… Tamangic languages The Tamangic languages2 are a small, uncontroversial grouping within the Bodic section of Sino-Tibetan spoken entirely within Nepal. Within Bodic, the Tamangic languages are placed within the Tibetic branch of the Bodish languages, along with Central Bodish [the Tibetan Complex and Tshangla] and Ghale. Proposed relationships of the languages within Bodic are displayed in Appendix 1. The languages within the Tamangic group and their relationships to each other are given in the figure below.
.╅ The abstract has been written posthumously for Michael Noonan by the editor. .╅ The work reported on in this paper has been supported by the following grants from the National Science Foundation: DBC-9121114, SBR-9600717, and SBR-9728369. I would like to thank the two anonymous reviewers for very helpful comments on this paper.
 Michael Noonan TAMANGIC Tamang complex W. Tamang E. Tamang
WESTERN Gurung
Manange-Nar-Phu complex Manange Nar-Phu
NORTHWESTERN
Thakali Complex Thakali
Chantyal Seke
Figure 1.╇ Proposed genetic relationship in Tamangic
People speaking Proto-Tamangic probably entered Nepal from Tibet between fifteen hundred and two thousand years ago, an estimate made on the basis of a number of facts, including the closeness of the relation of Tamangic to the Tibetan Complex and the relations among the Tamangic languages themselves. The center of their radiation is the upper Kali Gandaki Valley; from there the Tamangs and later the Gurungs moved east. The Tamagic languages are notable for the extensive use made of nominalized clauses and nominalizations generally within their grammars. This state of affairs is most fully developed in Chantyal and is discussed in detail in Noonan 1997: a brief summary is provided in Section 2 below. In this paper, I will discuss some issues concerning the historical development of nominalizations within the Tamangic family. In Section 2, I will provide a bit of background information concerning the form and function of nominalizations within this group, followed by a discussion in Section 3 of some historical developments affecting nominalizations.
2.â•… Tamangic verbal systems and nominalization In this section, I’ll provide some background information concerning Tamangic verbal systems and the functions of nominalizations.
2.1â•… Interclausal relations in proto-Tamangic Data from the grammars of the modern languages, together with comparative and areal data, yield information concerning many basic aspects of the interclausal syntax in Proto-Tamangic, though many problems remain. From a morpho-syntactic perspective, two sorts of clause-types can be distinguished, independent and dependent: independent clause types are capable of being integrated into discourse on their own,
Historical development of nominalizers in the Tamangic languages 
while dependent clause types depend on another clause for at least part of their interpretation. In Proto-Tamangic, the two sorts were also identifiable morphologically. In dependent clause types, the final verbal within the verb complex is overtly marked with a subordinator suffix; this verbal heads a nominalized or a converbal [i.e. adverbial] clause. The nominalization suffix is *pa, which goes back at least to Proto-Bodic. There were likely a number of converbal suffixes. All the languages evidence a sequential converb in *si, which is likely Bodish. There is also a certain amount of evidence for a simultaneous converb in *ma and perhaps one in *kay, and a manner converb in *na. The conditional is *la, which may also be ProtoBodic: this suffix is evidenced in a wide variety of functions in the contemporary Tamangic languages and may already have assumed some other functions in the Proto-Tamangic period.3 For independent clauses, negative, imperative, and interrogative affixes can be reconstructed back to Proto-Tamangic. (1) *-e interrogative *-(k)u imperative *ha- negative4 *tha- negative imperative
Tense-aspect morphology in the modern languges is mostly transparently derived from the non-finite morphology discussed above or from verbs. The deverbal affixes in (2) are universally attested in Tamangic.
(2) *-m(u) non-past/imperfective [< *mu ‘be’] *-ci(n) perfective [< *cin ‘finish’]5
These suffixes suggest a syntagm in which the verb complex could consist of juxtaposed bare verbals [i.e. where no verbal is marked with non-finite suffixes] and where the final verbal is also unsuffixed [i.e. not overtly marked for tense-aspect]. Both possibilities persist in the Tamangic languages, albeit as minority patterns.
.â•… The uses of the various converbs in Chantyal are discussed in detail in Noonan 1999. The reconstructions provided here are informal, lacking specifications for vowel length and, where applicable, tone. .â•… The negative can also occur in dependent clauses. .â•… The evidence for the origin of the Tamangic perfective in *cin ‘finish’ is quite robust. Reflexes of this affix range from -čin in Nar-Phu, to -jin ~ -ji in Seke, to -ci in Thakali and Tamang, to -ji ~ -i in Chantyal [in Chantyal, the former occurs after consonants, the latter intervocalically]. Semantically, Heine & Kuteva 2002 list a number of examples of a similar grammaticalization, and syntactically juxtaposed bare verbals are still attested in Tamangic, as explained in the text.
 Michael Noonan
Nar-Phu, for example, attests juxtaposed bare verbals [a.k.a. verb concatenation], in particular with motion verbs: (3) nôkyu cε-se tBuŋ.nâpraŋ čBâŋ cε pi tê l5¥-čin dog def-erg bee nest def go.fast fall do-past ‘the dog unwittingly knocked down the beehive’
Nar-Phu also attests unmarked final verbals, though only modal verbs allow this: (4) lakpε-se iŋliš pBi-ne čBur Lakpa-erg English speak-inf be.able ‘Lakpa can speak English.’
Mazaudon 2003 describes a similar pair of constructions for Eastern Tamang. Dependent clauses were, by definition, subordinate. Independent clauses could also be subordinate, but only as complements of ‘say’, a situation that persists in the modern languages except where the relevant constructions are transparently borrowed from Indo-European Nepali.6
2.2â•… Constructions with nominalizers The internal grammar of nominalized clauses is fairly consistent across the modern languages, and we can therefore assume that it has been unchanged since the Proto-Tamangic period. The argument array associated with the nominalized verbals is assigned the same cases as their independent clause counterparts.7 Compare the independent clause (5a) with its nominalized counterpart (5b) in the Chantyal sentences below: (5) a.
bBuluŋ-s6 gBwaral ca-i leopard-erg wild.goat eat-perf ‘The leopard ate the wild goat’
b. na-s6 [bBuluŋ-s6 gBwaral ca-wa] mara-i I-erg leopard-erg wild.goat eat-nom8 see-perf ‘I saw the leopard eat the wild goat’
Nominalized verbals clearly played an important role in the grammar of ProtoTamangic, judging by the role they currently play in the grammars of the modern languages. The functional range of nominalized verbals is fairly similar within the
.â•… See Noonan 2006 for an extended discussion of the discourse functions of complements of ‘say’ in Chantyal. .â•… This is true as well for converbal clauses. .â•… Chantyal -wa is a reflex of the Proto-Tamangic nominalizer *-pa.
Historical development of nominalizers in the Tamangic languages 
languages of the group, allowing us to infer that the proto language probably shared this functional range. These functions are listed in (6) below and are discussed in detail in Noonan 1997:
(6) Functions of forms with the nominalizer in Tamangic languages:
– nominalization in the strict sense: naming activities and states – verb complementation: clausal complements of verbs – purpose nominal: ‘in order to’ sense, with the locative: *-pa-ri < nominalizer+locative case – adnominal clause: clausal modifiers of nouns, possibly with the genitive [see §3.1 below] – non-relative attributive nominals: nominal, even case-marked nominal, modifiers of nouns, possibly with the genitive [see §3.1 below] – agent and patient nominals: nominals referring to the agent or patient roles within predications – as main clause verb with a mirative sense
All of these functions can be expressed with the same nominalizer, for example -wa in the case of Chantyal. For illustrations of the first three, see Noonan 1997. The others, cross-linguistically somewhat more unusual, are illustrated and discussed below.
2.2.1â•… Adnominal clauses The use of clauses whose heads are nominalized verbs, and which function as adnominal modifiers, is characteristic of the Bodic languages [Noonan 2008]. Within the Tamangic languages, this is the only native syntactic device for allowing clausal modification of nouns. Examples can be found in (7) and (8) below. The use of the genitive with these nominalizations will be discussed in §3.1. adnominal clauses (7) m6nchi-s6 ca-si-wa gay-ye sya person-erg eat-ant-nom cow-gen meat ‘the beef that the person ate’ (8) cá pxra-bá-e mxi jaga that walk-nom-gen person pl ‘those walking people’ (=sentries)
chantyal
gurung
2.2.2â•… Non-relative attributive nominals Rather less common within Bodic generally,9 though well attested in Tamangic, is the situation where the affix used to form nominalizations is suffixed to non-verbal roots
.╅ An anonymous reviewer reminded me that constructions similar to the Tamangic nonrelative attributive nominals are well-attested in Kiranti languages as well. My impression,
 Michael Noonan
[e.g. in (9)] and case-marked nouns [e.g. in (10)]. The resulting nominals can be used attributively, hence the designation ‘non-relative attributive nominals’. non-relative attributive nominals t6yla-wa saka yesterday-nom ancestor ‘yesterday’s ancestors’ [V101]10
chantyal
b. lig6«-wa samra-ye ph6lce back-nom thigh-gen muscle ‘back thigh muscle’ [I24]
chantyal
(9) a.
bana-r-bá-e sĩ forest-loc-nom-gen wood ‘trees from the forest’
gurung
b. nBa-ri-g6m-wa sya inside-loc-abl-nom meat ‘innards’
chantyal
(10) a.
c.
ram-siŋ-wa photo Ram-com-nom photo ‘Ram’s photo’ [i.e. a photo Ram owns]11
chantyal
In some languages, as in the Gurung example in (10a), these non-relative attributive nominals occur with the genitive: these are the same languages that use the genitive with nominalized clauses used adnominally. Two additional facts about these non-relative attributives should be noted. First, the forms affixed with the nominalizer are nouns: they can fill noun slots, be pluralized, and take case affixes. The following examples from Chantyal illustrate this. (11) m6ŋg6le-ri-wa gay-ma Mangale-loc-nom cow-pl ‘cows from Mangale’ (12) m6ŋg6le-ri-wa-ma-s6 syal mara-i Mangale-loc-nom-pl-erg jackal see-perf ‘the people from Mangale saw the jackal’
however, is that these constructions are not as common in the Tibetan Complex, in KhamMagar, and in the Newar languages. .â•… [V101] and similar notations below refer to published Chantyal discourses: Noonan et al. 1999; Noonan 2005. .â•… In Chantyal, a construction with the genitive, Ram-ye photo, is used to express ‘a photo of Ram’ [i.e. one having his image], though the latter can also be used to identify one he owns: see Noonan 1997.
Historical development of nominalizers in the Tamangic languages 
In (11), the non-relative attributive nominal modifies ‘cows’. By itself, however, it is also a noun meaning ‘those from Mangale’, and as such it can take a plural suffix and fill a clausal noun slot, as in (12).12 It should be noted that when simple, non-case-marked nouns like t6yla ‘yesterday’ in (9a) are not used adnominally, they are never found with the nominalization suffix. Second, suffixation with the nominalizer is recursive, the only limitations being those of sense and processability. Example (13) below shows a non-relative attributive nominal formed from a case-marked attributive nominal: (13) m6ŋg6le-ri-wa-ma-siŋ-wa photo Mangale-loc-nom-pl-com-nom photo ‘the photo belonging to the people from Mangale’
This form, too, may fill a nominal slot. In other words, an attributive nominal can be built off of another attributive nominal and fill a noun slot within a clause, as in (14): (14) na-s6 m6ŋg6le-ri-wa-ma-siŋ-wa-ra dekh6-i I-erg Mangale-loc-nom-pl-circ-nom-dat show-perf ‘I showed it to the owners from Mangale’
2.2.3â•… Agent and patient nominals Agent and patient nominals are also formed with the nominalizer in Tamangic languages like Chantyal.
agent and patient nominals
(15) Agent nominal a. na-s6 reysi thũ-wa-ye naku khway-k6y mu I-erg raksi drink-nom-gen dog feed-prog be.npst ‘I’m feeding the raksi-drinker’s dog’
chantyal
Patient nominal chantyal b. c6 l6ra pari-wa-ma g6tilo l6ra a-ta-si-n t6 that strip make.happen-nom-pl good strip neg-become-ant-sup fact ‘those strips that I made might not have become good strips’ [I110]
In (15a), the agent nominal reysi thũ-wa ‘raksi drinker’ is suffixed with the genitive marker and modifies ‘dog’. Agent and especially patient nominals are often referred to as ‘internally headed relative clauses’ in the literature on Bodic languages. In the context of some languages, that might be a reasonable analysis, but for many, perhaps most Bodic languages, these constructions are probably best analyzed as agent and
.â•… Nouns formed in this way are thus a species of the agent and patient nominals discussed in §2.2.3.
 Michael Noonan
patient nominals. For instance in (15b), the patient nominal c6 l6ra pari-wa-ma ‘those strips that I made’ is treated as a single nominal and the nominalized verb receives the plural suffix -ma, an unlikely scenario if the construction were an internally headed relative clause since we would expect the internal head to bear the plural in that case.13
2.2.4â•… Main clause verb with a mirative sense In Tamangic languages, when nominalizations appear as main clauses, the typical effect is one of mirativity, i.e. the sense that the predication so expressed is in some sense surprising, contrary to expectation, or in some way exasperating. (16)
main clause with mirative sense b6nnu-ye nal tato ta-si-wa gun-gen barrel hot become-ant-nom ‘The barrel of the gun had become hot!’ [R29]
chantyal
Noonan 2008 provides some discussion and references to the situation in other Bodic languages.
3.â•… Historical developments In this section I’ll discuss some historical developments of nominalizations in the Tamangic languages. These developments include (1) the issue of the genitive with adnominal nominalizations, (2) the development of tense distinctions in nomimalized clauses, (3) the creation of new nominalizers, and (4) the use of nominalized verbals in periphrastic constructions.
3.1â•… The genitive with adnominals It was noted in several places above that nominals used adnominally may or may not occur with the genitive in Tamangic languages. Compare, for example, the Chantyal and Gurung examples of nominalized clauses used adnominally in (7) and (8), and the Chantyal and Gurung examples of non-relative attributive nominals in (10). In this section we will take up the question of whether the genitive should be posited with this construction in Proto-Tamangic.
.â•… The patient nominal in (15b) can’t easily be analyzed as consisting of a head followed by a postmodifier since postmodifiers don’t otherwise occur in Chantyal.
Historical development of nominalizers in the Tamangic languages 
3.1.1â•… The use of the genitive in the modern languages The modern Tamangic languages present a mixed picture with regard to the use of the genitive with nominalizations used adnominally. A summary of the data from the modern languages can be found in (17): (17) chantyal: Never uses the genitive. thakali: Georg’s 1996 grammar makes no mention of the genitive with relative clauses. Hari & Maibaum 1970 assert that the genitive is optional, but it should be noted that Georg and Hari & Maibaum investigated different dialects of Thakali. seke: Isao Honda (personal communication) reports that the genitive is optional with nominalizations. manange: Hildebrandt 2003 reports that relative clauses are formed with the nominalizer -p∙ [ -pe. One problem with this interpretation is that the genitive in Manange is -l∙, not -i. Still, it is probable that at one point, Manange had a genitive in -i [< *-kyi],14 the modern genitive in -l∙ deriving historically from the dative, which frequently is -ra/la in other Tibetic languages. However the data from closely related Nar-Phu offers another interpretation of Manange -pe. nar-phu: In Nar-Phu, adnominal clauses with present senses use the nominalizer -pε [ nominalizer may have been followed. This analysis is possible for Nuosu Yi (this volume).
 Stephen Morey
(35) lam sawng wa hpe dai n măjaw la lam¹ soŋ¹ waa¹ phee⁴ dai¹ n⁴- mәjoo⁴ laa¹ [road first def a.ag] [that neg- good male wa i dun na gaw m. waa¹ ii⁴ dun⁴ naa⁴ go¹ m¹ def ag] lead seq top excl ‘And that bad fellow led the one who was first on the road.’ (Story of the Blind Men (SDM08-20050801-002) No. (46), told by Bhupeswar Ningda)
We can see from this example how a full noun meaning ‘father’ might develop into a kind of generic light noun meaning ‘the one who …’ which itself becomes eventually regarded as a marker of definiteness. Example (35) may be a case of structural ambiguity: wa could be interpreted as a nominalizer (i.e. sawng wa ‘leading, being first’) or as definiteness marker. This is one of the few cases where wa co-occurs with the anti-agentive hpe, but there are other examples from the same text where the element marked by wa hpe is already nominal One such example is n măjaw la wa hpe ‘to the bad man’ referring to the same individual mentioned in (35). The nominal in this latter case is la ‘male’ and so wa is not functioning as a nominalizer. By analogy, our analysis of the first NP In (35) is that wa does indeed function as a definiteness marker here, but perhaps also has the earlier, now infrequent, function of generic nominalizer. Synchronically wa is far more frequently encountered as a definiteness marker with virtually no nominalization function, as for example the second instance of it in (35). For this reason principally we have glossed wa as a definiteness marker in all cases, and suggest that for the Numhpuk Singpho speakers, wa is a single entity, whose most salient function is definiteness.
6.â•… Conclusion We have shown that Numhpuk Singpho has two main nominalization processes: that marked by hpa (Section 4), most often used to derive a new nominal from a verb but also found in other functions such as clausal nominalization and marginally relativization; and the clausal nominalization most often found co-occurring with wa (Section 5). Numhpuk Singpho differs from many Tibeto-Burman languages in that it shows at most marginal nominalization-relativization syncretism. The usual relative clause structure of realis marked clause preceding the head noun is unrelated to the nominalization discussed in this paper. The use of hpa for some uncommon relative structures as we have seen in (14) above does suggest a partial syncretism.
Nominalization in Numhpuk Singpho 
In Numhpuk Singpho, the ubiquitous presence of the definiteness marker wa in clausal nominalization constructions certainly suggests that it may have a history as a nominalizer or could be in the process of being re-analysed as a clausal nominalizer. The closest variety to Numhpuk Singpho about which we have information is Turung (see Footnote 8 above), where nominalized clauses do not co-occur with the definiteness marker wa. This might be a recent development in Turung, or a substrate influence of Tai languages on Turung, but it leads us to suggest that a similar type of umarked nominalization is occurring in Numhpuk Singpho as well. Thus we do not term wa a clausal nominalizer, albeit admitting that it might in the past have been one. Rather, we suggest that Numhpuk Singpho is an example of a language in which clauses can be nominalized without an overt marker, and then license the presence of NP markers, such as definiteness marker wa.
Abbreviations a.ag anti-agentive adv adverbialiser ag agentive benf benefactive caus causative cop copula cos the verb ‘return’ used as a change of state decl declarative def definite euph euphonic (sounding well) excl exclamation fut.imm future immediate gv generic or light verb (Dai 1992)
imp imperative loc locative neg negative nomz nominalization one one pn proper name poss possessive prt particle real realis seq possessive marker used as sequential sfp sentence final particle top topic
References Bickel, Balthasar. 1999. Nominalization and focus constructions in some Kiranti languages. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren G. Glover (eds), 271–296. Kathmandu: Royal Nepal Academy. Burling, Robbins. 2003. The Tibeto-Burman languages of Northeastern India. In The Sino-Tibetan Languages, Graham Thurgood & Randy LaPolla (eds), 169–191. London: Routledge.
 Stephen Morey Coupe, Alec. 2005. Teasing apart nominalization and relativization in Mongsen Ao. Australian Linguistics Society Annual Conference 2005, Monash University, Melbourne, 28–30th September. Dài Qìngxìa. 1992. Jĭngpōyŭ Yŭfă (Jinghpo Grammar) Beijing: Zhong yang min zu xue yuan chu ban she. Dai, Qingxia & Lon Diehl. 2003. Jinghpo. In The Sino-Tibetan Languages, Graham Thurgood & Randy J. LaPolla (eds), 401–408. London: Routledge. DeLancey, Scott. 1986. Relativization as nominalization in Tibetan and Newari. Paper presented at the Nineteenth Annual International Conference on Sino-Tibetan Languages and Linguistics, Ohio State University, Columbus. DeLancey, Scott. 1999. Relativization in Tibetan. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren G. Glover (eds), 231–249. Kathmandu: Royal Nepal Academy. DeLancey, Scott. 2008. Nocte and Jinghpaw: Morphological correspondences. Paper presented at the North East Indian Linguistics Society III, Gauhati University, January 2008. Genetti, Carol, Couple, Alec R., Bartee, Ellen, Hildebrandt, Kristine & Lin, You-Jing. 2008. Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himalayan Area. Linguistics of the Tibeto-Burman Area 31(2): 97–143. Hanson, Ola. 1896. A Grammar of the Kachin Language. Rangoon: American Baptist Missionary Press. Hanson, Ola. 1906. A Dictionary of the Kachin Language. Rangoon: American Baptist Missionary Press. (Reprinted in 1954). LaPolla, Randy J. 2008. Nominalization in Rawang. Linguistics of the Tibeto-Burman Area 31(2): 45–66 (Special Issue on Nominalization in Tibeto-Burman edited by Alec Coupe). Maran, LaRaw. 1971. Burmese and Jinghpaw: A Study of Tonal Linguistic Processes. OPWSTBL IV, x, 210p. (Rev. by Matisoff, JAS 32 (1973). 4: 741–3). Matisoff, James A. 1972. Lahu nominalization, relativization, and genitivization. In Syntax and Semantics, Vol. I, John Kimball (ed.), 237–57. New York NY: Seminar Press. Matisoff, James A. 1974a. The tones of Jinghpaw and Lolo-Burmese: Common origin vs. independent development. Acta Linguistica Hafniensia 15(2): 153–212. Matisoff, James A. 1974b. Verb concatenation in Kachin. Linguistics of the Tibeto-Burman Area 1(1): 186–207. Matisoff, James A. 2003. Handbook of Proto-Tibeto-Burman: System and Philosophy of SinoTibetan Reconstruction. Berkeley CA: University of California Press. Matthews, Peter. 1997. Concise Dictionary of Linguistics. Oxford: OUP. Morey, Stephen D. 2005. The Tai Languages of Assam – A Grammar and Texts. Canberra: Pacific Linguistics. Morey, Stephen D. 2010. Turung – A Variety of Singpho language Spoken in Assam, Canberra: Pacific Linguistics. Needham, J.F. 1889. Outline Grammar of the Singpho Language – As Spoken by the Singphos, Dowanniyas and Others Residing in the Neighbourhood of Sadiya. Shillong: Assam Secretariat Press. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Noonan, Michael. 2008. Nominalization in Bodic languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins.
Nominalization in Numhpuk Singpho 
Watters, David E. 2008. Nominalization in Himalayish Languages of Nepal. Linguistics of the Tibeto-Burman area. 31(2): 1–44. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins.
Nominalization in Nuosu Yi* Hongyong Liu & Yang Gu
South China Normal University/Chinese University of Hong Kong Nuosu Yi (a Tibeto-Burman language spoken mainly in the southwestern provinces of Sichuan, Yunnan, and Guizhou in China) has two sentential nominalizers. One is the morpheme ko33 (a third person singular pronoun in origin), which is used to mark verb complement clauses, topic clauses, temporal adverbial clauses, and conditional clauses. The other is the morpheme su33, which is found in gerundives, relative clauses, and sentential focus clauses. In this paper, we start with a descriptive distinction between derivational nominalization and syntactic nominalization in Nuosu Yi, and then move on to an examination of the nominalization strategies involved in the use of su33, whose grammatical functions range over marking definiteness, relativization, and focus complementation. We explore the mechanisms giving rise to structural reanalysis that has taken place in su33 nominalization. We also extend our analysis to the morpheme ko33 and discuss its similarities with and differences from su33 and how it develops from a pronoun to a sentential nominalizer.
*An earlier version of this paper was presented at the Linguistics Seminar of the Department of Linguistics and Modern Languages, the Chinese University of Hong Kong in 2008, and we thank the audience for their valuable feedback. We are very grateful to Da Wu, the coauthor of the book A Grammar of Yi, for his insight into Nuosu Yi, which is his mother tongue, and the numerous discussions we had with him which enhanced our understanding of the language. Some of the data presented here were collected by the first author during his field trips to Liangshan Yi Autonomous Prefecture in Sichuan, China. We are thankful to Hxi Lyrgursse, Qubi Atguop, Jienuo Juti, Aruo Zuoge, and Jili Youpo who served as our Nuosu Yi consultants. This paper would not have been possible without their assistance. We also extend our sincere thanks to Suhua Hu, for her suggestions on some of the data presented in this paper. Our special thanks go to Foong Ha Yap for the numerous insightful comments and useful references on grammaticalization she offered to us in the past year. We also thank the reviewers for their critical comments and suggestions, which helped us to reconsider some of the issues and strengthen our arguments in the revision. Finally we would like to acknowledge the support from the Graduate School and the Department of Linguistics and Modern Languages of the Chinese University of Hong Kong to the first author, including travel subsidies and a postdoctoral fellowship, and the research funding from the Chinese University of Hong Kong (Direct Grant #4450194) to the second author.
 Hongyong Liu & Yang Gu
1.â•… Introduction Yi, which was also called Lolo in the past in China, is mainly spoken in the southwestern provinces of Sichuan, Yunnan, and Guizhou in China. It is a member of the Lolo-Burmese (Burmese-Lolo) sub-branch of the Tibeto-Burman (TB) branch of the Sino-Tibetan language family. The total population of the Yi is around 6.57 million according to the national census of the year 1990. Yi has a large number of varieties across the three provinces. The variety studied in this work is Nuosu Yi, which is a subdialect of the northern dialect of Yi. It is also referred to as Nosu, Northern Yi, Sichuan Yi, or Liangshan Yi in the literature. This variety is mainly spoken in Liangshan Yi Autonomous Prefecture, Sichuan Province, China. The data in this paper are from Shengzha dialect, which serves as the lingua franca in Sichuan Liangshan Yi Autonomous Prefecture. It is the basis for the official Sichuan Yi orthography. The basic word order of Nuosu Yi is SOV, which means a verb always follows its arguments, but not necessarily in the sentence final position. Sentence final positions are taken by aspect, mood, evidential and other grammatical markers. Nominalization can occur at both the morphological level and the syntactic level in Nuosu Yi. At the morphological level, a nominalizer takes the form of a suffix, creating a noun from a verbal or adjectival root. At the syntactic level, Nuosu Yi shows a heterogeneous nominalization strategy. A clause can be nominalized by the definite article,1 classifiers, or the factual complementizer. The resultant nominalized clause can function as a sentential subject, a topic, a sentential object, an adverbial, a relative clause, a conditional clause, or a complement clause of the copula. In the following sections, we will explore all these nominalization strategies in Nuosu Yi. The structure of this paper is as follows. Section 2 examines different morphological nominalization patterns. Section 3 examines the nominalization strategy by the definite article su33, plus some discussion of the nominalization strategy by classifiers. We show how the interaction of the nominalizer su33 and classifiers specifies the referential and number properties of the head nouns in relative constructions. Section 4 examines the nominalization pattern with the factual complementizer ko33. We first describe the multiple functions of this nominalizer and then identify its grammaticalization path from a pronoun to a complementizer-type nominalizer.
.╅ One of the reviewers raises the question of whether the clause is nominalized by the definite article, or the presence of the definite article is licensed because the clause is already nominalized. The Nuosu Yi data show that the relative clause is nominalized by the definite article, and no other nominalizer is found in the Nuosu Yi relative construction. So the definite article in Nuosu Yi relative construction plays two roles at the same time: as a definiteness marker as well as a nominalizer. This will be detailed in Section 3.2.
Nominalization in Nuosu Yi 
Finally, the paper is closed with a brief summary of all the nominalization patterns we have observed in Nuosu Yi.
2.â•… Derivational nominalization Similar to many other languages, in Nuosu Yi, a verbal root with a nominalizing suffix can be the name of the activity or state designated by the verbal root, or may be one of the arguments associated with the verbal root (Comrie & Thompson 1980). There is an array of nominalizing suffixes that can be categorized under the label of morphological nominalizer whose function is to derive lexical words, hence derivational nominalizers.
2.1â•… Argument nominalization In Nuosu Yi, there are a number of suffixal morphemes which can derive argumentdenoting nominals. Based on the lexical semantic meanings they bear, we classify them as agent nominalizer (-su33), theme/patient nominalizer (-du33), abstract-noun nominalizer (-lu33), manner nominalizer (-thε33), and locative nominalizer (-dш33). In the following, we will examine them one by one.
2.1.1â•… Nominalization of theme/patient In Nuosu Yi, the affix -du33 can be suffixed to a verbal root to derive a nominal. When this happens, it derives a nominal which is theme/patient denoting. Hence we define -du33 as a patient nominalizer. As exemplified in (1a), the patient nominalizer -du33 can also be suffixed to a root verb such as dzш33 ‘eat’, and the resultant deverbal noun dzш33du33 means ‘food’, which can be understood as “something to be eaten”. Additional examples can be seen in (1b–f). (1) a.
dzш33du33 b. eat-nmz ‘food’
ga55du33 wear- nmz ‘clothes’
c.
s‹21du33 d. know-nmz ‘knowledge’
v‹33du33 buy-nmz ‘goods’
e.
bo21du33 f. possess-nmz ‘wealth’
mu33du33 do-nmz ‘task’
2.1.2â•… Abstract-noun nominalization The nominalizing suffix -lu33 in Nuosu Yi can be used to derive deverbal nouns designating abstract entities relating to the verbal root. In (2a), the nominalizer -lu33 is
 Hongyong Liu & Yang Gu
suffixed to the root verb ŋo21 ‘think’, and the resultant deverbal noun means ‘thought’, which can be understood as an abstract noun. The difference between -du33 and -lu33 is that the nouns derived by -du33 refer to “things that are V-ed”, while nouns derived by -lu33 refer to abstract entities associated with the verbal roots. For example, mo33 lu33 refers to a person’s point of view rather than what is seen.2 (2) a.
ŋo21lu33 b. think-nmz ‘thought’
mo33lu33 see-nmz ‘point of view’
c.
vu21lu33 d. sell-nmz ‘goods’
kш55 lu33 can-nmz ‘technology’
2.1.3â•… Manner and locative nominalization In Nuosu Yi, there is a suffix -thε33 which derives a noun from an action verb, yielding a nominal meaning “a way of V-ing”. We can regard -thε33 as a manner nominalizer. As shown in (3a), -thε33 is suffixed to the verbal root mu33 ‘do’, and the resultant deverbal noun means ‘way of doing’. (3) a.
mu33 thε33 b. do-nmz ‘way of doing’
dzш33 thε33 eat-nmz ‘way of eating’
There is another suffix -dш33 which derives a nominal from an action verb meaning “a place where ‘V-ing’ happens”, as in (4). The lexical origin of this locative nominalizer is “field, ground, or earth”. (4) a.
Ži34 dш33 b. sit-nmz ‘seat’
zo34 dш33 study-nmz ‘school’
c.
hi55 dш33 d. stand-nmz ‘job, post’
i55 dш33 live-nmz ‘home’
In (4a), the nominalizer -dш33 is suffixed to the root verb Ži34 ‘sit’, and the resultant deverbal noun means ‘seat’, literally “a sitting place” or “a place for sitting”.
2.2â•… Nominalization of predicates In Nuosu Yi, there exists a bound morpheme -su33. The lexical origin of this morpheme is the noun ‘person’ (Hu 2002). -su33 functions as a suffix which can derive a noun from .â•… As brought to our attention by Foong Ha Yap (pers. comm.), the contrast between -du33 and -lu33 is similar to the difference between the Japanese nominalizers mono and kodo.
Nominalization in Nuosu Yi 
a verb. For instance, the verb ‘labor’ is turned into a noun by the nominalizer -su33, as highlighted in (5), where the derived noun is used as the object of the main verb.3 Suffixation of -su33 may also enable the derived noun to refer to the agent argument of the verb. In this case, -su33 serves as an agentive nominalizer, as shown in (6).4 (5) ŋo21 kha34 di33 Ži33 ╛╛╛Žo21 bo21 su33 hш21 Œ‹33. we all [laborV-su]N attach importance to ‘We all attach importance to labor.’ (Chen & Wu, 1998: 68) (6) dzш33su33 ndo33su33 a34 ti33 lo34, mu33su3 Œo55su33 dŒε55a21 gш33. eat-nmz drink-nmz only Conj do-nmz work-nmz neg-have ‘There are only eaters and drinkers, but no doers.’ (Hu, 2002: 193)
The derived nouns dzш33su33, ndo33su33, mu33su33 and Œo55su33 in (6) refer to the agent argument of the respective verbs, dzш33, ndo33, mu33 and Œo55. In other words, -su33 can derive a noun with both the action/state reading and the participant reading, and the proper interpretation of a deverbal noun with this nominalizing suffix, to a large extent, is determined by the utterance context. For example, mu33su33 can mean either ‘doer’ or ‘deed’, as shown in (6) and (7) in the following, respectively.
.â•… In the discussion of the Chinese versatile nominalizer de, Zhu (1983) divides nominalizers into two types: self-referring and transfer-referring. These types are also used in Chen and Wu (1998), and Hu (2002) in their discussions about the Nuosu Yi nominalizers. A selfreferring nominalizer nominalizes a verb or an adjective to denote an action designated by the verb or a state designated by the adjective, e.g. Žo21bo21su33 ‘labor’ in (5). A transfer-referring nominalizer nominalizes a verb or an adjective to denote a participant/argument of the verb or the adjective, e.g. dzш33su33 ‘eater’, ndo33su33 ‘drinker’, and mu33su33 ‘doer’ in (6). .â•… Examples are presented in the following way: the first line is the transcription of the target language using standard IPA symbols. In presenting the tones, we adopt the tone letters devised by Chao (1930), where digits indicate the pitch value on a five-point scale, with 5 being the highest. Thus 55, 33, and 21 represents a high level, a mid level, and low falling tone, respectively. The second line is word-to-word glossing. The third line is the English translation. For expository purposes, some times individual Yi words in the examples may be glossed in a separate line when they are highlighted in the discussion. The abbreviations are as follows: 1,2,3 sg pl acc agent asp comp
person cl singular conj plural cop accusative def agent marker dsm aspect complementizer loc
classifier conjunction copular verb definiteness marker direct quotative speech marker locative marker
linker neg nom nmz prt sfp top
linker negative nominative nominalizer particle sentence final particle topic marker
 Hongyong Liu & Yang Gu
(7)
hi21 su33 mu33su33 dŒ‹34 t.ho33 zo34. say-nmz do-nmz agree should ‘speaker’ ‘behavior’ ‘A speaker should agree with his deeds.’
Different from (6), where mu33su33 is interpreted as ‘doer’, in (7) it is interpreted as ‘behavior’ or ‘deed’. In the same vein, hi21su33 in (7) is naturally understood as ‘speaker’ instead of ‘speech’.5 The morpheme -su33 can also derive a nominal from an adjective. When this happens, the derived noun can only refer to the argument of the adjective, as shown in (8) and (9).6 (8) a. tsh‹33 a33 nâ•›i33 su33 kha33, a33 t.hu33 su33 a21 kha33. ˚ 3sg red-nmz want white-nmz neg-want ‘red thing(s)’ ‘white thing(s)’ ‘He wants the red one(s), and doesn’t want the white one(s).’ b. ŋa33 li33 a33 ‰‹55 su33 Ži33 hε33 vu33, a34 li33su33 Ži33 hε33 vu33. 1sg top new-nmz also like old-nmz also like ‘new thing(s)’ ‘old thing(s)’ ‘I like both the new and the old (things).’ (9) a.
kш33su33 b. ga55 su33 stupid-nmz rich-nmz ‘blockhead’ ‘rich person’
c.
khf33 su33 d. gu33 su33 brave-nmz firm-nmz ‘hero’ ‘firm person’
We already learned from 2.1. that the suffix -du33 can nominalize verbs. To derive a de-adjectival nominal from an adjective, Nuosu Yi also uses -du33. It is well known that adjectives are “intransitive” in the sense that when they are used predicatively, they predicate on one argument which often bears the semantic role of theme of the predicate. In (10), -du33 is suffixed to the adjectival root o33 bu33 ‘smart’, and the resultant de-adjectival noun, serving as the complement of the verb dŒo33 ‘have’, refers to the property “smartness” rather than a person who is smart.
.â•… We thank Suhua Hu for mentioning this to us. .â•… Foong Ha Yap (pers. comm.) pointed out that (8) and (9) illustrated a lexical semantic extension of the light noun su33 from ‘person’ to ‘thing’, from animate to inanimate. This idea is also reflected in the grammaticalization of hua ‘speech/words’ in the Mandarin topic marker -de hua, as discussed in Jiang (2004). We return to this point in 4.2.
Nominalization in Nuosu Yi 
(10) tsho33 tsh‹34 ma33 o33 bu33du33 a21 dŒo33. person this cl smart-nmz neg-have ‘smartness’ ‘This person is not smart.’ (Lit: ‘This person does not have smartness.’)
3.â•… Nominalization by the definiteness marker In Section 2, we have seen that the morpheme su33 functions as a deverbal/de-adjectival nominalizer. As a matter of fact, su33 is categorically ambiguous. Aside from being a derivational nominalizer, it can also be used as a syntactic nominalizer to nominalize a verb phrase or a clause. What is more interesting about this morpheme is that it is the definiteness marker in this language. In this section, we will first discuss the use of su33 as a definiteness marker, and then go on to examine the sentential nominalizer function of this morpheme. Chen (1989) points out that su33 can encode definiteness of a nominal, and can be treated as a definite article.7 (11) tsho33 o33 bu33 ma33 su33 la33 o34. person clever cl def come asp ‘The clever person came.’
Dai and Hu (1999) investigate all the syntactic distributions of su33 and claim that the morpheme is a structural particle which can either mark the definiteness of a nominal or help form a nominalized construction. Liu and Wu (2005) argue that su33 has developed from a light noun (meaning ‘person’) to a determiner (D), and further developed from a determiner to a complementizer (COMP), as is shown in the cline: Light noun> Determiner > Complementizer.8 To wit, what is assumed in the literature is that su33 can be used either as a determiner or a complementizer.
.â•… Although it has been generally recognized that classifier languages, especially those in the Sino-Tibetan language family, are articleless languages in that they typically do not possess definite or indefinite articles, a number of languages in this family are found to possess (in) definite markers, for instance, the two definite markers /le/ and /te/ and the indefinite marker /ke/ in Qiang as documented in LaPolla and Huang (2003), the indefinite marker /mi/ in Jingpo analyzed by Gu (2009), and the definiteness marker /wa/ in Numhupk Singpho discussed by Morey (this volume). In this paper we would treat such markers roughly as articles. .â•… One of the reviewers puts forward two possible ways of grammaticalization for our consideration: (1) two pathways, LN>NMZ and LN>DEF, which converge; (2) one pathway LN>DEF>NMZ. We have no diachronic data to confirm the second one. What can be assured is that the origin of su33 is a light noun (meaning ‘person’) and now it is recruited as both a nominalizer and a definiteness marker. In this sense, the reviewer’s first suggestion is more grounded.
 Hongyong Liu & Yang Gu
When su33 goes together with a classifier, the [N+CL+su33] sequence renders a definite reading, as exemplified in (12). (12) i33ti34 gu34 su33 Œo33 si44 la33. coat cl def catch take come ‘Bring the coat (to me).’
Without the morpheme su33, the [N+CL] sequence gives rise to an indefinite reading. (13) i33ti34 gu33 Œo33 si44 la33. coat cl catch take come ‘Bring a coat (to me).’
In (12) and (13), there is no cardinal number in the noun phrases; they are singular. Nominals with cardinal numerals are plural noun phrases and they show the same pattern with the above singular examples. (14) i33ti34 sf33 gu34 su33 Œo33 si44 la33. coat three cl def catch take come ‘Bring the three coats (to me).’ (15) i33ti34 sf33 gu33 Œo33 si44 la33. coat three cl catch take come ‘Bring three coats (to me).’
From (14) and (15) we can see that the presence of su33 yields a definite reading. We hence treat su33 on a par with definite articles. Note that su33 is not a demonstrative. Demonstratives occupy the syntactic slot preceding a classifier, but su33 follows a classifier. In Nuosu Yi there is no number differentiation with singular and plural nouns. Demonstratives inflect for number, but su33 does not. There is an obligatory requirement for su33 to co-occur with a classifier, resulting in the sequence of [classifier+su33]; i.e. su33 must co-occur with a classifier to determine the reference of a noun. In this respect, the definite article in classifier languages is different from that of non-classifier languages like English. In English, a definite article can directly go with a noun and thus determine the referent of the noun, for instance, the book, the coat, etc. In Nuosu Yi, such an expression is ungrammatical, as shown in (16)–(17).9 This could be due to the fact that su33 still retains traces of its lexical meaning.
.â•… The contrast in (16) and (17) indicates that su33 still retains traces of its lexical meaning, analogous to the Korean nominalizer kes (‘thing’), but distant from the Mandarin Chinese nominalizer de which is more semantically bleached and hence more grammaticalized (See the discussion of Simpson & Wu, 2001).
Nominalization in Nuosu Yi 
(16) *i33ti34 su33 Œo33 si44 la33. coat def catch take come Intended reading: ‘Bring the coat to me.’ (17) *Œi33 su33 Œo33 si44 la33. water def catch take come Intended reading: ‘Bring the water (to me).’
A definite nominal in Nuosu Yi may be in one of the following three forms: bare noun, N+Dem+CL, and N+CL+ su33. If a definite nominal phrase is not a simple bare noun, it may assume either of the following two ways to express definiteness. The first way is to resort to definite article su33, as in (18). The second way is to use a demonstrative, as in (19). Whenever a demonstrative is used, the definite article su33 cannot occur. This is borne out in (20). (18) vε34 vε33 a33 nâ•›i33 Ži21 pu33 su33 ˚ flower red two cl def ‘the two red flowers’ (19) vε34 vε33 a33 nâ•›i33 a33 dz‹34 Ži21 pu33 ˚ flower red those two cl ‘those two red flowers’ (20) *vε34 vε33 a33 nâ•›i33 a33 dz‹34 Ži21 pu33 su33 ˚ flower red those two cl def
These examples confirm that su33 functions to mark definiteness. In the following sections, we will examine the nominalization patterns brought forth by the morpheme su33, and aims to account for the linkage between the function of a definiteness marker and that of a nominalizer.
3.1â•… Gerundives As we have already mentioned, su33 is a polysemous morpheme. On the one hand, it can function as a definiteness marker on the right periphery of a nominal phrase. On the other hand, it can go with a verb phrase. When this happens, su33 turns the verb phrase into a nominal analogous to a gerundive nominal, as shown in (21). Being a gerund, the nominalized verb phrase can serve as an argument of the main verb. (21) i55dŒo33mu33 [[VP mu34 dш33 t‰a55 si21 Œi33 tshu33] su33] t.i33 lo34. quickly field occupy to house build nmz stop ‘Quickly stop (the action of) occupying the field to build houses.’ (Chen & Wu, 1998: 51) (22) ŋa33 [[VP vi55 ga33 gu55] su33] zo33 t.hi33. 1sg clothes sew nmz learn want ‘I want to learn sewing clothes.’
(Chen & Wu, 1998: 69)
 Hongyong Liu & Yang Gu
As indicated in the gloss, su33 in (21) nominalizes the verb phrase ‘occupy the field to build houses’. The adverb ‘quickly’ modifies the transitive main verb ‘stop’, which takes the nominalized verb phrase as its object. su33 in (22) nominalizes the verb phrase ‘sew clothes’, and the verb ‘learn’ takes the nominalized verb phrase as its object.
3.2â•… Relative clauses So far we have seen that su33 functions as a definiteness marker in the right periphery of a nominal phrase. In addition, it follows a VP to nominalize it, yielding a gerundlike nominal. Are these two functions independent of each other or are they two sides of the same coin? In this section, we seek an answer to this question by examining the role of su33 in relative clauses where these two functions are shown to converge. In a relative clause of Nuosu Yi, the relative head noun can bear the thematic role of agent, like a34Žε33 ‘cat’ in (23), or the thematic role of patient, like vo55 ‘pig’ in (24). In the absence of a relative head noun, the nominalizer su33 can also help mark both subject headless relatives in (25a) and object headless relatives in (25b).10 (23) [a34 Žε33 [a34 hε33 Œu33 a21 kш55 su33]] tsho55 Œ‹21 Œ‹33. ╛╛↜渀屮cat ╛╛↜渀屮mouse catch neg-can su people laugh ‘People laugh at the cats that cannot catch mice.’ (Chen & Wu, 1998: 64) (24) [vo55 [mu33 ka55 xo33 ta33 su33]] a55ko21 kш21 si55 dzш33 o34. ╛╛↜渀屮pig ╛╛↜渀屮Muga feed asp su Atgop agent kill eat asp ‘Atgop killed and ate the pigs that Muga fed.’ (25) a. [thш21 Œ‹33 pi33 su33] li33 thi55 Ži33. ╛╛↜渀屮book read su top here sit ‘The ones who study at school sit here.’
(Chen & Wu, 1998: 65)
b. ŋa33 li33 [tsh‹33 v‹33 su33] kha33. 1sg top ╛╛↜渀屮3sgNom buy su want ‘I want what he bought.’
(Chen & Wu, 1998: 69)
It seems that the function of su33 in Nuosu Yi relatives is not unitary. It can introduce a relative clause and the resultant relative construction is definite.11 Evidence of Nuosu
.â•… These are different from the agent nominalization and patient nominalization discussed in Section 2.1. The verb ‘read’ in (25a) takes the object ‘book(s)’, and the scope of the morpheme su33 ranges over the whole VP, thш21 Œ ‹³³pi³³ ‘read books’, instead of the verb Œ ‹³³pi³³ ‘read’ alone. Furthermore, the scope range of the nominalizer in (25b) is more obvious. If the scope of the nominalizer is just the verb v ‹³³ ‘buy’, then the proper nominalizer should be -du33, as seen in (1d). .â•… The referential property of su33 in Nuosu Yi is [+definite] rather than [+specific]. A specific indefinite nominal phrase in Nuosu Yi is normally marked by classifiers, as can be observed in the contrast seen earlier between (14) and (15).
Nominalization in Nuosu Yi 
Yi relative clause being definite, as in other languages, is that they cannot occur in existential constructions, as shown in (26). (26) *[vo55 [mu33ka55 xo33 ta33 su33]] dŒo33. ╛╛↜渀屮pig ╛╛↜渀屮Muga feed asp su exist ‘There exists the pigs that Muga fed.’
Note that in this regard, the Nuosu Yi relative construction offers an excellent example for the relativizer typology made in Ouhalla (2004). According to Ouhalla, there are two types of relativizers cross- linguistically: the Complementizer-type (C-type) relativizer and the Determiner-type (D-type) relativizer. Languages like English have the C-type relativizer in that the relativizer that is the same as the normal complementizer of sentential complementation as in ‘John said that he would leave early’. In Amharic, the relativizer is different from the complementizer of sentential complementation. It is the definite article, as illustrated in (27).12 (27) lGj-u yä-gäddälä-w Gbaab boy-the linker-killed-the snake ‘the snake the boy killed’
(Mullen 1986: 386; Ouhalla 2004: 294)
Amharic is a language with N-final relatives. In Amharic, the relative clause precedes the definite article (relativizer) -w, which itself precedes the relative head noun, yielding an order: RC>D>NP, as shown in (27). The definite article morpheme -w, introduces the relative clause ‘the boy killed’. At the same time, it marks the relative head noun as definite. Like Amharic, Nuosu Yi has the determiner-type relativizer su33, which is different from the normal sentential complementizer ko33, as highlighted in (28).13 (28) ŋa33 tha21 ko33 vo55 tsh‹34 ma33 Ži21 ha33 t.i21 dŒo33 o34. 1sg guess comp pig this cl two hundred half-kg have asp ‘I guess that this pig weighs two hundred kilograms.’ (Chen & Wu, 1998: 185)
Along the lines of Kayne’s (1994) and Ouhalla’s (2004) typological analysis of relative clauses, we propose that Nuosu Yi is a language having a D-type relativizer. The morpheme su33 evolves from a lexical noun, meaning ‘person’ to a light noun, playing the
.â•… The linker yä can also occur in possessive constructions as the possessive marker, as in (38) and (39). It reflects the genitivization and relativization syncretism observed in many Tibeto-Burman languages, as first described by Matisoff (1972) of the Lahu morpheme ve. .â•… For relevant discussion of the Nuosu Yi complementizer ko33, interested readers are referred to Liu and Gu (2008). The sentential complementizer ko33 is grammaticalized from a 3rd person pronoun. We will further examine this morpheme ko33 in detail in Section 4.
 Hongyong Liu & Yang Gu
role of a nominalizer, turning a relative clause into a nominal, and the nominalized relative clause stands in apposition with the relative head noun.14 This grammaticalization pathway of a light noun being recruited as a nominalizer is widely attested in Tibeto-Burman languages. What is special about Nuosu Yi relative construction is that the nominalizer is also responsible for marking the definiteness of the relative construction. Here we need to figure out the relationship between the nominalizer function and the definiteness marker function of the morpheme su33. One of the reviewers raises the question whether the morpheme su33, as a definiteness marker, is responsible for nominalizing the relative clause, or the nominalizer function is a prior function, which then developed the definiteness marker function. Before we answer this question, we summarize the functions and the corresponding syntactic distributions of su33 in the table below so as to have a more comprehensive picture. Table 1.╇ The function and distribution of the morpheme su33 Function
Nominalizer
Nominalizer & Definiteness marker
Definiteness marker
Syntactic Distribution Syntactic Structure
Gerundives
Relative Constructions
N+CL+ su33
[VP]- su33
NR.HEAD [[RC]-(CL) su33]
N [(Numeral)-CL su33]
We propose that the nominalizer function of the morpheme su33 is a prior function, which then developed the definiteness marker function. The relative construction is a bridging context linking the nominalizer function and the definiteness marker function. The emergence of the definiteness marker function in relative construction is due to the peculiar syntactic structure of Nuosu Yi genitive construction and appositive construction, a point which will become clearer in the following discussion. With respect to the relative construction in Nuosu Yi, it can be divided into the following three types according to the grammatical number feature and the referential property of the head nouns:
.╅ One of the reviewers raises the question of whether the clause is nominalized by the definite article, or the presence of the definite article is licensed because the clause is already nominalized. The Nuosu Yi data show that the relative clause is nominalized by the definite article, and no other nominalizer is found in the Nuosu Yi relative construction. So the definite article in Nuosu Yi relative construction plays two roles at the same time: as a definiteness marker as well as a nominalizer. This will be detailed as we proceed.
i.
Nominalization in Nuosu Yi 
RC + classifier
(29) vo55 mu33 ka55 xo33 ta33 ma33 a55 ko21 kш21 si55 dzш33 o34. pig Muga feed asp cl Atgop agent kill eat asp ‘Atgop killed and ate a pig that Muga fed.’
ii. RC + su33 (30) vo55 mu33 ka55 xo33 ta33 su33 a55 ko21 kш21 si55 dzш33 o34. pig Muga feed asp def Atgop agent kill eat asp ‘Atgop killed and ate the pig(s) that Muga fed.’
iii. RC + classifier+su33 (31) vo55 mu33 ka55 xo33 ta33 ma33 su33 a55 ko21 kш21 si55 dzш33 o34. pig Muga feed asp cl def Atgop agent kill eat asp ‘Atgop killed and ate the pig that Muga fed.’
The three types of relatives differ significantly in terms of referential properties of the relative head. By referential properties here we mean (in)definiteness and number features realized on the relatives. For the first type, as in (29), the relative head is indefinite and singular. For the second type, as in (30), the relative head is definite, singular or plural.15 For the third type, as in (31), the relative head is definite singular. These feature patterns can be summarized as follows: Table 2.╇ Referential property of the relative head nouns in three types of Nuosu Yi relatives Types
Definiteness
Number feature of the relative head
i. Relatives with classifier ii. Relatives with su33 iii. Relatives with classifier + su33
Indefinite Definite Definite
Singular Singular/Plural Singular
From Table 2, we can see that the major function of classifiers in relative clause is to mark the grammatical number feature,16 i.e. the presence of a classifier indicates singularity unless it is accompanied by a cardinal numeral larger than one, and the presence of su33 indicates definiteness, while the absence of su33 indicates indefiniteness.
.â•… The relative head noun in (30) can be singular only in the case where Muga fed only one pig. In this case, “all the pig that Muga fed” refers to the only one pig that Muga fed. What matters here is the key concept of “inclusiveness” associated with the semantics of the definite article. See Lyons (1999) for details on the semantics of definiteness. .â•… See Borer (2005) for a detailed discussion for the dividing function of classifiers, and how the dividing function is related to number marking in English and some other languages.
 Hongyong Liu & Yang Gu
In (30), since there is no classifier preceding su33, the default interpretation of vo55 is plural, unless it was the case that Muga only fed one pig. Notice that in a Nuosu Yi relative structure, a cardinal numeral cannot be added directly to the relative head noun to express plurality. As shown by the contrast in (34) and (35) below, the relative head noun in Nuosu Yi is always bare. This property of Nuosu Yi relatives stands in sharp contrast with that of its Mandarin Chinese counterparts. It has been long noticed that in Mandarin Chinese there exist two kinds of relatives, giving rise to two different word orders. According to Li & Thompson (1981: 124), the orders result from the following two schemas (32) and (33). The relative clause in (32) modifies the head noun book; but the relative clause in (33) modifies the demonstrative-Cl-N sequence. i. (associative phrase) + classifier/measure phrase + relative clause + (adjective) + noun (32) na ben [wo mai de] shu that cl ╛╛↜渀屮i buy de book ‘that book, which I bought’
ii. (associativephrase) + relative clause + classifier/measure phrase + (adjective) + noun (33) [wo mai de] na ben shu ╛╛↜渀屮I buy de that cl book ‘that book that I bought’
Unlike Mandarin Chinese, which has the [Dem-CL-RC-N] (32) and [RC-[Dem-CLN]] (33) pair, Nuosu Yi does not have the counterpart of (33), i.e. the relative head in the Nuosu Yi relative construction must be a bare noun, as is evidenced in the following contrast: (34) vo55 mu33ka55 xo33 ta33 so33 ma33 su33 a55ko21 kш21 si55 dzш33 o34. pig Muga feed asp three cl def Atgop agent kill eat asp ‘Atgop killed and ate the three pigs that Muga fed.’ (35) *[vo55 so33 ma33] mu33ka55 xo33 ta33 su33 a55ko21 kш21 si55 dzш33 o34. ╛╛↜渀屮pig three cl Muga feed asp def Atgop agent kill eat asp Intended meaning: ‘Atgop killed and ate the three pigs that Muga fed.’
In the following, we will offer a syntactic analysis of Nuosu Yi relatives and explain why the relative head in Nuosu Yi relative constructions can only be assumed by bare nouns. To account for this, we need refer back to Ouhalla’s (2004) analysis of the Amharic relative clause exemplified earlier in (27), and repeated below as (36). (36) lGj-u yä-gäddälä-w Gbaab boy-the linker-killed-the snake ‘the snake the boy killed’
(Mullen 1986: 386; Ouhalla 2004: 294)
Nominalization in Nuosu Yi 
Ouhalla proposes that (36) has the following internal structure: (37) lGj-u yä-gäddälä-w Gbaab [DP [DP [TP boy-the linker-killed]-the] snake]
The definite article -w is analyzed as the head D of the inner DP relative clause; it is not the head of the outer DP. Ouhalla argues for this position by comparing Amharic relative construction with Amharic possessive construction. The comparison between (38) and (39) shows that the definite article -u goes with the possessor phrase, rather than with the whole possessive construction, i.e. “the boy’s” instead of “the … notebook”. In other words, däbtär ‘notebook’ is not marked with a definite article. In (39), the possessor phrase is a proper name, which is inherently definite; therefore, the definite article -u does not show up after BGrhanu. If the definite article in (38) is marking the whole possessive construction, then the definite article should also occur in (39), since there is no semantic redundancy between the definite article and the common noun wGšša ‘dog’. (38) yä- lGj-u däbtär linker-boy-the notebook ‘the boy’s notebook’
(Mullen 1986, 307; Ouhalla 2004, 295)
(39) yä-bGrhanu wGšša linker-Birhanu dog ‘Birhanu’s dog’
(Ouhalla 2004, 295)
If the possessor NP in (38) is replaced with a relative clause, the structure of possessives and that of relatives become the same, as shown in (40). This supports the analysis given in (37), i.e. -w marks the DP relative clause.17 (40) a.
yä- lGj-u däbtär [DP [DP [NP linker-boy ]-the ] notebook ]
b. lGj-u ╛╛↜渀屮yä-gäddälä-w Gbaab [DP [DP [TP boy-the linker-killed] -the ] snake]
From (40) we can observe that the definite article -w as the head of the DP relative clause actually takes a tense phrase (TP) as its complement. This D-TP analysis for relative clauses is different from the standard analysis of English relative clauses which states that English relative clauses are introduced by the complementizer that or wh-words. The Amharic relative construction does not involve the grammatical
.â•… In Ouhalla’s analysis, the relativized head noun Gbaab ‘snake’ is raised out of the DP relative clause, a derivation that does not concern us here. Some functional accounts (e.g. Shibatani 2009) appeal to gapping (without movement) as a mechanism triggering nominalization.
 Hongyong Liu & Yang Gu
category of complementizer. Ouhalla (2004) predicts that if the relative clause in a language is a DP with the [D-TP] structure, the language lacks relative pronouns, and that the relativizer introducing relatives and the complementizer introducing sentential complements in this language must be two different morphemes. Here we need to ask what it means by saying that the definite article takes a TP as its complement. To our understanding, this process is, in effect, a process of nominalization. The determiner, a nominalizer in the first place, turns a clause into a nominal phrase. The resultant nominalized relative clause, by default, refers to the whole set of referents associated with the predicate in the relative clause, unless other means like plural morphology and classifiers are used to quantify the denotation set to clearly specify the number and referential features of the relative head noun. The default whole-set meaning meets the inclusiveness requirement of the definite article. The inclusiveness reading of the nominalized relative clause hence offers a possible link between a nominalizer and a definiteness marker, and this can be considered as the semantic trigger facilitating a nominalizer to further evolve into a determiner. In the following discussion, we will treat the morpheme su33 in Nuosu Yi relative constructions as a determiner. Notice this grammatical categorial label, by no means, denies its nominalizer function in the first place. Let us return to Nuosu Yi relatives. The pattern seen in (34) and (35) in fact leads us to the conclusion that the Nuosu Yi relative clause has a [D-TP] structure. It is not uncommon that languages may make use of aspect to anchor tense.18 Given that many Tibeto-Burman languages are rich in aspect (Dai 1990, 1998) while lacking tense morphology, we assume that the TP structure in Nuosu Yi takes an aspect head in the sense that tense is anchored by an aspect marker in a relative clause in this language. Nuosu Yi relatives in effect show a [D-AspP] structure, wherein the relativizer is the determiner su33. This determiner takes an aspect phrase as its complement, and consequently nominalizes the aspect phrase to form a DP relative clause, as in (41), analogous to the Amharic definite article -w which nominalizes a TP.
.â•… Recent research on tense and aspect indicates that languages without overt tense marking may rely on structural domains different from that of “tensed” languages for anchoring tense. In Smith (2007) and Rosen (2007), for instance, the authors argue independently that languages like Mandarin Chinese and Lolo, a Tibeto-Burman language, make use of view point aspect for tense interpretation. As argued by Rosen, when (speaker’s) point of view is made use of in structuring an event, e.g. telicising an event, the CP layer, i.e. the discourse domain is involved, in interpreting tense. This amounts to saying that grammatical aspect (in the sense of Comrie (1976) can anchor tense. Whether the aspect marker under discussion, e.g. ta³³ in (41) below, has actually evolved into a tense marker in Nuosu Yi is beyond the scope of this paper. We leave the issue here for future exploration.
Nominalization in Nuosu Yi 
(41) vo55 [DP [AspP mu33 ka55 xo33 ta33] [D su33 ]] pig Muga feed asp def ‘the pig(s) that Muga fed’
As reflected in (41), the determiner su33 heads the relative clause to its left.19 All we want to say at this juncture is that the D-TP analysis of relatives allows us to capture two important facts regarding the Nuosu Yi relatives. First, with some minor modification, we are able to work out the structural relation between the determiner head and its complement clause, i.e. the determiner su33 and the aspect phrase for the relative structure as seen in (41). Other technical details aside, this follows the word order of Nuosu Yi, which is a head final language.20 Second, and more importantly, it allows us to approach the question brought up earlier in this section in (35), repeated below in (42): why can’t a classifier appear in a relativized noun phrase in Nuosu Yi?. (42) *[vo55 so33 ma33] mu33ka55 xo33 ta33 su33 a55ko21 kш21 si55 dzш33 o34. ╛╛↜渀屮pig three cl Muga feed asp def Atgop agent kill eat asp Intended meaning: ‘Atgop killed and ate the three pigs that Muga fed.’
To arrive at an answer to this question, let us consider some related facts. In Nuosu Yi when two nominals are put together, there are actually two interpretations: possession and apposition, as in (43) and (44), respectively. (43) [mu33 ka55]i [ma55mo21]j/i* ╛╛↜渀屮Muga ╛╛↜渀屮teacher ‘Muga’s teacher(s)’ (44) [ma55mo21]i [mu33 ka55]i ╛╛↜渀屮teacher ╛╛↜渀屮Muga ‘Teacher Muga’
.â•… Foong Ha Yap (pers. comm.) pointed out that the definiteness interpretation may come from the appositive relationship of vo55 ‘pig’ and the nominalized clause, yielding a “headed” relative structure. Since nominalizer-turned-relativizer su33 is at the right-edge of the relative clause, it could be reinterpreted as a definiteness marker. .â•… Derivationally, what we have in mind is that su33 selects the aspect phrase following the [D-AspP] analysis. Due to the morphological property of su33as a clitic, it has to cliticize onto a phrasal constituent. This triggers its complement phrase, i.e. the aspect phrase, to move up to the specifier position of D to derive the surface word order of AspP-D, with the relativized head noun moving out of the relative clause, as in (41).
 Hongyong Liu & Yang Gu
These two structures can be differentiated as follows. In (43), the two nominals denote two different referents, so the possibility of co-indexation is eliminated. In (44), co-indexation of the two nominals is possible; therefore, the two nominals denote the same referent. Notice that in each of these two examples, one of the nominals is a bare noun. The possession or apposition reading is determined by the interpretation of the common noun ma55mo21 ‘teacher’: in (43) it is interpreted as somebody other than Muga, and in (44) it is interpreted the same as Muga. Hence the common noun has either different or the same reference with regard to the proper name it co-occurs with: different reference yields a possessive relation, whereas co-reference an appositive relation. This kind of reference is made possible because bare nouns in classifier languages are argumental (Chierchia 1998) in the sense that they have the same semantic function, i.e. the same referential function to be used as arguments. Following this line of thinking, let us refer back to (41), repeated below in (45). The part of the nominalized relative construction in (45) could have the same reference as the relative head noun on its left: (45) vo55 [DP [AspP mu33ka55 xo33 ta33 ] [D su33 ]] pig Muga fed asp def ‘the pig(s) that Muga fed’
We therefore conclude that the two nominals, vo55 and the entire DP, mu33ka55 xo33 ta33 su33, are in an apposition relation, analogous to what we have seen in (44). This offers a natural explanation to the problem regarding (42) where a classifier cannot appear in a relativized noun phrase. The relevant part of data is repeated in (46): (46) *[vo55 so33 ma33] mu33 ka55 xo33 ta33 su33 ╛╛↜渀屮pig three cl Muga feed asp def
The problem in (46) arises from the conflict of reference of the two nominals in the relative construction. The relativized nominal phrase vo55 so33 ma33 contains a string of [N+Num+CL], which yields an indefinite reading ‘three heads (of) pigs’. The relative clause, on the other hand, is headed by su33, hence a DP yielding a definite reading. The reference of these two nominals is essentially in conflict. As a result, an apposition relation between the two cannot be established. We are then left with the option of a possession relation, with vo55 so33 ma33 ‘three heads (of) pigs’ being the possessor, a situation seen in (43). But the interpretation arising from such a relation is nonsensical, i.e. “three pigs’ what Muga fed (pigs)”. Consequently, neither option is available for (46). Our analysis predicts that a classifier phrase can occur within a relative clause only as in (47). (47) vo55 [mu33 ka55 xo33 ta33 so33 ma33 su33] pig ╛╛↜渀屮Muga feed asp three cl def ‘the three pigs that Muga fed’
Nominalization in Nuosu Yi 
The nominal structure in (47) has an appositive interpretation, something that is difficult to get in English, “(the) pigs, the three (pigs) that Muga fed”. This eventually explains the contrast between (34) and (35).
3.3â•… Stand-alone nominalization/focus clause Apart from its occurrence in the relative construction, su33 can also occur in the sentence final position, sometimes followed by a copular verb, as in (48). (48) Œo33 tш21 la33 o34 su33 (ŋш33). sheep alive come asp su cop ‘(It is the case that) the sheep came alive.’
This instance of su33 has been analyzed as a mood marker, whose function is to add more assertive emphasis to a declarative clause (Dai & Hu 1999; Kokado 2000). Kokado (2000) argues that su33 marks the speaker’s comment on a realis event. In other words, su33 has something to do with past tense, perfective aspect, or stance marking. Given our discussion of su33 as a nominalizer in 3.2, we continue to pursue that the sentence final su33 has a role to play in nominalization. Following Watters (2008) and LaPolla (2006), a nominalized clause can convey focused (new as opposed to presupposed) information, and it can be predicated of by a copula. In languages with a tense system, the copular is needed as it realizes the tense morphology. In tenseless languages, the copular could be phonetically null, due to the absence of the tense morphology. As we have pointed out in 3.2, Nuosu Yi is a language which to a large extent anchors tense on aspect, so it is plausible to assume that the sentence final copula is susceptible to phonological reduction due to its insignificance in realizing tense. With these lines in mind, we can readily explain why su33 in (48) cannot be a stance marker as proposed in Dai and Hu (1999), because it is not simply adding more assertive force to a declarative clause. The following contrast shows that the two sentences have different information structure; the first one has a focus structure, highlighting a fact or a state of affairs, whereas the second one is a declarative structure which depicts a dynamic event. (49) Œo33 tш21 la33 o34 su33. sheep alive come asp su ‘(It is the case that) the sheep came alive.’ (50) Œo33 tш21 la33 o34. sheep alive come asp ‘The sheep came alive.’
 Hongyong Liu & Yang Gu
In other words, (48) and (49) pattern alike.21 They have a cleft-like structure, with the copula or its phonetically null counterpart as the predicate taking a nominal complement which is marked by su33. (50), on the other hand, does not have a cleft structure.22 There is a further piece of evidence in support of our view that the sentence final 33 su is not a mood marker nor an assertive stance marker, contra Dai and Hu (1999), Li and Ma (1981a & b), and Kokado (2000). These authors hold the view that the sentence final su33 is a stance marker, which can be used to show speaker’s assertive comment on a realis situation. However, we find data which do not show the speaker’s assertive force. As shown in (51), su33 is not confined to declarative clauses; it is also found in interrogative sentences such as the following.23 (51) A: nш33 khш21thш33 na33 su33? you when ill su ‘When did you fall ill?’ B: i21Ži21 mu33 thi33 ko33 na33 su33. today morning in ill su ‘I fell ill this morning.’
More accurately, (51) should have the interpretation: “A: When was it that you fell ill?” and “B: It was this morning that I fell ill”. (51) clearly shows that part A is not
.â•… One of the reviewers points out to us that not all stand-alone nominalization constructions serve cleaving functions or have strong focus interpretations, and that, quoting Matisoff (1972), even though we can understand stand-alone nominalization constructions as being cleft, this does not mean that there is actually a higher clause or a cleft structure. In other words, nominals can stand alone, but predicates cannot stand alone, as they must go with arguments. This is why there is no 0-place predicate. At this stage, we are not entirely clear why Tibeto-Burman languages allow stand-alone nominalization constructions, and the real function of the nominalizer as well as the optional copula call for further research. One thing plausible is that we are experiencing some kind of structural changing process. See also our discussion surrounding (52) in the following. .â•… Notice without the sentence-final particle su³³, the copula verb cannot be added to a declarative sentence. This indicates that there is a selection requirement between su³³ and the copula ŋш³³. In other words, ŋш³³ must take a complement marked by su³³. For us, this simply means ŋш³³ takes a nominalized complement to form a cleft construction, a common focus structure cross-linguistically.
Œo33 t 21 la33 o34 ŋ sheep alive come asp be
33.
m
m
(i)
.â•… The example was cited in Kokado (2000: 26), originally from Li and Ma (1981a). More confirmation data are desired to determine whether the sentence final su³³ in utterance A of (51) carries any prosodic cues which encode speaker stance.
Nominalization in Nuosu Yi 
an assertion. Even if it is the case that the nominalizer carries a certain stance of the speaker, it remains to be worked out what the stance is.24 Based on these observations, we conclude that the sentence final su33 is not as simple as an assertive stance marker or a mood marker.25 In the present case, it consistently functions to nominalize a clause. With this in mind, we can have at least two syntactic analyses for (48), as represented below in (52) and (53), respectively.26 For simplification, we will assume that the copula is phonetically null as indicated by the parentheses. Note that although being phonetically null, the copula is by no means non-existent, a point which will become clearer as we proceed. Here we differ from Matisoff (1972) and Watters (2008) in that they do not posit a null copula in the stand-alone nominalization construction. One of the reviewers mentions that the stand-alone nominalization construction is somehow syntactically incomplete, but can be licensed by marked prosody. As far as Nuosu Yi is concerned, no such marked prosody is observed. Our Nuosu Yi consultants consistently tell us that there is a null copula in the sentence final position. (52) [DP Œo33] [DP [AspP tш21 la33 o34] su33] (ŋш33). sheep alive come asp su cop ‘The sheep is what came alive.’ (53) [[AspP[DP Œo33] tш21 la33 o34 ] su33] (ŋш33). sheep alive come asp su cop ‘It is the case that the sheep came alive.’
As understood, the bare nominal Œo33 ‘sheep’ in (52) can be interpreted as a definite nominal. The rest of the sentence is a nominalized relative clause, and more specifically,
.â•… Suhua Hu (pers. comm.) pointed out to us interrogative sentences in Nuosu Yi do not in general end with su³³. But in Kokado (2000), quite a number of instances of the sentence final su³³ were found in interrogative sentences. .â•… Here we are viewing the function of the sentence final particle from a purely synchronic perspective. From a diachronic perspective, there may exist a link between the nominalizer function and the function of a pragmatic stance marker. See Yap, Matthews and Horie (2004) for discussion of the grammaticalization process from a nominalizer to a stance marker found in a number of East Asian languages. .â•… There exists a semantic/pragmatic difference between the equative copula construction analysis and the sentential focus construction analysis, represented by the syntactic structures in (52) and (53). The difference is due to the different functions of the copula verb. It needs more diachronic research to confirm the hypothesis that the nominalizer in stand-alone nominalization construction is grammaticalized from the relativizer-type nominalizer, although it is not difficult to observe a highly plausible structural reanalysis shown in (52) and (53) from a purely synchronic perspective.
 Hongyong Liu & Yang Gu
it is a headless relative clause. su33 takes the headless relative clause on its left as its complement. The structure in (52) is actually a familiar one, similar to what we have seen in (45): apposition with two juxtaposed nominals denoting the same referent. The crucial difference between (52) and (45) is the optional use of the copula verb in (52). Here the copula equates a bare definite noun with a nominalized headless relative. The whole construction yields an equative copular structure. Alternatively, (48) may have a structure as in (53), where the semantic subject 33 Œo and the rest of the clause are not in appositive relation, but in subject-predicate relation, and su33 nominalizes the whole clause. Nominalization by su33 leads to a sentential-cleft construction. The nominalized proposition serves as an argument of the copula verb, yielding the meaning as in (53). To sum up, the morpheme su33 has dual functions, a nominalizer and a definiteness marker (i.e. determiner.). These two functions seem to conflate in the relative constructions we have examined so far. For the sentence-final su33, we have shown in (52) that a headless relative clause formed by su33 is used as the second nominal in an equative construction, and we have also shown in (53) that a nominalized clause formed by su33 is used as the complement of the copula verb forming a cleft construction. The morpheme su33 nevertheless plays the same role of being a nominalizer in (52) and (53), although the nominalized clauses in (52) and (53) have different syntactic functions.
4.â•… Nominalization by the complementizer ko33 There is yet another morpheme ko33 in Nuosu Yi identified by Chen and Wu (1998) as a nominalizer. As a lexeme, ko33 is a third person singular accusative pronoun. It also functions to nominalize clauses. Unlike the D(eterminer)-type nominalizer su33, ko33 is a C(omplementizer)-type nominalizer. In terms of grammatical functions, ko33 is in complementary distribution with su33. For example, ko33 cannot occur in relative constructions, while su33 can; ko33 can introduce a factual complementation clause, while su33 introduces an eventive complement clause; ko33 can turn a proposition into a topic clause or a conditional clause, while su33 occurs in a focus construction. The functions of ko33 are summarized as follows: (54)
i. ii. iii. iv. v.
locative/directive morpheme third person singular accusative pronoun factual complementizer topic clause marker conditional clause marker
Cross-linguistically, it is not uncommon for locative/directional morphemes to develop into demonstratives and person pronouns, as there is a strong tendency for linguistic
Nominalization in Nuosu Yi 
forms expressing deictic spatial notions to be grammaticalized into terms expressing locative concepts and/or references with cardinal directions (Heine & Kuteva 2001). In this section, we will focus on the functions of ko33 listed in (iii)–(v) of (54), and demonstrate how they can be subsumed under the function of sentential nominalizer.
4.1â•… Verb complementation clauses with ko33 A factual clause can be introduced by ko33. As shown in the following, the verb tha21 ‘guess’ in (55) and the verb hш21 ‘look’ in (56) take their respective complement clauses introduced by ko33. ko33 is most frequently found immediately following speech verbs such as hi21 ‘say’, as shown in (57). (55) ŋa33 tha21 ko33 vo55 tsh‹34 ma33 Ži21 ha33 t.i21 dŒo33 o34. 1sg guess ko pig this cl two hundred half-kilo have asp ‘I guess that this pig weighs two hundred kilograms.’ (Chen & Wu, 1998: 185) (56) ŋa33 hш21 ko33 a33 nâ•›a55mu33 hш21 a21 sa55. ˚ 1sg look ko more look neg good ‘I think that it does not look good.’ (57) su33 hi21 ko33 nш34 Ži33 ko33 t.ho33 di34. others say ko 2sg also 3sg.acc join dsm ‘Others said that you had also participated in it.’
(Chen & Wu, 1998: 186)
(Hu, 2002: 246)
The sentence final morpheme di34 in (57) is a speech marker, indicating the preceding information up to the point of ko33 following the speech verb hi21 is a quoted (direct) speech. Given the fact that Nuosu Yi is a strict SOV language, the word order in (57) and all other similar examples is unusual, as it shows an SVO order. While the force triggering the emergence of this SVO word order is still unknown, what is known is that ko33 frequently occurs after speech verbs and verbs of cognition and perception, such as say, think, guess, and feel which take factual complements.27 This suggests that the functions of ko33 as conditional clause marker and topic clause marker as listed
.╅ Although the puzzling SVO word order in Nuosu Yi calls for an explanation, it is clear that Nuosu Yi has been largely influenced by its contact with Mandarin Chinese. Most of Nuosu Yi people are bilinguals of Nuosu Yi and the Southwestern Mandarin Dialect. More evidence for the language contact comes from the fact that Tibeto-Burman languages are typical ergative languages, but Nuosu Yi does not show an ergative case-marking pattern. In Nuosu Yi, there are no case markers. However, there is a third person pronoun ko33, which can only be used as objects, but never subjects; i.e. the 3rd person pronoun object ko33 of a transitive verb can never be the subject of an intransitive verb. This is not in conformity with the ergative case-marking pattern commonly found in Tibeto-Burman languages, showing that Nuosu Yi is not an ergative language. As pointed out by one of the reviewers, the fact that Nuosu Yi is not an ergative language may be due to language contact with Chinese.
 Hongyong Liu & Yang Gu
in (54) may stem from its ability to introduce factual complementation. The semantic core of these three functions, i.e. factual, conditional and topical, is made possible by nominalization, since a factual complementizer can pack a proposition into an information unit and enable the packed information to be used as arguments of verbs of perception and cognition. Nominalization makes a sentential topic transparent, and conditional clauses are, in essence, sentential topics (Haiman 1978). The difference between a sentential topic and a conditional clause is that the latter has an unrealized condition as the topic. Hence in essence, they share the same core function of nominalization. In the following sections, we will examine the functions of the morpheme ko33 as a nominalizer in more detail.
4.2â•… Nominalization of clauses with ko33 In Nuosu Yi, a proposition can be nominalized by ko33, and the resultant nominalized clause can be a topic, as exemplified in (58)–(60).28 (58) dza34 ŋgш33 a34 Œi33 t‰a33 ko33 xi55dŒu21 sf33 Œε55 dŒo33. rice chew child feed ko drawbacks three kind have ‘As for feeding the children with chewed rice, there are three kinds of drawbacks.’ (Chen & Wu, 1998: 64) (59) tsh‹21 mu33 ko33 t‰ш55. in such a way do ko viable ‘Acting in this way is viable.’
(Hu, 2002: 205)
(60) tsh‹33 bu33ma33zo33 ko33 hε33 tsu55 dŒ‹33 hε33 tsu55. 3sg study ko absorbed very absorbed ‘With regard to his studying, he is very absorbed.’
ko33
(Hu, 2002: 203)
marker.29
Sometimes, is seen to co-occur with a topic According to Hu (2002), 33 ko is a temporal adverbial particle, but (61) suggests the actual temporal adverbial clause marker is ko33 plus the topic marker nш33. (61) a.
mu33 thi33 la33 ko33 nш33, i55Ži33 ku33 o34. sky bright come when fall sleep asp ‘When it was dawn (the sky becomes bright), he fell asleep.’
.â•… See Jiang (2004) on the Mandarin Chinese -de hua for relevant discussion about this view. .â•… Foong Ha Yap (pers. comm.) pointed out that the presence of the topic marker nш³³ after ko33 strongly suggests the function of ko33 as a sentential nominalizer. She also drew our attention to the possibility that some kind of “repackaging of information” in the sense of Givón (2009) may have caused the third person pronoun ko33 to get reinterpreted as a head-initial complementizer seen earlier in (55)–(57), and a head-final complementizer in (58)–(60). To us, more specifically, the so-called “repackaging of information” results in structural reanalysis and structural change as extensively discussed in Roberts and Roussou (2003).
b. i55 tш21 la33 ko33nш33, ma55 So21 o34. get up come when afternoon asp ‘When he got up, it was afternoon.’
Nominalization in Nuosu Yi 
(Chen & Wu, 1998: 265)
The compounded ko33 and nш33 can also introduce conditional clauses, similar to the English when used as if in some cases, the Japanese tara used as ‘when’ and ‘if ’, and the German wenn in both temporal adverbial clauses and conditional clauses.30 (62) ma33 ha33 a21 dŒi21 ko33 nш33, ŋa33 a21 nd�‹33mu33 bo33. rain neg-fall ko top 1sg surely go ‘When/If it doesn’t rain, I will surely go.’ (Chen & Wu, 1998: 212)
Our hunch is that the function of ko33 nш33 as a conditional clause marker may be derived from the nominalization function of ko33. In Nuosu Yi, a verb phrase cannot be used directly as a topic. Hu (2004) observes that a verb phrase can be a topic only if it is nominalized or reduplicated.31 The topic marker nш33 can be optional, because topic markers in many languages can be omitted. When nш33 is omitted, a nominalized clause without the topic marker is still interpretable as a conditional clause, as shown in (63).32 (63) no21 nd�a55mu33 Žo21a21bo21 ko33, dzш33du33 kha55 bi55 la33? 2pl hard neg-work ko food where from come ‘If you don’t work hard, how can you have food?’ (Hu, 2002: 206)
At this juncture, let us digress on a similar phenomenon in Mandarin Chinese. Interestingly, the conditional topic marker in (68) is -de hua, the first of which is the versatile nominalizer -de, and the second element is a noun hua meaning ‘word/speech’. (64) ni bu nuli gongzuo de hua, zenme hui chenggong? you not hard work de speech how will succeed ‘If you don’t work hard, how can you succeed?’
Jiang (2004) offers a detailed analysis for the emergence of this topic marker in Mandarin Chinese.33 Based on diachronic data, she argues that -de hua originates from the structure of shuo VP de hua (‘say VP de words’). VP de hua form an appositive
.â•… See Haiman (1978) for a detailed discussion of conditions and topics. .â•… Reduplication is a common morphological process in Tibeto-Burman languages. One of its functions is to make verbal categories into topics, and the essence of such reduplication is nominalization. See for reference, Dai and Gu’s (2003) analysis on reduplication in topicalization of Jingpo. .â•… See Dai and Gu (2003: 58) for discussion of a similar phenomenon in Jingpo. .â•… The same pattern is shown with Malay cause/reason subordinator punya pasal ‘because’ (See Yap, this volume).
 Hongyong Liu & Yang Gu
relative relation, with VP-de and hua standing in an apposition, and hua being the relative head. The relative head noun hua becomes light due to semantic bleaching, and gradually fades out from the appositive relative construction. During the bleaching process, the head noun hua absorbs or pied-pipes onto the nominalizer (relativizer) de. This offers an opportunity for the two elements belonging to different syntactic constituents to merge together, resulting in the compounded de hua. The most crucial step for de hua to become a topic marker is to get free from being controlled by the verb shuo ‘say’. This is made possible by topicalizing the constituent VP de hua to the clause initial position. The stranded verb shuo ‘say’ then becomes elidable as well. This account in effect supports the development or structural change of an appositive clause to a conditional clause. There are strong parallels between Jiang’s analysis for the Mandarin Chinese -de hua and the compounding of nominalizer ko33 and topic marker nш33 in Nuosu Yi. Notice the most frequent context where ko33 occurs is the factual complementation structure, as seen in (57), repeated here as (65).34 (65) su33 hi21 ko³³ [nш34 Ži33 ko33 t.ho33] di34. others say comp(3sgACC) ╛╛↜渀屮2sg too 3sg.acc join dsm ‘Others said that you had also participated in it.’ (Hu, 2002: 246)
The sentence in (65) can be seen as a case where the factive complementizer, which emerges from a pronoun ko33, is reanalyzed as a head-initial complementizer. It stands in apposition to the speech clause to refer to the factual complement. Notice ko33 is a clause-initial complementizer in (65), but a clause-final complementizer in the part of the conditional clause in (63). This word order contrast can be accounted for if we turn to the parallelism between the Mandarin Chinese topic marker -de hua and the Nuosu Yi topic marker ko33nш33. It is plausible that the bracketed speech complement in (65) escapes from the control of the ‘say’ verb to the clause initial position via topicalization. The speech marker di34 then automatically drops out, since the topicalized clause now loses its identity as a speech. The stranded subject and verb su33 hi21 ‘others say’ become elidable as well. All these lead to structural reanalysis. In the process of restructuring, the third person singular pronoun stands out as a good candidate for the nominalizing function, since Nuosu Yi does not allow a VP to be used as a topic. The changes are illustrated in (66):
.â•… According to Chen and Wu (1998: 97), the citation tone of the 2sg pronoun is tone 33, just as the topic marker nш³³. The 34 tone of the 2sg pronoun in (65) is caused by a phonological rule which regulates two adjacent 33 tones will result in a 34–33 tone sequence, with the first 33 level tone changing to a 34 sandhi tone. It remains unclear whether the topic marker in Nuosu Yi is grammaticalized from the 2sg pronoun, and surely more research is needed here.
(66)
Nominalization in Nuosu Yi 
a.
su³³ hi²¹ ko³³ [ … … ] di³4 others say (3sgACC) ‘Others said that ….’
b.
[ … … ] su³³ hi²¹ ko³³ (3sgACC)
di³³
As ko33 can nominalize a clause in the topic position, it merges with the topic marker nш33. Together they mark conditional clauses, and further mark temporal adverbial clauses as seen in (61) and (62). Different from the grammaticalization of su33 from a general light noun meaning ‘person’ to a nominalizer/relativizer, and further to a definiteness marker, in which process the notion of definiteness has come to play a central role, the grammaticalization of ko33 from a third person accusative pronoun to a nominalizer is mediated by an apposition construction involving speech/cognition contexts.
5.â•… Conclusion This paper discusses the various forms and uses of nominalization in Nuosu Yi at both morphological and syntactic levels. Our analysis shows that when su33 marks a relative clause, the relative clause has the status of a DP. When it occurs in the sentence final position, it still plays a nominalization role, turning the preceding verb phrases or clauses into complex nominal structures, yielding equative copula constructions or sentential cleft constructions. The genitive construction in the form of “NP+NP” and the obligatory phrase final position of the D-type relativizer su33 account for the absence of classifiers in relative head noun phrases, and the absence of relative stacking in Nuosu Yi. Empirical data from Nuosu Yi confirm the typological generalization proposed by Ouhalla (2004) that in some languages (e.g. Hebrew and English) relatives are introduced by complementizers, while in some other languages (e.g. Amharic and Arabic), relatives are introduced by determiners. Mechanisms of syntactic reanalysis for the new grammatical categories are explored with the morpheme ko33 which suggests a grammaticalization path from a third person accusative pronoun to a nominalizer. Since ko33 is a deictic pronoun in origin and the most typical contexts in which this pronoun is used are that of speech and cognition, it is easy to establish an apposition relation between this pronoun and the complement of speech/cognition verbs, and then the pronoun becomes semantically light, losing its person and case features, and develops into a nominalizer to demarcate the complement of speech/cognition verbs as a nominal argument.
 Hongyong Liu & Yang Gu
Both su33 and ko33 develop into a nominalizer via the bridging context of apposition. However, there exists a sharp contrast between them, as shown in (67). (67) a. np+[(clause)-su33] b. vspeech/cognition+ko33+[direct speech clause/…] c. [topic clause] +ko33
(clause-initial c) (clause-final c)
In (67a), the NP stands in apposition with the following nominalized clause [(clause)-su33], whereas in (67b) the third person accusative pronoun ko33 stands in apposition with the following direct speech clause if the verb in the construction is a verb of speech. This difference may be attributable to their lexical origin. The nominalizer su33 has a light noun origin; therefore, it is more likely to appear in relative construction, serving as an operator, binding the variable of the gap-argument in the relative clause.35 The nominalizer ko33 has a deictic pronoun origin; therefore, it is more likely to appear in complementation construction, on the one hand, serving as the complement of the verb, and on the other hand, having a high chance of becoming a complementizer, which changes a quotative direct speech clause into a reportative indirect speech clause, if the verb involved is a verb of speech. The clause-final complementizer ko33 is shown likely to have derived from the topicalization of a complement clause.
References Borer, Hagit. 2005. In Name Only. Oxford: OUP. Chen, Kang & Wu, Da. 1998. A Grammar of Nuosu Yi. Beijing: Central University for Nationalities Press. Chen, Shilin. 1989. Specificity and definiteness in Nuosu Yi. Minzu Yuwen (Ethnic Linguistics) 2: 21–28. Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language Semantics 6: 339–405. Comrie, Bernard. 1976. Aspect. Cambridge: CUP. Comrie, Bernard & Thompson, Sandra A. 1985. Lexical nominalization. In Language Typology and Syntactic Description, Vol. III: Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 349–398. Cambridge: CUP.
.â•… It is interesting to note that like the lexical meaning of the nominalizer su33 in Nuosu Yi, the relativizer/nominalizer in many other Tibeto-Burman languages has the lexical origin of the light noun meaning ‘person’. In English, the relative pronoun who in “John saw the person who he admired very much”, which serves as the wh-operator binding the gap argument, is a combination of the sense of “wh-” and “person”.
Nominalization in Nuosu Yi 
Dai, Qingxia. 1990. Studies on Tibeto-Burman Languages, I. Kunming: Yunnan Nationality Press. Dai, Qingxia. 1998. Studies on Tibeto-Burman Languages, II. Kunming: Yunnan Nationality Press. Dai, Qingxia & Hu, Suhua. 1999. On the structural particle su33 in Nuosu Yi. Zhongguo Yuyanxue de Xin Tuozhan (Recent Advances in Chinese Linguistics), Feng Shi & Wuyun Pan (eds), 205–215. Hong Kong: City University of Hong Kong Press. Dai, Qingxia & Gu, Yang. 2003. Remarks on the grammatical category of topics: A morphological analysis of topics in Jingpo. In Modern Linguistic Theories and Studies on Minority Languages in China, Qingxia Dai & Yang Gu (eds), 50–69. Beijing: Nationality Press. Givón, Talmy. 2009. The Genesis of Syntactic Complexity. Amsterdam: John Benjamins. Gu, Yang. 2009. langai, mi yu Jingpo yu shuminci jiegou zai xi (langai, mi and the nominal structure of Jingpo revisited). Yuyan Kexue (Linguistic Sciences) 8: 225–238. Haiman, John. 1978. Conditionals are topics. Language 54: 564–589. Heine, Bernd & Kuteva, Tania. 2002. On the evolution of grammatical forms. In The Transition to Language, Alison Wray (ed.), 376–97. Oxford: OUP. Hu, Suhua. 2002. Research on Yi Structural Particles. Beijing: Central University for Nationalities Press. Hu, Suhua. 2004. The topic construction in Liangshan Yi. Ethnic Linguistics (Minzu Yuwen) 3: 9–15. Huang, C.-T. James. 1982. Logical Relations in Chinese and the Theory of Grammar. Ph.D. dissertation, MIT. Huang, Shizhe. 2007. Property theory, adjectives, and modification in Chinese. Journal of East Asian Linguistics 15: 343–369. Jiang, Lansheng. 2004. On the lexicalization of the cross-structural combination “de hua”. Zhongguo Yuwen 5: 387–400. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kokado, Norio. 2000. The function of mood particle su33 in Liangshan Yi. Journal of Southwest University for Nationalities 21: 22–26. LaPolla, Randy J. 2006. Nominalization in Rawang. Paper presented at 39th International Conference on Sino-Tibetan Languages and Linguistics. University of Washington. LaPolla, Randy J. & Huang, Chenglong. 2003. A Grammar of Qiang with Annotated Texts and Glossary. Germany: Mouton de Gruyter Press. Li, Charles N. & Thompson, Sandra A. 1981. Mandarin Chinese – A Functional Reference Grammar. Berkeley CA: University of California Press. Li, Ming & Ma, Ming. 1981a. 600 Conversational Sentences of Nuosu Yi. Chengdu: Sichuan Nationality Press. Li, Ming & Ma, Ming. 1981b. A Grammar of Nuosu Yi (Liangshan Yiyu Yufa). Chengdu: Sichuan Nationality Press. Liu, Hongyong & Wu, Da. 2005. On the structure of YI “N+Num+Cl+ su33”. In Research on Classifiers in Sino-Tibetan Languages, Jinfang Li & Suhua Hu (eds), 83–101. Beijing: Central University for Nationalities Press. Liu, Hongyong & Gu, Yang. 2008. On the quotative speech marker and evidential marker in Nuosu Yi. Minzu Yuwen (Ethnic Linguistics) 2: 16–23. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Morey, Stephen. This volume. Nominalization in Numhpuk Singhpo. In Yap, Grunow-Hårsta & Wrona (eds).
 Hongyong Liu & Yang Gu Mullen, Dana. 1986. Issues in the Morphology and Phonology of Amharic: The Lexical Generation of Pronominal Clitics. Doctoral dissertation, University of Ottawa. Ouhalla, Jamal. 2004. Semitic relatives. Linguistic Inquiry 35: 288–300. Roberts, Ian G. & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Rosen, Sara Thomas. 2007. Structured events, structured discourse. In The Oxford Handbook of Linguistic Interfaces, Gillian Ramchand & Charles Reiss (eds), 181–268. Oxford: OUP. Simpson, Andrew & Wu, Xiu-Zhi Zoe. 2001. The grammaticalization of formal nouns and nominalizers in Chinese, Japanese and Korean. Language Change in East Asia, Thomas E. McCauley (ed.), 250–283. Richmond: Curzon. Smith, Carol. 2007. Reference time without tense. In Recent Advances in the Syntax and Semantics of Tense, Aspect and Modality, Louis de Saussure, Jacques Moeschler & Genoveva Puskás (eds), 229–250. Berlin: Mouton de Gruyter. Watters, David E. 2006. Nominalization in Himalayish languages. Ms, RCLT. Yap, Foong Ha, Mathews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline: The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Zhu, Dexi. 1983. Zi zhi he zhuanzhi (Self-referring and transferring). Fangyan 3: 81–100.
Finite structures from clausal nominalization in Tibeto-Burman Scott DeLancey
University of Oregon Nominalization has long been recognized as one of the driving processes of Tibeto-Burman syntax and syntactic change. A pervasive and recurrent process in the family is the replacement of old finite clause structures with new constructions based on nominalizations. TB languages repeatedly innovate new, marked clausal constructions with a nominalized verb and finite copula. The process is exemplified through case studies from Tibetan, Sunwar (from the Kiranti branch of TB), and the Kuki-Chin branch. Frequently such constructions lose their marked status and become the ordinary finite construction, resulting in the creation of new verbal categories and systems. Many TB verbal systems transparently reflect this origin, for example Modern Tibetan tense/aspect forms like -pa yin, -pa red, both consisting of the nominalizer -pa in construction with an equational copula. As such a construction becomes more opaque, a language may be left with a semantically empty final particle, reflecting an erstwhile nominalizer or copula. The prevalence of such final particles in the family is evidence of the pervasiveness of the phenomenon which is described in this paper.
1.â•… Introduction Tibeto-Burman languages illustrate a typological syndrome found in a number of families throughout the world, characterized by verb-final constituent patterns and syntax, which is substantially centered on verb serialization (Matisoff 1969; DeLancey 1991) and nominalization (Matisoff 1972). The use of nominalized clause constructions to form relative clauses has been a topic of considerable interest to students of Tibeto-Burman syntax (Matisoff 1972; Mazaudon 1978; DeLancey 1999, 2005; Genetti 1992a; Genetti et al. 2008; Noonan 1997, 2008, inter alia). In this paper, I will discuss another manifestation of the importance of nominalization in Tibeto-Burman syntax, which is the regular development of new finite structures from clausal nominalization constructions. We will see that clausal nominalization in construction with a copula is, indeed, a major – source of new finite clause constructions throughout the family.
 Scott DeLancey
2.â•… The Phenomenon The frequent phenomenon of nominal morphosyntax associated with the finite verb has been noted since the earliest days of Tibeto-Burman studies. Konow (1908, 1911) and other early investigators were insistent that the verbs of Tibeto-Burman were not truly verbs in the sense found in Indo-European, but essentially nominal: The different classes of words are not clearly distinguished, and many instances occur in which a word can be used at will as a noun, as an adjective or as a verb. The verb can, on the whole, be described as a noun of action, and we find phrases such as “my going is” instead of “I go.” (Konow 1911: 929)
Konow was referring in particular to phenomena like the use of possessive proclitics as verb agreement in Kuki-Chin, which I will discuss below. As we will see, the stigmata of nouniness which these scholars perceived in the morphosyntax of Tibeto-Burman verbs reflect the fact that in many Tibeto-Burman languages the finite construction of the verb reflects an earlier construction in which the sentence or verb phrase is nominalized. The construction often includes a copula, of which the nominalized sentence is then an argument, but the copula may be dropped over time, and sometimes nominalizations are simply used as finite clauses. The relationship between nominalization and finiteness in Tibeto-Burman was re-introduced to the modern field by Matisoff ’s seminal article (1972), in which he points out that the Lahu nominalizer ve occurs not only in familiar nominalization contexts, but marking relative clauses, and sometimes even occuring in non-subordinated main clauses. This phenomenon is very widespread in Tibeto-Burman; in fact I would impressionistically claim that languages which show clear examples of a nominalized clause, with or without a final copula, functioning as a main clause are far more common than those (if in fact any could be found) which do not. To begin with a concrete example, most of the verb forms in the modern Tibetan dialects (for the standard Lhasa dialect see, e.g. DeLancey 1991 or Denwood 1999) obviously originated as NOMZ + COP constructions (see Saxena 1992, 1997). The general structure of a finite verb construction in modern Tibetan is the lexical verb with a nominalizer and a copula, as in these Lhasa Tibetan examples in Table 1:1
.╅ These are presented in transliterated orthographic form; in speech -pa yin, in particular, is subject to substantial phonological reduction, to the point where a non-literate speaker may not recognize the etymology. Educated speakers, however, are quite able to explain the structures of the forms, which are a recent development in the language, as we will discuss below.
Finite structures from clausal nominalization in Tibeto-Burman 
Table 1.╇ Examples of the general structure of a finite verb construction in modern Tibetan
nga nga nga nga nga
Stem
Nom
Copula
zas za za za za
-pa -gi -gi -rgyu -mkhan
yin yod yin yin yin
‘I ate.’ ‘I am eating.’ ‘I will eat.’ ‘I still have to eat.’ ‘I am going to/intend to eat.’
The various nominalizers contribute their own semantics to the tense/aspect/ modality constructions formed with them. The nominalizer -pa is the oldest Bodish, and Tibeto-Burman, nominalizer, and is associated with perfectivity in most of Bodish (DeLancey 2005). The non-perfective gi/kyi and rgyu are apparently related. Kyi does not occur anywhere outside of the tense aspect paradigm, but rgyu is used productively as a nominalizer for irrealis event predications. Finally, mkhan is originally an agentive nominalizer (DeLancey 1999), so that the last sentence could be literally (or, at least, etymologically) parsed as ‘I am an eater’. This construction – a nominalized clause finitized by a copula (which in many languages may be optional or even absent altogether) – is one of the two basic sources of clausal structure in Tibeto-Burman languages. I will refer to it as a nominalized clause construction; its schematic manifestation in a verb-final language is STEMNOMINALIZER (COPULA). Several scholars have discussed the common Tiberto-Burman phenomenon in which a formally nominalized clause is used as a main clause, i.e. where a construction which is formally non-finite has acquired a finite function (see especially Bickel 1999). I will be discussing primarily examples of grammaticalization of the construction with the copula. Such a construction is, of course, already finite, and it originates as a means of expressing a marked pragmatic or semantic category, such as aspect, evidentiality, register or style, or strength of assertion. My purpose, however, is not to examine the functions of such constructions in the beginning of their careers, but to document how, over time, they work themselves into the core finite system of the language, often eventually displacing and replacing older finite constructions. Obviously the development of new finite constructions from simple nominalizations, with no copula, represents the same diachronic phenomenon.
3.â•… Three case studies In this section, we will examine three examples of the development of finite structures from nominalized clause constructions. Two recently-innovated finite con-
 Scott DeLancey
structions from the Western Kiranti language Sunwar show how easily a nominalized clause construction can find a place in the finite tense/aspect system of a TibetoBurman language. And the history of the modern Kuki-Chin finite clause gives an illustration of how easily a nominalized clause construction can completely replace earlier finite morphosyntax.
3.1â•… Sunwar: Nominalization in action I will present here a set of data originally reported in DeLancey(1992b), which show that Sunwar turns to nominalized clause constructions repeatedly as it innovates new tense/aspect/evidentiality categories.
3.1.1â•… A recently-innovated finite construction The Sunwar copula system has four etyma: locative/existentials tsha and ‘baak, and an equational na, with a suppletive allomorph ho. The equational na is the oldest of all, with a robust Tibeto-Burman etymology (Matisoff 2003). Tsha and ho, on the other hand, are transparent borrowings from Nepali. The last existential, ‘baak, has the semantics of a recently-grammaticalized locative copula, in that it still occurs with the lexical sense ‘stay, live at’. Moreover, it has no evident copular cognates elsewhere in Bodic or Tibeto-Burman. Sunwar’s nearest relatives, Hayu and Bahing, each has only a single copula, which is clearly cognate to Sunwar na. Thus its copula function appears to be a relatively recent development. The Sunwar verb has three paradigms, illustrated with 1st person singular forms of the verb la ‘go’ in Table 2: Table 2.╇ Paradigms of the Sunwar verb “go” Tense/aspect
Sunwar verb “go”
simple past non-past
la-ŋa là-ti laî-nu-ŋ
Note that, with a vowel stem verb like la, the non-past has an expanded stem form, adding an off-glide /i/ and a falling tone, which elsewhere in the language occurs only when an underlying coda consonant is deleted. Consonant-final verb stems are identical in all three conjugations. The simple conjugation occurs in negated verbs, and a few other constructions. Both this and the past paradigm are widely shared throughout Kiranti, and reconstruct for Proto-Kiranti. The third paradigm, the non-past, is a very new development, not shared even with Bahing or Hayu, Sunwar’s closest Northwest Kiranti kin. Its status
Finite structures from clausal nominalization in Tibeto-Burman 
as a recent innovation is equally evident on language-internal grounds, as we will see directly. The non-past paradigms of the Sunwar non-past verb endings are given in Table 3: Table 3.╇ Non-past paradigms of the Sunwar non-past verb endings Singular
1st 2nd 3rd
-nuŋ -neye -ba
Dual
Plural
Inclusive
Exclusive
Inclusive
Exclusive
-nase -nisi -nise
-nasku -nisi -nise
-nai -nini -nim
-niki -nini -nim
And Table 4 shows the conjugation of the equational copula na: Table 4.╇ Conjugation of the equational copula na Singular
1st 2nd 3rd
na-ŋ ‘na-ye ho
Dual
Plural
Inclusive
Exclusive
Inclusive
Exclusive
na-se na-si na-se
na-sku na-si na-se
na-i na-ni na-m
na-iki na-ni na-m
The correspondence of these two paradigms is evident on simple inspection; the non-past paradigm is manifestly a morphologization of the na paradigm. Note that the non-past paradigm, like that of na, has an intrusive form in the 3rd singular (originally the Bodic nominalizer *ba), which must reflect the same gap in the original paradigm as the borrowed form in the copular paradigm. This leaves the problem of the non-past stem, which, given the transparency of the suffixes, must likewise have a fairly shallow analysis. Indeed, the alternation which we see between la- and laî- has an exact parallel in the behavior of verbs with an underlying coronal stop final (Genetti 1992b; DeLancey 1992b). Compare the conjugated forms of ta- ‘see’ and sad- ‘kill’ in Table 5: Table 5.╇ Conjugated forms of ta- ‘see’ and sad- ‘kill’
Imperative Past (1sg → 3sg) Past (3sg → 1sg) Non-Past (1sg → 3sg)
ta- ‘see’
sad- ‘kill’
tau taa-ta-ŋ ta-ti taî-nu-ŋ
sadu saî-ta-ŋ saî-ti saî-nu-ŋ
 Scott DeLancey
The behavior of ta in the non-past would be explained if there were an underlying coronal obstruent coda *-t or *-s in the non-past stem. Thus, by quite shallow internal reconstruction, we can reconstruct the original form of the non-past construction as *STEM-t na-AGR, which instantiates our STEM-NOMINALIZER COPULA formula.
3.1.2â•… New perfects I have noted above the evidence that ‘baak is a recent innovation as a copula. The Nepali borrowing tsha is clearly a very recent adoption into Sunwar, so fresh that it retains its full Nepali inflectional paradigm. The contrast between ‘baak and tsha is epistemological, marking a distinction between what I have elsewhere (DeLancey 1986, 1990) called “old” and “new” information, a peculiar Tibeto-Burman manifestation of the wider mirative category (DeLancey 1992a, 1997). Tsha represents the unmarked category of information, and ‘baak is mirative, representing information which the speaker would have no way of knowing save through information channels to the outside world. However, for our present purposes, the main point about this distinction is that it is an innovation in the language, so I will not go into its semantic content in detail. For further discussion, see DeLancey (1990, 1992a, b, 1997, 2001). Some versions of this distinction are quite widespread in the Himalayas and adjoining areas, but there is no basis for reconstructing it to any great time depth. It is widespread in Bodic, except where it appears to have been borrowed directly (as from Central Tibetan into Cuona Monpa). It is also expressed in the various languages in which it occurs by forms which clearly are not cognate, and sometimes not even homologous (in that, for example, it is expressed in Kathmandu Newar by verb endings, in Tibetan dialects by copular distinctions). Thus the category as such doesn’t seem to be reconstructable to Proto-Bodic or even to the level of Proto-Tibetan (DeLancey 1992a). So when we find it expressed in Sunwar by an opposition between two copulas both of which are evidently new developments in the language, we can infer that the category itself is a relatively new introduction into Sunwar, representing that language’s somewhat belated entry into the mini-Sprachbund, which is characterized by the “egophoric” or “conjunct/disjunct” pattern (DeLancey 1992a). Again, this inference is supported by the lack of any such category in Hayu or Bahing. Sunwar, as a Tibeto-Burman language, has a predilection for nominalized constructions of various kinds. Thus, in typical Tibeto-Burman fashion, the existence of this semantic distinction in the copula automatically leads to its incorporation into the verb system. All three copulas occur with various non-finite forms of the verb in constructions which form a significant part of a complex system of tense/aspect/evidential categories. These include innovative constructions consisting of a clause marked with
Finite structures from clausal nominalization in Tibeto-Burman 
the nominalizer -šo2 in constructions with ‘baak or tsha, which bring their evidential/ mirative senses with them. In example (2), we get the clearly mirative sense of ‘baak, indicating direct visual evidence of a fact, which otherwise would have been unknown to the speaker. This is contrasted with the unmarked semantics of tsha in example (1): (1) kyaršε ‘saî-šo tshaa goat kill-nomz exist/3s/nonpast ‘He is killing a goat/goats.’ (2) kyaršε ‘saî-šo ‘baâ-ta goat kill-nomz exist-3spast ‘[I saw] he was killing a goat/goats.’ [example: when I discovered him]
Borchers (2008) describes a variety of Sunwar in which we can already see further grammaticalization of this last construction. Her data include the mirative construction (which she calls “expression of unexpected action”) in (2), and also examples where ‘baak is suffixed directly to the verb stem in (3): (3) goi das ghanta hir-baa-ti-ni you(polite) ten hour walk-unex.-past-2pl ‘You walked for ten hours?!’
She also reports the use of bare -šo-nominalized clauses as finite main clauses. In analyzing these constructions in Sunwar, I have consistently argued that they have to be quite recent developments: on comparative grounds, they must postdate the divergence of Sunwar and Bahing, which is a very shallow relationship; on internal grounds, the transparency of the relationship between the non-final paradigm and the copula, and the corresponding regularity of the paradigm (in contrast to the opacity and irregularity of the older past paradigm) show that the non-past construction is a recent innovation, while the perfect constructions can hardly be more than a few centuries old. The nominalized clause construction is not necessarily as fertile in other Tibeto-Burman languages as it seems to be in Sunwar, but it is productive everywhere.
3.2â•… Kuki-Chin: Reconstructing a nominalized clause construction A somewhat older nominalized clause construction is illustrated by the Kuki-Chin languages, a branch of Tibeto-Burman spoken in western Myanmar and Northeast India. The Kuki-Chin verbal system is an anomaly within Tibeto-Burman. In general, Tibeto-Burman languages tend toward suffixing rather than prefixing morphology. Suffixal verb agreement is found in a substantial number of languages, and can be
.â•… For the etymology of this form see Noonan 2008: 227–8.
 Scott DeLancey
reconstructed for Proto-Tibeto-Burman (Bauman 1975; DeLancey 1989; van Driem 1993; inter alia). Some languages have agreement systems with only suffixes, others include a few prefixes in the paradigm. The Kuki-Chin languages are unique in the family in having innovated a new, exclusively prefixal agreement paradigm. The history of this development is written in the structure of the system. The paradigm is evidently a recent development. In at least some languages of the branch, the agreement morphemes are not fully morphologized, but remain phonologically somewhat independent of the verb, so that many authors describe them as clitics rather than prefixes. The first piece of evidence for their history is the fact that the clitic forms used for verb agreement are also the possessive proclitics used with nouns. Note the uses of the 1st person proclitic ka3 in the following examples from the Northern Chin language Sizang (Stern 1963): (4) (ké:i) ká pài: hî: (I) 1st go FINAL ‘I go/went.’ (5) kâ mei: 1st tail ‘my tail’ (6) kà káp hî: 1st cry.I FINAL ‘I cry’ (7) kà ká: nà: 1st cry.II nomz ‘my crying’
The second piece of evidence is the sentence-final particle hi, which occurs in the verbal examples. This is homophonous with the equational copula. Both can occur together in (8): (8) ama khua:-pui:-te a hi: hi: he villager 3sg be FINAL ‘He is a villager.’
Notice that, in this example, the first occurrence of the morph is preceded by a proclitic agreeing with the 3rd person subject, and is thus behaving as a verb. Thus, on grounds of morphology and syntactic distribution, the copula and the final particle
.╅ The tonal alternation in the proclitic is phonologically conditioned by the tone of the following verb or noun stem.
Finite structures from clausal nominalization in Tibeto-Burman 
are synchronically distinct. Nevertheless, the diachronic derivation of the latter from the former can hardly be doubted. We can see the system at work in (9) (Stern 1984): (9) a.
na-lá:i hong thák ka-ngá: a: 2-letter cis send 1-receive NONFINAL ‘I having received your letter which [you] sent to me …’
b. k-ong thûk kí:k lâ-lê:u hî: 1st-cis reply again once.more FINAL ‘I in turn reply to you.’
Here we see the proclitics functioning both as possessives (na-lá:i ‘your letter’) and as agreement markers (ka-ngá: ‘I receive’, k-ong thûk ‘I reply [to you]’). We can also see the synchronic function of the final particle: it contrasts with non-final markers like the a: which ends the first clause, and marks the end of a clause chain. In TibetoBurman languages without a final particle construction this function is marked by a finite verb. These two facts together indicate that the synchronic finite construction of the Kuki-Chin verb is a reanalysis of a former nominalization construction. The use of possessive clitics with a verb stem implies nominal status for that stem. If we hypothesize that, in the structure ká pài: hî:, the verb stem was historically nominalized, and the structure reflects an original *‘It is my going’, we neatly explain why the agreement proclitic and the final particle are homophonous with, respectively, the possessive and the copula. The original construction must have been a nominalized clause construction with a nominalized verb stem, its subject expressed as a grammatical possessor, and the copula as the highest verb. A further fact about the Northern Chin languages lets us fill in the story. In most of Kuki-Chin, the new finite construction with proclitic agreement and a copular sentence final particle has completely replaced the older, inherited Tibeto-Burman finite verb. In Northern Chin, however, some of the old suffixed forms remain as an alternate finite paradigm, so that the verb may have either agreement proclitics or agreement suffixes, but not both, as in (9) and (10): (10) pài: ke-ŋ go neg-1sg ‘I don’t go.’ (11) ká pài: kei hî: 1st go neg FINAL ‘idem.’
As in the nearby Jinghpaw-Konyak languages (DeLancey 2010, 2011), in Northern Chin, the agreement suffixes do not attach directly to the lexical verb stem, but to a set of highly grammaticalized elements derived from old auxiliary verbs.
 Scott DeLancey
The suffixes, 1st person -ŋ and 2nd -tεô, are both well-attested in other branches of the family, and clearly represent inheritance from Proto-Tibeto-Burman (Bauman 1975; DeLancey 1989). Thus we have to reconstruct two distinct main clause constructions for Proto-Kuki-Chin: the older finite form, where the end of the sentence is marked by an auxiliary conjugated with agreement suffixes, and the newer, originally nominalized construction, where the lexical verb looks like a noun possessed by its subject, and the end of the sentence is marked by the grammaticalized copula. We have descriptions of the alternation for only two Northern Chin languages, Sizang (Stern 1963) and Tiddim (Henderson 1957, 1965). In both languages, the use of the nominalized construction is more formal, and the older suffixed forms informal. So, pending data from other members of the branch, we can tentatively reconstruct a Proto-Kuki-Chin sociolinguistic pattern in which the old, inherited finite construction is used in everyday informal situations, and the innovative, nominalized structure is used for more formal or public styles. (Henderson notes that in Tiddim the nominalized structure is used in storytelling and narrative). In Northern Chin the alternation remains; in the other branches of Kuki-Chin, the more formal construction has completely replaced the original. Proto-Kuki-Chin, then, was a language like those discussed by Noonan (1997, 2008) and Bickel (1999), in which a nominalized construction is used instead of the ordinary finite construction for some marked purpose. This interpretation is strengthened by the fact that the suffixes are in complementary distribution not only with the proclitics, but with the entire construction – the final particle occurs only on clauses with proclitic agreement, not with the suffixal paradigm. The available reports suggest that the function of the alternation between the old finite and the nominalization clause constructions may have been stylistic, with nominalization associated with more formal style.
4.â•… The evidence of final particles As we have noted, the occurrence of a sentence-final particle in declarative sentences is a peculiarity of many Tibeto-Burman languages. In languages with such a particle, it marks the end of a clause chain; in languages which lack this construction, such as modern Central Tibetan, the final verb in a clause chain requires a finite verb (which may itself be a nominalized clause construction). I cannot provide etymologies for all of the final declarative particles in the family, but where their history has been traced, they derive from old nominalizers or copulas, and represent the final stage of grammaticalization of a nominalized clause construction. We will look at only two examples here, from Classical Tibetan and Lolo-Burmese, but similar data can be found in many other branches of the family as well.
Finite structures from clausal nominalization in Tibeto-Burman 
4.1â•… Classical Tibetan Classical Tibetan has a quite different system, with a grammaticalized sentence-final particle ‘o, which geminates a preceding final consonant, as in (12): (12) nam.mkha′-la ‘phur te ‘gro’ o sky╅╅╅╇ loc fly NONFINAL go FINAL ‘[He] went flying (around) in the sky. (13) bya de-s srog thar-pa-r byas so bird dem-erg life save-nomz-loc did FINAL ‘The bird acted to save his life.’ (14) khyim.bdag-gi bu-s ‘di skad ces gsol to4 householder-gen boy-erg this word thus ask FINAL ‘The householder’s son made this request.’
If we hypothesize that declarative final particles in general originate in the nominalizer or the copula of a nominalized clause construction, as in Kuki-Chin, then it is significant that equational sentences frequently occur with no copula, i.e. that the final particle can serve duty as an equational copula (examples from Hahn 1974: 39): (15) bram.ze de dbul.po zhig go Brahmin that pauper a FINAL ‘That Brahmin was a pauper.’ (16) khyim de chen.po-’o house that bigâ•…â•…â•›FINAL ‘That house is big.’
By the time of its attestation in Classical Tibetan, ‘o is nothing but a final particle; there is no homophonous but distinct copula as there is in Kuki-Chin. It probably reflects a widespread Proto-Tibeto-Burman copular root *way which we will see below turning up in other branches. As far as I know, this etymon has no other reflex in modern varieties of Tibetan, in all of which the oldest copulas are forms of Classical Tibetan yin and yod. However, it is well attested elsewhere in the Bodish branch, and Bodic subbranch, to which Tibetan belongs. In Kurtöp, representing the East Bodish languages which are the closest cousins of Tibetan, the copulas are na and wen, the latter < *wayna (DeLancey 2008). We find the root as a copula in West Himalayan, the closest to Tibetan of the non-Bodic branches, as in Rangpo hwә- (Zoller 1983: 68), and in Kiranti, representing East Himalayan, as in Limbu existential wa· (van Driem 1987).
.╅ The initial t in to in ex. (14) is a fossil of a lost perfective suffix on the verb (Przyluski & Lalou 1933).
 Scott DeLancey
Since *way then clearly reconstructs for each of the ancestors of Tibetan, from the nearest node all the way back, its loss in Tibetan calls for some explanation; the explanation is that it persisted into Classical Tibetan as the sentence final particle, and subsequently abandoned its function as a copula. The idea of obligatory final particles which do nothing but mark the end of a sentence may seem a bit redundant to speakers of European languages. But they actually do have a function in Tibeto-Burman languages, all of which have clause-chaining structure. A literary example is (17): (17) a.
ltad.mo lta-ba-╅╇ ‘i lam-du rgyal.po-‘i bu.mo dang phrad nas sights see-nomz-gen way-loc king-gen daughter with meet abl ‘on the way to see the sights [he] met the king’s daughter.’
b. rgyal.po-‘i bu.mo la chags te king-gen daughter loc desire NONFINAL ‘desiring the king’s daughter.’ c.
bu.mo dang nyal nas daughter with lie.down abl ‘[he] slept with the girl.’
d. phyi gsang shor te after secret come.out NONFINAL ‘later the secret getting out.’ e.
rgyal.po-s zin nas king-erg seize abl ‘the king seized him.’
f.
gsod-du bcug-pa las kill-dat/loc command-nomz abl ‘and as he ordered him to be killed.’
g. khyim.bdag gi bu-s ‘di skad ces gsol to householder gen boy-erg this word thus ask FINAL ‘the householder’s son asked thus.’
The non-final constructions are the equivalent of conjunctive participles in Indic languages; the to at the end of this sequence of clauses effectively marks the end of an episode. This sort of thing is normal in Tibeto-Burman narrative, where the break – the end of the clause chain – is marked by a finite verb. When a language innovates a new finite construction out of a nominalized clause construction, the resulting sentence final particle takes over the functions of the older finite construction, especially that of marking the end of a clause chain.
4.2â•… Lolo-Burmese The first serious study of a final particle construction in Tibeto-Burman is Matisoff ’s (1972) description of ve in the Loloish language Lahu. Matisoff (1985) identifies this,
Finite structures from clausal nominalization in Tibeto-Burman 
as well as the Jinghpaw final particle ai, and others, as a reflex of the same ProtoTiberto-Burman copula *way, which we identified in the previous section as the source of the Classical Tibetan final particle. In fact, this particle is found throughout Lolo-Burmese. Bradley (1979: 376–7) independently reconstructs a declarative verb particle *way3 for Proto-Loloish; this is obviously the same etymon. It is reflected in Burmish as well as Loloish, in the “mark of assertion” /ε/ which Vittrant (2002) shows can be internally reconstructed from the final particles of Burmese. Burmese has two “modal” final particles, realis te and irrealis me. There are also two clausal nominalizers, realis ta and irrealis hma, which mark many complement constructions, “cleft” constructions, and, most interestingly, exclamatory independent main clauses. These have exactly the same modal values as te and me. Nominalizers which encode aspectual or realis/irrealis distinctions are not uncommon in Tibeto-Burman (DeLancey 2005; Noonan 2008). So, if we interpret Vittrant’s /ε/ as analogous (even if not cognate, although it would certainly appear to be) to Proto-Loloish *way3, we have here once again a synchronic system of finite main clause marking which derives directly from reanalysis of a nominalized clause construction, with nominalizers ta and hma plus a sentence-final copula.5
5.â•… Summary Previous work on the topic of nominalized clause constructions in Tibeto-Burman has dealt substantially with synchronic issues surrounding the use of nominalized clauses in finite contexts. My concern here has been diachronic, and to some degree typological. We have examined a number of examples of finite constructions in various Tibeto-Burman languages which have finite verb forms or clause constructions which demonstrably developed from nominalized clause constructions. I have not given a survey of languages across the family, but anyone familiar with any Tibeto-Burman language can provide analogous examples. The Sunwar perfects have numerous parallels throughout Bodic, the best known being the Tibetan constructions exemplified by the Lhasa data discussed at the beginning of this paper. The Central Tibetan system exemplified there is no more than a few centuries old: in the vernacular 15th century biography of Milarepa we find the
.â•… Simpson (2008) proposes a quite different history for these forms, in which the nominalizers -ta and -hma derive from the fusion of a construction in which the genitive-marked final particles te. and me. modified haa ‘one’ in a “headless” relative clause, i.e. represent an innovative nominalization construction. I find Vittrant’s account more convincing. In any case, Simpson agrees that the modern finite constructions in Burmese represent old nominalizations.
 Scott DeLancey
nominalized clause constructions which make up the modern system, but they do not appear to be fully grammaticalized, and have not yet completely replaced the older system (Saxena 1992, 1997). The older Classical system itself appears to be an old nominalized clause construction, so that the contemporary Tibetan languages are now on their second round of reinventing their finite clause structure. Parallel developments can also be found in the Western Himalayan languages (Saxena 1992, 2000), Tshangla (Andvik 1999), and elsewhere throughout the family. The innovation of new agreement prefixes from possessive proclitics on nominalized verbs, as we saw in Kuki-Chin, is not quite so common, but it has parallels in Kham (Watters 2002: 411ff.) and elsewhere. The final particle phenomenon which we saw in Kuki-Chin and Classical Tibetan is particularly important; if these are fossils of old nominalized clause constructions, then we have prima facie evidence for the ubiquity of this grammatical development throughout the family. We saw that Proto-Lolo-Burmese had a nominalized clause construction which already was used as a main clause, and which is the source of the primary finite main clause type in the daughter languages. So we see again, as in Kuki-Chin, the finitization of a nominalized clause construction as the basis for the syntactic organization of an entire branch of the family. The same may well be true of the Naga and other less well-documented groups, but it is not possible to make definitive statements without a more secure sub-grouping of these languages than we have at present. We can infer that Proto-Kuki-Chin, Proto-Lolo-Burmese, very likely ProtoBodic, and, for that matter, no doubt Proto-Tibeto-Burman itself, were typical TibetoBurman languages, like the modern languages discussed by Bickel, Genetti, Noonan, and others, in which one or more nominalized clause constructions coexisted with an older finite system. A nominalized clause construction with the copula *way is very widely represented, and very likely existed in Proto-Tibeto-Burman – though at that level it presumably would have had the transparency of the modern Sunwar perfects (or, at least, of the slightly older Central Tibetan forms), and the *way would still have been a copula, which function has been largely lost in modern reflexes of the root such as Classical Tibetan ‘o, Jinghpaw ai, and Lolo-Burmese *way3. It is also possible that the proclitic construction in Kuki-Chin is ultimately rooted in a nominalization construction of Proto-Tibeto-Burman provenience (DeLancey 1988; Watters 2002: 406), although if so it must have undergone considerable alteration and regularization in Proto-Kuki-Chin. But we cannot simply explain the proclivities explored in this paper by reconstructing a particular source construction to ProtoTibeto-Burman. My main point in this paper is not comparative, but typological – it is not simply the case that many Tibeto-Burman languages have undergone such a development, as documented in our examination of Kuki-Chin, Classical Tibetan, and Lolo-Burmese. The striking fact is that they continue to do so, over and over as
Finite structures from clausal nominalization in Tibeto-Burman 
we have seen in Sunwar and modern Tibetan. This is a consistent, repeated pattern across the family, synchronically and diachronically, and serves as another example of how nominalization is the primary engine driving Tibeto-Burman syntax and syntactic change.
Abbreviations abl cis dat erg gen loc neg nomz unex
ablative cislocative dative ergative genitive locative negative nominalizer unexpected
References Andvik, Erik. 1999. A Grammar of Tshangla. Leiden: Brill. Bauman, James. 1975. Pronouns and Pronominal Morphology in Tibeto-Burman. Ph.D. dissertation, University of California, Berkeley. Bickel, Balthasar. 1999. Nominalization and focus in some Kiranti languages. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 271–296. Kathmandu: Royal Nepal Academy. Borchers, Dörte. 2008. A Grammar of Sunwar. Leiden: Brill. Bradley, David. 1979. Proto-Loloish [Scandinavian Institute of Asian Studies Monograph 39]. London: Curzon. DeLancey, Scott. 1986. Evidentiality and volitionality in Tibetan. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe & Johanna Nichols (eds), 203–213. Norwood NJ: Ablex. DeLancey, Scott. 1988. On the origins of the Kuki-Chin agreement prefixes. 21st International Conference on Sino-Tibetan Languages and Linguistics, Lund, October 5–9. DeLancey, Scott. 1989. Verb agreement in Proto-Tibeto-Burman. Bulletin of the School of Oriental and African Studies 52: 315–333. DeLancey, Scott. 1990. Ergativity and the cognitive model of event structure in Lhasa Tibetan. Cognitive Linguistics 1: 289–321. DeLancey, Scott. 1991. The origin of verb serialization in modern Tibetan. Studies in Language 15: 1–23. DeLancey, Scott. 1992a. The historical status of the conjunct/disjunct pattern in Tibeto-Burman. Acta Linguistica Hafniensia 25: 39–62.
 Scott DeLancey DeLancey, Scott. 1992b. Sunwar copulas. Linguistics of the Tibeto-Burman Area 15(1): 31–38. DeLancey, Scott. 1997. Mirativity: The grammatical marking of unexpected information. Linguistic Typology 1(1): 33–52. DeLancey, Scott. 1999. Relativization in Tibetan. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren W. Glover (eds), 231–249. Kathmandu: Royal Nepal Academy. DeLancey, Scott. 2001. The mirative and evidentiality. Journal of Pragmatics 33(3): 371–384. DeLancey, Scott. 2005. Relativization and nominalization in Bodic. Tibeto-Burman Linguistics: Proceedings of the 28th Annual Meeting of the Berkeley Linguistics Society, 5–72. Berkeley CA: Berkeley Linguistics Society. DeLancey, Scott. 2008. Kurtoep and Tibetan. In Chomolangma, Demawend und Kasbek: Festschrift für Roland Bielmeier zu seinem 65. Geburtstag, Brigitte Huber, Marianne Volkart, & Paul Widmer (eds), 29–38. Halle: International Institute for Tibetan and Buddhist Studies GmbH. DeLancey, Scott. 2010. Towards a history of verb agreement in Tibeto-Burman. Himalayan Linguistics 9.1: 1–38. DeLancey, Scott. 2011. Nocte and Jinghpaw: Morphological correspondences. In North East Indian Linguistics, Volume 3, Gwendolyn Hyslop, Stephen Morey & Mark Post (eds), 61–75. New Delhi: Cambridge University Press India. Denwood, Philip. 1999. Tibetan [London Oriental and African Language Library 3]. Amsterdam: John Benjamins. van Driem, George. 1987. A Grammar of Limbu. Berlin: Mouton de Gruyter. van Driem, George. 1993. The Proto-Tibeto-Burman verbal agreement system. Bulletin of the School of Oriental and African Studies 16(2): 292–334. Genetti, Carol. 1992a. Semantic and grammatical categories of relative clause morphology in the languages of Nepal. Studies in Language 16: 405–427. Genetti, Carol. 1992b. Segmental alternations in the Sunwari verb stem: A case for the feature [front]. Linguistics 30: 319–358. Genetti, Carol, Bartee, Ellen, Coupe, Alec, Hildebrandt, Kristine & Lin, You-Jing. 2008. Syntactic aspects of nominalization in five Tibeto-Burman languages of the Himalayan area. Linguistics of the Tibeto-Burman Area 31(2): 97–143. Hahn, Michael. 1974. Lehrbuch der klassischen tibetischen Schriftsprache. Bonn: Michael Hahn. Henderson, Eugénie J.A. 1957. Colloquial Chin as a pronominalized language. BSOAS 20: 323–327. Henderson, Eugénie J.A. 1964. Tiddim Chin: A Descriptive Analysis of Two Texts. London: OUP. Konow, Sten. 1908. Tibeto-Burman Family, Vol. III of George Grierson, Linguistic Survey of India. Calcutta, Office of the superintendent of government printing. (Reprint Delhi: Motilal Banarsidass, 1967–1968). Konow, Sten. 1911. Tibeto-Burman languages. Encyclopedia Britannica, 11th edn. Matisoff, James. 1969. Verb concatenation in Lahu: The syntax and semantics of ‘simple’ juxtaposition. Acta Linguistica Hafniensa 12(2): 69–120. Matisoff, James. 1972. Lahu nominalization, relativization, and genitivization. In Syntax and Semantics, Vol. 1, John Kimball (ed.), 237–257. New York NY: Seminar Press. Matisoff, James. 1985. God and the Sino-Tibetan copula, with some good news concerning selected Tibeto-Burman rhymes. Journal of Asian and African Studies (Tokyo Foreign Languages University) 29: 1–81. Matisoff, James. 2003. Handbook of Proto-Tibeto-Burman [University of California Publications in Linguistics 135]. Berkeley CA: University of California Press.
Finite structures from clausal nominalization in Tibeto-Burman 
Mazaudon, Martine. 1978. La formation des propositions relatives en tibétain. Bulletin de la Société de Linguistique de Paris 73: 401–414. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. In Honor of T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Noonan, Michael. 2008. Nominalizations in Bodic languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins. Przyluski, Jean & Lalou, Marcelle. 1933. Le da-drag tibetain. Bulletin of the School of Oriental Studies 7: 87–89. Saxena, Anju. 1992. Finite Verb Morphology in Tibeto-Kinnauri. Ph.D. dissertation, University of Oregon. Saxena, Anju. 1997. Aspect and evidential morphology in Standard Lhasa Tibetan: A diachronic study. Cahiers de Linguistique, Asie Orientale 26(2): 281–306. Saxena, Anju. 2000. Diverging sources of new aspect morphology in Tibeto-Kinnauri: External motivation or internal development. In Historical Linguistics 1995, Vol. 1: General Issues and Non-Germanic Languages, John Charles Smith & Delia Bentley (eds), 361–375. Amsterdam: John Benjamins. Simpson, Andrew. 2008. The grammaticalization of clausal nominalizers in Burmese. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 265–288. Amsterdam: John Benjamins. Stern, Theodore. 1963. A provisional sketch of Sizang (Siyin) Chin. Asia Major 10: 222–278. Stern, Theodore. 1984. Sizang (Siyin) Chin texts. Linguistics of the Tibeto-Burman Area 8(1): 43–58. Vittrant, Alice. 2002. /Ta/ et /Ma/, deux nominalisateurs particuliers du birman. In Aspects de la prédication, Sarah Leroy & Aleksandra Nowakowska (eds), 335–351. Montpellier: Université Paul Valéry. Watters, David. 2002. A Grammar of Kham. Cambridge: CUP. Zoller, Claus. 1983. Die Sprache der Rang pas von Garwhal. Wiesbaden: Harrassowitz.
part iii
Iranian languages
Linker, relativizer, nominalizer, tense-particle On the Ezafe in West Iranian* Geoffrey Haig
University of Bamberg The Ezafe particle is well known from recent work on Persian syntax, but its historical origins and developmental pathways have received less attention. This chapter redresses the balance by considering the Ezafe in a related West Iranian language, Northern Kurdish. In Northern Kurdish, the Ezafe has largely retained its demonstrative/relativizer origins, and also occurs as a nominalizer in the sense employed in this volume. In Modern Persian, however, these characteristics have entirely disappeared, and the Ezafe is merely a NP-internal Linker. I suggest that there is no reason to afford Persian any special priority when evaluating the developmental pathways of the Ezafe particle as a whole within the West Iranian languages. I also discuss an additional development within Northern Kurdish, where the Ezafe particle has become part of the predicate complex, echoing similar developments in a number of unrelated languages.
1.â•… Introduction The term ‘Ezafe’ is adopted from Arabic grammar (id↜āfat), where it means ‘addition, supplement’. Within Iranian linguistics it is used to refer to an unstressed vocalic particle which occurs between a noun and an adjective or other nominal modifier. Ezafes are found in a large number of West Iranian languages, where they exhibit
*This paper had its seeds in a presentation at the Second International Conference on Iranian Linguistics (Haig 2007), and I am extremely grateful to the audience there for much stimulating feedback. In particular I owe a debt of gratitude to Candice Cheung for drawing my attention to related phenomena in Southeast Asian languages, to Foong-ha Yap for her informed criticism and encouragement, and to three anonymous reviewers for their constructive contributions. One of them, Richard Larson, abandoned his anonymity and engaged me in an extremely challenging discussion on a number of semantic issues that I had neglected. As many of these points concern the synchronic analysis of the Kurdish Ezafe, rather than its history, it was not possible to give all of them the coverage they merit. Needless to say, none of the people mentioned here bear any responsibility for the remaining shortcomings.
 Geoffrey Haig
striking parallels to the “highly polysemous” morphemes that have been in the focus of research on nominalizers in certain languages of East Asia, for example -wa in Chantyal (Noonan 1997), or Mandarin Chinese di/de (Yap and Matthews 2008). Although these parallels have not gone unnoticed (Larson & Yamakido 2005; Cheung 2006; Dikken & Singhapreecha 2004), the prevalent research bias towards Persian has led to a reductionist account of the historical developments of the Ezafe within its broader Iranian context. This chapter will focus on the Ezafe in Northern Kurdish, where the Ezafe has retained many of the features of its Old Iranian ancestor, including relativizer, demonstrative and nominalizer functions. The Northern Kurdish Ezafe has also extended its distribution from the nominal into the verbal domain, echoing developments in some languages of East Asia, where erstwhile nominalizers have become so-called “stance particles” in the predicate (Yap, Choi & Cheung 2010). In stark contrast to Northern Kurdish, the Ezafe in Persian (Farsi) has undergone an almost diametrically opposed development: It has withdrawn from the syntactically more autonomous anaphoric and demonstrative functions, now surviving as just a phonologically atrophied blob of phonetic form with a highly abstract – and still hotly debated – function, perhaps best circumscribed with the term ‘Linker’. This chapter is organized along the following lines: In Section 2, the essential features of the Ezafe in Northern Kurdish are outlined, concentrating on those features where it differs significantly from the better-known Persian variety. In Section 3, the Persian Ezafe is presented together with three current approaches to its analysis. In Section 4, a diachronic scenario is formulated, contrasting the developments in the two languages. Section 5 draws some more general conclusions and relates the Iranian data to ongoing work on nominalization particles in unrelated languages.
2.â•… The Ezafe in Northern Kurdish The term Kurdish refers to a bundle of closely related languages and dialects spoken across a large area of the Middle East, with its epicentre lying at the intersection of the Iraqi, Iranian and Turkish borders. Three main dialect groups are distinguished: Southern Kurdish, Central Kurdish (or Sorani Kurdish), and Northern Kurdish, also called Kurmanji, abbreviated here as NK. Northern Kurdish is the largest dialect in terms of numbers of speakers (approx. 20 million); it is spoken in parts of North Iraq, in Syria, a large part of Eastern Turkey, and in West Iran around Lake Urmiya. There are also additional pockets in the Caucasus, and in East Iran (Khorasan). Kurdish is traditionally classified as a Northwest Iranian language, as opposed to Persian, which belongs to the Southwest Iranian branch of Iranian. All varieties of Kurdish are OV, though verb arguments expressing Recipients and Goals often occur after the predicate. However, within the NP, most lexical modifiers follow,
The Ezafe in West Iranian 
while determiners and quantifiers (demonstratives, numerals) precede the head. Northern Kurdish, unlike Central and Southern Kurdish, and unlike Persian, has preserved grammatical gender (masculine and feminine) as well as an inherited two-way case opposition between an unmarked Direct case, and a marked Oblique. NK has split ergativity, with ergative constructions confined to clauses based on past transitive verb forms (Haig 1998; Dorleijn 1996; Haig 2008). In the ergative construction, subjects take the Oblique case. In NK the Ezafe is one of the most frequent grammatical morphemes and occurs in a number of partially overlapping functions. The NK Ezafe, unlike its cognate in Persian (see Section 3), inflects for gender (masculine vs. feminine) and number (singular vs. plural).1 The actual forms are shown in Table 1, in standard Roman-based orthography (see Haig & Matras 2002); their distribution is discussed below: Table 1.╇ The forms of the definite Ezafe in Northern Kurdish Masculine
Feminine
Plural
-(y)ê
-(y)a
-(y)ê(n)/yêt (Bahdînî)
The numerous functions of the Ezafe in NK can be conveniently divided into two broad groups: the adnominal linking function, and the demonstrative/anaphoric function. In addition, one dialect of NK uses the Ezafe as an aspectual particle (see Section 2.3). Let us begin with the adnominal linking functions.
2.1â•… The Ezafe in adnominal linking constructions In this construction, the Ezafe links a post-nominal modifier to the head noun. The gender and number of the head noun determines the choice of Ezafe particle. For example, the head in (1) is dest ‘hand’, which is grammatically masculine; in (2) mal is feminine, while in (3) heval is plural. The gender/number of the modifier, on the other hand has no effect on the form of the Ezafe: (1) dest-ê te hand(m)-ez.m 2s:obl ‘your hand’ (2) mal-a mezin house(f)-ez.f big ‘big house’
.╅ The form of the Ezafe is also sensitive to the definiteness of the noun it refers to, but these complications are ignored here, as is the considerable regional variation in the phonological form of the Ezafe.
 Geoffrey Haig
(3) heval-ên keçk-ê friend(pl)-ez.pl girl-obl ‘friends of the girl’
Examples (1) to (3) illustrate typical types of adnominal modifiers in the linker construction: possessor in (1) and (3), and adjective in (2). The Ezafe can also link a PP to the head noun, as in (4): (4) zilam-ê li ber derî man-ez.m in.front.of door:obl ‘the man in front of the door’
Ezafe particles are also required when a head noun is modified by more than one successive element, as in the following: (5) keçk-a min a piçûk daughter-ez.f 1s:obl ez.f young(er) ‘my young(est) daughter’
As can be seen, the second Ezafe is still sensitive to the gender and number of the head. I will refer to sequences of more than one Ezafe-linked modifier as a modifier chain. In any modifier chain, a possessor always precedes any other elements, as in (5). The Ezafe can also be used to introduce a relative clause. The next two examples are from the Bahdînî dialect and show a relative clause linked to the head noun solely with an Ezafe. Note that here the masculine singular and the plural form of the Ezafe are homophonous (I have adopted the transcription of the source for these examples, hence the minor differences to the preceding ones): (6) tišť-Äfi [min day-av hinga]RC thing-ez.pl ╛╛1s:obl give:pst-postv 2pl:obl ‘The things [I gave to you (pl).]’ (7) aw k´as-Äfi [awwilī b-ē-t]RC dem person-ez.m ╛╛↜渀屮first subj-come:pres-3s ‘that person [who shall come first.]’
(MacKenzie 1961a: 203)
In most dialects of NK, however, and particularly in the written language, the head noun takes the Ezafe and additionally, the relative clause is introduced by a complementizer, ku: (8) çîrok-a [ku wî ji min re got]RC story-ez.f ╛╛↜渀屮compl 3s.obl adp 1s.obl adp say:pst.3s ‘The story [that he told me.]’
Finally, note that the Ezafe in NK is compatible with overtly indefinite nouns. When nouns are singular and indefinite, they take the suffix -ek (related to the numeral yek ‘one’). To this suffix, the Ezafe may be added, as in:
The Ezafe in West Iranian 
(9) jin-ek-a kurd woman-indef-ez.f Kurdish ‘a Kurdish woman’
Furthermore, the Ezafe can be used following a demonstrative pronoun, as in the following example, where the Ezafe links the demonstrative to a relative clause: (10) ev-ê ko hat dem-ez.m compl come:pst(3s) ‘this(one), who came’
(Bedir Khan & Lescot 1986)
To sum up, in the adnominal linking function the Ezafe particle is generally prosodically dependent on the immediately preceding word. In many descriptive and pedagogical studies (e.g. Blau & Barak 1999), it is actually treated as part of nominal inflectional morphology. It appears to be merely a type of agreement, mechanically replicating features of the head noun and required by a particular syntactic configuration, thus resembling a prototypical inflectional morpheme. However, such an analysis runs into difficulties when confronted with the Ezafe in modifier chains such as (5), where the Ezafe is actually separated from its head noun by other constituents. Here, the relation between Ezafe particle and head noun comes closer to one of anaphora rather than agreement. The next function we examine quite clearly belongs in the realm of anaphora, as we shall see. Corbett (2003) notes that the distinction between agreement and anaphora can be profitably considered a continuum rather than a categorical one. It is not unreasonable to consider the Ezafe in modifier chains as an intermediate point in a continuum between anaphora and agreement.
2.2â•… The demonstrative/anaphoric function of the Ezafe The Ezafe in NK can be, and regularly is, used independently of a head noun. I will refer to this usage as a demonstrative or anaphoric function (MacKenzie (1961a) refers to a ‘demonstrative’, other authors refer to a ‘secondary’ or ‘absolute’ form of the Ezafe). In this function, it comes closest to being a nominalizer in the sense of Noonan (1997). The crucial point is that the demonstrative/anaphoric Ezafe, unlike the linking Ezafe, occurs outside of the phrase in which its antecedent occurs. In fact, it is the pronominal head of its own phrase (though the precise analysis remains debatable). Some examples are the following: (11) Tu kijan hesp-î di-bîn-î? 2s which horse-obl ind-see:pres-2s ‘Which horse did you see?’ Yê(ez.m) Soro/yê min/yê te … ‘Soros’s/mine/yours (Lit: that-of Soro/that-of me/that-of you)’
 Geoffrey Haig
(12) şev-ên zivistan-ê dirêj-in, yên havîne kurt-â•… in night-ez.pl winter-obl long-cop:pl ez.pl summer short-cop:pl ‘The nights of winter are long, those of summer are short.’ (Bedir Khan & Lescot 1986: 199) (13) yÄfi dwē …â•…â•›yÄfi sēyē ez.m second…ez.m third ‘The second one … the third one’
(Bahdînî dialect, MacKenzie 1961a)
The demonstrative/anaphoric Ezafe achieves the same ends as the English pronoun one: it takes its reference from a contextually inferable set of entitities, to which an additional feature is added to ensure full interpretability. The phrase the green one is used in English to refer to a single element out of a contextually recoverable set; the attribute green is singled out as the most salient for identifying the intended referent from among those potentially available. Thus it is often part of the answer to the question ‘Which …?’, as in (11) above, or it is used in contrastive contexts.2 In principle, this type of usage is quite distinct from the adnominal linking function discussed in the preceding section. All traditional accounts of Kurdish take pains to distinguish the two terminologically, coining various additional terms for the anaphoric/demonstrative Ezafe (absolute Ezafe, demonstrative Ezafe, l’ezafe tonique etc.). However, there is a very close association between the demonstrative/ anaphoric Ezafe and the use of the Ezafe in modifier chains, discussed above. In some contexts, it is probably impossible to decide which of the two is intended. Consider for a moment a very frequently used phrase with a modifier chain: (14) [bira-yê min ê mezin]NP(subj) hat ╛╛↜渀屮brother-ez.m 1s:obl ez.m big came ‘My older brother came.’
.╅ An anonymous reviewer has questioned my analysis of this kind of Ezafe as pronominal, suggesting that the anaphoric function that appears to be associated with it could be attributed to the construction itself. His main criticism of my analysis is that it requires the postulation of more than one Ezafe in EK, differentiated according to function. While I appreciate that this approach is theoretically less satisfying, I am nevertheless at pains to point out the underlying historical unity behind the different types of Ezafe in contemporary NK, which have all developed from what was clearly once a pronominal element. Furthermore, the different labels used here (linking Ezafe, Demonstrative Ezafe etc.) are no more than approximate functional labels, with no great theoretical status implied; in practice, there are fluid transitions and indeterminate cases that defy ready classification. The historical account seems to me best suited to account for this messy data. Indeed, the present variation is no more than the reflex of related changes playing themselves out at somewhat different rates. I do of course concede that this mode of presentation does not allow for a very elegant and readily formalizable synchronic analysis.
The Ezafe in West Iranian 
There is little doubt that the second Ezafe particle should be interpreted as part of the entire NP headed by bira ‘brother’. However, if the same sequence of words occurs as the subject of a copula, at least two different interpretations are possible: (15) bira-yê min ê mezin-e brother-ez.m 1s:obl ez.m big-cop:3s
The pragmatically neutral interpretation of (15) is: ‘(It) is my big brother’. But given a particular context, where the speaker is asked to identify his brother among a group of boys, this clause could have the reading: ‘my brother is the big one’. In that case, the sequence ê mezin would be interpreted as the complement of the predicate, along the following lines: (16) [bira-yê min]NP(subj) [ê mezin]NP(predicate complement) -e ╛╛↜渀屮brother-ez.m 1s:obl ╛╛↜渀屮ez.m big -cop:3s
On this reading, we would have to classify the Ezafe in ê mezin as an example of the demonstrative/anaphoric Ezafe. Thus in practice there are overlaps between adnominal Ezafes, and independent ones. Examples (17) and (18), from Bedir Khan and Lescot (1986: 198–199), provide further evidence of structural ambiguity (the translations given reflect those of the source): (17) re-ya me a dur e road-ez.f 1pl:obl ez.f far cop:3s ‘Our road is long/is a long one’
The structure of (17) would actually allow the interpretation: ‘(it/that) is our long road’, which would leave the phrase a dur as a chained modifier, part of the NP headed by ‘road’. But this is apparently not intended in the context. On the reading provided in the source, a dur is syntactically disjunct from the preceding NP, and actually part of the predicate phrase. The next example is fully comparable; the phrase ê spî is a demonstrative/anaphoric Ezafe, not a chained modifier: (18) xani-yê me ê spî-ye house-ez.m 1pl:obl ez.m white-cop:3s ‘Our house is the white one’
Such contexts provide a classical case of bridging constructions for language change: structurally ambiguous strings whose precise interpretation is coerced from context. Over time, one of the possible interpretations may become conventionalized to the extent that the others are no longer available. In this case, a contextually-driven interpretation becomes a grammatical construction with a fixed pairing of form and meaning. It is tempting to consider the emergence of the anaphoric/demonstrative Ezafe in this light, that is, as emerging from one of the potentially available readings of the Ezafe in a modifier chain. However, the historical evidence suggests that this is not
 Geoffrey Haig
the case; if anything, the development went the other way (see Section 4). Nevertheless, constructions of the type (17) and (18) almost certainly played a role in the emergence of what is a genuine Kurdish innovation, the use of the Ezafe as a tense/aspect operator, to which we now turn.
2.3â•… The Tense Ezafe In the North Kurdish dialect of Bahdînî, spoken around the townships of Zakho, Dohuk and Amediye in North Iraq, the Ezafe morpheme is used in an additional construction, which I refer to as the Tense Ezafe. The Tense Ezafe is briefly discussed in Haig (2007) and Haig (2008), but has otherwise been ignored in the literature on the Ezafe. As the name implies, the Tense Ezafe is part of the predicate, rather than being associated with nominal syntax. In the terms of Roberts and Roussou (2003), it has crossed from the D and N domain to the T domain – a development that is actually not predicted in their framework. Although a development of this nature appears to be unique in Iranian, it does display striking parallels to the development of so-called “stance particles” in a number of languages of East Asia (Yap, Choi & Cheung 2010), and to the emergence of various kinds of copular particles, which we look at below. To convey some notion of how the Tense Ezafe works, a number of examples from various sources are given below. Many of the examples contain several exponents of the Ezafe, but the one we are interested in at present is bold-faced: (19) xuşk-a min ya çuy-î sîk-ê sister-ez.f 1s.obl ez.f go:pst-ptcpl market-obl ‘My sister has gone to the market’ (20) Sofi Mistefa-y kiçk-ek a hey3 Sofi Mistefa-obl daughter-indef ez.f existent.cop(3s) ‘Sofi Mistefa has a daughter’
(own fieldwork)
(Blau 1975: 70)
(21) Got-ê ku šah-ê wan yê mir-î say:pst-to.him that king-ez.m 3pl.obl ez.m die:pst-ptcpl (He) said to him that their King had died.’ (MacKenzie 1961a) (22) Hal-ê wî ê keftî-ye condition-ez.m 3s:obl ez.m fall:ptcpl-cop:3s ‘He is poor (Lit.: ‘his condition/state has fallen)’
(Hassan 2006: 14)
.╅ In the source from which this example is taken, the Ezafe-particle is actually written as part of the preceding word. I have separated here for ease of comparison. There is in fact considerable variation in the orthographic rendering of these particles.
(23) Ew (y)êt kurd-in dem ez.pl Kurd-cop:pl ‘They are Kurds.’
The Ezafe in West Iranian 
(own fieldwork, Zakho 2006)
Just like the adnominal Ezafe, the Tense Ezafe inflects for number and gender of the closest preceding NP in the absolute case (this can, in many cases, be equated with the ‘subject’, but there are complications in the past tenses which I will not enter into here): examples (19–20) are feminine, (21–22) are masculine, while (23) is plural. The phonological forms of the Tense Ezafe are almost identical to the corresÂ� ponding forms of the demonstrative/anaphoric Ezafe. There is thus no doubt that etymologically we are dealing with the same morpheme; this degree of identity in the paradigm can hardly be coincidence, and there simply is no other morpheme in the language that could plausibly have provided the source. Note also that the Tense Ezafe is not restricted to third person NPs, but is fully compatible with personal pronouns of the first or second person, where it is also sensitive to gender: (24) Ez ya/yê kurd-im 1s ez.f/ez.m Kurd-cop:1s ‘I am a Kurd (uttered by female speaker/male speaker)’
In all the above examples, the use of the Tense Ezafe is the normal and pragmatically unmarked means of expressing the sentences given. In other dialects of Northern Kurdish, however, the same propositions would be expressed without the Ezafe particle. Thus for (24), we would simply have Ez kurd-im ‘I am a Kurd’, with no means of differentiating female from male speakers. Now occasionally, BK speakers may use this form too. Whether this is interference from the neighbouring, more prestigious dialects (now widely heard in Kurdish-language media), or whether there is actually a semantic difference between the two forms is extremely difficult to establish at present, for reasons I will discuss below. For the time being, I will be content with providing an account of the environments in which they regularly and consistently occur in spoken texts (e.g. in Blau 1975). There are certain types of predicate which, in present tenses, are generally associated with the Tense Ezafe: state, existential and locative predicates (be there, exist, have), cf. (20, 23, 24). In particular, the Tense Ezafe frequently occurs with resultative participles in -ī (as in (21, 22). In other words, the Tense Ezafe is typically associated with stative/resultative predicates rather than with dynamic ones, and is generally associated with a present-tense form of the copula. When combined with a participle, the Tense Ezafe imparts a sense of completed action (resultative), but with relevance to the present. For example, (19) would be uttered only in a situation where the girl concerned was actually gone, and no longer visible.
 Geoffrey Haig
According to MacKenzie (1961a and 1962), the two types of predicate just mentioned are the sole ones found with the Tense Ezafe. However, more recent sources show that it also occurs with truly verbal predicates: (25) Ez ê di-bêj-im 1s ez.m ind-say:pres-1s ‘I am saying (right now)’ (French transl.: Je suis en train de dire, Blau 1975: 40) (26) Ez yê xwarin-ê çê-di-k-im 1s ez.m meal-obl prev-ind-do:pres-1s ‘I am making/preparing a meal (right now)’
(own field work, Zakho 2005)
When combined with a finite present tense verb form, as in (26), the Tense Ezafe implies that the action concerned occurs at the time of speaking. At present, little more can be said with certainty on the semantics of the Tense Ezafe, and it has proved methodically very difficult to gain consistent judgements from native speakers. Part of the problem lies in the complex sociolinguistics of Bahdînî Kurdish within Northern Kurdish: Northern Kurdish has an emergent written standard, promulgated on the internet and over satellite television, and currently the object of intense debate among native speakers. Due to decades of linguistic repression, these ongoing controversies are emotionally loaded in the extreme. Now the Tense Ezafe is not part of that emergent standard language, thus it is to a certain extent stigmatized as a Bahdînî localism. As a result, metalinguistic discussions with native speakers on such forms are inevitably sullied with evaluative issues of standardization.4 A reliable description of the semantics and pragmatics of the Tense Ezafe in Bahdînî Kurdish thus remains a major desideratum in Kurdish studies. However, the available corpora of spoken Bahdînî Kurdish (Blau 1975 and MacKenzie 1961a, 1962), and my own fieldwork show quite clearly that it is a ubiquitous feature of the spoken language, though its precise semantic contribution to the predicate remains elusive. There are, however, two robust syntactic/pragmatic constraints on the use of the Tense Ezafe: it is restricted to affirmative, declarative clauses; in questions or negated clauses it is not found. These two facts are extremely relevant for understanding the history of the Tense Ezafe, as we shall see. The starting point for the development of the Tense Ezafe must have been the demonstrative and anaphoric use of the Ezafe introduced above. This type of Ezafe is
.â•… I am grateful to Sadiq Basid, Newzad Hiruri and Neçirvan Hiruri for much enlightening discussion on the Tense Ezafe and related issues.
The Ezafe in West Iranian 
anaphoric in the sense that it refers back to a contextually recoverable NP, with which it agrees in gender and number. Consider now (22), repeated here as (27): (27) Hal-ê wî ê keftî-ye condition-ez.m 3s:obl ez.m fall:ptcpl-cop.3s ‘He is poor’ (lit.: ‘his condition/state has fallen’)
(Hassan 2006: 14)
In modern Bahdînî Kurdish, (27) is the pragmatically unmarked way of expressing the proposition ‘he is poor’. The function of the Ezafe in this clause cannot be considered adnominal, and hardly anaphoric; if anything, it modifies the predicate ‘has fallen, be in a fallen state’. But it is fairly easy to understand how this construction came about: it must have arisen through constructions where the initial NP was a left-dislocated topic, and the Tense Ezafe was an anaphoric/demonstrative referring back to that topic:5 (28) [His state]TOPIC
[thatDEM/ANAPH-SUBJ is one that has fallen/which has fallen]S
Over time, this construction lost its pragmatically marked status and became the unmarked means of making such statements. The left-dislocated topic was thus reanalyzed as the grammatical subject of the clause. At this point, the anaphoric/ demonstrative Ezafe is rendered superfluous, at least as far as its anaphoric function is concerned. It is then reanalyzed as part of the predicate, where it becomes a marker of perfective aspect, used primarily with non-verbal predicates. From there, it has extended its range to co-occurrence with more dynamic verbal expressions and taken on the meaning of immediacy (cf. examples (25) and (26) above). Support for this account of the emergence of the Tense Ezafe comes from the pragmatic restrictions on its use: it is not possible in interrogative or negated clauses. In other words, it is not fully grammaticalized as a tense marker, but still carries some traces of its origins in a pragmatically marked structure, which almost certainly would have been restricted to affirmative, declarative clauses. Further support comes from well-documented parallels in unrelated languages. Li and Thompson (1977) describe several cases of how copulas develop from erstwhile anaphoric/demonstrative elements, the best-documented among them being Mandarin and Hebrew. Less well-known in the literature is the development of the
.â•… An alternative analysis is that original construction was not a topic+clause, but a NP with a relative clause: ‘His state, which is fallen’. Such an analysis is quite possible, given the relativizing function of the Ezafe, but would not explain how the resultant construction achieves its illocutionary force. I must leave this issue unresolved for the moment; it does not directly impinge on the broader claim that the Ezafe has left the nominal domain to become part of the predicate.
 Geoffrey Haig
predicate marker i in Tok Pisin. This particle is always required before a predicate that has a third person subject: (29) a.
em i nogut 3s pred bad ‘He is bad’
b. Jon i bin wok asde John pred pst work yesterday ‘John worked yesterday’
There is little doubt that the predicate marker evolved from the English pronoun he (often pronounced [i:]), although the model for the grammatical pattern appears to have been provided by one of the substrate languages (see Mosel 1984). Another particularly well-documented case of demonstratives developing into tense markers is Panare (Carib), described in Gildea (1993), and similar, though somewhat different developments are discussed for the Amazonian language Chamicuro in Parker (1999). There is thus a striking cross-linguistic pattern for elements originally clustering in the domain of demonstratives and anaphors to shift sideways into the predicate. What is most intriguing in all the cases mentioned above – Kurdish, Hebrew, Mandarin, Tok Pisin, Panare, and the stance particles discussed in Yap, Choi and Cheung (2010) – is the often elusive semantics of the resultant particles. Furthermore, they are often subject to pragmatic or other restrictions in their predicative functions, i.e. they are not fully grammaticalized. For example, in Panare, the erstwhile demonstratives only occur as copulas with third person subjects, clearly betraying their demonstrative origins. It seems that these elements are often only partially incorporated into the predicate domain, and retain both functional and grammatical reflexes of their nominal origins for a considerable period of time. Formulating a formal analysis in terms of discrete category membership for this kind of in-between element is thus often extremely problematic. In at least one dialect of Bahdînî Kurdish, the developments have been accompanied by phonological erosion and the loss of the nominal categories of gender and number. In the region of T ↜ur ‘Abdin in Southeastern Turkey, there are about eight villages inhabited by members of the the religious minority of the Yezidis, a Kurdishspeaking religious group with its centre among the Bahdînî Kurdish speakers of North Iraq. Bailey (2005: 29–30) refers to this dialect as Êzîdiyê Torê (ÊT). ÊT has a particular form of the “present perfect tense”, formed with a participle and “a clause with the particle î/wî.[footnote ommitted]”. Bailey provides the following example of this particle, glossed here simply part: (30) belê, min î masî girtî-n yes, 1s:obl part fish(pl) catch:ptcpl-pl ‘Yes, I have caught fish’ (Bailey 2005: 30, original translation, gloss modified)
The Ezafe in West Iranian 
Although she does not propose a diachronic connection to the Tense Ezafe of Bahdînî Kurdish, it seems certain that the particle î in (30) is a continuation of the latter. In favour of this interpretation are four facts: (i) the position and function of the particle correspond exactly to the perfective/resultative use of the Tense Ezafe discussed above; (ii) the close historical ties of the speakers of these dialects with speakers of Bahdînî Kurdish; (iii) the particle does not appear in the corresponding negated clause quoted by Bailey (2005: 29), which again echoes the constraints on the Tense Ezafe in Bahdînî Kurdish; (iv) even in Bahdînî Kurdish, the form of the particle is often unstable and the gender distinction is not always faithfully reflected, leading to a consistent spelling in some sources as -î. Thus it is phonologically already extremely close to the ÊT particle. I therefore assume that this particle is a continuation of the Tense Ezafe, which has lost the nominal categories of number and agreement and is now reduced to a single form. The developments are summarized in Table 2. Table 2.╇ From Demonstrative/Anaphoric to Tense Particle in Northern Kurdish Dem./Anaph.ezafe
Tense Ezafe
-(y)a
-(y)a
-(y)ê
-(y)ê
-(y)ên
-(y)ên
(Common NK)
(Bad. K only)
Perfect particle in ÊT -î/wî
(ÊT only)
Before closing this section, I should briefly address the issue of the position of the Tense Ezafe. It may appear odd that a particle which generally follows the subject NP, and is thus quite distant from the lexical predicate, should nevertheless be analyzed as part of the predicate. But within an Indo-European context, such a development is actually quite commonplace, and indeed is also characteristic of English: Tense/Aspect clitics occur primarily clause initially, and can be separated from the lexical verb by other items: (31) I’ve always liked linguistics.
Part of this tendency can be attributed to the second-position preference for clitics generally. Northern Kurdish actually has another tense particle, the future marker dê/ wê, which is likewise bound to the clause-second position (ignoring some additional complications): (32) Şevîn dê wî bi-bîn-e Şevîn fut 3s:obl irr-see:pres-3s ‘Şevîn will see him.’
(Rizgar 1996: 144)
 Geoffrey Haig
2.4â•… Summary of Northern Kurdish While the Ezafe in Northern Kurdish shares adnominal linking functions with the Ezafe in Persian, it also exhibits some very remarkable divergent developments. Putting it most generally, the Ezafe in NK is characterized by greater phonological autonomy and substantive content: it inflects for number and gender, and is not prosodically dependent on a preceding host. Its functional distribution basically falls into two broad categories: adnominal linker, that is, linking simple modifiers to head nouns, and demonstrative/anaphoric, where it shares some characteristics of the English pronoun one. In the latter function, it comes closest to what is generally understood as a nominalizer: it transforms some form of modifying phrase, such as a possessive attribute or an adjective, into a NP. From this function an additional extension has occurred in some dialects, most notably in Bahdînî, where the demonstrative/anaphoric particle is reanalyzed as part of the predicate, yielding the Tense Ezafe.
3.â•… The Ezafe in Modern Persian Modern Persian, the sole official language in the state of Iran, belongs to the Southwest branch of the Iranian branch of Indo-European. Like Northern Kurdish, it is verb-final and within the NP, most lexical modifiers follow the head noun. Determiners and quantifiers, however (demonstratives, numerals) precede the head. Nominal inflectional morphology is greatly impoverished in comparison to Old Iranian: there is a single case marker, enclitic -râ, expressing direct objects (and marginally other functions that can be ignored here), while other case relations are expressed through prepositions, or syntactically. There is no gender, and no trace of ergativity. In Persian, the Ezafe is generally transcribed with -(y)e. However, in standard Persian orthography (based on the Arabic script) it is not indicated at all, except indirectly after certain letters. Thus in the written language, the presence of an Ezafe can usually only be inferred from the context. The Persian Ezafe is restricted to the adnominal linking function: It follows a noun modified by an adjective (33), a noun expressing a possessor (34), a noun with a sortal or type-specifying sense (35), or certain prepositional adjuncts (36). All examples in this section are from Samvelian (2007), unless stated otherwise: (33) lebâs-e zibâ dress-ez beautiful
‘the beautiful dress’
(34) lebâs-e maryam dress-ez Maryam
‘Maryam’s dress’
(35) lebâs-e arusi dress-ez wedding
‘the wedding-dress’
The Ezafe in West Iranian 
(36) lebâs-e bi âstin dress-ez without sleeve ‘the dress without sleeves’
The postposed modifer can be more complex, as in the following bracketed constituent: (37) qehremân-e [rande šode az mihan-aš] hero-ez ╛╛↜渀屮driven:ptcpl become:ptcpl from homeland-poss.3s ‘The hero driven from his homeland’
When a complex NP with an Ezafe-linked modifier is the direct object, then the object marker -râ will follow the last item in the NP: (38) Agar sabr kon-id zan-e [ziba-tar va tannâz-tar If patience do:irr-2pl woman-ez ╛╛↜渀屮beautiful-comp and charming-comp az ân]=râ peydâ xâh-id kard. from dem=acc finding fut.aux-2pl do:pst.3s ‘If you are patient you will find a woman [more beautiful and more charming that this]’
As in Northern Kurdish, more than one modifier may be linked to the head noun (modifier chains). Within such chains, each modifier is preceded by an Ezafe particle (alternatively, modifiers of the same category can be linked by the conjunction va ‘and’, but we ignore these issues here): (39) ketâb-e târix-e sabz-e bi arzeş-e maryam book-ez history-ez green-ez without worth-ez Maryam ‘The green, worthless history book of Maryam’s.’
There is an important difference between chained modifiers in Northern Kurdish, and in Persian. In Northern Kurdish, a possessor is generally the first item in such a chain (cf. (13) above). In Persian, on the other hand, a possessor is always the final element of a modifier chain. Thus to express ‘my little room’, only (40b) is possible: (40) a. *otaq-e man-e kučak room-ez 1s-ez small b. otaq-e kučak-e man
A second important difference between the Ezafe in NK and in Persian is the following: as we have seen, relative clauses can perfectly easily be linked to head nouns with the Ezafe in NK. In Persian, however, this is impossible. The contrast is illustrated in (41): (41) a. *Zan-e [ke man dîd-am] woman-ez ╛╛↜渀屮that I see:pst-1s Intended reading: ‘The woman that I saw’ b. Jin-a [ku min dît] woman-ez.f ╛╛↜渀屮that 1s:obl see:pst(3s) ‘The woman that I saw’
(Persian)
(NK)
 Geoffrey Haig
The constraint against combining relative clauses with Ezafes has been a central issue in the recent literature on the Persian Ezafe (see Section 3.1). However, the data from Northern Kurdish show that it is actually a language-specific constraint, hence the analyses of the Ezafe in Persian cannot be extended wholesale to related languages with Ezafes.6 A further constraint on the Persian Ezafe is that the Ezafe particle cannot be hosted by nouns bearing certain suffixes, in particular the indefiniteness suffix -i: (42) a. i.
zan woman
ii. zan-e zibâ woman-ez beautiful ‘the beautiful woman’
but: b. i.
zan-i woman-indef ‘a woman’
ii. *zan-i-ye zibâ woman-indef-ez beautiful Intended for ii: ‘a beautiful woman’
To express (42b), we require a paraphrase with a preposed indefinite yek (identical to the numeral one): yek zan-e zibâ. Alternatively, the indefiniteness suffix can follow the modifier, as in zan-e zibâ-i.7 Finally we must note that the Persian Ezafe is prosodically reliant on a preceding host. Unlike the Northern Kurdish Ezafe, it cannot introduce a phrase. Compare (43) with the Northern Kurdish example (11) above: (43) *ye Maryam ez Maryam Intended reading: ‘the one of Maryam, Maryam’s.’
In sum, the Persian Ezafe is a particle required by a certain syntactic configurations obtaining within NPs, but it is subject to the following syntactic and prosodic restrictions: .╅ Richard Larson has suggested that the difference between NK and Persian in this respect may actually go deeper than an isolated fact concerning the distribution of the Ezafe and complementizer; it may be linked to the degree of finiteness of the clauses concerned. I think this observation is correct, and that even in NK, the use of the Ezafe without a complementizer is more acceptable when the relative clause contains a less finite verb form (participle etc.). However, this would lead to a discussion on the finer points of finiteness in the languages concerned, which goes beyond the scope of the present paper. .╅ I am grateful to an anonymous reviewer for pointing this out to me.
The Ezafe in West Iranian 
1. It cannot link the head noun to a finite relative clause introduced by the complementizer ke 2. A possessor is always the final element in any chain of parallel Ezafes 3. An Ezafe is prosodically dependent on a preceding host 4. An Ezafe cannot be hosted by a noun bearing the indefiniteness suffix.
3.1â•… Recent analyses of the Persian Ezafe While all scholars agree that it is undoubtedly part of a nominal projection, be it a DP, an NP or an AP, the Ezafe is not readily accountable in terms of conventional X-bar phrase structure, and the associated repertoire of functional categories. In a sense, all the proposals outlined in this section are faced with the same dilemma: how to fit the Ezafe particle into a theoretical framework which provides no category that readily accomodates it. Most more recent work draws on Samiian (1983), who formulates an analysis in terms of X-bar structures. More recent scholars have successively modified Samiian’s proposals, but, with the exception of Samvelian (2007), have largely been content with explaining a similar set of data. In this section I will briefly discuss the proposals of Ghomeshi (1997), Larson and Yamakido (2005) and Samvelian (2007). Ghomeshi (1997) suggests that Persian nouns are inherently non-projecting, thus lack both a specifier-position, and complements. They may, however, still be accompanied by modifying material, as we have seen above. This is achieved by stipulating that N-modifiers are adjuncted to the N0-level. However, and this is pivotal in the argument, the only modifiers permitted in this position are non-projecting, or non-phrasal. Ghomeshi’s formulation is intended to reflect the fact that there are restrictions on the type of phrases that can be included under an Ezafe, as noted above with respect to relative clauses. Possessor modifiers, on the other hand are in the specifier position of the DP. The latter move accounts for the fact that possessors are the final members of a complex DP (cf. (40) above), thus outside of any other modifiers. Another important motivation for this analysis are definiteness effects: Ghomeshi suggests that the presence of a possessor Ezafe generally renders the entire phrase definite, thus they resemble specifiers. As far as the combination of N with non-possessive modifiers is concerned, on Ghomeshi’s account they actually wind up looking rather like compounding structures. The Ezafe itself is reduced to a purely structural linking element, introduced into PF at a late stage of the derivation by a rule of “Ezafe insertion”. The question of its categorial status is thus essentially side-stepped. A number of empirical problems with Ghomeshi’s account are pointed out by Samvelian (2007). She notes that contrary to Ghomeshi’s claims, phrasal material is possible under the Ezafe, as in (37) above. Thus the claim that Persian nouns are non-projecting is seriously challenged. Furthermore, Samvelian suggests that indefinite readings of possessed NPs
 Geoffrey Haig
are possible, casting doubts on the validity of the Specifier-analysis of the possessor. Samvelian claims that the apparent definiteness effects can be readily accounted for in terms of morphological constraints on suffix combining, a point we return to below. A radically different view of the Ezafe is suggested by Larson and Yamakido (2005), who take up a suggestion first put forward by Samiian (1994). On their account, the Ezafe is in fact a case marker. Their claims also build on the observation that the Ezafe is required to precede certain types of modifiers, but not others (for example, relative clauses). Basically, they suggest that the modifiers which do require the Ezafe bear the feature [+N], while those that do not (relative clauses, certain types of PP) lack this feature. They also claim that many of these modifiers actually have complement status. They are, for example, possessors, or the semantic complements of action nouns (44), or the complements of adjectives (45): (44) taxrîb-e šahr destruction-ez city ‘the destruction of the city’ (45) nagarân-e bache-hâ worried-ez child-pl ‘worried about the children’
For these and similar examples, the case-analysis has some intuitive appeal. The problem with the case analysis arises when we come to explain why the Ezafe should also occur with adjectival modifiers: in what sense can they be considered case-assigned by the head noun? To achieve this, the authors propose extending the shell-theory of the VP, originally proposed in Larson (1988), which works on the assumption that the VP is universally right-branching, with additional constituents introduced into the specificer position of some verb. The details of these proposals cannot be discussed here; the crucial point for the Ezafe is that Larson and Yamakido (2005) suggest that within the DP, it is the D that is the head, and and hence may assign case. But note that ‘case’ in this sense actually refers to quantifier scope effects: the NP that combines with D saturates the quantifier restriction of the D. A simple determiner such as English every assigns a single case to its NP, more complex expressions such as every … except include additional roles. The net result of this additional machinery is that nominal modifiers – adjectives for example – are not adjuncts, but “oblique complements” which combine with the head prior to other arguments. A consequence of this move is that all modifiers are base-generated in a post-head position, and in a language like English, adjectival modifiers are subsequently moved to pre-head position in the course of the derivation. What consequences would this revised view of the DP have for the Persian Ezafe? Larson and Yamakido (2005) suggest that Persian includes in its case system
The Ezafe in West Iranian 
a “generalized genitive preposition”, which is inserted to check case on [+N] complements of D inside the DP. On this account, the Ezafe heads its own X-bar phrase, with the modifier as complement. However, for apparently purely prosodic reasons, phonologically it attaches to the preceding item. The advantage of this approach is that it readily accounts for the chains of Ezafe modifiers because it assigns the Ezafe structurally to the modifier it precedes, rather than to the nominal head of the phrase. However, the disadvantages are considerable: it rests entirely on the assumption that a shell-structure, with universal right-branching, is not only viable for the VP, but can be meaningfully applied to DPs. But the shell-analysis of the VP itself is by no means uncontroversial (see Culicover & Jackendoff (2005: 50–56) for a summary critique). The extension of the notion of “case” to adjectival modifiers appears a little post hoc. Furthermore, there are West Iranian languages such as Māzandārani which regularly have prenominal adjectival modifiers, and yet they occur with what the authors analyze as a “case marker”: (46) xajir-e rikā nice-case boy ‘the nice boy’
(Borjian & Borjian 2007: 204)
In Turkmen Balochi, attributive adjectives are likewise prenominal, and are obligatorily accompanied by what Axenov (2006: 87) refers to as the “attributive suffix” -ēn: (47) rašīd-Äfin jinēnzāg-ā bi d↜igār-ā yēr kurt slender-attr woman-obl to ground-obl down put(pst.3s) ‘He put down the slender woman (from the camel)’
Now according to Larson and Yamakido (2005), the Ezafe on the postnominal adjectives licenses them for case, thus allowing them to remain in their base-generated position. Without case, they would be obliged to move to a prenominal position, as in English. The presence of overtly marked prenominal adjectives is therefore not predicted on this theory, yet this constellation does occur. A second drawback is that Larson’s theory predicts that the Ezafe should not occur with relative clauses containing finite verbs, because they are not [+N]. But as we have seen above, Ezafes readily occur with relative clauses in Northern Kurdish. A very recent contribution to the debate is Samvelian (2007). She rejects both the case-approach of Samiian (1994) and Larson and Yamakido (2005), and the approach of Gomeishi (1997). According to her, the facts of the Ezafe are best accounted for in morphological rather than syntactic terms. She claims that the Ezafe is a “phrasal affix”, exhibiting both affixal and clitic-like properties. However, it is morphological rather than syntactic in the sense that it forms a distributional class with certain other affixes. Many of the constraints on the use of the Ezafe which had previously been explained via syntax receive on her account a more natural explanation in terms of
 Geoffrey Haig
slot competition. As for the function of the Ezafe, she adopts a diametrically opposed stance to Larson and Yamakido (2005). According to Samvelian, the Ezafe is an affix attaching to a nominal head in anticipation of a following complement or modifier. On the head-marking vs. dependent-marking typology of Nichols (1986), the Ezafe is thus an instance of head-marking morphology. On this account, the Ezafe forms a constituent with the N both prosodically as well as functionally. Note how this account contrasts with Larson’s, who sees the Ezafe as forming a syntactic constituent with the modifier. Samvelian’s approach also differs from Ghomeshi’s in assigning Ezafe-linked possessor phrases to a complement position, rather than the specifier position. However, Samvelian is forced to adopt additional stipulations to account for the use of the Ezafe with chained modifiers, which, as shown above, are separated from their putative “bases” by other elements. In a sense, all three proposals are primarily geared towards accounting for the constraint on relative clauses, which is somehow deemed to require explanation. Ghomeshi (1997) states that relative clauses cannot combine with the Ezafe because they contain phrasal material; Larson and Yamakido (2005) attribute the constraint to the feature [−N] on relative clauses, while Samvelian (2007) interprets the same facts as evidence of a morphological constraint on the co-occurrence of the Ezafe with another suffix. But against the background of West Iranian languages like Northern Kurdish (see above), or Zazaki (Paul 1998: 145), where the Ezafe regularly combines with relative clauses, the Persian facts emerge as very much language-specific particularities, which do not actually tell us very much about the Ezafe as a broader phenomenon in West Iranian. Table 3 sums up the differences and similarities between the Persian and the NK Ezafe: Table 3.╇ Functions of the Ezafe in Persian and in Northern Kurdish Function
Persian
Northern Kurdish
Inflects for gender, number Links simple adjectives to head Links Possessors to head
no yes
yes yes
yes, possessor final in modifier chain some restrictions yes no no no
yes, possessor initial in modifier chain yes yes yes yes yes
no
yes (Bahdînî dialect and ÊT dialects)
Links modifying PPs to head Used for modifier chains Links relative clause to head Can be hosted by an indefiniteness suffix Occurs without overt head noun in nominalizing function (‘the one …’) Tense marker
The Ezafe in West Iranian 
4.â•… Th e forerunners to the Ezafe in Old Iranian: The relative pronoun hypothesis Nothing is known about the possible lexical source of the Ezafe. However, it can be reliably traced back to certain extremely frequent and polyfunctional grammatical elements of Old Iranian. Two contributions to the analysis of the Ezafe incorporate historical data into their analyses: Haider and Zwanziger (1984) and Karimi (2007), the latter dealing with Central Kurdish. Both analyses assume that the forerunner of the Ezafe particle was basically a relative pronoun. The general line of reasoning is that in Old Iranian relative clauses, an overt copula was often lacking, so the construction with the relative pronoun also looked much like the combination of noun with simple adjective or possessor. Over time, the Ezafe lost its complementizer function, which was taken over by different particles (e.g. Persian ke, presumably from an old interrogative item). It then underwent further phonological and functional erosion, until it was ultimately reduced to a kind of generalized “genitive case marker” (see Larson & Yamkido 2005), a mrker of a “predication phrase” (Karimi 2007), or a “linker” (Dikken & Singhapreecha 2004), depending on one’s analysis. The relativepronoun hypothesis is particularly attractive for those scholars who restrict themselves to Persian or Central Kurdish but is less so for the Northern Kurdish data, or for Zazaki (Paul 1998). Furthermore, the two most detailed studies of the Old Iranian ancestors of the Ezafe, Kent (1944) and Seiler (1960), show quite clearly that in origin, the Ezafe was not simply a relative pronoun in the conventional sense. In this section, I will very briefly examine the Old Iranian data before relating them to the phenomena we have been investigating. There are two main sources for Old Iranian: Avestan, and Old Persian. Old Persian is the direct ancestor of modern Persian (Farsi), and is attested in stone inscriptions dating from 6–4 century bc. Avestan on the other hand is the language of the earliest sacred texts in the Zoroastrian religion. Two varieties of Avestan are distinguished: Old Avestan is the oldest, and is extremely close to to the Sanskrit of the R↜渀屮gveda: the oldest poems are estimated to have been composed around 1500 bc, although the dating remains controversial. Young Avestan on the other hand was spoken presumably in the first millennium ad. Texts written in Avestan are collectively termed the Avesta. They are of an arcane religious nature, and were transmitted orally by specially trained priests across centuries before being committed to writing sometime around the middle of the first millennium ad. It should be evident that both the interpretation and dating of the Avesta remains a very delicate undertaking. Turning first to Old Persian, Kent (1944) systematically examines the then available 388 instances of the Old Persian particle hya, the presumed ancestor of the Ezafe. Kent relates hya to to the demonstrative sya-s of Vedic Sanskrit. The Old Persian
 Geoffrey Haig
hya (which inflected for case, number and gender) appeared in a large number of constructions. Kent classifies the extant examples according to function and concludes that about 70% of them can be analysed as relatives, while the rest are what he somewhat misleadingly terms ‘articles’. Two criteria are investigated by Kent in defining relatives. The first concerns the finiteness of the clause introduced by hya: (48) ima [tya adam akunavam] this [hya:neut.acc 1s:nom do:pst.1s] ‘this (is) [that.which I did]’
In (48), hya introduces a clause headed by a finite verb, hence is classified by Kent as a relative pronoun. The finiteness criterion runs into difficulties, however, because in many cases, the phrase concerned lacks a verb altogether. In such cases, a second criterion may be invoked, namely the case form of hya. Kent considers it a property of the relative to take its case from the relative clause it introduces, rather than from its antecedent. Thus the following are also considered by him to be relative clauses: (49) […] Dārayavaum [hya manā pitā] ╅╇╛╛↜Darius:acc ╛╛↜渀屮hya:m.nom 1s:gen father ‘Darius [who (was) my father]’ (50) hacā Sakaibiš [tyai para Sugdam] from Scythian:pl:abl ╛╛↜渀屮hya:pl.nom beyond Sogdiana ‘from the Scythians [who (are) beyond Sogdiana]’
In (49) and (50) hya is in the nominative, as required by the relative clause, rather than in the accusative and ablative of their antecedents. Kent therefore counts these examples as relative clauses, despite the lack of a finite verb. The importance attached to the case form of hya has tended to obscure another parameter, at least as revealing of the status of the particle: the definiteness of the head noun. In all the above examples, the putative ‘head noun’ is definite, for example a proper name, or qualified by a demonstrative. In this type of context, the supposed relative clause is not a restrictive relative clause, because it does not contribute to the identification of the referent. Surely a more natural translation of (49), for example, would be ‘Darius, my father’. A substantial number of the phrases introduced by hya are thus non-restrictive, or appositive. In the earlier typological literature on relative clauses, this type was actually excluded from the definition of relative clause (Keenan & Comrie 1977). The important point here is simply that what appears to be a relative clause syntactically is functionally more of a loose appositive construction. I believe the primary function of the particle was in fact to introduce such appositive phrases. The case-criterion is nevertheless revealing because in other examples, hya takes its case from its antecedent:
The Ezafe in West Iranian 
(51) avam Gaumātam [tyam magum] dem:acc Gaumāta:acc ╛╛↜渀屮hya:m.acc Magian ‘(I struck down) that Gaumāta, [the Magian]’
Kent accepts that many examples cannot be reliably classified either as relatives, or appositives, or in his terminology, articles, and I entirely concur with this finding. There was a very strong tendency – almost a rule of Old Persian – for relative clauses expressing generic senses (English he who, that which) to be headless in Old Persian, as in (48) above, and there are also examples of hya serving to turn an adjective into a NP (comparable to English one, as in the red one): (52) hya tauvīyā tyam skauθim naiy jatiy m:nom strong hya:m.acc weak:acc neg may.strike ‘(that) the strong (one) may not strike the weak (one)’
Examples such as these reflect what Kent assumes is the origin of hya, namely as an amalgamation of an erstwhile demonstrative with a relativizer, basically ‘that-which’. If we accept this view, and I am unaware of more plausible alternatives, then it is clear that the Ezafe in Old Persian was already a kind of nominalizer, and not merely a relative pronouns.8 It is in fact remarkably similar to what I suggested above for the BK Ezafe, although it lacks the Tense Ezafe function. It is therefore reasonable to assume that the forerunners of Northern Group Kurdish had a system of relative/demonstrative particles quite close to that attested in Old Persian. The facts from Avestan are largely compatible with those of Old Persian, although the particle itself is etymologically somewhat different. Seiler (1960) notes that both Old and Young Avestan were characterized by the extremely frequent usage9 of a particle ya-, which inflected for gender, number and case, though the shape of the forms is not at issue here. It was used in the sense of a relative pronoun, introducing clauses with overt predicates as well as clauses lacking them: of the 330 examples investigated by Seiler, almost 60% (197) lack a finite verb in the relative clause (Seiler 1960: 57). In many examples, the particle ya- does
.â•… Note that in the Digor dialect of Ossetic, there is a prenominal marker i, used as a marker of ‘definiteness’ (Thordarsson 1989: 468), which evidently goes back to the Old Iranian relativizer *ya(h). Again this example demonstrates that there is no necessity for the old relativizer to develop into a linker of the Persian type; it may instead (or additionally) become a determiner of some sort. .â•… Seiler (1960) does not supply figures on the frequency of ya-. Frequency data on the Avesta is now available in Doctor (2004), but regrettably, the author fails to supply a lemmatized frequency list, so that the total frequency of ya- must be extracted by totalling the individual frequencies of all the members of the paradigm. My preliminary efforts in this direction indicate that forms of ya- are certainly among the 20 most frequent words in the Avesta.
 Geoffrey Haig
show the agreement properties expected of a relative pronoun (Seiler (1960: 75); simplified glosses have been supplied): (53) apaši vazaite arštiš [yąm aŋhayeiti avi.miδriš] backwards fly:pres:3s lance(nom) ╛╛↜渀屮ya:acc hurl:pres.3s Mithra-Foe ‘the lance flies backwards, which the Mithra-Foe hurls’
Here yąm takes its case (Accusative) from the relative clause verb aŋhayeiti ‘hurls’, not from its antecedent arštiš. However, elsewhere it agrees in case with its antecedent, rather than taking case from the embedded verb. The following example contains two instances of ya-, the first corresponding to a headless relative (‘he who …, whosoever’). The second is of greater interest: (54) yō miδram aiwi.družaiti yim vacahinәm ya:m:nom contract:acc break:pres ya:acc through.word:acc ‘Whoever breaks the contract which (is) through word’ (i.e. a spoken agreement)’
Two points are important here: first, the phrase introduced by the second ya- is separated from its antecedent (miδram), and this is in fact very common. Second, it agrees with its antecedent in case. Both in terms of distance from its antecedent, and of case agreement, it does not conform to the expectations of a relative pronoun. This type of construction is referred to by Seiler (1960: 175) as an “appositive construction”; the phrase introduced by ya- is a kind of parenthetic insertion, only loosely integrated into the clause as a whole. While the facts from Old Iranian remain controversial, we can conclude that neither in Avestan, nor in Old Persian are the presumed ancestors of the Ezafe simply relative pronouns. In both languages there is a strong tendency to couple relativizer with demonstrative functions, reflected in their frequent use in appositive senses, and as ‘headless’ relatives. These particles thus included a nominalizing component, which has been faithfully retained in Northern Kurdish. In Persian, however, we find a reduction, involving the complete loss of autonomous demonstrative and nominalizing functions, leaving the particle now solely as a linker, dependent on a co-phrasal head noun.
5.â•… Summary and conclusion Although we do not know what the direct ancestor of the Northern Group of Kurdish was, we can reasonably assume that the ancestor language exhibited a demonstrative/ relative very similar to the Old Persian one. The developments from an assumed protoNorthern Kurdish down to present-day Northern Kurdish are summed up in Figure 1:
The Ezafe in West Iranian 
Old Iranian
NK
Rel. pron/dem./anaph. (inflects for case, number, gender)
NP-internal linker Rel. pronoun Dem./anaph. (number, gender) (number, gender) (number, gender) Tense Ezafe (BK) (number, gender)
Figure 1.╇ Functional diversification of the Ezafe from Old Iranian to Northern Kurdish
What we find in NK is that all the original functions have been retained (although morphological case distinctions on the Ezafe have been lost), and in the dialect of Bahdînî Kurdish, the Ezafe particle has expanded its functional range ‘sideways’ as it were into the domain of the Tense system.10 Otherwise, the situation in NK is remarkably reminiscent of the Old Persian system, where the Ezafe exhibited a hybrid mix of Demonstrative with Relative, as MacKenzie (1961b: 82) had already noted. The developments in Persian are shown schematically below, where we see that the polyfunctional element hya has basically lost most of its positive content, and is now reduced to the so-called ‘linking’ function in Modern Persian. The diachronic developments in Persian can roughly be characterized as the complete loss of (traditional) functional categories, leading in a sense to a functional dead end, but not complete disappearance of the particle. The developments are summed up in Figure 2: Old Iranian
Rel. pron/dem./anaph. (inflects for case, number, gender)
Mod. Persian
NP-internal linker (uninflected)
Figure 2.╇ Functional atrophy of the Ezafe from Old Iranian to Modern Persian
The history of the Ezafe in West Iranian is a good illustration of Lass’s notion of exaptation. The term is a metaphor from evolutionary biology, used by Lass (1990) to describe instances of once-productive morphology which, through changes Â�elsewhere .â•… The additional development in ÊT discussed in connection with (30) above are not included in Figure 1; graphically they would involve an additional arrow from the Tense Ezafe to a particle which has lost gender and number features.
 Geoffrey Haig
in the system, become functionally redundant. Such productive morphology, particularly if it is frequent, will generally not simply disappear. Instead, it attracts as it were a new function. The development of the Ezafe in Persian can be seen in this light: with the intrusion of a more general complementizer ke, and the wholesale loss of gender and case distinctions in the language, the Ezafe lost its relativizing, its anaphoric, and its complementizer functions, yet it remained in situ in the NP, becoming what can be informally circumscribed as a “Linker”, though the nature of its function continues to be hotly debated, as we saw in Section 3.1. Indeed, the question of what function it has may in fact be beside the point; it is there simply as a historical relic. Likewise, the use of the Tense Ezafe in Bahdînî Kurdish can be seen in this light: the tense/aspect value conveyed by the marker was not previously available, and is not part of the tense/aspect system of related dialects. But when the pragmatically-marked construction (left-dislocated topic) became the pragmatically unmarked means of making certain stative/locative statements, the Ezafe marker lost its anaphoric function and was literally left stranded in the resultant construction (see Section 2.3). And again, it came to be associated with a new and hitherto unavailable functional distinction. Why did the Ezafes in two closely related languages undergo such contradictory developments? There is no certain answer to this question, although it is almost certainly linked to the presence vs. lack of number and gender distinctions. An Ezafe particle that is capable of expressing gender and number has a greater chance of retaining anaphoric and demonstrative functions than one that has lost the features of gender and number. There is presumably some measure of critical phonological and functional mass that enables a polyfunctional particle either to expand its domain, or, when it drops below that threshold, to withdraw into a restricted syntactic and prosodic niche, as is the case of the Ezafe in Persian.
Abbreviations abl acc adp aux comp compl cop def dem ez
Ablative Accusative Adposition Auxiliary Comparative Complementizer Copula Definite Demonstrative Ezafe particle
f fut ind indef irr m neg neut nom obl
Feminine Future tense Indicative Indefinite Irrealis Masculine Negation Neuter Nominative Oblique
pl poss pred pres prev pst ptcpl s
Plural Possessive Predicate Present Preverbal particle Past Participle Singular
The Ezafe in West Iranian 
References Axenov, Serge. 2006. The Balochi Language of Turkmenistan. A Corpus-based Grammatical Description. Uppsala Universitet: Acta Universitatis Upsaliensis. Bailey, Denise. 2005. A Comparative Study of Grammatical Relations in Northern Kurdish. MA thesis, Seminar für Iranistik, University of Göttingen. Bedir Khan, Emir Djeladet & Lescot, Roger. 1986. Kurdische Grammatik. Kurmancî Dialekt. Bonn: Verlag für Kultur und Wissenschaft. (Translation of the French original from 1970). Blau, Joyce & Barak, Veysi. 1999. Manuel de kurde kurmanji. Paris: L’Harmattan. Blau, Joyce. 1975. Le Kurde de ‘Amādiya et de Djabal Sindjār. Paris: Klincksieck. Borjian, Habib & Borjian, Miryam. 2007. Marriage rites in South Caspian villages: Ethnographic and linguistic materials from Māzandarān. Oriental Archive. Quarterly Journal of African and Asian Studies 75: 191–214. Cheung, Chi Hang Candice. 2006. Chinese ‘modifying’ de as a case marker. Paper presented at WECOL, October 27–29, 2006. Corbett, Greville. 2003. Agreement: The range of the phenomenon and the principles of the Surrey Database of Agreement. Transactions of the Philological Society 101: 155–202. Culicover, Peter & Jackendoff, Ray. 2005. Simpler Syntax. Cambridge: CUP. den Dikken, Marcel & Singhapreecha, Pornsiri. 2004. Complex noun phrases and linkers. Syntax 7(1): 1–54. Doctor, Raiomund. 2004. The Avesta: A Lexico-statistical Analysis. Leuven: Peeters. Ghomeshi, Gila. 1997. Non-Projecting nouns and the Ezafe construction in Persian. Natural Language and Linguistic Theory 15(4): 729–788. Gildea, Spike. 1993. The development of tense markers from demonstrative pronouns in Panare (Cariban). Studies in Language 17(1): 53–73. Haider, Hubert & Zwanziger, Ronald. 1984. Relatively attributive. The eä zāfe-construction from Old Iranian to Modern Persian. In Historical Syntax, Jacek Fisiak (ed.), 137–172. Berlin: Mouton. Haig, Geoffrey & Matras, Yaron. 2002. Kurdish linguistics: A brief overview. Sprachtypologie und Universalienforschung 55: 3–14. Haig, Geoffrey. 1998. On the interaction of morphological and syntactic ergativity: Lessons from Kurdish. Lingua 105: 149–173. Haig, Geoffrey. 2007. The emergence of the Tense Ezafe in Bahdînî Kurdish. Paper presented at the Second International Conference on Iranian Linguistics, University of Hamburg, August 17–19. Haig, Geoffrey. 2008. Alignment Change in Iranian Languages: A Construction Grammar Approach. Berlin: Mouton. Hassan, Dawud. 2006. Polysemy in Kurdish. Lalesh 5: 6–19. Hiruri, Newzad. Series of articles on Bahdinî Kurdish. Nefel Karimi, Yadigar. 2007. Kurdish Ezafe constructions: Implications for DP structure. Lingua 117: 2159–2177. Keenan, Edward & Comrie, Bernard. 1977. Noun phrase accessibility and Universal Grammar. Linguistic Inquiry 8: 63–99. Kent, Roland. 1953. Old Persian. Grammar, Texts, Lexicon. New Haven CT: American Oriental Society. Kent, Roland. 1944. The Old Persian relative and article. Language 20: 1–10.
 Geoffrey Haig Larson, Richard & Yamakido, Hiroko. 2005. Ezafe and the deep position of nominal modifiers. Paper presented at the Barcelona Workshop on Adjectives and Adverbs. Universitat Pompeu Fabra, Barcelona, March 18. Lass, Roger. 1990. How to do things with junk: Exaptation in language evolution. Journal of Linguistics 26: 79–102. Li, Charles N. & Thompson, Sandra A. 1977. A mechanism for the development of copula morphemes. In Mechanisms of Syntactic Change, Li, Charles N. (ed.), 419–444. Austin TX: University of Texas. MacKenzie, David. 1961a. Kurdish Dialect Studies, Vol. 1. London: OUP. MacKenzie, David. 1961b. The origins of Kurdish. Transactions of the Philological Society 63(1): 68–86. MacKenzie, David. 1962. Kurdish Dialect Studies, Vol. 2. London: OUP. Mosel, Ulrike. 1984. Tolai Syntax and its Historical Development. Canberra: ANU Press. Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 62: 56–119. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Parker, Steve. 1999. On the behavior of definite articles in Chamicuro. Language 75(3): 552–562. Paul, Ludwig. 1998. Zazaki. Grammatik und Versuch einer Dialektologie. Wiesbaden: Reichert. Rizgar, Baran. 1996. Learn Kurdish. A Multi-level Course in Kurmanji. London: Lithosphere. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: CUP. Samiian, Vida. 1994. The Ezafe Construction: Some Implications for the Theory of X-bar Syntax. In Persian Studies in North America, Medhi Marashi (ed.), 17–41. Maryland: Iranbooks. Samvelian, Pollet. 2007. A (phrasal) affix analysis of the Persian Ezafe. Journal of Linguistics 43: 605–645. Seiler, Hans-Jakob. 1960. Relativsatz, Attribut und Apposition. Wiesbaden: Harrassowitz. Thordarsson, Fridrik. 1989. Ossetic. In Compendium Linguarum Iranicarum, Rüdiger Schmitt (ed.), 456–479. Wiesbaden: Reichert. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], Maria Jose López-Couso & Elena Seoane, (eds), in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins. Yap, Foong Ha, Choi, Pik-ling & Cheung, Kam-siu. 2010. Delexicalizing di: How a Chinese noun has evolved into an attitudinal nominalizer. In Formal Evidence in Grammaticalization Research [Typologial Studies in Language 94], An Van linden, Jean-Christophe Verstraete & Kristin Davidse (eds), 63–92. Amsterdam: John Benjamins.
part iv
Korean and Japanese languages
Nominalization and stance marking in Korean* Seongha Rhee
Hankuk University of Foreign Studies This paper investigates the stance-marking function of nominalizers in Korean. This function is a product of grammaticalization processes whereby the nominalizers acquired special meanings from subjectification and intersubjectification of the interlocutors. Diverse stances are marked by these nominalizer-derived endings. Of the diverse types of stances, epistemic stances, which indicate the proposition’s status with respect to the speaker’s knowledge state, are the most prominent type, including addressee confirmation, approval/suggestion giving, conviction, self-assurance, prediction, conjecture, exclamation, etc. Also widely used are attitudinal stances, which signal the speaker’s attitude toward the addressee, including friendliness, promissive, intention, etc. On the other hand, the category of emotional stances seems to be a minor one as there is only one such stance, i.e. regret.
1.╅ Introduction Over the years, the large number and variable status of Korean nominalizers have led to controversy over the exact number of nominalizers in Korean. There are currently believed to be about twenty forms in Modern Korean that are generally recognized to function as nominalizers. Each form has a different level of productivity, but some carry doubts as to their status in this category. Since the category of nominalizers was first discussed in earnest from a modern linguistic viewpoint in Joo (1910: 102), there has accumulated a large body of literature describing the characteristics of these multiple forms that seemingly overcrowd a single functional category. Despite much attention to nominalizers, however, the �analyses
*╇ This research was supported by a Hankuk University of Foreign Studies research fund. Special thanks go to Foong Ha Yap, Karen Grunow-Hårsta, Janick Wrona, Shaun Manning, Lyman McLallen and anonymous reviewers for their comments and constructive criticisms. Some issues and suggestions could not be fully addressed due to space limitation. All Â�remaining errors are mine.
 Seongha Rhee
have been largely restricted to synchronic descriptions (but see Hong 1983a & b; Hong 1957; Kim 1978 for studies from a diachronic perspective) with a few more recent works from a panchronic perspective (e.g. Park 1999; Seo 2003; Shin 2005a; Rhee 2008). There has been no research that investigated nominalization with respect to stance marking,1 and this paper intends to fill this gap by addressing the issues involving the newly emerging stance-marking function in the course of grammaticalization of these nominalizers. The organization of this paper is as follows: Section 2 reviews the state of affairs of the nominalizers in historical and contemporary Korean; Section 3 discusses the referring and expressive functions carried by nominalizers; Section 4 discusses the functions of speaker’s stance carried by each nominalizer and the mechanisms involved in their emergence; and Section 5 summarizes the discussion and concludes the paper.
2.â•… Nominalizers in Korean 2.1â•… Nominalizers in history In the history of Korean, numerous nominalizers came into existence, with many either becoming defunct or shifting to other functions (Rhee 2008). According to Kang (1993) there were twenty-two to twenty-six nominalizing suffixes (plus their allomorphs and simple phonological variants) through Late Middle Korean (15th–16th century) and Early Modern Korean (the 17th–19th century). There were a particularly large number of nominalizers that derived the names referring to concrete entities. However, these have generally been excluded in the discussion of nominalizers, and despite their obvious merits for discussion, an exploration into this special group of nominalizers is beyond our immediate concern. The nominalizers that have been frequently discussed are -l, -m, -n, -i, -ki, -ti, -ci, and -kes. Historically there have been changes in productivity. For instance, -m was the most productive until the beginning of the 20th century, and -ki acquired primacy after -m (Rhee 2008). In Modern Korean -kes is the most productive nominalizer. It is also to be noted that -ki, -ti, and -ci are historically related as phonological variants. In Modern Korean, however, -ti is defunct (except for certain dialectal uses) and -ki and -ci are functionally separate forms with different distributions, the latter being restricted to negative sentences and the former elsewhere.2 .â•… Shin (2005b) and Yap et al. (2004) are notable exceptions. They discuss nominalizers and stance-marking from crosslinguistic perspectives including Korean. .â•… By virtue of their productivity the discussion in this paper focuses primarily on the functions of two nominalizers, -ci and -kes in Sections 3 and 4.
Nominalization and stance marking in Korean 
Unlike the nominalizers -m and -ki mentioned above, which are phonologically simple and do not require complex morphosyntactic operations for derivation of a nominal form, -kes is phonologically more complex and requires complex morphosyntactic operations for nominalization, i.e. it requires the use of adnominalizers. This is because kes started its life as a full-fledged noun denoting ‘thing, skin, surface’ (Hong 1983a), which became a defective noun from Middle Korean (Huh 1983; Jeong 1987; Lee 1988), labeled as such because it is so semantically weak that it normally cannot be used independently. Though semantically bleached, syntactically it still functions as the head noun of a nominalized string. In addition to -kes, there are many forms that belong to this subcategory of nominalizers that acquired the function as a result of bleached meaning while they were defective nouns. With the exception of -kes, these nominalizers are relatively unproductive. Despite extensive semantic bleaching, many of these nominalizers still retain vestiges of their nominal semantics and thus their use is restricted to specific contexts, often forming collocations. For this particular class of nominalizers the use of adnominalizers is mandatory because the verbal predicates need to be turned into modifiers of the defective noun. There are three adnominalizers in modern Korean: -n for anterior, -nun for simultaneous, and -l for prospective.
2.2â•… Nominalizers in Modern Korean In Modern Korean, nominalizers in active use are -m, -ki, -ci, and -kes. By virtue of their high productivity, these forms may be called the primary nominalizers. These primary nominalizers, with the exception of -kes, are mostly opaque with respect to their source lexemes and they share a characteristic that they derive nominals directly from verbs and adjectives. On the other hand, the secondary (or less productive) nominalizers are the defective nouns that function as nominalizers (N.B. -kes is also a defective noun but is used very productively).3 The defective nouns that participate in the formation of the secondary nominalizers, i.e. those of lower productivity, are numerous, as shown in the partial list in (1) with their lexical meanings.4
.â•… As an anonymous reviewer points out, tey ‘place’ is productive. Despite its relative productivity, however, the token frequency of tey in the 13 million word KAIST KORTERM Corpus is only about 11,200 (inclusive of its functions of defective noun and of being part of grammaticalized markers), whereas that of kes and its variants in comparable functions amounts to about 150,800. .â•… The number of defective nouns varies depending on the researcher’s subjective judgment on defectiveness of the ‘nouniness’, or nominal characteristics, of the forms concerned. The defectiveness criteria are usually based on morphosyntactic autonomy, but researchers use the autonomy with differing levels of rigor and thus the number of the members in this
 Seongha Rhee
(1)
cikyeng phan seym nawi cham thong tey
‘domain’ ‘venue/situation’ ‘calculation’ ‘margin’ ‘time’ ‘case, barrel’ ‘place’
nolus phok the palam ttaymwun yang cek
‘role’ ‘width’ ‘lot/foundation’ ‘wind’ ‘reason’ ‘appearance’ ‘time’
cengto phwum moyang pep mwulyep chek tongan
‘degree’ ‘appearance’ ‘appearance’ ‘law’ ‘vicinity’ ‘pretence’ ‘duration’
Even though the lexical meanings of these nouns are substantive, when they are used as nominalizers, their meanings are considerably general. The level of semantic generality as a nominalizer varies, and their respective productivity is widely different. When these defective nouns are used as nominalizers, their typical constructions are as exemplified in (2).
(2) Typical Constructions: XP-Adnominalizer-Defective Noun-Copula
XP-n moyang-i-: [XP-ant.adn-appearance-cop] XP-nun seym-i-: [XP-simul.adn-calculation-cop] XP-l cikyeng-i-: [XP-pros.adn-domain-cop]
‘be of the appearance of XP, seem to be XP’ ‘be equal to XP’ ‘be in the state of XP’
Despite the formal complexity of the construction, the semantics of the entire construction is barely more than that of a copula.5 Even though the kind of adnominalizers employed contribute to making semantic differences on the meaning of the construction, i.e. focus on the effect of the past event in the case of anterior adnominal -n, the on-going event or state with the simultaneous adnominal -nun, and future-orientation with the prospective adnominal -l, there is a considerable neutralization of temporal relationship when the adnominalizers become a part of nominalizing constructions.6
category differs. For instance, Huh (1995) lists as many as 99 defective nouns, whereas Ko (1989[1970]), Lee (1988), Ahn (2001) and others list around 50 defective nouns. The nomenclature for the defective noun category also varies: some, highlighting the absence of their semantic content, label it ‘formal nouns’, whereas some others, highlighting their morphosyntactic dependency, label it ‘dependent nouns’. .â•… As Foong Ha Yap (p.c.) points out, this phenomenon is intriguing in that these constructions are basically equative constructions, where the clause is nominalized and is treated as identifiable with an entity referred to by the defective noun. The development of these equative constructions from syntactic constructions involves syntactic upgrading whereby the subject of the embedded clause became reanalyzed as the main clause subject (see 4.3.2 for a similar phenomenon involving -kes). .â•… There is often fossilization in choices of adnominalizers depending on the defective nouns: cikyeng ‘domain’, nawi ‘margin’, the ‘lot/foundation’, etc. tend to exclusively take the prospective adnominalizer -l, and phan ‘venue’, palam ‘wind’, thong ‘case/barrel’, etc. tend to exclusively take the simultaneous adnominalizer -nun.
Nominalization and stance marking in Korean 
The semantic differences between the copula and the sentential ending with the nominalizer are that these have added nuances derivable from the adnominalizer and the defective noun.
3.â•… Functions of nominalizers As is well discussed in Yap and Matthews (2008), Korean nominalizers are highly polyfunctional. Such diverse functions may be classified into two major groups: referential functions and expressive functions. Their referential functions may be categorized in terms of the ontological dimensions of the semantics of the nouns derived through nominalizing processes, i.e. the abstractness of the nominal semantics, following Lyons (1977) and Dik (1997). Expressive functions largely refer to their functions as sentential endings and connectives marking the speaker’s epistemic stances.
3.1â•… Referring functions Nominalizers -i, -m, -ki and -kes can be used to derive nouns designating first order entities (i.e. spatial entities such as individuals, things and places).7 (3)
khokkil-i kuli-m ponpo-ki mek-ul-kes
‘elephant’ (Lit. ‘one (with) the nose being long’) ‘picture’ (Lit. ‘one of drawing’) ‘example’ (Lit. ‘one to see (as) a model’) ‘food’ (Lit. ‘thing to eat’)
Nominalizers -i, -m, -ki, and -kes are used to derive nouns designating second order entities (i.e. temporal entities such as actions, processes and states). (4)
tonpel-i talli-m tenci-ki cwuk-nun-kes
‘earning, money-making’ ‘running’ ‘throwing’ ‘dying, to die’
The nominalizers that originate from defective nouns tend to refer to second order entities (states). This has to do with the fact that the defective nouns, though consider-
.â•… In addition to -i, nominalizers that derive first order entities are particularly numerous. Most of these nominalizers lost productivity and the derived nominals are highly fossilized. cf. Kang (1993: 369–391) for examples of twenty-six nominalizers in Early Modern Korean. To limit the research focus, this paper does not address the issues involving such nominalizers.
 Seongha Rhee
ably bleached in terms of their semantics, tend to be semantically neutralized to mean ‘situation’ or ‘state’ as shown in the following examples.8
(5) a.
-lcikyengi- (< cikyeng ‘domain’) b. -nnolusi- (< nolus ‘role’)
ku-nun cwuk-lcikyengi-ta he-top die-end-dec ‘He is in trouble.’
c.
-lphani- (< phan ‘venue’)
ton-ul ta nalli-lphani-ta money-acc all fly-end-dec ‘(He) is about to lose all money.’
cengmal hwangtangha-nnolusi-ta truly be.embarrassing-end-dec ‘It is really embarrassing.’
d. -nphoki- (< phok ‘width’) ku-nun wancenhi cwuk-unphoki-ta he-top completely die-end-dec ‘He is as good as dead.’
The meaning of example (5a), where the defective noun cikyeng ‘domain’ is used, is something like “He is in the situation of trouble,” derived from the literal meaning of “(It) is the domain in which he will die.” Likewise, (5b) is “I’m really embarrassed” from the literal translation of “(It) is (my) role of being really embarrassed.” Here the words cikyeng ‘domain’ and nolus ‘role’ do not have much lexical content. The absence of such lexical meaning is well illustrated by the fact that there is no subject-predicate correspondence, i.e. in examples that have an overt sentential subject, such as (5a) and (5d), the subject ‘he’ can be equated with neither ‘a domain’ nor ‘a width’, respectively. These nominalizing defective nouns refer to a state in which the sentential subject is situated. Nominalizers -m, -ki and -kes can also derive nominals that designate third order entities, i.e. propositions. (6) a.
na-nun ku-ka mwucoyha-m-ul cheum-pwuthe I-top he-nom be.innocent-nomz-acc beginning-from
kwutkey mit-ess-ta firmly believe-pst-dec
‘I firmly believed from the beginning that he was innocent.’
b. ku-ka phyenhi calcinay-koiss-ki-lul pala-n-ta he-nom comfortably get.along-prog-nomz-acc hope-pres-dec ‘(I) hope that he is living comfortably.’
.╅ The sentential endings glossed as end in these examples containing nominalizers are periphrastic and polymorphemic, and thus, as was suggested by a reviewer, could be further broken down for morphemic gloss. Since they function as penultimate sentential endings they were treated in glossing as if they were single units. Since the penultimate endings have not been differentiated from ultimate endings, some peculiarities such as duplicated function markers (-end-end) occur as well.
Nominalization and stance marking in Korean 
c.
nay-ka ne-lul salangha-nun kes-un motwu I-nom you-acc love-simul.adn nomz-top everyone
ta a-n-ta all know-pres-dec
‘Everyone knows that I love you.’
In example (6a) the nominalizer -m transforms the sentence into a nominal proposition, i.e. ‘he was innocent’ into ‘that he was innocent’. The fact that this constituent is of nominal character is evidenced by the replaceability of the nominalized proposition into the structures of different syntactic levels with absolute semantic equivalence, i.e. ku-ka mwucoyha-m [he-nom be.innocent-nomz] ‘that he is innocent’ with ku-uy mwucoyha-m [he-gen be.innocent-nomz] ‘his being innocent’ or even with ku-uy mwucoy [he-gen innocence] ‘his innocence’.9 The similar syntactic flexibility largely applies to the nominalizers -ki and -kes in (6b) and (6c).
3.2â•… Expressive functions Expressive functions carried by the nominalizers involve grammaticalization of these nominalizers into sentential endings. The functions of the sentential endings are further divided into two subcategories: regular and ‘bullet-point’ endings.
3.2.1â•… Regular sentential endings The emergence of sentential endings from nominalizers seems to be a very recent development. The issue of whether some of these forms are full-fledged sentential endings is potentially controversial, with the exception of -ci-endings that are considered full-fledged sentential endings. In other words, from a morpho-syntactic perspective, some of these are simply nominalized constituents, analogous to English expressions No smoking, or No loitering. Examples of the -m-endings are as follows:10 (7) a.
cinipha-ci mos ha-m b. ilpangthonghayng-i-m enter-comp neg do-end one.way.thoroughfare-cop-end ‘Do not enter./No entering.’ ‘(This road) is one-way.’
Example (7a), marked with the nominalizer -m, resembles English No entering or No trespassing in many ways: in morphosyntax in that both are verb-derived gerundival .â•… The verb ha- in mwucoyha- ‘be innocent’ is a light verb semantically equivalent to the English be or do. Even though mwucoyha- in (6a) is in non-finite form and thus compatible across tenses, it may be finite, e.g. mwucoyha-ess-m with the past tense marker -ess, in which case it specifically refers to the past state of affairs. .â•… Incidentally, the complementizer -ci in (7a) also functions as a nominalizer, the objective complement of the verb ha- ‘do’ and may be optionally followed by an accusative marker -lul.
 Seongha Rhee
nominals, and in pragmatics in that both are used as posted signs for warning or directive (see below). In the same manner, example (7b) is literally ‘Being a one-way thoroughfare’, and is semantically equivalent to the English One way. This type of nominalized sentence commonly occurs as a posted sign. As posted signs, the illocutionary force of prohibition or demand of compliance is fully conventionalized as part of the semantics of this -m ending. Alternatives include the use of pure nominals e.g. Cinip kumci ‘Entrance prohibition’, or Ilpangthonghayng ‘Oneway thoroughfare’, etc., and the use of -kes ending, which carry stronger illocutionary forces than the -m ending (cf. discussion below). The counterpart sentences with a regular declarative ending, e.g. -pnita, lack such force, and the sense of prohibition in them is available only through inference, especially with (7b). A similar situation is observed with the nominalizer -ki as shown in the two examples in (8). (8) a.
nolli-ki eps-ki b. ssuleyki an peli-ki ridicule-nomz not.exist-end waste neg throw.away-end ‘Don’t ridicule (me)!/No ridiculing.’ ‘No littering.’
The functions of the two examples in (8) are warning or prohibition, similar to those in (7). The difference between nominalizers -m in (7) and -ki in (8) is that the -m-derived concept is more concrete and the forms using it carry more direct force, whereas the -ki-derived concept is more abstract and the forms using it carry more indirect force. Therefore, prohibition by using -m in (7) is more strict and inarguable and demands compliance (especially with the prohibitive imperative -mos ham), whereas the prohibition using -ki in (8) is more suggestive, soliciting co-operation. In contrast with -m and -ki, -ci has the established status as a sentential-ending. Its distinct status as an unequivocal sentential ender leads some scholars to doubt that the sentential ending -ci may not be of nominalizer origin. This is understandable considering that use of a nominalizer as a sentential ender does not seem natural unless a dynamic view of grammar, such as one in the grammaticalization framework, is adopted. Rhee (2004), drawing upon historical and dialectal data, shows that the nominalizers -ki and -ci and the prohibitive connective -ci are of a common origin. Since some of the -ci-marked examples are undoubtedly of prohibitive sentence origin (cf. -ci in (7a) above), thus establishing a connection between prohibitive connective (or complementizer), nominalizer and sentential ending, -ci as a sentential ending places itself in the category of sentential endings that originated from nominalizers. The examples in (9) illustrate the use of -ci as a sentential ending. (9) a. nay mal an tut-ko mos payki-ci my word neg listen-and neg bear-end ‘You can’t endure without following my instruction./You can’t help but obey me.’
Nominalization and stance marking in Korean 
b. ku salam cham coh-un salam-i-ci the person truly be.good-adn person-cop-end ‘He is a nice person indeed!’ c.
ese cip-ey ka-ci quick house-to go-end ‘Why don’t you go home right away!’
The examples in (9) contain the sentential ending -ci and carry emphatic meaning as compared with the counterparts with a regular declarative sentential ending. Emphasis is derivable from the speaker’s conviction. For instance, (9a) is what a conceited speaker would say: the speaker is so thoroughly prepared there can be no disobedience on the part of the addressee. Similarly, his being a nice person in (9b) is unarguable, and the speaker’s suggestion to the addressee to return home in (9c) is undoubtedly the best option available for the addressee (see 4.2 for a discussion of stance marking). The next category is the end markers involving -kes and its phonologically reduced variant -ke. As indicated earlier, since -kes is originally a noun and still carries the overt morphosyntactic traces as such, it requires an adnominalizer to precede it. There are two types of sentential endings that involve -kes: one where the sentence ends with -kes as in (10), and the other where -kes simply constitutes a part of the constellation of sentential ending morphemes as in (11) (note the optional deletion of /s/ from kes.)11 (10) a.
nayil chayk sa-lkes b. wuhoycen ha-cima-lkes tomorrow book buy-end right.turn do-proh-end ‘Buy a/the book tomorrow.’ ‘Do not turn right.’
(11) a.
nayil-un pi-ka o-lkesi-ta b. ta cwuki-epeli-lkei-a tomorrow-top rain-nom come-fut-dec all kill-perf-end-end ‘It will rain tomorrow.’ ‘I’ll kill you/them all!’
c.
kuttay ku-ka o-nkei-a then he-nom come-end-end ‘Right then he came.’
Examples in (10), where sentences end with -kes, resemble the uses of -m in (7), i.e. they carry strong illocutionary force. On the other hand, examples in (11), in which -kes occurs with the copular i- ‘be’ (note that the morphemic gloss is given unanalytically), carry diverse meanings: -lkesi- in (11a) has a prediction meaning;
.â•… The category of ‘sentential endings’ is heterogeneous in that it includes diverse grammatical categories that form constellations of ultimate and penultimate verbal morphology for tense, aspect, modality, sentence type, honorification, politeness, etc.
 Seongha Rhee
-lkei- in (11b), an intention, promise or announcement meaning; and -nkei- in (11c) an emphatic meaning.12 Still another type of -kes-derived ending is the one where the nominalizer -kes occurs with the accusative marker -l at the end. The fact that this type of sentential ending contains the accusative marker as the final element of the sentential-ending constellation suggests that this usage is developed from main clause ellipsis,13 leaving only the constituent marked by the accusative marker -l. The following examples show the use of -kes-derived sentential endings.14 (12) a.
pap-ul ta mek-unkel food-acc all eat-end ‘(I) ate it all! (What can I do since there’s nothing left?)’
b. kuttay yelsimhi ha-lkel then diligently do-end ‘(I) should have studied harder then.’
Example (12a) has a helplessness meaning, and (12b) has a regret meaning, even though the two meanings are conceptually closely related. The emergence of such meanings has to do with their syntactic origin, i.e. main clause ellipsis. From a morphosyntactic point of view, these sentences are merely sentential fragments, i.e. accusativemarked arguments serving as the theme of the elided main clause verb. Native speakers tend to reconstruct the elliptical structures as those that may be translated as ‘What should I do with…’ (where the with-phrase counterpart can be marked by an accusative in Korean), or ‘I should have dealt more wisely with the situation of…’ (where the constituent led by ‘with the situation of ’ can be marked by an accusative in Korean).
3.2.2â•… ‘Bullet-point’ sentential endings The primary nominalizers -m, -ki, and -kes (but not -ci) have a special usage that resembles regular sentential endings, yet differ in that this usage is restricted to a par.â•… The meaning of the -nkei- ending in (11c) is very elusive, but undoubtedly has the nuance of mirativity (DeLancey 2002), roughly translating into “You know what? What happened is he walked in right then!” This type of mirative meaning is directly related to “It is that…,” “It was that…,” and “It will be that…,” which are literal renderings of–kes endings, -nunke(s)i-, -nke(s)i-, and -lke(s)i-, respectively. .â•… See Ohori (1995) and Shibasaki (2007) for similar phenomena involving suspended clauses in Japanese. .â•… The historical and conceptual connection between the conservative -l+ke(s)+l [pros. adn+thing+acc] and the ending -lkel derived from it is evident in the fact that Koreans frequently make mistakes with respect to spacing: the former needs an orthographic space between the prospective adnominal -l and -kes, whereas the latter does not (cf. Lee & Lee 2003[2001]: 371–372).
Nominalization and stance marking in Korean 
ticular writing style, i.e. summary writing, or ‘bullet-point’ style, as shown in the following examples.15 (13) a.
salinpem-i kamok-eyse tomangchi-m murder.convict-nom prison-from flee-end ‘A murder convict flees from the prison.’
b. John-i Mary swukcey-lul tow-acwu-ki John-nom Mary homework-acc help-benef-end ‘John (needs) to help Mary finish her homework.’ c.
thoykun kil-ey seythakmwul chac-a-o-lkes after.work way-at laundray pick.up-nf-come-end ‘(I/You…) need to pick up laundry on the way home from work.’
Even though English translations are given in full sentence forms, given the absence of unique bullet-point styles in English, these Korean sentences are commonly perceived as quasi-sentences, as evidenced by the fact that they are not likely to occur in spoken language. Despite the apparent incompleteness, these sentences are optimal in enumeration of similar items. For instance, example (13a) may be followed by a listing of -m-ending sentences describing events involving a prison-break in chronological order. Likewise, example (13b) may be found among the -ki-ending sentences that describe, for instance, tasks for a group of individuals including John; and example (13c) may be one of many items written in a notebook as a reminder of what to do today. Similar to posted signs ending with the primary nominalizers (cf. 3.2.1), sentences with bullet-point endings may alternatively be substituted, though often not preferred, by sentences with pure nominal endings, in which case the nouns used preferably are Sino-Korean nouns. For instance, the last word tomangchi-m in (13a) may be naturally replaced by a noun thalchwul ‘escape’, and, though less naturally, chacao-lkes in (13c) by hoyswu ‘retrieval’. However, this option is not available when the main verb does not have a semantically synonymous and stylistically equivalent Sino-Korean counterpart nominal as in (13b). These pure-nominal endings are often used as titles of newspaper articles for brevity. Among notable aspects of this usage of bullet-point sentences is the fact that the endings show a division-of-labor phenomenon in terms of temporal distinctions and differential levels of illocutionary force. As for the temporal distinctions, the ending
.╅ The bullet-point endings typically involve uninflected verb forms. However, the -m-ending is perfectly compatible with verbs in past tense form for its semantic congruity with past event designation; -ki-ending is less so; and -lkes ending is not compatible (see discussion below).
 Seongha Rhee
-m is employed in enumerating past time events, as for instance, in listing historical events. Therefore, example (13a) describes an event that already took place. On the other hand, the bullet-point ending -kes is used for listing future events. Even though the -kes ending can be combined with all three temporally distinguished adnominalizers (i.e. anterior, simultaneous, and prospective) elsewhere, its use in bullet-point endings is restricted to the ending that combines with the prospective -l only. This has to do with the fact that -kes ending (i.e. -lkes) is employed in enumerating future events, such as those that may appear on a to-do list. The bullet-point ending -ki is relatively independent from temporal restrictions, but the time reference with the -ki ending tends to be non-past, since it tends to carry imperative overtone soliciting compliance (cf. 3.2.1). Due to this characteristic -ki is more common in future time reference, even though such tendency has not been fully conventionalized. The uses of -ki and -kes, both used in future time reference, exhibit differences in terms of their illocutionary force: the level of obligation on the sentential subject is higher with -kes than with -ki. Therefore, a list of things to do for children in a nursery may employ the -ki ending in order to avoid an authoritative and imposing tone; whereas a list of things to do on a job description may employ the -kes ending in order to promote the maximal compliance on the part of the employees, as shown in the following examples. (14) a.
in a nursery toilet
kkaykkushi son ssis-ki thoroughly hand wash-end ‘Wash hands thoroughly’
b. in a restaurant toilet congepwen-un yongpyen-hwu son-ul ssis-ulkes employee-top toilet.use-after hand-acc wash-end ‘Employees must wash hands after using toilet (= before returning to work).’
4.â•… Speaker-stance marking In the preceding discussion we have looked at the diverse functions carried by the nominalizers in the referring and expressive domains. The development of the nominalizers into sentential endings, a syntagmatic location where diverse modality markers occur, gave rise to an important function, the speaker-stance marking, as shall be explicated in the following discussion.
Nominalization and stance marking in Korean 
4.1â•… Speaker-stance The term ‘stance’ is used to refer to diverse concepts. Englebretson (2007: 2–3), in expressing terminological indeterminacy, notes that the understanding of ‘stance’ that emerges is indeed heterogeneous and variegated, but always intimately bound up with the pragmatic and social aspects of human conduct. In addition to the multiplicity of what stance means, there is diversity in nomenclature for similar concepts. For instance, Biber et al. (1999), Dancygier and Sweetser (2005), and many others use the term ‘stance’; Nuyts (2001) calls it ‘addressee orientation’; and Smith (2002) calls it ‘point-of-view’. Generally speaking, these scholars included attitudinal and epistemic components in stance. The notion of evidentiality, which indicates the source of information, is also included in the discussion of stance marking. The notion of stance, as used here, covers a wide range of speaker’s emotional, attitudinal, epistemic, evidential states. Attitudinal stance largely refers to the speaker’s attitude toward the addressee and is thus interactional. Epistemic stance relates to the speaker’s knowledge state regarding the veracity of the proposition. Emotional stance is related to the speaker’s positive, negative, or neutral emotion toward the proposition or the event denoted by it. Evidential stance refers to the source of information, or how the speaker acquired the information. Within each category diverse terms have been used as grammatical labels of the markers. Some of the labels are used mutually interchangeably, while others have overlapping functions. The labels of grammatical markers that occur in the discussion of stance in the literature may be listed as in Figure 1 according to their subcategories.
Stance
Attitudinal
Epistemic
Emotional
Evidential
Cold Friendly Enthusiastic Indifferent Helpless Promissive Intentional Directive Encouraging
Certain Likely Possible Impossible Conviction Suppositive …
Positive Negative Neutral …
Direct Indirect Inferential Reportative Nonvisual Witness Non-witness Sensory Assumed …
Figure 1.╇ Subcategories of stance
 Seongha Rhee
4.2â•… Stance-marking by sentential endings 4.2.1â•… -ci-endings: -ci, -ciyo, and -cyo One of the most distinctive functional characteristics of the -ci-ending is that the speaker’s friendliness toward the addressee is added to the sentence meaning. This friendliness is interactional and belongs to attitudinal stance-marking. Within the category of endings involving the nominalizer -ci are -ci, -ciyo, and -cyo, the last being an orthographically contracted form of -ciyo, consisting of sentential ending -ci plus -yo, the politeness ending. These two variant forms are often indistinguishable in spoken language but these differences become visible in written language. In written language the reduced form -cyo has a slightly different meaning than -ciyo. Among the differences is the fact that -cyo is distinctively feminine, if not exclusively so.16 The functions of -ci and -ciyo/–cyo endings are so diverse, subtle and interrelated that their typology cannot be easily established. The labels given here, therefore, should not be considered as those of clearly delineable categories. The most common function of -ci and -ciyo/-cyo endings is addressee confirmation as shown in (18).17 (15) Addressee Confirmation [=I know this, and I know you know this, too.] a.
kyelhonha-nikka hayngpokha-ci? marry-reas be.happy-end? ‘You are happy because you got married, right?’
b. kusalam cham chakha-cyo s/he very be.nice-end ‘S/he is a nice person indeed.’
Addressee confirmation is not conspicuously discernible from form-based translation. According to the real meaning of -ci and -cyo in these examples in free translation, the speaker is saying “I know this, and I know you know this, too.” The addressee confirmation function is applicable not only to interrogative sentences like (15a), but also to a declarative as in (15b). Since the speaker assumes shared knowledge about a proposition, these -ci/-cyo marked sentences generally have a friendly tone (more so with -cyo for its association with feminine speech). However, confirmation is the primary function of this form, and, therefore, friendliness may be overridden and -ci- or -ciyo-marked endings may be used in non-friendly contexts such as
.â•… An anonymous reviewer points out, accurately and insightfully, that even in spoken language -ciyo sounds more polite than -cyo and that the use of -ciyo is more acceptable to seniors than -cyo. .â•… See Lee (1991: 436–454) for discourse functions of -ci, which he labeled as ‘Committal’ suffix.
Nominalization and stance marking in Korean 
police interrogations. The interrogator in this case is using the common ground strategy to have the interrogated acknowledge that the interrogator already has knowledge of it (see below for its relation to conviction). The next category is the function of marking feigned or exaggerated friendliness or companionship. This is closely related to the first function, but what is assumed is not shared knowledge, but shared interest. It is as if the speaker is saying, “I’m telling you this, because we are friends.” Some examples are given in (16). (16) Feigned/Exaggerated Friendship [=I’m saying this, because we are friends.] a.
way amwu-to an o-cyo? why anyone-even neg come-end ‘Why is nobody coming?’
b. yesnalyesnal-ey han namwukkwun-i sal-ass-ci old.day.old.day-at one wood.cutter-nom live-pst-end ‘Once upon a time, there lived a wood-cutter.’ c.
ku tongan-to motwu cal cinay-si-ess-ciyo? That time-also all well live-hon-pst-end ‘You and yours have been all doing fine?’
Some Korean dictionaries for learners of Korean as a second/foreign language (e.g. Lee & Lee 2003[2001]: 795–796) call this function “adding a friendly tone.” In addition, it has a delicate shade of meaning that may be translated, for (16a), into “Why is nobody coming? Isn’t it strange? I know you are wondering about it, too.” Example (16b) is typically used in storytelling. By employing the -ci-ending, the speaker is saying, “As you would expect, once upon a time, there lived a wood-cutter.” The child listening to this story is invited to the vivid storyline about to unfold, as if the story-teller knew that the child has a shared interest and shared expectation. As the essential function of this -ci is attracting the addressee into the world of the speaker/writer, letter writers often use this ending in one of the first sentences of salutation in a letter, as in (16c), to create a shared world between the writer and the reader. The next category is approval- or suggestion-giving. The meaning added by this stance marker can be phrased as: “I know it is a good option for you and I agree with you,” as shown in the following examples. (17) Approval/Suggestion-Giving [=I’m telling you because I know it is a good option for you.] a.
kulehkey ha-si-cyo b. han can ha-ko ka-si-cyo thus do-hon-end one glass do-and go-hon-end ‘Do so.’ ‘Why don’t you have a drink before you leave?’
 Seongha Rhee
The meaning with this type is not entirely uniform: the approval meaning as in (17a) is more of agreement or approval as its core meaning and, typically, the speaker assumes that he or she already knows the addressee’s intention; whereas the suggestion meaning as in (17b) does not do so. The next category is conviction. The morpheme -ci, when used either as a nominalizer, connective, or a sentential ending, is strongly associated with the speaker’s epistemic stance of conviction and prediction. The determinativeness of conviction in the epistemic domain also leads to the development of the attitudinal stance of the promissive. The conviction meaning is closely related to the previously discussed addressee-confirmation as shown in the following examples that contrast the stancemarking -ci-ending and the regular -e-ending: (18) a.
ney-ka ku-lul cwuki-ess-ci? you-nom he-acc kill-pst-end ‘You killed him, didn’t you!’ (I already know you did!)
b. ney-ka ku-lul cwuki-ess-e? you-nom he-acc kill-pst-end ‘Did you kill him?’
Closely related to the conviction meaning is self-assurance. Since this marker of epistemic stance of certainty is used in monologue style only, there is no -ciyo or -cyo counterpart, which is understandable considering that -ciyo and -cyo are intersubjective forms containing the politeness marker -yo. The following are examples of -ci ending with the self-assurance function. (19) Self-Assurance [=I’m saying this simply because I want to remind myself.] a.
nay-ka ile-l cwu-l al-ass-ci I-nom be.so-adn nomz-acc know-pst-end ‘I knew that it would turn out like this.’
b. nay-ka wusan kacyeo-ass-ci?! I-nom umbrella bring-pst-end ‘I brought my umbrella, didn’t I! (I knew that it would rain…)’
This usage is similar to the previously mentioned addressee-confirmation function. This is also a kind of confirmation directed to the speaker himself or herself. This usage is common in sentences that may be translated as “I knew it! I knew it!” as in (19a), or in self-confirming as in (19b). Employing a regular declarative ending in place of a self-assurance ending either results in the complete loss of such a self-assurance meaning, or renders such sentences unacceptable. The next category within the category of epistemic stance is prediction, which also carries the speaker’s stance of certainty about the truth of the proposition being
Nominalization and stance marking in Korean 
presented. It is worth noting incidentally that some scholars consider that conjecture, doubt, or lack of confidence is the basic meaning of the morpheme -ci. For example, Martin (1992) calls it a ‘suspective’ marker. This is in direct contrast with another group of scholars (Ko 1976; Jang 1973; Suh 1984; Rhee 2008) who think that conviction is its core meaning.18 The apparent lack of confidence is due to the suppositive morpheme -kyess- that often gets attached to -ci.19 The following example is one in which -ci is used with the suppositive -keyss- (the future marker), which weakens the degree of certainty. (20) Prediction [=I’m saying this because I know this will be the case.] chacha ichye-ci-kyess-ci gradually forget-pass-fut-end ‘(It) will gradually be forgotten.’
Example (20) seems to have the conjecture meaning, but it still contains a considerable degree of the speaker’s conviction. If it is contrasted with the same sentence without the final -ci, it becomes a plain declarative sentence without any direct speakerinvolvement in terms of projection of the speaker’s epistemic stance of conviction. The following pair of examples illustrates the point. (21) a.
kulen il-un kot ichi-eci-ci such matter-top soon forget-pass-end ‘Such things get forgotten soon.’ (I can tell this from my experience…)
b. kulen il-un kot ichi-eci-e such matter-top soon forget-pass-end ‘Such things get forgotten soon.’
The next category is the function of promissive, the speaker’s attitudinal stance. The development of this function is related to the speaker’s self-affirmation, in the sense that the speaker, by employing this speech act, is affirming his or her intention, which
.â•… Rhee (2004) shows that -ci, in its sentential end-marking function, encodes the speaker’s emotion, belief, or determinative attitude, which corresponds to the speaker’s epistemic and attitudinal stances under discussion here. .â•… It is noteworthy that the conviction meaning of -ci is prominent when it is used in non-future tense sentences. This force weakens when it is used with -kyess-, a suppositive and future tense marker. Similarly, when the futuristic morpheme (the prospective adnominalizer -l) is used, especially with the know-class verbs (i.e. al- ‘know’ and molu- ‘not know’), the emphatic meaning also seems to decrease. Such impression seems to be due to the indeterminacy inherent in any future (or prospective) event, not to the semantics of -ci. See below for more discussion.
 Seongha Rhee
binds the speaker himself or herself to carry out the proposed action. This marker -ci is simply the marker of self-determination. This usage, however, is different in that it is fundamentally interactional, and therefore, substitution of this ending with a regular ending makes the sentences awkward and devoid of such interactional meaning. The following are some examples of Promissive marking by -ci and -ciyo. (22) Promissive (Determinative) [=I’m telling you this because I’m determined to do it.] a.
nay-ka tow-acwu-ci b. cey-ka thaywu-etuli-ciyo I-nom help-benef-end pol.I-nom carry-hon.benef-end ‘I will help you.’ ‘I will give you a ride (Polite).’
The final category is exclamation. Exclamation is an utterance about something the speaker has newly perceived and is note-worthy, and therefore it is fundamentally based on the speaker’s epistemic stance. This newly-perceived and noteworthy event creates a degree of conflict with the previous knowledge state, and the degree of conflict determines the degree of exclamation. When exclamatory sentences are marked with a -ci-ending, it is usually used without the politeness marker at the end as shown in (23), but if there is a politeness marker, the function of which is turning a sentence into an interactional one, the sentence is perceived as one that seeks the addressee’s confirmation. (23) Exclamation [=I’m saying this, because it is truly amazing.] a.
cip-un elmana coh-un kos-i-nci! home-top how be.good-adn place-cop-end ‘What a good place a home is!’
b. elmana hwangtangha-n il-i-nci! how be.embarrassing-adn matter-cop-end ‘How embarrassing it is!’
This exclamatory use of -ci typically involves the use of wh-words. In terms of its source construction, the exclamatory sentence originated from a rhetorical negative sentence translatable as “I don’t know how…,” from which the main clause subject and verb, “I don’t know,” are omitted. In other words, only the theme argument of the sentence survives the ellipsis and constitutes an exclamatory sentence.20
.â•… As a reviewer points out the verb in the elided main clause may not be restricted to al- ‘know’ as nominalized structures with -ci frequently collocate with verbs of cognition and utterance. It is true that the source structures may vary depending on the context.
Nominalization and stance marking in Korean 
4.2.2â•… -kes-endings Nominalizer -kes is very versatile in function and highly productive in use in Modern Korean. The token frequency of the substantive kes, inclusive of nominal and nominalizer uses, ranks first (National Academy of Korean, 2002).21 As for the sentential endings, there are diverse forms of -kes because it cannot be used alone but must be combined with adnominalizers and other morphemes to form such endings. There is variation with respect to phonological shapes: kes and ke, even though each of these variants has a level of fossilization in certain constructions and thus cannot always be used interchangeably. The sentential endings involving -kes frequently use the prospective adnominal -l in the construction. The most basic function of -lkes, the combination of -l and -kes, is to mark the future. Therefore, most meanings listed here are closely related to the temporal notion of futurity. Among the diverse stance-marking functions of -kes sentential ending, the epistemic stance marking of conviction is the primary one. In particular, the function of marking the speaker’s evaluation of the state of affairs, as shown in the following examples, occurs with a high frequency. (24) a.
Conjecture
b. Prediction
ku-nun acik ca-koiss-ulkei-a he-top still sleep-prog-end-end ‘He should still be sleeping.’
ku salam mos o-lkel that person neg come-end ‘I bet he cannot come.’
As shown in the examples above, the first category of the -kes functions has to do with the speaker’s estimation with differing levels of conviction: from mere conjecture by -lkei- to prediction with conviction by -lkel. In the case of conjecture, as in (24a), the sentence does not seem to carry the speaker’s conviction prominently. Example (24b), on the other hand, has strong conviction meaning. Since conviction exists to greater or lesser degrees, the strength of conviction may be modified by degree adverbials such as celtaylo ‘absolutely’, ama(to) ‘perhaps’, etc. As was pointed out earlier, the original form of the ending -lkel has the Accusative marker -l at the very end (cf. 3.2.1). This is an elliptical structure which says, “With the fact that he will not be able to come, you think there is a possibility,” where the main clause is omitted and the counterpart construction of the with-phrase occurs with an accusative marker.
.╅ The frequency survey by the National Academy of Korean is based on a corpus of about 1.5 million word texts taken from Korean textbooks, literary works, newspaper, magazines, drama texts, TV talk shows, etc. The source texts are predominantly literary, spoken data accounting for less than seven percent.
 Seongha Rhee
This type of sentential ending is closely related to the next functional category: regret-marking, typically in past-referring context, as shown in the following example. (25) Regret naccam-ina ca-lkel nap-select sleep-end ‘I should have taken a nap!’
Also an ellipsis-based construction, the reconstructed sentence of example (25) is one that can be translated as “I did something else in a situation where I should have taken a nap,” where the phrase that can be translated into ‘in a situation…’ surfaces as an accusative marked constituent. Both prediction-marking and regret-marking functions marked by -lkel develop from elliptical structures, as is obvious from its final element -l, the accusative marker (Ahn 2000: 429). By ending an utterance with the accusative-marked sentential fragment, the speaker invites the addressee’s reconstruction of the elided main clause, and the variability in inferences gives rise to diverse meanings, such as prediction and regret. This type of semanticization of inferred senses from ellipsis is widely discussed in Rhee (2002).22 The next category has to do with the determinative attitude of the speaker: intention and promise (Park 1999), as exemplified in the following examples. (26) a.
Intention (Determinative)
b. Promissive (Determinative)
kkok sengkongha-lkei-a surely succeed-end-end ‘I will surely succeed.’
nay-ka towacwu-lkey I-nom help-end ‘I will help you.’
In the above example, the endings -lkei- and -lkey mark the speaker’s determinative attitude. It is interesting to note that historically -lkey originated from -lkeia (thus explaining why -lkey cannot be further followed by the poly-functional ultimate sentential ending -a). However, these forms are separately specialized and, thus, they are no longer interchangeable. Furthermore, these endings cannot be replaced with other declarative endings (such as -ta or -pnita) without the loss of the function of marking determinative attitude. Though similar in that they both encode determinative attitudinal stance, one fundamental difference is that assertions of intention tend
.╅ Ending an utterance with an accusative-marked sentential element is a very unusual speech act in this verb final language, and the addressees are naturally invited to look for what remains unsaid.
Nominalization and stance marking in Korean 
to be directed to the speaker himself or herself, whereas assertions of promissive are directed to the addressee. Finally, a -kes-derived sentential ending can carry the function of announcing information in an addressee-friendly tone, a function of marking the speaker’s attitudinal stance, as shown in the following example. (27) Announcement (Addressee-Friendly) hwacangsil-ey tanyeo-lkey restroom-to visit-end ‘I’ll be back after I use the restroom.’
It is obvious that this addressee-friendly announcement function is closely related to the previously discussed functions of marking intention and promissive. However, this announcement function is different from them in that this is employed in an interactional situation (unlike intention) and without making a promise (unlike promissive). Since it is possible that the speaker makes a promise about, for instance, making a trip to the restroom and returning in a hypothetically constructed situation, it seems plausible to plot promissive and announcement on a single continuum of interaction with varying degrees of addressee-directed obligation.
4.3â•… Mechanisms 4.3.1â•… -ci: Emphasis and friendliness The functions of the nominalizer -ci, when used as a sentential ending, can best be characterized as the markers of emphasis and friendliness. The development of such stance meanings seems to have followed the track shown in (28). (28)
NEGATIVE nomz
CONTRASTIVE nomz conn
SUGGESTIVE conn end
CONVICTION end
Historically, the nominalizer -ci first appears in negative constructions, at first interchangeably with -ki. It seems that -ci absorbed the emphatic meaning from this negative construction. The status of -ci in terms of grammatical category is definitely that of a nominalizer, which makes the -ci-marked constituent the theme argument of the main verb, as evidenced by the fact that it is often marked by an accusative marker -l.23 Examples of -ci in negative constructions are as follows.
.â•… The main verb anh- in (29b), glossed here as ‘be not’ (containing a copula) was in fact derived from ani ha- [neg do], where ha- ‘do’ is a light verb whose semantics has bleached
 Seongha Rhee
(29) Negative Construction a.
kekcengha-ci ma-seyyo b. tangsin-ul salangha-ci anh-ayo worry-nomz stop-end you-acc love-nomz be.not-end ‘Don’t worry.’ ‘I don’t love you.’
Then, later -ci is found in contrastive constructions. Because of the contrast effect in the context, the emphasis meaning seems to have been fortified. It is due to the fact that for two items to be placed for contrast, the items need to be well-defined to avoid one blending into the other. This is well illustrated in the following examples, where ‘staying home’ vs. ‘coming,’ and ‘violence’ vs. ‘play,’ being contrasted are distinct (at least in the conceptualization of the speaker). (30) Contrastive Construction a.
cip-ey iss-ci way o-ass-eyo? home-at exist-conn why come-pst-end? ‘Why did you come, not staying home?’ Lit. ‘Stay home, why did (you) come?’
b. kuke-n phokhayng-i-ci cangnan-i ani-ta it-top violence-cop-conn play-nom be.not-dec ‘It is violence, not a playful action.’
In a contrastive construction the focus is on the first clause, marked by the connective -ci, and the second clause is often merely an appendage, just added for stylistic reasons of contrast. This is very prominent in such doublet phrases as saltunci cwuktunci ‘either you live or die,’ mektunci maltunci ‘either you eat or not’, satunci maltunci ‘either you buy it or not’, etc., where the speaker’s concern is always on the first element, not on the second. The next step is the suggestive or encouragement function, usually conveying the speaker’s strong conviction for the benefit from the proposed action. In other words, the speaker presents one of the two contrasted items as the more favorable option, as in the following examples (‘resting’ over ‘doing something’ or ‘divorce’ over ‘staying in wedlock’).
to that of a copula. It is also to be noted in this context that, due to the structural similarity of -ci with other non-finite connectives such as -a, -key, and -ko, and its exclusive use in negative constructions, -ci is sometimes classified as a negative connective (or prohibitive connective). It seems not uncommon for a nominalizer to take up the connective or linker function as attested by other nominalizers (e.g. Yap & Wang this volume, Yap & Kwok 2005; Yap & Matthews 2008; Kratochvíl this volume; Noonan 2008, among others).
Nominalization and stance marking in Korean 
(31) Strong Encouragement a.
com swi-ci [kulay(yo)(?)] a.little rest-conn [do.so(end)] ‘Why don’t you take some rest?’
b. ihon-ul ha-ci [kulay-ss-e(yo)(?)] divorce-acc do-conn [do.so-pst-end] ‘Why didn’t you get a divorce?’
From this stage, the nominalizer -ci may appear as the sentential ending or with an appended simple expression like “do you do so?” at the end (indicated by angled brackets), which later seems to have been elided. This interrogative sentence becomes obsolete when the question becomes obvious from the first part already through repetition. With this process through which the main clause of the interrogative sentence becomes unnecessary, the grammatical status of the connective -ci also undergoes change. In other words, the previous [A or B] contrastive construction is routinized with [A or not A]. When the routinization has proceeded further, the construction [A or] can be construed as [A or not A], a process wherein the connective becomes reanalyzed as a sentential ending. When the emphasis meaning becomes the primary meaning of the sentential ending, it comes to be used in many different contexts simply for emphatic value as noted in the preceding discussion. The issues of subjectification and intersubjectification involved in the emergence of diverse stance-marking functions deserve our attention. When we use the nominalizer, we are basically claiming, as a rough generalization, that “X is a thing” at a certain conceptual level, be it a first-order, second-order, or third-order entity. It appears, from the progression of semantic change involving the nominalizer -ci, that we attribute some subjective meaning to it, “The thing X is what matters to you.” This process seems to be operative in the development of addressee-confirmation function and self-assurance function, the latter being typically used in a situation where the speaker talks to himself or herself as if he or she is the addressee. Furthermore, we attribute more intersubjective and interactional meaning to a linguistic form as if saying, “X is what matters to us.” This is apparent in the development of friendliness becoming associated with a form. Through this path of subjective and interactional meaning attribution, the nominalizers become the markers of emotive or affective stance marker. This can be diagrammatically presented as (32).24
.╅ An anonymous reviewer points out that the subjective and intersubjective stages are not clearly separable in that subjectivity inevitably involves interactive meaning. The distinction is admittedly non-discrete but the separation as intended here is solely based on the fact that intersubjective meanings have to do more with the mutual interaction of both interlocutors,
 Seongha Rhee
(32) Objective > Subjective > Intersubjective (Interactional) “X is a thing” “X is what matters to you” “X is what matters to us”
4.3.2â•… -kes: Emphasis, conviction, intention and commitment A development similar to the one that involved -ci can be observed with the nominalizer -kes as well. The nominalizer -kes has many semantic and functional characteristics such as emphasis, conviction, intention, and commitment. As was noted earlier, the lexical meaning of kes is ‘thing’. When -kes is used to nominalize a proposition, among others, we are claiming an identity relation between the proposition and a thing, by virtue of the adnominalizers that connect a proposition with the nominalizer -kes, which in many instances still retains its lexical semantics ‘thing’ (see for example, Simpson & Wu 2001). Syntactically, the copula can be in the form of either present tense -i- or past tense -iess-. This has to do with the fact that the future tense is marked with a form that involves -kes-, i.e. -lkesi-, and thus if a future marking on the this copula were allowed, the form would be somewhat awkward like -kesilkesi- in which kes is repeated. The fact that only present and past forms are allowed with -kes- has to do with the speaker’s desire to present the ‘thing’ as a concrete entity that either exists or existed, in contrast to something that will exist. In these structures even the thing to exist in the future is expressed as one that already exists. This point is illustrated in the following diagrammatic presentation, given in English structure for succinctness. (33)
a. b. c. d. e. f.
It is a thing that was…â•… [Present-tense copula + Anterior Adnominal] It is a thing that is…â•… [Present-tense copula + Simultaneous Adnominal] It is a thing that will…â•… [Present-tense copula + Prospective Adnominal] It was a thing that had been…â•… [Past-tense copula + Anterior Adnominal] It was a thing that was…â•… [Past-tense copula + Simultaneous Adnominal] It was a thing that would…â•… [Past-tense copula + Prospective Adnominal]
As is evident from (33), the ‘thing’ as a substitute of an entity, an event, or a proposition is expressed as one existing in the present or the past. This state of affairs, i.e. regarding the nominalized entity as something in existence and not as one of future possibility must have contributed to the creation of such emphatic meanings as emphasis, conviction, etc. and of the speaker’s present psychological states such as intention, commitment, etc. This type of objectification or reification process can be illustrated by such a simple example as “It will rain” in (34).
whereas subjective meanings largely arise from shifting objective meanings to abstract, mental, and evaluative ones, in the mind of the speaker/writer.
Nominalization and stance marking in Korean 
(34) a.
pi-ka o-ri-la rain-nom come-fut-dec ‘It will rain.’
b. pi-ka o-l-kes-i-ta rain-nom come-pros.adn-thing-cop-dec ‘It will rain.’ (Lit. ‘(It) is the thing that the rain will come.’)
Example (34a) is the typical future tense sentence before the modern future tense marker -lkesi- came into existence. The complex morphemic gloss of -l-kes-i- is considered a single future tense marking morpheme in Modern Korean. Incidentally, this is an excellent example of syntactic upgrading, where the subject of the embedded clause becomes the main clause subject through reanalysis (Heine et al. 1991: 169–170).25 The objectification or reification process can be diagrammatically presented as in (35). (35)
“It will rain” Proposition IDENTITY Abstract RELATION Subjective Judgment ––––––––>
“a thing” Entity Tangible Objective Description of Existence
The diagram in (35) shows that what the speaker wants to say is that according to his or her judgment, it will rain. But in the real world, the future belongs to the unknown, existing only as a world of possibility, and thus nothing that belongs to the future is a reality, so it is not certain if it will truly rain. Here, the speaker objectifies the abstract situation as if it were a concrete, tangible object that could be referred to as ‘a thing’. The emphasis effect is obtained from the act of regarding a proposition as a concrete entity. This type of borrowing of a linguistic form denoting ‘a concrete entity’ to mark emphasis may be conceptually well-motivated, and is attested in other languages, as shown by the use of thing in English. For instance, the grammaticalizing constructions in English The thing is that…, Thing is that…, Thing is… etc. are all newly emerging means of expressing emphasis (Kim 2003). As a matter of fact, there has been a claim that the Korean -kes sentential ending originated from the English It-is-that construction through calquing by translators (Choi 1994). Furthermore,
.â•… This type of reanalysis changes a sentence like It is not Peter (who) wants a dog into Peter does not want a dog in Teso, an Eastern Nilotic language. In exact parallelism, the Korean sentence It is that he will come [originally: Ø is the thing that he will come] has been reanalyzed as He will come, as shown in the present discussion. Syntactic upgrading refers to the status change of the embedded subject into the main clause subject, or the shift from the subordinate clause to the main clause.
 Seongha Rhee
concreteness carrying over to the grammatical meaning is also observed in the usage of English relative pronouns: the relative pronoun that, originated from the demonstrative pronoun that, which, as opposed to other pronouns like which, who, etc., is required in the structure where the antecedent is pinpointed by such devices as the restricting adverb the only, the superlative, etc. (e.g. the only thing that…, the fastest car that…, etc.). There is a subjectification and intersubjectification process as well in the development of -kes-derived stance marking. Through objectification the meaning of emphasis is derived (cf. It will rain vs. Thing is that it will rain, in the preceding discussion). This emphatic meaning develops into conviction. Conviction inherently makes reference to the speaker’s belief state, and thus the acquisition of the conviction meaning involves subjectification. From a cursory search of the corpus data from the 21st century Sejong Project, it is found that from Korea’s first modern daily newspaper, The Independence (Toklipsinmwun), in 1896, -kesi- with an adnominalizer occurs as the sentential ending frequently. The earliest instances of these tend to have emphatic meanings and conviction meanings as shown in the following examples. (36) a.
pyeng-i cyenyem-i toy-nankesi-la disease-nom contagion-nom become-end-end ‘(and thus) the disease becomes transferred’ (1896, Toknipsinmwun 6769)
b. philkyeng mwusam skatalk-i is-nankesi-la absolutely certain reason-nom exist-end-end ‘There must be some reason.’ (1986, Toknipsinmwun 6858)
When conviction further develops into intention, the speaker is adding a more subjective meaning of conation. When the sentential subject is sentient, and the event denoted by the clause is emphatic or of conviction, especially when the sentence makes reference to a future event (by using the prospective adnominal -l), inferring a conative meaning seems to be a natural process. Furthermore, when the sentential subject is the first person, and involves interaction with the addressee, the intention becomes commitment. This last stage involves intersubjectification. The speaker uses his intention with respect to the real-world situation from the perspective of the speaker and the addressee. The -kes-ending with the promissive meaning is exemplified by the following example. (37) cal po-acwu-lkesi-ni yemnye mal-ko well see-benef-end-as worry stop-and ‘Don’t worry because I will take care of you, and…’ (1896, Toknipsinmwun 6972)
The foregoing discussion on the genesis of diverse meanings involving the -kes-ending can be diagrammatically presented as in (38).
Nominalization and stance marking in Korean 
(38) EMPHASIS
CONVICTION
INTENTION
COMMITMENT
5.â•… Conclusion The Korean language has had many nominalizers throughout its history, and each of them has had its own journey of rise and fall through time. They came from diverse lexical sources, including defective nouns, but they constitute a single grammatical category called nominalizers. This paper examined the extended functions of nominalizers in the speaker’s stance-marking, largely as sentential endings, in addition to their primary referring functions of deriving first order, second order, and third order entities. Sentential endings involving nominalizers consist of two major categories: regular sentential endings and ‘bullet-point’ sentential endings. Stances of the speaker, as proposed here, include such subcategory stances as attitudinal, epistemic, emotional, and evidential stances. The stances marked by the sentential endings derived from nominalizers are diverse, encompassing epistemic, attitudinal, and emotional stances. They do not seem to involve evidential stances. Epistemic stances, which indicate the proposition’s status with respect to the speaker’s knowledge state, seem to be the most prominent type of stances marked by the sentential endings involving nominalizers. They include such meanings as addressee confirmation, conviction, approval/suggestion giving, selfassurance, prediction, conjecture, exclamation, etc. Also widely used are attitudinal stances, which indicate the speaker’s attitude toward the addressee, including friendliness, promissive, intention, etc. On the other hand, the category of emotional stances seems to be a minor one as there is only one such stance, regret.26 This paper also shows that the diversity of the functions is due to the contributions of the participating linguistic forms and the idiosyncrasies of the source constructions, and at the same time, to the interactions of the nominalizer with the participating adnominalizers that are sensitive to the aspectual-temporal distinctions. Furthermore, it is suggested that subjectification and intersubjectification that operated in the process of semantic-functional extension also enabled the emergence of diverse functions of the sentential endings as stance-markers. .â•… The claim that the category of emotional stances is minor is simply based on the lack of functional diversity, i.e. the number of subcategories, not on overall use (i.e. token) frequencies. A quantitative analysis involving natural discourse data is called for to determine the extent of usage beyond the mere number of subcategories.
 Seongha Rhee
Abbreviations acc adn ant benef conn cop dec end fut hon neg nf nom
accusative adnominalizer anterior benefactive connective copula declarative sentential ending future honorific negative non-finite nominative
nomz pass perf pol pres prog proh pros pst reas select simul top
nominalizer passive perfective polite present progressive prohibitive prospective past-tense reason selective simultaneous topic-marker
References Ahn, Hyo-Kyung. 2001. Hyentaykwukeuy Uyconmyengsa Yenkwu (A Study on Bound Nouns in Modern Korean). Seoul: Yeklak Publishing. Ahn, Jeong-A. 2000. Hyentaykwukeuy mwunpephwa hyensang (Grammaticalization in Modern Korean). In Hyentaykwukeuy Hyengsengkwa Pyenchen 1 (Development and change of Modern Korean), Jongseon Hong (ed.), 415–466. Seoul: Pakiceng Publishing. Biber, Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan & Finegan, Edward. 1999. Longman Grammar of Spoken and Written English. London: Longman. Choi, Kyu-Il. 1994. Hankwuke Kes-uy sayongkwa uymi kinung (3) (Uses and semantic functions of kes in Korean (3)). Jeju National University Journal of Humanities and Social Â�Sciences 38: 17–58. Dancygier, Barbara & Sweetser, Eve. 2005. Mental Spaces in Grammar: Conditional Constructions. Cambridge: CUP. DeLancey, Scott. 2002. The mirative and evidentiality. Journal of Pragmatics 33(3): 369–382. Dik, Simon C. 1997. The Theory of Functional Grammar, Part 1: The Structure of the Clause [Functional Grammar Series 20]. Berlin: Mouton de Gruyter. Englebretson, Robert. 2007. Stancetaking in discourse: An introduction. In Stancetaking in Discourse [Pragmatics and Beyond New Series 164], Robert Englebretson (ed.), 1–26. Amsterdam: John Benjamins. Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1991. From cognition to grammar – Evidence from African languages. In Approaches to Grammaticalization, Vol. 1 [Typological Studies in Language 19], Elizabeth Closs Traugott & Bernd Heine (eds), 149–187. Amsterdam: John Benjamins. Hong, Jongseon. 1983a. Myengsahwa emiuy pyenchen (Historical change of nominalizing particles). Kwukekwukmwunhak (Korean Language and Literature) 89: 31–89. Hong, Jongseon. 1983b. Myengsahwa emi -um-kwa -ki (Nominalizing particles -um and -ki). Language 8(2): 241–272.
Nominalization and stance marking in Korean 
Hong, Ki-Moon. 1957. Cosenelyeksamwunpep (Historical Linguistics of Korean). Pyongyang: Kwahakwen Publishing. Huh, Woong 1995. 20-seyki Wulimaluy Hyengthaylon (20th Century Korean Morphology). Seoul: Saem Publishing. Huh, Woong. 1983. 15-seyki Wuli Yeysmalpon (15th Century Korean Grammar). Seoul: Saem Publishing. Jang, Seok-Jin. 1973. A generative study of discourse in Korean: On connecting sentences. Language Research 9(2): 1–145. Jeong, Howan 1987. Hwuki Cwungseykwuke Uyconmyengsa Yenkwu (A Study on Defective Nouns in Late Middle Korean). Seoul: Hakmunsa Publishing. Joo, Si-Kyung. 1910. Kwuke Mwunpep (Korean Grammar). Seoul: Bakmunsekwan Publishing. Kang, Eun Kook. 1993. Cosene Cepmisauy Thongsicek Yenkwu (A Diachronic Study of Korean Suffixes). Seoul: Sekwanghakswulcalyo Publishing. Kim, Hijean. 2003. ‘The-Fact-Is-That’ and Related Constructions: A Usage-based Approach and Implications in English Education. Ph.D. dissertation, Hankuk University of Foreign Studies, Seoul. Kim, Seung-Gon. 1978. Hankwuke Cosauy Thongsicek Yenkwu (A Historical Study of Korean Particles). Seoul: Tayceykak Publishing. Ko, Yong-Kun. 1976. Hyentaykwukeuy mwuncheypepey tayhan yenkwu (A study on the stylistics in Modern Korean). Language Research 12(2): 16–53. Ko, Yong-Kun. 1989[1970]. Kwuke Hyengthaylon Yenkwu (A Study of Korean Morphology). Seoul: Seoul National University Press. Kratochvíl, František. This volume. Discourse structuring functions of Abui demonstratives. In Yap, Grunow-Hårsta & Wrona (eds). Lee, Hee Ja & Lee, Jong Hee. 2003[2001]. Emi Cosa Sacen (Dictionary of Particles) Seoul: Hankuk Publisher. Lee, Hyo Sang. 1991. Tense, Aspect, and Modality: A Discourse-pragmatic Analysis of Verbal Affixes in Korean from a Typological Perspective. Ph.D. dissertation, University of↜California, Los Angeles. Lee, Ju-Haeng. 1988. Hankwuke Uyconmyengsauy Thongsicek Yenkwu (A Diachronic Study on Korean Defective Nouns). Seoul: Hansaem Publishing. Lyons, John. 1977. Semantics. Vols. 1 & 2. Cambridge: CUP. Martin, Samuel. 1992. A Reference Grammar of Korean. Tokyo: Charles E. Tuttle. The National Academy of Korean 2002. Hyuntay Kwuke Sayong Pinto Cosa (A Survey of Use Frequency in Modern Korean). Noonan, Michael. 2008. Nominalization of Bodic languages. In Rethinking Grammaticalization: New Perspectives (Typological Studies in Language 76). LÓpez-Couso, María José & Elena Seoane (eds) in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins. Nuyts, Jan. 2001. Subjectivity as an evidential dimension in epistemic modal expressions. Journal of Pragmatics 33: 383–400. Ohori, Toshio. 1995. Remarks on suspended clauses: A contribution to Japanese phraseology. In Essays in Semantics and Pragmatics in Honor of Charles J. Fillmore [Pragmatics and Beyond New Series 32], Masayoshi Shibatani & Sandra A. Thompson (eds), 201–218. Amsterdam: John Benjamins. Park, Joo Young. 1999. A Study of Grammaticalization of kut. MA thesis, Sang Myung University, Seoul. Rhee, Seongha. 2002. From silence to grammar. Paper presented at New Reflections on Grammaticalization 2 Conference, University of Amsterdam, April 4–6.
 Seongha Rhee Rhee, Seongha. 2004. When stop goes further: From malta ‘stop’ to auxiliary verbs in Korean. Korean Language Research 13: 309–339. Rhee, Seongha. 2007a. Colloquial variation and its implications in grammaticalization theory. LACUS Forum 33: 237–246. Rhee, Seongha. 2007b. On interrelation of instrumental and ablative. Discourse and Cognition 14(3): 131–153. Rhee, Seongha. 2008. On the rise and fall of Korean nominalizers. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 239–264. Amsterdam: John Benjamins. Seo, Eun A. 2003. Change aspect of the postposition in the Korean nominalizer ‘-m, -ki’. Korean Language Research 10: 83–114. Shibasaki, Reijirou. 2007. Ellipsis and discourse-syntactic structures in Japanese interview discourse: The emergence of the evidential marker to. Language and Linguistics 8(4): 939–966. Shin, Mi Kyong. 2005a. Bridging the worlds of adnominal and pronominal: The case of Korean -n geot. Paper presented at New Reflections on Grammaticalization 3, Santiago de Compostela, Spain, July 17–20. Shin, Mi Kyong. 2005b. Development of nominalizers in some East Asian languages. MPhil thesis, Chinese University of Hong Kong. Simpson, Andrew & Wu, Xiu-Zhi Zoe. 2001. The grammaticalization of formal nouns and nominalizers in Chinese, Japanese and Korean. In Language Change in East Asia, Thomas E. McAuley (ed.), 250–283. Richmond: Curzon. Smith, Carlotta. 2002. Accounting for subjectivity (Point of view). In The Legacy of Zellig Harris. Language and Information into the 21st Century, 1: Philosophy of Science, Syntax, and Semantics. Festschrift für Zellig Harris [Current Issues in Linguistic Theory 228], Bruce E. Nevin (ed.), 137–163. Amsterdam: John Benjamins. Suh, Jung-Soo. 1984. Contaypep Yenkwu (A Study on Honorification). Seoul: Hanshin Publishing. Yap, Foong Ha & Kwok, Bit-chee. 2005. Non-finitization and nominalization: A dichronic look at shifts in grounding in Classical and Modern Chinese. Paper presented at the 3rd International Conference on New Reflections on Grammaticalization 3, University of Santiago de Compostela, July 17–20. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76]. María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 309–343. Amsterdam: John Benjamins. Yap, Foong Ha, Matthews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Yap. Foong Ha & Wang, Jiao. This volume. From light noun to nominalizer and more: The grammaticalization of zhe and suo in Old and Middle Chinese. In Yap, Grunow-Hårsta & Wrona (eds).
A case of non-derived stand-alone nominalization Evidence from Japanese Janick Wrona
Kyoto University It has been argued that stand-alone nominalizations derive from a copula-type nominalization (Yap, Matthews & Horie 2004; Yap & Matthews 2008). This paper takes a detailed look at the history of a stand-alone type main-clause nominalization in Japanese and argues that there is no evidence for this type of development.Instead it is proposed that the stand-alone type is one of several uses of nominalizations. In light of evidence that stand-alone nominalizations need not be derived, several languages for which the copula-type>stand-alone development has been proposed are re-examined and the conclusion is that the non-derived view for stand-alone nominalizations should be seriously considered in the absence of evidence to the contrary.
1.â•… Introduction In numerous languages in (East) Asia, the predicate of a main clause can be nominalized with interesting semantico-pragmatic effects.1 Two ways in which this can be done are illustrated in (1) from Lahu and in (2) from Japanese.
(1) Lahu
yf la tu ve he come fut nmlz ‘it is the case that he will come’
(Matisoff, 1972: 247)
.╅ I would like to thank three anonymous reviewers for their detailed and constructive comments. I have attempted to accommodate these as much as possible. Comments from the audiences at Tohoku University (December 2006) and the Chinese University of Hong Kong (January 2007) where versions of this paper were presented also helped improve the final version. The results reported in this paper were supported by a grant-in-aid from the Japan Society for the Promotion of Science which I hereby gratefully acknowledge.
 Janick Wrona
(2) Japanese
(I woke up very early this morning). Denwa-ga natta-no dearu Phone-nom rang-nmlz cop ‘The phone rang (you see).’ (Miyazaki, Adachi, Noda & Takanashi, 2002: 230)
In Lahu, nominalising a main clause has the effect of making it a statement of fact. In Japanese a nominalized main clause can be used as an explanation as in (2), but similar constructions in other languages may have more unpredictable semantic effects. Yap, Matthews & Horie (2004) have suggested the term stance-marking to cover the range of semantico-pragmatic effects of such constructions. Apart from the semantico-pragmatic difference between (1) and (2), there is also an important syntactic difference. In Japanese, the nominalized clause is embedded under a copula; that is, the nominalized clause is strictly speaking a predicate nominal. In the example from Lahu, it is truly a nominalized main clause with no copula involved. Following Watters (2008) I will use the term stand-alone type to refer to main-clause nominalization not embedded under a copula, and the term copula type to refer to nominalized clauses embedded under a copula. Some languages opt for the copula type, others for the stand-alone type while still others may adopt both strategies, sometimes with a resulting semantico-pragmatic difference. In a diachronic context, it has been suggested that the stand-alone type of main clause nominalization, illustrated in (1), may develop from the copula type as in (2) (Noonan, 1997; Yap & Matthews, forthcoming; Yap et al., 2004). The main objective of this paper is to show that the stand-alone type of main clause nominalization need not be diachronically derived from the copula type. I illustrate this by examining the history of a stand-alone type main clause nominalization in Japanese. This type of nominalization has been associated with stance-marking since the earliest times, but there is no evidence that it was ever embedded under a copula. The conclusion is that, cross-linguistically, there must be at least two different diachronic sources for construction types like the one in (1), and it cannot be assumed a priori that they developed from copula constructions. In fact I will make the stronger claim that there is no solid evidence for the copula type>stand-alone type development. The organization of this paper is as follows: Section 2 introduces the type of nominalization that this paper is concerned with from the perspective of modern Japanese, while Section 3 deals, in some detail, with its (early) history. Section 4 offers an explanation as to why the Japanese type of nominalization discussed in this paper is exclamative. Section 5 illustrates main-clause nominalizations in a number of (East) Asian languages and in Section 6, I will re-evaluate some proposed developments of stand-alone main-clause nominalization.
A case of non-derived stand-alone nominalization 
2.╅ -sa derivatives in modern Japanese In Japanese the derivative -sa attaches to stems of morphological adjectives and converts them into nouns. Table 1 gives some examples from modern Japanese. Table 1.╇ The formation of sa-derivatives Adjective
Stem
-sa derivative
-ku adjectives
takai ‘is high’ hukai ‘is deep’
takahuka-
taka-sa ‘height’ huka-sa ‘depth’
-i adjectives
utukusii ‘is beautiful’ muzukasii ‘is difficult’
utukusimuzukasi-
utukusi-sa ‘beauty’ muzukasi-sa ‘difficulty’
adjectival nouns
kiree da ‘is pretty’ oroka da ‘is stupid’
kiree oroka
kiree-sa ‘prettiness’ oroka-sa ‘stupidity’
The derivational process is highly productive in Modern Japanese, but only applies to (morphological) adjectives. The resulting noun denotes a state. These state nominalizations on -sa are used as nouns, as in examples (3)–(8).
(3) -sa nominal as subject
utukusi-sa-ga hosii beautiful-sa-nom want ‘(I) want beauty’
(4) -sa nominal as direct object
Sono ningen-rasi-sa-o kanzi-ta dem human-like-sa-acc feel-pst (he) felt his humanness’
(Shibatani, 1990: 220)
(Martin, 1988: 909)
(5) -sa nominal as oblique object
sekkusu-no saityuu-ni sono hagesi-sa-ni awasete ongaku-o sex-gen middle-in dem intense-sa-dat match music-acc nagasu kondoomu-ga kaihatu-sa-reru play condom-nom develop-do-passive ‘A condom that plays music that matches the intensity of the intercourse will be developed’ (Google)
(6) -sa nominal in locative phrase
Keerin… no yo-sa wa…. benri-sa ni aru bicycle race gen good-sa top convenient-sa loc exist ‘The good things about bicycle race lies in the convenience’ (Martin, 1988: 910)
 Janick Wrona
(7) -sa nominal in genitive construction
detokkusu-wa utukusi-sa-no kagi detox-top beautiful-sa-gen key ‘detox(ing) (is) the key to beauty’
(Google)
(8) -sa nominal in exclamative construction
Nan to iu utukusi-sa! what comp say beautiful-sa ‘How beautiful!’
sa-derivatives are cases of lexical nominalization. Lexical nominalization is a situation where a noun that has been derived from a verb or an adjective can function as the head of a noun phrase (Comrie & Thompson, 1985). The noun resulting from this derivation has the same morpho-syntactic characteristics as non-derived nouns. Clausal nominalization, as opposed to lexical nominalization, does not exhibit any evidence that the head is a lexical noun (Comrie & Thompson, 1985: 392). On the contrary, what is nominalized is a whole clause. Though lexical nominalization and clausal nominalization are in principle distinct phenomena, it may be difficult to determine whether a given type of nominalization in a language is one or the other (cf. Comrie, 1976; Koptjevskaja-Tamm, 1993). However, sa-nominalizations are unmistakably cases of lexical nominalization, not clausal nominalization. With respect to stand-alone type main clause nominalization, there does not appear to be any difference between clausal and lexical nominalization; both types of nominalization are found as stand-alone type main clause nominalization (e.g. Lahu/ Chantyal versus Japanese/Korean). Like lexical nominalizations in general, it is possible to express the arguments of the sa-nominalized predicate. Here there are two possibilities. If the nominalized predicate is an adjective, the subject (S) is obligatorily genitivized as shown in (9). (9) nedan-no (*ga) taka-sa price-gen (*nom) high-sa ‘the highness of the price’
However, since the negative and the desiderative suffix in Japanese inflect like adjectives, a negative and desiderative predicate can be nominalized using -sa. In some cases, the arguments (S and P) will still obligatorily genitivize as in (10). (10) [saikin-no gakusei-no (*ga) hon-no (*o) ╛╛↜渀屮late-gen student-gen (*nom) book-gen (*acc) yom-ana]-sa-ni akire-ta read-neg-sa-dat surprise-pst ‘I was surprised that students nowadays do not read books’
But if the nominalized predicate is causativized (or passivized) as well, the agent (A) and the direct object (P) are expressed as in finite clauses as shown in (11).
A case of non-derived stand-alone nominalization 
(11) [sensee-ni piano-o hik-ase-ta]-sa ni ╛╛↜渀屮teacher-by piano-acc play-caus-desid-sa ‘out of desire to have the teacher play piano’
In Koptjevskaja-Tamm’s (1993) typology of the expression of arguments in action nominal constructions (a subtype of lexical nominalization), this means that (modern) Japanese sa-nominalizations instantiate Type 1 (the sentential type) and Type 4a (the double-possesive nominal type).2
3.â•… The history of -sa derivatives To my knowledge, there have been no serious attempts to posit a lexical source for the derivative -sa. The dictionary of Old Japanese Jidaibetsu Kokugo Daijiten (Zdb, 1967 s.v. -sa) suggests that it goes back to the noun sama meaning ‘appearance, direction’, but Martin (1987: 804) points out that it is not clear whether -sa is the output derived from phonological erosion or whether a noun -ma ‘room’ is suffixed to -sa. -sa derivatives of lexical adjectives are attested in the earliest texts in Japanese, but desiderative -tai and negative -nai as morphological adjectives did not come into existence until later so one does not find examples equivalent to (10) and (11) in Old Japanese.
3.1â•… Old Japanese (8th century) In Old Japanese (OJ) only lexical adjectives are derived by -sa.3 Even though desiderative -tai and negative -nai as morphological adjectives did not exist in Old Japanese, there are other auxiliaries with modal meaning that inflect like adjectives (e.g. -masi (DUB), -besi (NECC)). Whether the limitation of -sa derivatives is a genuine restriction or the result of a gap in the texts is difficult to say, but since the same limitation holds in the ensuing period, it is likely to be genuine. From the few examples found in Old Japanese, it seems that the S argument of the nominalized predicate obligatorily genitivizes. There are, however, two (near-identical) examples involving the pragmatic particle mo. (12) [imwo-ga kokoro-no subye]-mo subye na-sa ╛╛↜渀屮wife-gen heart-gen means-prt means not exist-sa ‘There is no means to my wife’s heart!’ (Man’yôshû 5.796; 18.4106)
.╅ Comrie (1976) and Comrie & Thompson (1985) also deal with the typology of argument expressions in action nominals. .╅ It should be noted that there is another Old Japanese suffix -sa that follows the Conclusive form of verbs. It is found in e.g. M 20.4514. Its meaning/function seems to be different from that of the derivative -sa and it will not be treated here.
 Janick Wrona
A construction nominalized by -sa has three functions in Old Japanese: as a subject, an object and, most commonly, as an exclamative. (13) below illustrates the subject usage. (13) aki-no ywo-no momo-ywo-no naga-sa ari-kose-nu kamo autumn-gen night-gen 100-night-gen long-sa exist-ben-neg kprt ‘Won’t you let the autumn night be 100 nights long’ (Man’yôshû 4.546)
Here nagasa ‘length’ is the subject of ar- ‘exist’. It is important to note that there is no graphic representation of -sa in (13); the reading of the character 長 as nagasa is dictated only by the reading tradition.4 Therefore, one cannot be certain that the example is genuine and that -sa derivatives can function as subjects in Old Japanese. (14) koyopi-no naga-sa ipo-ywo tugi-koso tonight-gen long-sa 500-evening continue-ben.imp ‘please continue the length of tonight for 500 nights’
(Man’yôshû 6.985)
In this example nagasa ‘length’ is the object of tugi ‘continue. Here the derivative -sa is phonographically written. In the remainder of cases of OJ -sa, the -sa derivative is used sentence-finally as an exclamative. Some examples are given below. (15) [puna-pito-wo miru]-ga tomosi-sa ╛╛↜渀屮boat-people-acc see.adn-gen enviable-sa ‘How enviable it is to see the boat-people!’
(Man’yôshû 15.3658)
(16) [ima sara-ni apu-beki yosi-no naki]-ga sabusi-sa ╛╛↜渀屮now again-temp meet-pot.adn means-gen not exist.adn-gen sad-sa ‘How sad it is there is once again no means to meet!’ (Man’yôshû 15.3734)
The examples in (15) and (16) are sentence final, and they represent the most common function of -sa-nominalized constructions in Old Japanese. The nominalized predicates tomosisa ‘enviable-SA’ and sabusisa ‘sad-SA’ in (15) and (16) take nominalized sentential arguments (in square brackets) in the genitive case (ga). These nominalized clauses correspond to the subject complement clause of the equivalent finite clause (without -sa). It is very clear that in (15) and (16), the nominalized predicate is not embedded under a copula; rather it is the syntactic main predicate. Firstly, there are no examples of -sa-nominalized clauses embedded under a copula. One possible counter-argument here is that Old Japanese does not have a (morphological stable) copula and that there is therefore a covert copula involved in constructions like (15) and (16). However, and this is the second argument, it is
.╅ Most Old Japanese texts consist of a mixture of logographic and phonographic characters. In general, only phonographically written sequences can be used as linguistic evidence. For OJ, phonographically used characters are italicized in this paper.
A case of non-derived stand-alone nominalization 
difficult to see how such a copula construction could yield an exclamative meaning. A third argument comes from a comparative perspective. It is interesting to note that in Classical Ryukyuan -sa is used solely as an exclamative and never followed by a copula. (17) mi-kutu ma-faya-sa hon-word int-fast-sa ‘the august words [are] really fast’
(Omoro Sôshi III: 88.)
In modern Ryukyuan dialects, the Conclusive form of adjectives often end in -san (e.g. Shuri-Naha takasan ‘is high’), historically a contraction of the nominalizer -sa+ar‘exist’. However, this is an innovation (Vovin, 2005), just as the Conclusive-Adnominal distinction in the Ryukyuan verbal paradigm is an innovation.
3.2â•… Early Middle Japanese (9th–12th century) In Early Middle Japanese, one still only finds -sa-nominalization of lexical adjectives. Presumably this is a genuine restriction; since there is a fair amount of material dating from this period making it less likely that the absence is due to lacunae. As in Old Japanese, one finds -sa-nominalized constructions as subjects and objects. (18) osorosi-sa-fa me nare-tare-ba horrible-sa-top eye get used to-perf-because ‘because I am used to horrors’ (Ochikubo Monogatari, maki 1, p. 56) (19) miru koto-no uresi-sa-wo tutume-ba see that-gen happy-sa-acc wrap up-if ‘If I wrapped up my happiness of seeing (you)’ (Ochikubo Monogatari, maki 2, p.125 )
In addition, one finds genitives, adverbial clauses and predicate nominals. (20) oya-no notamafu koto-wo fitaburu-ni inabi-maus-an koto-no parent-gen say(hon). thing-acc outright-adv refuse-say.conj thing-gen itoosi-sa ni tori-gataki mono-wo feel bad-sa because take-difficult think-acc ‘Because (I would) feel bad to refuse outright what my parents said, (I chose) something that was difficult to get’ (Taketori Monogatari, p. 18) (21) uki mi-kara-fa onazi nagekasi-sa ni koso miserable.adn body-abl-top same lament-sa cop.inf kprt ‘from (my) miserable self (there) is the same lamenting’ (Genji Monogatari, Miwotsukushi, p. 116)
Just after 1000 A.D., when Genji Monogatari was written, the two (historical) constituent parts of copula (nari), namely ni (glossed COP.INF above) and ari ‘exist’nari were,
 Janick Wrona
not infrequently, preserved. This is the situation in (21) with the further twist that the second part ar- ‘exist’ has dropped.5 It is important to note that, even though (21) is an example of a -sa-nominalization as a predicate nominal, it is not a copula-type main clause nominalization. Basically, (21) has the structure [NPSUBJ NP COP], whereas copula-type main clause nominalizations have the structure [main clause]NP COP. Therefore, examples like (21) are very different from the type suggested as the diachronic source of stance-marking constructions, like (1). More importantly, there are quite a few examples of exclamatives in Early Middle Japanese. (22) negafi-wo kanafuru koto-no uresi-sa. request-acc meet comp-gen happy-sa ‘How happy (I am) that my request will be met!’
[9th century] (Taketori Monogatari, p. 46)
(23) okina-no haberu yo simo kau old man-gen exist(pol) night prt thus yami-tamafu-ga wabisi-sa. stop-hon-gen terrible-sa
[10th century]
‘How terrible that the nights the old man is here will stop in this way!’ (Ochikubo Monogatari, maki 2, p. 114) (24) munasiku naru-tamafi-nuru kanasi-sa yo. [12th century] empty become-hon-perf sad-sa emp ‘How sad that (someone esteemed) has died (become empty=die)!’ (Konjaku Monogatari, p.265)
3.3â•… Late Middle Japanese (13th–17th century) The following usages are attested in Late Middle Japanese: subject, object, genitive and adverbial clauses. A single example is given below. (25) uresi-sa kagiri nasi happy-sa limit not-exist ‘There is not limit to (my) happiness’
[13th century] (Ujishûi Monogatari, p. 149)
Again, one finds an exclamative usage of -sa-nominalized constructions. This is illustrated in (26) below. (26) kaku ofasimasi-taru koto-no uresi-sa. thus be-perf comp-gen happy-sa ‘The happiness of (you) being that way’
[13th century] (Ujishûi Monogatari, p. 250)
.╅ The precise form of the existential predicate would be are, the so-called izenkei-form of ar-. It is the focus particle koso that calls for this particular form.
A case of non-derived stand-alone nominalization 
(27) fito-no koto-no asamasi-sa people-gen affair-gen shameful-sa ‘The shamefulness of people’s affairs’
14th century] (Tsurezuregusa, p. 277)
(28) mi-kage-wo wogamu arata-sa yo hon-shape-acc worship new-sa emp ‘(we) worship your divine shape anew!’ (Takasago)
[15th century]
(29) fito-ni aise-raruru koto-no uresi-sa yo [16th century] person-dat love-pass.adn fact-gen happy-sa emp ‘Oh, how happy (I am) that (I am) loved by people’ (Isoho Monogatari, p. 462)
3.4╅ Summary Table 2 below summarizes the usages of -sa derived nominal constructions over a period of almost a thousand years. Table 2.╇ Summary of usages of -sa derived nominal constructions
Subject Object Exclamative Genitive Adverbial Predicate nominal
Old Japanese
Early Middle Japanese
√ √ √
√ √ √ √ √ √
Late Middle Japanese √ √ √ √ √
From the perspective of this paper, there are two important things to note. Firstly, there is no evidence of a diachronic development leading from a -sa-nominalized clause embedded under a copula to the stand-alone nominalization (illustrated in (15) and (16) for Old Japanese). If fact, there is not attestation of -sa-nominalizations embedded under a copula at all throughout the history of the Japanese language. I argued that examples like (21) (with a -sa-nominalization as the predicate nominal) attested in Early Middle Japanese were quite different. This situation contrasts the developmental patterns described in Yap, Matthews & Horie (2004) and Yap & Matthews (forthcoming) where the stand-alone type is derived from copula constructions. Secondly, there is no evidence in Japanese that the stance-marking (exclamative) meaning of sa-nominalization as in (15) and (16) are secondary and derived. In fact, as far as one can judge from the data, -sa-nominalization seems inextricably linked to the exclamative function. In the next section I will show that it is possible to explain
 Janick Wrona
why nominalized clauses are associated with stance-marking without appealing to diachronic developments from other constructions.
4.â•… Nominalization, nominals and stance-marking In Old Japanese, there are quite a large number of sentences that consist of just a relative construction. These are assumed to be exclamative in nature. (30) patu-pana ni kwopwisiki wa-ga se first-flower cop.inf is beloved.adn I-gen husband ‘my husband, like the first flower…..’
(Man’yôshû 20.4443)
This observation was made nearly 100 years ago by the Japanese philologist Yoshio Yamada (1929[1908]), but Japanese is by no means unique in this respect. Similar observations for English have been made by Elliott (1974) and Grimshaw (1979) and in more detail by Portner & Zanuttini (2005). An example is given in (31). (31) The silly questions the students ask! (= (25) in Portner & Zanuttini, 2005)
As Portner & Zanuttini (2005) point out, the presence of a relative clause is essential to the exclamative interpretation of noun phrases. ‘Exclamative’ is here (and in Portner & Zanuttini’s work) taken as a sentence type, i.e. the conventional pairing of particular form with a particular meaning. In order for a relative construction to be interpreted as an exclamative it has to contain a full clause; otherwise they do not qualify as sentence types. Therefore, noun phrases not containing a relative clause may function as exclamations, but not as exclamatives. Nominalizations are nominal in nature while at the same time preserving clausal properties e.g. an argument-predicate structure (cf. Comrie, 1976; KoptjevskajaTamm, 1993). This dual nature endows them with the potential to function as exclamatives in a similar vein to relative constructions. I would like to suggest that the observation made for relative constructions can be extended to -sa-nominalizations if the nominalized clause fulfil the following two conditions: (1) they have a possible argument-predicate structure and (2) they are used in syntactic isolation, i.e. in nonembedded contexts. Thus, sa-nominalizations fulfilling these two conditions are conventionally associated with exclamatives as a sentence type.
5.â•… Stance-marking nominalization in other (East) Asian languages There are numerous other languages in (East) Asia where main clauses are nominalized with interesting semantico-pragmatics effects. Some examples are given in (32)–(39).
A case of non-derived stand-alone nominalization 
(32) Chantyal bәnnu-ye nal tato ta-si-wa gun-gen barrel hot become-ant-nmlz ‘The barrel of the gun had become hot!’
(Noonan, 1997: 381)
(33) Classical Tibetan gnam sdug-pa-la sky.is beautiful-nmlz-locus particle ‘What a beautiful sky’
(Beyer, 1992: 384)
(34) Lahu
f la tu ve he come fut nmlz ‘He will come’/’it is the case that he will come’
(Matisoff, 1972: 247)
(35) Korean palo kuttay cenhwa-ka o–n kes-i-ta right hen phone-nom come-ant.adn nmlz]-cop-dec ‘Right then the phone rang.’ (Rhee, 2008: 249) (36) Takale Kham sohmlo o-ra-kәi-wo three 3s-3p-eat-nmlz ‘He ate three of them’
(Watters, 2008: 32)
(37) Old Japanese wa-ga koromode-no puru toki-mo naki I-gen sleeve-gen dry.adn time-prt not exist.adn ‘My sleeves never dry!’ (Man’yôshû 10.1994) (38) Old Japanese pama-no kiywokyeku beach-gen.is clear.nmlz ‘Oh, the clearness of the beach!’
(Man’yôshû 7.1239)
(39) Early Middle Japanese nami-no tatu naru koto” to urufe-if-ite wave-gen stand cop nmlz comp worry-say-ptc “saying worriedly “the waves are really rising!”’
(Tosa Nikki, p. 33)
(40) Modern Japanese denwa-ga natta-no dearu phone-nom rang-nmlz cop ‘The phone rang (you see)’
(Miyazaki et al., 2002: 230)
The semantico-pragmatic effects of the examples are summarized in Table 3.
 Janick Wrona
Table 3.╇ Stance-marking function of nominalizations in East Asian languages No.
Language
Nominalization
Function
Source
(32) (33) (34) (35) (36) (37) (38) (39) (40)
Chantyal Classical Tibetan Lahu Korean Takale Kham Old Japanese Old Japanese Early Middle Japanese Modern Japanese
Clausal (-wa) Clausal (-pa) Clausal (-ve) Clausal (-n kes) Clausal (-wo) Clausal (adn) Clausal (-ku) Clausal (koto) Clausal (noda)
Mirative Exclamative Statement of fact Assertion of fact Backgrounding Exclamative Exclamative Exclamative Explanation
(Noonan 1997) (Beyer, 1992) (Matisoff 1972) (Rhee, 2008) (Watters, 2006) (own research) (own research) (own research) (Miyazaki et al. 2002)
With the exception of the Korean example in (35) and the modern Japanese example in (40), all of the examples above are of the stand-alone type.
6.â•… The diachronic development of stand-alone nominalizations In Section 3, I presented evidence that in Japanese, stand-alone main-clause nominalization is not derived from copula constructions. In this respect, the scenario in Japanese runs counter to recent publications arguing that the diachronic development from copula-type main-clause nominalization to stand-alone main-clause nominalization constitutes a robust grammaticalization path (or chain) (Yap & Matthews, 2008; Yap et al., 2004). At the very least then, one has to consider two possible diachronic developments of stand-alone main-clause nominalizations for a given language. One possibility is that the stand-alone type developed from the copula type as Yap & Matthews (2008) and Yap, Matthews & Horie (2004) argue is the case in many East Asian and Tibeto-Burman languages. Another possibility is that the stand-alone type is an original construction and that it cannot be considered derived and secondary as I argued for Japanese above. In this section, I want to re-examine the evidence that has been put forth in support of the grammaticalization path from copula type to standalone type to see if the data is incompatible with the non-derived view of stand-alone nominalization put forward in this paper. My conclusion is that there is no strong evidence in favour of the development from copula type to stand-alone type.
6.1â•… Some theoretical prerequisites Before proceeding to evaluate the evidence for a development from copula-type to stand-alone type, it is necessary to spell out some assumptions and theoretical prerequisites. Firstly, I believe that two issues concerning the diachronic development of nominalizers and their functions can and should be separated.
A case of non-derived stand-alone nominalization 
The first issue is the ontogenesis of nominalizers. Here the main concern is to try to trace the etymology of nominalizers, i.e. the development of a functional category from a lexical category. This may reveal interesting insights into the lexical sources of nominalizers and is of great interest to grammaticalization studies. However, since I see ‘nominalizer’ as inherently a-categorial (cf. Simpson & Wu, 2001) and regard nominalization as a morpho-syntactic process that turns verbs and adjectives into nominals (cf. Givón, 2001), it cannot be the point of departure of a grammaticalization chain that maps out a functional development of the nominalizers in question. On this view it does not make sense to say that a nominalizer develops a complementizer or a stance-marking function. Secondly, I basically agree with Andersen (to appear) and Lehmann (1995, 2004) that grammaticalization involves a complex of interrelated changes at the content level, the semantactic level, the morphosyntactic level and the expression level, and that it must be specified how each level is affected in a change scenario (Andersen, to appear). Furthermore, I agree with Lehmann that only historical facts count as evidence of a grammaticalization chain and that a grammaticalization chain based on synchronic variation is not proof of a given grammaticalization chain. Since none of nominalization patterns discussed in the following sections involves a complex of interrelated changes and/or are based on synchronic evidence, I do not consider them instances of grammaticalization. With these issues out of the way, let us proceed to look at the evidence for a development from a copula type to a stand-alone type main clause nominalization in various languages in Asia.
6.2â•… Chantyal As far as I am aware, Noonan (1997) was the first to suggest the diachronic development from a copula type main-clause nominalization to a stand-alone type. An example was given in (32). Noonan suggested the following diachronic development of the usages of Chantyal nominalizer -wa (taken from (53) in Noonan (1997: 381)). He is careful to point out that these developments are not grammaticalization chains in the sense that this term is used in grammaticalization theory, but he does assume that the schema in (41) reflects a historical reality. (41)
verb complementation
> noun complementation
purpose nominalization
agent & patient nominals
> attributive nominal
relative clause
> non-relative attributive
verb periphrasis
> main verb
 Janick Wrona
From the perspective of this paper, it is the development from verb periphrasis to main verb (stand-alone) that is interesting. In Chantyal, there is a large number of constructions involving the nominalizer -wa and either the identity copula -B in or the locational/attributive copula -mu. These constructions express tense and aspect distinctions in Chantyal (Noonan, 2008). Noonan (1997) assumes that the main verb use of -wa is derived historically from these uses of the nominalizer in verb periphrasis. More specifically, he assumes that the main verb (or stand-alone) usage came about through ellipsis of one of the copula verbs. This seems like a reasonable assumption, but it is not one that is historically verifiable from a syntactic point of view. Semantically, it is not entirely clear either how a construction expressing tense and aspect distinctions came to express mirativity. It would be simpler to assume that the stand-alone type did not “develop” at all, but was an original usage of -wa nominalization, much like Japanese sa-nominalizations. This explains both the syntactic and the semantic aspects of -wa nominalization.
6.3â•… Japanese There have been a number of nominalizers throughout the history of the Japanese language. In the following I will deal with the most important ones.
6.3.1â•… no Yap, Matthews & Horie (2004) suggest that the copula-type>stand-alone type development is attested for the Japanese nominalizer no. The copula-type is exemplified in (40) and (42) illustrates the stand-alone type. Both examples are from modern Japanese. (42) sonna fukanoo na koto iw-anai no such impossible cop thing say-neg stance ‘Don’t say such an impossible thing’
(Cook, 1990: 432)
no is probably the nominalizer in Japanese that has received the most attention (e.g. Hashimoto, 1969; Iwai, 1973; Iwai, 1974; Nishi, 2006; Nishina & Yoshimura, 2005; Yoshikawa, 1950), but it is not well-understood how its stand-alone usage developed. In my own research (Wrona, forthcoming-a), I have argued that the nominalizing uses of no emerged in the Azuchi-Momoyama period (1573–1603), but that the n(o) da construction involving the copula da is not attested until the Edo period and even then, only sporadically (cf. Iwai, 1974; Kinsui, 1995). One factor that has been completely overlooked in the literature on the development of the stand-alone type no is the existence of the homophonous, but etymologically unrelated, exclamative particle no (stand-alone type is thus not available. From a semantic perspective, one may note that the proposed development seems to fit since both the copula-type no da and the stand-alone no express assertion. However, it is quite possible that when no came to be used as a nominalizer, it acquired at the same time the potential distribution of a nominalized clause and the semantico-pragmatic interpretations associated with this distribution. This means that the view put forward in this paper that the stand-alone is non-derived is compatible with the data whereas the development from copula-type>stand-alone type is not supported.
6.3.2â•… koto There are a couple of instances of the nominalizer koto used as a stance-marker (exclamative) in the Old Japanese texts. Some relevant examples are given below. (44) inisipe-ni ar-ikyeru waza-n kusubasiki olddays-in exist-mpst.adn event-gen mysterious koto-to ipi-tugu nmlz-comp say-handdown ‘It is said and handed down: “Events in the old days were (really) mysterious!” (Man’yôshû 19.4211) (45) karakuni-wo ika-ni pu koto-so Land of Kara-excl why say nmlz-foc ‘Oh, the land of Kara! Why is it called so!’
(Nihonshoki kayô 99)
In Early Middle Japanese the usage appears to have gained currency. The examples in (44) and (45) are of the stand-alone type and not zero copula constructions. Firstly there are no attestations of a copula following stance-marking koto in Old Japanese. Secondly, they seem to be semantically different from copula constructions with koto attested later in Japanese. While it is difficult to judge on the basis of the few Old Japanese examples, it is relatively clear that the Early Middle Japanese construction corresponding to (44) and (45) are exclamatives, something that the corresponding copula construction can never be. So there are reasons to believe that the stand-alone nominalization in (44) and (45) did not develop from copula constructions.
6.3.3â•… Inflectional nominalization In addition to the two nominalizers koto and no with relatively clear lexical and grammatical sources, there are two inflectional nominalizers, the Nominal form and the Adnominal form. Both are used in stance-marking (exclamatives) stand-alone constructions.
 Janick Wrona
The Adnominal form as a stand-alone main-clause nominalization is illustrated in (38) above. One never finds this construction embedded under a copula in Old Japanese, so it would be difficult to argue that the stand-alone construction came about through a historical development from a copula construction. As for the Nominal form (glossed below as NMLZ), very little is known of its prehistory (but see Wrona, 2008). It is possible that one has to distinguish between two types. (46) ipa-naku nomwi so [a-ga kwopuraku]-pa say-neg.nom prt kprt ╛╛↜渀屮i-gen love.nmlz -top ‘I have not said it, that I am love-longing’ (47) [pama-no kiywokyeku] ╛╛↜渀屮beach-gen is clear.nmlz ‘Oh, the clearness of the beach!’
(Man’yôshû 14.3560)
(Man’yôshû 7.1239)
(46) seems to be a pseudo-cleft of a sort. a-ga kwopuraku-pa ‘that I am love-longing’ is a right-dislocated topic and the presupposed subject clause. The underlined predicate ipanaku ‘(I) have not said’ is the focused constituent as evidenced by the (stacked) focus particles nomwi so. To the extent that (46) is a pseudo-cleft one could argue that there is a covert copula involved (see Wrona, 2005) and occasionally focus particles in Old Japanese are found where the copula is found in later stages of Japanese. (47) is different since there is no evidence that a copula, covert or otherwise is involved. So it cannot be claimed that the structure exemplified in (47) developed from (46). In addition, (46) is the only example of this type, making it difficult to draw any conclusions. To sum up the discussion of stand-alone nominalization in Japanese, there is no evidence of the copula-type>stand-alone development. The data is, however, entirely compatible with the non-derived view of stand-alone nominalization.
6.4â•… Korean As in Japanese, there is no shortage of nominalizers in Korean (see Rhee, this volume). The ones that can be said to have a stand-alone usage are kes, -m, -ki and -ci, but as pointed out by Rhee (this volume), the sentence-ending status of the first three is not uncontroversial since there are severe pragmatic restrictions on the contexts in which kes, -m and -ki can occur. The stand-alone usage of these nominalizers is limited to posted signs of prohibition or “bullet-point” stand-alone nominalizations. There does seem to be evidence that the constructions with -ki and kes developed from a copula construction (Rhee p.c). However, it is also clear that these nominalizations do not qualify as genuine stand-alone main-clause nominalization due to their pragmatic limitations. In contexts not limited to “bullet-point” stand-alone constructions, the Korean nominalizer kes does not occur in stand-alone main-clause nominalization;
A case of non-derived stand-alone nominalization 
it is obligatorily embedded under a copula (Rhee, 2008). Concerning the nominalizer -ci, Rhee (this volume) argues that it does have a genuine stand-alone main clause nominalization function; however, -ci nominalizations are not found embedded under a copula, so one cannot argue that they derive from such constructions. The various sentence-ending markers of speaker conviction -lkei-, -nkei- and -nunkei- involve a reduced form of kes and the copula -i- It is extremely interesting to note that Choi (cited in Rhee, forthcoming) attributes the rise of speaker conviction -lkei-, -nkei- and -nunkei- to language contact with English and that they are calques of English ‘the thing is that…’. The sentence-ending markers of a speaker’s feeling of helplessness -nkel, -nunkel- and -lkel also involve a phonologically reduced form of kes. As noted by Rhee (2008) it is quite clear that the latter set -nkel, -nunkel- and -lkel involve verb ellipsis since the accusative case marker -l is involved. It cannot, however, be copula ellipsis, since the copula does not have any case assigning property. In short, there is evidence that the constructions with -ki and kes developed from a copula construction, but is not clear that these examples are genuine stand-alone main clause nominalizations.
6.5â•… Chinese Yap & Matthews (forthcoming) show that nominalization in the Chaozhou dialect, Cantonese and Mandarin Chinese participate in copula-type as well as stand-alone type nominalizations, but there is no direct evidence in favour of the copula-type>standalone type development. The Mandarin stand-alone type is illustrated below. (48) ta bu hui lai de 3sg not will come stance ‘S/he won’t come’
(Yap et al., 2004: 13)
As pointed out by Yap, Matthews & Horie (2004), the development of the stancemarking function is not easy to date for Mandarin de, Cantonese ge and Chaozhou kai. So there is no positive evidence in Chinese that the stand-alone type is derived from the copula-type.
6.6â•… Classical and Lhasa Tibetan As described in Yap & Matthews (forthcoming) both the copula-type and the standalone type nominalization is attested in Classical and Lhasa Tibetan. However, I have been unable to find any information relating to the diachronic development of -pa as a stance-marker. The relevant construction was illustrated in (33) from Classical Tibetan. According to Beyer (1992: 384) examples like (33) -pa (la) are exclamatives, but he makes no diachronic statements pertaining to the development of the usages of -pa.
 Janick Wrona
The same holds for the work of DeLancey (1986, 2003). Thus none of the sources available to me show that a copula-type>stand-alone type development took place in Classical and Lhasa Tibetan.
6.7â•… Rawang The Tibeto-Burman language Rawang is worth mentioning because it appears that the stand-lone type and the copula type are in synchronic variation (though it is not clear to what extent). According to LaPolla (2006) the Rawang nominalizer -wē (stand-alone type development either. The situation in Rawang raises a more general question. Stassen’s (2005) survey of 386 languages shows that in 175 languages of the sample zero copula is allowed to some extent. The existence of zero copulas may be a problem for any claim that makes a distinction between copula type and stand-alone type main-clause nominalizations. If the copula can be covert, how can we tell if a given constructions is a case of copula type main-clause nominalization with a zero-copula or a genuine case of stand-alone type main-clause nominalizations?. There are two factors that may help determine which construction is the right one. Firstly, if a given language has an overt copula, but there is no evidence of this with main-clause nominalizations, then one can conclude that it is a case of standalone type main-clause nominalization. Secondly, if the semantico-pragmatic effects of main-clause nominalization without an overt copula are unexpected (e.g. exclamative or mirative) given the cleft-like structure of the copula type main-clause nominalization, then, presumably, one can conclude that it is a case of stand-alone type main-clause nominalization.
7.â•… Some consequences As I have shown in Section 6, there is no solid evidence of the copula-type>stand-alone type development in any of the languages mentioned above. Even in the languages that allow access to historical sources (Japanese, Korean, Chinese, Tibetan), this development cannot be verified. Given the lack of evidence, the status of the development as a grammaticalization path is doubtful.
A case of non-derived stand-alone nominalization 
It is important to stress that I am not trying to argue that a copula-type>standalone type development is impossible, but I am saying that the evidence needs to be closely examined in order to make such a claim. In particular, since the evidence from Japanese discussed in Section 3 shows that stand-alone type main-clause nominalization does not have to be derived; there was no reason to assume that -sa nominalizations were ever embedded under a copula. Therefore the possibility that stand-alone nominalizations are not derived ought to be at least considered for other languages too, including the ones mentioned in Section 6. This is particularly pressing in languages where the stand-alone type has semantico-pragmatic properties that are not obviously derived from a copula-construction. For example, it is not easy to see how dropping a copula could explain how a standalone type becomes an exclamative as in Japanese and Tibetan.
8.â•… Conclusion In this paper, I looked at the history of a stand-alone type main-clause nominalization in Japanese. This type of stand-alone nominalization was used as exclamatives. I argued that there was no evidence in Japanese that this stand-alone type mainclause nominalization was derived from a copula-type nominalization as proposed by Yap, Matthews & Horie (2004) and Yap & Matthews (forthcoming) for some East Asian languages. Instead I proposed that the stand-alone type is one of several uses of nominalizations. Given the evidence that stand-alone nominalizations need not be derived, I took another look at the languages for which the copula-type>stand-alone had been proposed, and my conclusion was that the non-derived view for stand-alone nominalizations should be seriously considered in the absence of evidence to the contrary.
References Andersen, Henning. To appear. Grammaticalization in a speaker-oriented theory of change. In Historical Linguistics and the Theory of Grammar. The Rosendal Papers, Thorhallur Eythorsson (ed.). Amsterdam: John Benjamins. Beyer, Stephen V. 1992. The Classical Tibetan Language. New York NY: State University of New York Press. Comrie, Bernard. 1976. The syntax of action nominals: A cross-language study. Lingua 40(2–3): 177–201. Comrie, Bernard & Thompson, Sandra A. 1985. Lexical nominalization. In Language Typology and Syntactic Description, Vol III: Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 349–398. Cambridge: CUP. Cook, Haruko Minegishi. 1990. An indexical account of the Japanese sentence-final particle no. Discourse Processes 13: 401–439.
 Janick Wrona DeLancey, Scott. 1986a. Relativization as nominalization in Tibetan and Newari. Paper presented at the 19th International Conference on Sino-Tibetan Languages and Linguistics, Ohio State University, September 11. DeLancey, Scott. 2003. Classical Tibetan. In The Sino-Tibetan Languages, Graham Thurgood & Randy J. LaPolla (eds), 255–269. London: Routledge/Curzon. Elliott, Dale E. 1974. Toward a grammar of exclamations. Foundations of Language 11(2): 231–246. Givón, Talmy. 2001. Syntax: An Introduction, Vol II. Amsterdam: John Benjamins. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10(2): 279–326. Hashimoto, Shinkichi. 1969. Joshi, jodôshi no kenkyû. Tokyo: Iwanami Shoten. Iwai, Yoshio. 1973. Nihongo Hôshi (Muromachijidaihen) (The History of Japanese (The Muromachi period)). Tokyo: Kasama Shoin. Iwai, Yoshio. 1974. Nihongo Hôshi (Edojidaihen) (The History of Japanese (The Edo period)). Tokyo: Kasama Shoin. Kinsui, Satoshi. 1995. Nihongoshi kara mita joshi (Particles: From a Historical Perspective). Gengo 24(11): 78–84. Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. LaPolla, Randy J. 2006. Nominalization in Rawang. 39th International Conference on Sino-Tibetan Languages and Linguistics. Univeristy of Washington, Seattle. Lehmann, Christian. 1995. Thoughts on Grammaticalization. Munchen: Lincom. Lehmann, Christian. 2004. Theory and method in grammaticalization. Zeitschrift für Germanistische Lingvistik 32(2): 152–187. Martin, Samuel E. 1987. The Japanese Language Through Time. New Haven CT: Yale University Press. Martin, Samuel E. 1988. A Reference Grammar of Japanese. Tokyo: Charles E. Tuttle. Matisoff, James A. 1972. Lahu nominalization, relativization, and genitivization. In Syntax and Semantics, Vol. 1, John P. Kimball (ed), 237–257. Tokyo: Taishukan. Miyazaki, Kazuhito, Adachi, Tarô, Noda, Harumi & Takanashi, Shino. 2002. Setsumei no modaritii (The modality of explaining). Modaritii (Modality), 230–260. Tokyo: Kuroshio Shuppan. Nishi, Yumiko. 2006. The emergence of the complementizer no in Japanese revisited. In Japanese/ Korean Linguistics 14, Timothy Vance & Kimberly Jones (eds), 127–137. Stanford CA: CSLI. Nishina, Akira & Yoshimura, Noriko. 2005. Hobunhyôshiki no shutsugen no no rekishiteki henka (The historical development of the complementizer no). Journal of International Relations and Comparative Culture 13(2): 75–89. Noonan, Michael. 1997. Versatile Nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Noonan, Michael. 2008. Nominalization in Bodic languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins. Portner, Paul & Zanuttini, Rafaella. 2005. The semantics of nominal exclamatives. Studies in Linguistics and Philosophy 81: 57–70. Rhee, Seongha. 2008. On the rise and fall of Korean nominalizers. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 239–264. Amsterdam: John Benjamins.
A case of non-derived stand-alone nominalization 
Shibatani, Masayoshi. 1990. The Languages of Japan. Cambridge: CUP. Simpson, Andrew & Wu, Xiu-Zhi Zoe. 2001. The grammaticalization of formal nouns and nominalizers in Chinese, Japanese and Korean. In Language Change in East Asia, Thomas N. McCauley (ed), 250–283. Richmond, Surrey: Curzon. Stassen, Leon. 2005. Zero Copula for predicate nominals. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds.) Oxford: OUP. Vovin, Alexander. 2005. A Descriptive and Comparative Grammar of Western Old Japanese. Folkestone: Global Oriental. Watters, David. 2008. Nominalization in the Kiranti and Central Himalayish languages of Nepal. Linguistics of the Tibeto-Burman Area 31(2): 1–43. Wrona, Janick. 2005. Specificational pseudo-clefts in Old Japanese. Folia Linguistica Historica 26(1/2): 139–157. Wrona, Janick. Forthcoming. The early history of no as a nominalizer. In Papers in Japanese and Korean Linguistics, Janick Wrona, Bjarke Frellesvig & Jieun Kiaer (eds). Aarhus: Aarhus University Press. Wrona, Janick. 2008. The nominal and adnominal forms in Old Japanese: Consequences for a reconstruction of pre-Old Japanese syntax. In Proto-Japanese [Current Issues in Linguistic Theory 294], Bjarke Frellesvig & John B. Whitman (eds), 193–215. John Benjamins. Yamada, Yoshio. 1929[1908]. Nihon Bunpôron (A Theory of Japanese Grammar). Tokyo: Hôbunkan. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins. Yap, Foong Ha, Mathews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Yoshikawa, Yasuo. 1950. Keishiki meisi “no” no seiritsu (The emergence of the formal noun “no”). Nihonbungaku Kyôshitsu 3: 29–38. Zdb. 1967. In Jidaibetsu Kokugo Daijiten (Jodaihen) (Unabridged Period Specific Dictionary of Japanese (Old Japanese)), H. Omodaka (eds). Tokyo: Sanseidô.
Nominalization in Okinawan From a diachronic and comparative perspective* Rumiko Shinzato
Georgia Institute of Technology This paper demonstrates that the four functions originally tied to rentaikei in Old Japanese (OJ), and its sister-language, Old Okinawan (OOk) have been carried on into Modern Okinawan (MOk). In particular, the new rentaikei maintains adnominal, exclamative and cleft functions (kakari musubi), while the new nominalizer si, takes on a headless relative clause/complementizer function. It is argued here that the rentai and shūshi distinction was indeed vital for the preservation of kakari musubi. It is further argued that a complementary distribution between Japanese and Okinawan such that OJ loses its kakari musubi system, prompting Modern Japanese (ModJ) to develop no desu, while MOk renews its kakari musubi system, and consequently does not develop no desu. This paper supports an implicitly assumed claim that the no desu construction is the ModJ counterpart of the OJ kakari musubi. From the point of view of grammaticalization, the development of stance marking functions from kakari musubi (cleft) in Okinawan substantiated Yap et al.’s claim that a cleft construction serves as a crucial link between a nominalizer and a stance marker. In a cross-linguistic perspective, this paper claims that the OJ and Okinawan rentaikei shows versatility seen in various Tibeto-Burman languages such as Chantyal and Lahu.
1.╅ Introduction This paper analyzes nominalized constructions in Okinawan from a diachronic and comparative perspective with its sister language, Japanese.1 By Okinawan, this paper *I have benefited greatly from the discussions and comments from Charles A. Quinn, Leon A. Serafim, Janick Wrona, Foong Ha Yap, Atsuko Izuyama, and Hisako Sokei. Thanks are also due to two anonymous reviewers for their invaluable comments and thoughtful advice. Any shortcomings of this paper are my sole responsibility. I would also like to express my gratitude to North East Asian Council, Association of Asian Studies for providing me with a research travel grant to Okinawa in 2007, which made the recording of conversations in Shuri dialect possible. .╅ Okinawan and Japanese are sister languages (Chamberlain 1895), which are hypo� thesized to have split from their ancestor, Proto-Japanese around the 6th century (Hattori 1976; Hokama 1977).
 Rumiko Shinzato
considers languages of Omorosōshi and Kumiodori, and its direct descendant, Modern Shuri dialect.2 By nominalized constructions, this paper refers to adnominals, nominalizers, and sentence-final nominalized forms. Specifically, it delineates the change in the predicate paradigm (the loss of the adnominal/finite distinction and the renewal of this distinction by the creation of new adnominal/finite forms), and its effects on the nominalized constructions as a whole in Okinawan syntax. In this diachronic analysis of the nominalized constructions, this paper points out the continuity of the original functions of the rentaikei (adnominal form) and their redistribution throughout the new nominalized constructions such as new rentaikei, nominalizer si, and kakari musubi, a cleft-like construction in Okinawan. In addition, it provides an opportunity to reexamine the diachronic change which took place in Japanese history in a new comparative light with Okinawan. This, in turn, substantiates the often implied claim of the no desu (nominalizer + copula) construction as the successor of kakari musubi (=KM) in Japanese traditional linguistics (Onoe 1982; Yanagida 1985). From the perspective of grammaticalization, the Okinawan history offers valuable data to be added to Yap et al.’s (2004) cross-linguistic diachronic study of genitives and adnominals/nominalizers. Furthermore, the elucidation of the distinct paths taken by Okinawan and Japanese, which are genetically related languages, will illuminate the intricate workings of nominalized constructions. In addition, Okinawan and Japanese data may enlighten issues concerning the syncretism between nominalization, relativization and genitivization seen widely in Tibeto-Burman languages (Matisoff 1972; DeLancey 1986 and Noonan 1997). Subsequently, this paper vouches for the benefit of examining the nominalized constructions in a larger context.
2.â•… Historical background 2.1â•… Old Japanese rentaikei and the rise of the Early Modern nominalizer no Old Japanese (=OJ, ~8th century) and Middle Japanese (=MJ, 9th~16th century) had an inflectional distinction between rentaikei (adnominal forms) and shūshikei (finite
.â•… Dialects of Amami, Northern Okinawa (e.g. Nakijin), Miyako and Yayeyama are beyond the scope of this study. Hybrid languages of Japanese and Okinawan such as Uchinā Yamatuguchi, or Yamatu Uchināguchi, which emerged among younger generation Okinawans in recent years present language interference both at lexical and syntactic levels (see Nohara, 1996; Takaesu 2002), and thus are purposely avoided to maintain the same Okinawan lineage (from Omorosōshi to Modern Shuri dialect) vis-À-vis Japanese.
Nominalization in Okinawan 
forms). For instance, the rentaikei appears in (1a), while the shūshikei form is used in (1b) at the end of the sentence.3 But this distinction was lost in Modern Japanese (=ModJ, 17th century~).4 (1) a.
mi-wataseba yanagi sakura wo kokimazete miyako zo haru look-across willows cherry obj blend capital kp spring
no nishiki nari-keru5 gen brocade be-pst (rentaikei)
‘I see willows and cherry trees blended, and (I realize) it is the capital that is the spring brocade.’ (Kokinshū 2: 90) b. furusato to nari-ni-shi Nara no miyako ni mo former.home comp become-perf-pst Nara gen capital in even
iro ha kahara-zu hana ha saki-keri colors top change-not flowers top bloom-pst (shūshikei)
‘Even in Nara, now my former home, why, the flowers still blossom in the colors of the yesteryear!’ (Kokinshū 1: 56)
As predicted from its name, ‘end stop’, shūshikei was used sentence-finally to complete a sentence. In contrast, rentaikei had four functions (cf. Quinn 2001; Nomura 2002), as shown in Table 1.6
.â•… Those verbs which take quadri-grade conjugation have formally identical rentaikei and shūshikei. However, they can still be distinguished by the existence and non-existence of an accompanying kakari particle (see Example 5 below), and by the subject marking (genitives ga and no vs. topic wa/Ø). The former choice is taken if the ending is rentaikei, while the latter is selected for shūshikei. .â•… To avoid confusion, I will use the Japanese words rentaikei and shūshikei when I strictly refer to forms, but will use English words such as adnominal, nominalizer/complementizer, etc. when I refer to functions. This is because rentaikei has multiple functions as seen in Table 1, and the adnominal function is only one of them. Thus, it is best not to have the English term ‘adnominal’ represent rentaikei. .â•… Since the focus of this paper is on the syntax and semantics of nominalized constructions, all of the examples are written logographically rather than phonetically, unless otherwise stated. .â•… These functions correspond to the conventional classification of clause types in traditional Japanese linguistics: adnominal:: rentai-ku (adnominal clause); S-Pronominal (headless relative clause)/complementizer:: juntai-ku (nominalized clause including headless relatives); exclamative:: kantai-ku (exclamatory clause), and nominalization following focus:: kakari musubi (cleft construction).
 Rumiko Shinzato
Table 1.╇ Functions of rentaikei7 Functions
Syntactic environments
Adnominal S-pronominal/complementizer Exclamative Nominalization following focus
___ N7 ___ Particle ___ ## In Kakari Musubi (N + KP (focus) ….___ ##)
The following four examples illustrate the four functions mentioned above: adnominal (2); S-pronominal (headless relative clause)/complementizer (3); exclamative8 (4b) and kakari musubi (5). The corresponding shūshikei (finite form) is added in the parentheses for a reference. (2) furusato to nari-ni-shi (ki) Nara no miyako ni mo former.home comp become-prf-pst Nara gen capital in even ‘Even in Nara, which has become my former home…’ (Kokinshū 2: 90) (3) …kimi ga ie ni ike to ohi-shi (ki) ha itari-kemu ka mo ╇╛╛╛↜you gen house to go.imp qt send-pst top reach-inf q sp ‘I wonder if (the cuckoo) that I sent to your house reached you…’ (Manyōshū 1505) (4) a.
wa ga koromode ha furu toki mo nashi I gen sleeves top dry time even not.exist ‘There has been no time when my sleeves are dry.’
b. wa ga koromode no furu toki mo naki (nashi) I gen sleeves sub dry time even not.exist ‘(Why) is there no time when my sleeves get dry!?’
(MYS 704)
(MYS 1994)
(5) minato no ashi no uraha wo tare ka tawori-shi (ki) estuary gen rush gen tip obj who kp cut-pst ware zo tawori-shi (ki) I kp cut-pst ‘Who is it that cut the tip of the rushes at the estuary? …It is I who cut them.’ (MYS 1288)
.╅ There are cases where no may intervene between the rentaikei and a head noun like example (43). .╅ The rentaikei in kantai-ku (exclamatory clause), has traditionally been translated into Modern Japanese with the nominalizer koto, but Quinn (2001: 314) convincingly argues no instead of koto is closer to the truth.
Nominalization in Okinawan 
At this point, a word of clarification is in order with regard to the syntactic position of rentaikei in exclamative (see example (4b)). This is because it is normally the shūshikei which appears sentence-finally, but here rentaikei is used instead of the expected shūshikei. This off-norm use of rentaikei produces a special rhetorical effect, namely, exclamation, which shūshikei cannot (cf. English translation).9 Rentaikei could also appear at the end of the sentence in the kakari musubi construction, as in example (5) above. As Quinn (1997) eloquently puts it, kakari musubi is so named for its two components, a focused phrase (the kakari ‘one in charge’) and a specially inflected predicate ending the sentence (the musubi ‘binding, closure’).10 This is a syntactic phenomenon functioning like a focus it-cleft (Whitman 1997; Quinn 1997; Shinzato 1998), in which the special kakari particles (KP) call for rentaikei (Ono 1993; Sakakura 1993).11 Clefts are often assumed to have originated in equational sentences, in which two nominals, or nominalized clauses are connected by a copula (Heine & Reh 1984; Harris & Campbell 1995; and Harris 2001). Thus, Japanese kakari musubi (KM) conforms to this generalization (see Serafim & Shinzato 2005). In its transition from Middle Japanese to Modern Japanese, a paradigm change occurred, with rentaikei ousting and replacing shūshikei, thereby abolishing the rentai/shūshi distinction (Kawabata 1982). It is generally assumed that this led to major syntactic changes in Japanese syntax, which divide OJ/MJ and ModJ. Among these, three changes relevant to the nominalized constructions are the establishment of the genitive particle ga as the subject marker (Yanagida 1985); the demise of the kakari musubi construction (Onoe 1982);12 and the recruitment of the nominalizer no, which is traditionally connected to the genitive case marker (see Horie 1998, and also Yap et al. 2004 and Nishi 2006).13 At a later stage, no developed into a sentence-final
.â•… In addition to this primary function, it is also noted that rentaishuūshi bun (adnominalending sentence) gives an explanation in connection to the utterance or the context. See Onoe (1982), Sakakura (1993), Iwasaki (1993), and Shinzato (1998). This meaning will be taken up in Section 4. .â•… This syntactic phenomenon is referred to as ‘agreement’ (Quinn 1997), ‘concordance’ (Hendriks 1998), and ‘government’ (Janick Wrona, p.c. 10/4/2005). .â•… This KM rule has one exception: the KP koso calls for izenkei (realis form) instead of rentaikei. But for a counterargument, see Serafim and Shinzato (2005). Also, for an alternative view that rentaikei triggers the KP, see Shinzato and Serafim (2003) and Serafim and Shinzato (2005). .â•… For a different perspective, see Hendriks (1998), which argues that there is no available evidence to back the inter-relatedness of the two, namely, the breakdown of the rentai-shūshi distinction and the demise of kakari musubi. .â•… In contrast to this genitive-complementizer link (see Horie’s account following Figure 1), Jorden (1962: 99) traces no to a lexical meaning of ‘matter, fact, case’. The fact that
 Rumiko Shinzato
particle or combined formulaicly with the copula da desu to form the so-called no (desu) construction (Ujiie 1992), a Japanese structural analog of the English cleft (it is X that~; it is that ~) to express the speaker’s stance, as below. (6) sā odekake-naharu nda14 ip go.out-hor it’s.that ‘You go out. (Lit. ‘It’s that you go out.’)
(Shunshoku Ume Goyomi 1837)
(7) kawaī deha nē. kawaisohu da to itta nda cute be not pitiful be qt said it’s.that ‘It’s not ‘cute’. It’s that I said ‘pitiful’.’ (Shunshoku Ume Goyomi 1837)
Regarding the development of the Japanese nominalizer and stance marker no, Yap et al.’s formulation below is informative. On an important note, they also identified the cleft construction as a crucial link from a nominalizer to a stance marker (but see Wrona, this volume for an alternative analysis). See Figure 1.15 Genitive/ NP-pronominal no > S-pronominal no > Complementizer no > Stance no Headed relative clause marker no (brief phenomenon) Figure 1.╇ Grammaticalization Path of Japanese no (Yap et al. 2004: 148)
Horie (1998: 172 Yap et al. 2004: 144) sees “family resemblance” among different no’s as below: Genitive no and NP-pronominal no are both ‘preceded by a dependent noun in dependent-head construction…[they] are, however, distinguishable in terms of the presence and absence of the lexical head noun.’ NP-pronominal no and S-pronominal no ‘both functions as pronouns’ but the former is modified by a noun phrase, the latter by a sentential clause. S-pronominal no and
the other genitive ga cannot form an appositive construction, while no can, might be related to this etymology (Ono T 1978). Yet on another account, Wrona (forthcoming), following Frellesvig (2001), hypothesizes that no originated in the adnominal form of the copula (also see Nakajima 1987). .â•… Onoe (2001 [1979]) points out that the three consecutive instances of sit down in a movie were subtitled in Japanese as: Osuwari. Suware. Suwaru-nda (the first two are imperative forms decreasing in the politeness levels, and the last one is the no da construction). This example clearly shows the stance-marking, and intersubjective nature of this construction. .â•… As noted in Yap et al. (2004: 161), positing the ‘headed relative clause’ usage is not without controversy.
Nominalization in Okinawan 
complementizer no are both preceded by a clause, however, ‘the former no serves as an external head to the preceding relative clause, while the latter no serves as a sentential nominalizer preceded by a complement clause.
2.2╅ Okinawan rentaikei; Okinawan si In this paper, three stages of the Okinawan language are considered as follows: Table 2.╇ Three stages of the Okinawan language Stage
Period
Contemporaneous with
Pre-Old Okinawan (Pre-OOk) before Omorosōshi Old Japanese (OJ) Old Okinwan (OOk) Omorosōshi (12th–17th century) Middle Japanese (MJ) Modern Okinawan (MOk) after Omorosōshi Modern Japanese (ModJ)
Omorosōshi is the oldest document in Okinawan, which was compiled during the 16th~17th century, but is said to represent the language from the 12th century to its completion (Hokama & Saigō 1972). This is a collection of poems and songs in praise of the gods and rulers sung at rituals and transmitted orally through generations. Parallel to Old Japanese (OJ), Old Okinawan (OOk) also had the rentai/shuūshi distinction and the same rentai functions, as listed in Table 1. The following examples illustrate all these functions: adnominal (8), S-Pronominal/complementizer (9), exclamative (10), and kakari musubi (11). Shūshikei counterparts are placed in parentheses. (8) Nakagusuku a-tsuru (tsu) uratoyomu tsudumi… Nakagusuku exist-prf famous drum ‘The famous drum which has been at Nakagusuku…’
(Omosōshi 52)
(9) kamishimo no mimoro-suru (su) kiyora ya everyone sub admire-do beautiful sp ‘…the one which everyone admires is beautiful.’ (10) ina chahana sacharu (sachiari). already big flowers bloom ‘Aren’t the big flowers already in bloom!’
(ibid 59)
(ibid 981, 770)
(11) Shiyori furu ame ya sudemidu do furiyoru (furiyori) Shuri fall rain top purified.water kp fall ‘Speaking of the rain falling in Shuri, it is purified water that falls.’
(ibid 370)
What is different about OOk from OJ is that it evinces the transition from the old predicate paradigm (OJ cognates) to a new one (cognates of the Modern Okinawan system) created by a suffixation of the existential verb yoru (‘be’) to a verb stem
 Rumiko Shinzato
around this period. Apparently, the old paradigm was dying out, perhaps due to the same leveling effect which occurred in MJ, leading to the loss of the rentai/shūshi distinction. However, OOk took a separate path from OJ: it created new rentai/shūshi forms, and renewed the distinction (see Hattori 1959; Mamiya 1983; Uchima 1994). Example (11) shows the old and new rentaikei side by side. The first one furu, is the old rentaikei identical to the OJ cognate, while furiyoru is the new rentai form (furi ‘stem’ + yoru ‘be’). Expectedly, this change triggered new developments in Modern Okinawan (MOk), which are distinct from MJ/ModJ in three respects. First, the new paradigm contributed to the survival of kakari musubi in Okinawan.16 The rentaikei-ending, or a nominalized clause, is a prerequisite for the kakari musubi construction because of its cleft structure. A generally held view is that rentaikei was felt to be the nominalized form only in opposition to shūshikei (Funaki 1987). Therefore, with the merger of rentaikei and shūshikei in ModJ, this uniqueness of rentaikei was also lost, which unavoidably led to the loss of the construction altogether. In contrast, Okinawan has preserved it to this day. Given the Japanese historical scenario, it is not difficult to imagine that the creation of the new rentaikei/shūshikei was the underlying force in the preservation of Okinawan kakari musubi. Second, the counterpart of the ModJ no desu did not emerge in Okinawan, perhaps due to the preservation of kakari musubi, which can express similar stance marking as Japanese no desu (more discussion in the next section). Third, the nominalizer si, was recruited in MOk, aruguably because the new rentaikei did not have the S-pronominal (i.e. headless relative clause)/complementizer functions, which the old rentaikei (OJ cognate) had. One may wonder why the new rentaikei was not equipped with this function. A clue to this puzzle may be found in the fact that the new rentaikei was mainly used in the sentence-final positions (ex. furiyoru in11) in Omorosōshi, while the old rentaikei was more frequently used in the sentence-medial adnominal positions (exx. atsuru in 8, and furu in 11). This distributional difference may suggest that the persistence of the adnominal/nominalizer function in the old rentaikei blocked the new rentaikei from making inroads into the sentence-medial positions, and thus a chance to develop into an S-pronominal (headless relative clause)/complementizer. In addition, it may be because the new rentaikei still retained the progressive aspectual meaning, and thus was not as light semantically
.╅ For a detailed comparative account and reconstruction of Proto-Japonic progenitors of three kakari musubi, see Serafim and Shinzato (2000); Shinzato and Serafim (2003) and Serafim and Shinzato (2005). There were two other kakari musubi,in Old Japanese: ya and namu. The alleged Okinawan cognate for ya only appears as a sentence-final particle function; there is no Okinawan cognate for namu.
Nominalization in Okinawan 
as the old rentaikei.17 When the old rentaikei disappeared from the system, the new rentaikei apparently failed to make inroads into the medial positions associated with the S-pronominal/complementizer, arguably because the new rentaikei was not perceived to have as strong nominal characteristics as the old rentaikei, and thus a need for a fresh nominalizer emerged in the new system. This new Okinawan nominalizer, as noted above, is si. Summarizing the above, the nominalized constructions in MOk had the following division of function. Note that all these four functions were once monopolized by rentaikei in OJ and perhaps by its Okinawan counterpart, pre-OOk. Compare Tables 1 and 2. Table 3.╇ Division of Function in MOk Nominalized Forms Functions
Syntactic environment
MOk counterpart
Adnominal S-pronominal/Complementizer Exclamative Nominalization following focus
___ N ___ Particle ___ ## N + KP (focus)….___ ##
new rentaikei si new rentaikei new rentaikei
Here are some examples of these four functions from 18th century theatrical scripts, called kumiodori.18 (12) tsibudi wuru hana nu tsiyu chata gutu in.bud be flower sub dew met like ‘It is like the flower that was in bud met the dew (to bloom)’ (Iriko Wodori no Tokiuta) (13) a.
tsibudi tsiyu machu-si… in.bud dew wait-nmlz ‘…the one in bud waiting for dew is…’
(Iriko Wodori no Tokiuta)
b. waka-aji nu kuma nyi kakuriyayi ime-si chichitsikiti… young-lord sub here in hide be.hor-nmlz hear ‘Hearing that the young lord is hiding in here…’ (Chūshin Migawari no Maki) (14) sati sati gukafu yi kutu d’ayabiru. well well good.job great thing be ‘Well, well, good job. It’s a great thing!’
(Ōkawa Adauchi)
.â•… I am grateful to Foong Ha Yap for calling my attention to this insight (p.c. 10/2/2008). .â•… Here I adopted Ifa’s transcription of the original scripts written in hiragana (Hattori et al. 1974).
 Rumiko Shinzato
(15) nama nu gutu yari ba fukurasha du ayuru now gen circumstance be cond happy kp cop ‘Given the present circumstance, I am HAPPY (not sad).’ (Ōkawa Adauchi)
At this point, a question might arise as to the origin of the nominalizer/complementizer si: Was it derived from the genitive element like the ModJ nominalizer no,19 or from some other source? According to Okinawa Kogo Dai-jiten [Dictionary of Old Okinawan Words] (Hokama eds. 1995), this new nominalizer/complementizer si, developed from a formal noun (keishiki meishi), more specifically su (phonetically, [si]), meaning ‘people, thing’, as shown in the OOk example (16). Note that the form appearing here is the old rentaikei. Compare this with (17=13) repeated below. What comes before si is the shortened form of the new rentaikei. (16) Ubudama ha inoru su du yokeru. Ubudama top bless person kp govern ‘The person who blesses Ubudama (life-generating ball) will govern.’ (Omorosōshi 102) (17) a.
tsibudi tsiyu machu-si… in.bud dew wait-nmlz ‘…the one in bud waiting for dew is…’
(Iriko Wodori no Tokiuta)
b. waka-aji nu kuma nyi kakuriyayi ime-si chichitsikiti… young-lord sub here in hide be.hor-nmlz hear ‘Hearing that the young lord is hiding in here…’ (Chūshin Migawari no Maki)
Going one step further, Serafim and Shinzato (2005) claim that the new nominalizer/ complementizer si is also related to the kakari particle su in Omoro sōshi, and offer a Proto-Japonic form as their progenitor. In addition, in their subsequent paper, (Serafim & Shinzato 2009), they hypothesize the cognation of this nominalizer/complementizer si with the nominalizers su and so in Northern Kyushu and extreme western Honshu dialects, as well as the nominalizer tsu/to in Fukuoka. Based on these accounts, it is rather obvious that MOk si has no genetic relation with the OOk genitive nu, unlike the Mod J nominalizer no, which is traceable to the genitive no (cf. Figure 1).
.╅ Janick Wrona (p.c. 10/4/2005) pointed out to me that in Toyama dialect, it was the OJ genitive ga which developed into a nominalizer, and provided me with the following references: Hirayama et al. (1993) and Murasugi (1991).
Nominalization in Okinawan 
2.3â•… Okinawan rentaikei and Okinawan si in natural spoken discourse In order to supplement the written data in Modern Okinawan in the previous section, this section offers an analysis of natural spoken discourse data on the usage of rentaikei and the nominalizer si. The data source is a 52-minute private conversation exclusively in Shuri dialect, recorded on June 1, 2007 at a private home in Shuri in which one of the conversants (male, 83 years of age) was born and raised. Topics of conversation include World War II, the pre-war legal systems, Okinawan language, history and culture, and people’s life styles and food, etc. What becomes evident from the spoken discourse data is that the same rentaikei and si usage seen in the kumiodori scripts as in examples (12)–(15) has continued into the present. Example (18) follows the speaker’s recollection of his association with his elder brother, who was 7 years older, and drafted when the speaker was in grade school. Here, the rentai form, (i)ru is preceding the noun maziwaē. (18) sījā ?uya nu gutu du ?umu-taru. chōdē ndi-(i)ru maziwaē big.bro. ╛╛╛parent gen like kp ╛╛╛think-pst sibling qt-call interaction zenzen nēn. at.all not ‘It’s that I only thought of my big brother like a parent. I never had associations as a sibling.’
In examples (19a) and (19b) si is used as an S-pronominal (headless relative clause). The si+ya is often contracted as sē. The si is preceded by a negated verb in (19a) and an adjective in (19b). (19) a.
nkashi-n-chu nu tabako sū-ran sē old days-gen-people sub tobacco smoke-not
(<si ya) uimisōra-n. ╇╛╛╛↜nmlz top exist.hor.-not
‘In the old days, there weren’t people who did not smoke.’
b. mata kuchijuke nchi, ari ga yassa sē (<si ya) also kuchijuke called that sub cheap ╇╛╛╛↜nmlz top
yō, Akatsuki ndi si ga a-tan yō. sp Akatsuki qt nmlz sub exist-pst sp
‘Also, (as for) kuchizuke (type of tobacco with mouth piece), for a cheap one of that, there was tobacco called Akatsuki.’
In the examples in (20), si functions as a complementizer. In (20a), the speaker talks about various forms of punishment in his childhood. He remarks that slapping a child’s face is not a traditional Okinawan punishment. In (20b), dwelling on a similar topic, he talks about a stick made of bamboo called waiguchi, used to discipline students. Its loud noise scared the kids more than its contact hurting them. Thus, it is different
 Rumiko Shinzato
from hitting students with a bamboo sword. In (20c), the speaker mentions seminars on the Okinawan legal systems, held biweekly, to which every head of household was required to attend. He explains that the purpose of these seminars was to prevent people from violating the law inadvertently. (20) a.
chira sugui- sē (<si ya) ?uchinā nēn dō. face slap ╇╛╛╛↜nmlz top ╛╛╛Okinawa not.existent sp
?urē yamatu-fū du yaru. ╛╛╛this Japan-style kp is ‘There is no face-slapping in Okinawa. It’s that it is a Japanese style (punishment).’ b. shinai si tatachu-si tō ( ModJ nominalizer OJ genitive ga > ModJ subject marker Pre-Ok genitive no > OOk subject marker Pre-OOk genitive ga > OOk subject marker
What the above developmental paths lead to is the observation that OJ no is the one recruited for the nominalizer, and oddly enough, is also the one which did not develop into a subject marker for a main clause predicate.22 Given this comparative perspective, it may not be too unreasonable to assume that the emerging subject marking function somehow inhibited genitive case markers from being recruited as nominalizers. This view is also substantiated on syntactic and semantic/cognitive grounds. Okinawan example (29) is already ambiguous between the two readings since nu/ga can be either the genitive like (29a), or the subject marker as in (29b). If the genitive nu had been recruited for the nominalizer, we would have further complications as in (30) since Modern Okinawan does not have an accusative marker. That is, the nominalized phrase, chikai-nu, could be interpreted as the object as in (30a) and (30b), or the subject as in (30c). (Example (30) is a hypothetical sentence, and ungrammatical in reality, so the asterisk is added). (29) Jirā nu/ga chikai-si a. b.
gen sub
kōyun.
use-nmlz buy use-nmlz buy
‘I will buy what Jirā uses (Jirā’s using).’ ‘Jirā will buy what someone else uses.’
(30) *Jirā nu/ga chikai-nu kōyun. a. b c.
gen sub gen
use-nmlz buy ‘I will buy what Jirā uses (Jirā’s using).’ use-nmlz buy ‘Jira will buy what someone else uses.’ use-sub buy ‘Someone whom Jirā uses (as his employee) will buy it.’
.╅ All of them naturally served as a subject in a subordinate nominalized clause.
 Rumiko Shinzato
To make the matter worse, in Okinawan there is yet another formally identical nu,23 which is attached to the adjective stems to yield cause/reason subordinate clauses as in (31a), but not nominalization constructions as in (31b): (31) takasa-nu
kōra-n
a. expensive-because buy-neg ‘I won’t buy it since it’s expensive.’ b. expensive-nmlz buy-neg *‘I won’t buy the expensive one.’
If nu had been recruited as a nominalizer, then another ambiguity would result. These add an extra burden to the sentence-parsing. It may be due to this cognitive reason that subject marking and complementizer-marking are kept separate (cf. Footnote 21). Interestingly, according to Kobayashi (2002), the subject marker no was used in the dialects of Western Honshū (in particular the area facing the Japan Sea), Kyushu and Okinawa. Coincidentally, these are the areas where the nominalizers su, so and si, but not no, are used (cf. Section 2.2 above). This correlation may not be a pure accident, and may be in line with the discussion above (see Serafim & Shinzato 2009 for further discussion).
3.4â•… Stance marking The clausal nominalizer > stance marker pathway was manifested in ModJ with the sentence-final particle no, the final stage of the OJ genitive no chain (see examples (6) and (7)). However, OOk si never developed this sentence-final particle usage.24
.â•… According to Chamberlain (1895), this nu corresponds to OJ mi, which is attached to the adjective stems forming a causal clause just like (24a). He (ibid: 118) states: “Having regard, however, to the frequent correspondence of Luch. [i.e. Luchuan, or Okinawan] n to Jap. m, there can be little doubt that these Luch. nu forms are the surviving relatives of the long extinct mi forms known to us only through a limited number of examples of Classical Japanese poetry…”. .â•… One of the reviewers brought the following examples extracted from female college students’ conversations as possible cases of the sentence-final si in Modern Okinawan: (The glosses and translations are the reviewer’s.): a.
tsumara katta ssi↜! b. boring pst si.assertion ‘(It) was boring!’
muri ya ssi! unacceptable be?/sp? si.assertion ‘(I) dislike (skinny guys)!’
It has to be acknowledged that these are examples of uchinā yamatuguchi (see Footnote 3) in which Japanese (e.g. muri) and Okinawan (e.g. ya ssi) syntax are mixed, and thus it is not uncontroversial to consider them to be the direct descendant of Old Okinawan in the fist place. Besides, this geminated ssi does not seem to be the cognate of the Shuri si. Just to mention a few arguments: (1) Shuri si is attached to a shortened rentai form without gemination.
Nominalization in Okinawan 
In the natural spoken discourse data (see Section 2.3), there are 106 tokens of si with the following make-up: S-pronominalizer (headless relatives) (91) and complementizer (15). However, there is no instance of si used as a sentence-final particle, standing alone sentence-finally. The only instance which may appear to come close to the sentence-final particle usage is (32) where si is followed by the sentence-final particle yō. In this example, the speaker is explaining the difference between the Japanese word, tenugui and its commonly mistaken Shuri counterpart, tīsāzi. His conclusion is that the Okinawan tīsāzi is a thing to use as a headband. After (32), he cites the famous scholar Kunio Yanagida and mentions that the Japanese tenugui can be used to wipe feet, but tīsāzi can never be used for that purpose. This particular si in (32) is more properly characterized as a pronominal, than as a sentence-final particle. (32) tīsāzi tu tenugui ya zenzen chigain dō. bichi. ‘tīsāji’ and ‘tenugui’ top totally different sp different tīsāzi ndi sē (<si ya) sāzisu-si yō. ‘tīsāzi’ qt ╇╛╛╛↜nmlz top put.on.a head-nmlz sp ‘Tīsāzi and tenugui are totally different. Different. A thing called tīsāzi is something to put on a head (as a headband).’
The absence of a sentence-final usage in si is also corroborated by another recording (13 minutes 33 seconds) of a conversation between two Shuri speakers. The recording was done at a studio of a radio station, and broadcast on January 23, 1955. The topic of the conversation was their memories of New Year’s Eve and New Year’s Day (January 23 was a New Year’s day according to the lunar calendar). The tape was first obtained by late Shiro Hattori at the time, and was recently transcribed and translated by Izuyama (2006). In the transcribed texts, there are 25 tokens of si, but they are either
The gemination suggests a fusion with -ru of the original rentai form. The Japanese conjugational ending katta never occurs with ru as in tumaran kattaru. Also no gemination occurs with ya (e.g. no -ya ssi ya, nor -ya ssiga); and (2) Shuri si, if it ever occurs near sentence-finally with a sentence-final particle (see ex. 32), occurs with a present-tense verb. Tsumaran katta is not only an adjective, but also a past tense form. In this regard, Nohara’s (1996) account seems more tenable. Analyzing the following hybrid examples similar to the reviewer’s b, Nohara (ibid: 281) points out semantic and functional similarities of ssā and sshī, and concludes that ya ssā is an Okinawan-based form, and ya ssī is its new variant. According to my informants, the short si is also possible. c.
omae tensai ya ssā d. you genius be sp ‘Aren’t you genius!’
aitsu atama ī ya sshī he brain good be sp ‘Isn’t he smart!’
 Rumiko Shinzato
S-pronominal, or complementizer, with no single instance of si used as a sentence-final particle. Even an alleged sentence-final si like example (32) is absent. I would like to attribute the absence of stance-marking si to the existence of kakari musubi, which is a nominalized focus construction itself, and already equipped with stance marking functions (Shinzato 1998). The ModJ no desu sentences (6) and (7) can be translated into Okinawan with kakari musubi as below, and the same stance of implied directive as in (6) and dismissive as in (7) are also transferred: (33) uri ?icha-bī du suru. ip ╛╛╛go.out.hor kp do ‘You go out (request).’ (34) chimuganasan ya aran, chimugurisan ndi du ?icharu cute be not pitiful qt kp ╛╛╛said ‘I didn’t say “cute”, I said “pitiful”.’
The complementary nature of the Japanese no desu and Okinawan kakari musubi (KM) constructions is also evident in the reverse translation: Japanese translation of Okinawan KM is often done by no desu. For instance, in speaker B’s utterance in Shuri example (35), kakari musurbi is translated as no by Izuyama (2006: 33). (35) A: Ok.: …jahari ?unu cimu du jaibīru-i ╇╛╛╛↜渀屮as expected ╛╛╛this.kind.of meaning kp cop-q ‘Is it this kind of meaning as expected?’ B: Ok.: ?urē ?an du jaru. that ╛╛╛so kp cop J:
sore wa sōna no that top so it’s.that ‘That is so.’
The use of the no desu counterpart is also appropriate for example (36). An interesting thing about this utterance is that it has a similar make up to example (34) in that the first part is the denial of the expected situation, and the second part is the emphatic confirmation and insistence of its complete opposite. Illuminating in this regard is that the original Japanese version of (34) is a no desu sentence as in (7). (36) Ok: nūn ?ara-n. hun di ?jā ni hwē ?īhōi du saru. anything ╛╛╛be-not hun qt ╛╛╛say by fly ╛╛╛repel kp did J:
nan.de.mo ari-masen. hun to itte hae wo opparatta ndesu. anything be-not hun qt say fly obj repeled it’s.that ‘It’s nothing. It’s that I expelled a fly saying “hun”’ (Oka no Ue no Ipponmatsu)
Nominalization in Okinawan 
The following is an example from the natural spoken discourse data. Speaker A explains that in the old days, boys had to clean up ashtrays in the morning and lamp chimneys in the evening. Speaker B who was born after World War II and unfamiliar with life without electricity, asked if people were really using oil lamps. To this, speaker A responds emphatically (see the interjections) that electricity was not at all available, so people used oil lamps. The kakari musubi here shows a strong assertive stance of speaker A. (37) A: tabaku nu sōji. yurō ( Snow. Snow!
 Rumiko Shinzato
I contend that cleft structure and its focusing function (which highlights the normally ‘unexpected’), stance marking function, and explanation function (see Section 3.4) can be all tied in this line of thought.
5.â•… Conclusion This paper was a diachronic as well as a synchronic study of Okinawan nominalized constructions from a functional perspective. It demonstrated that the four functions originally tied to rentaikei in OJ, and its sister-language OOk have been carried on to present-day Okinawan, through the multiple nominalized constructions in a complementary manner: the new rentaikei maintains adnominal, exclamative and cleft functions, while the new nominalizer si, takes on a S-pronominal (headless relative clause)/complementizer function. In a comparative perspective, Okinawan data offer new insight into the understanding of the intricate working of the Japanese nomininalized constructions from both diachronic and synchronic perspectives. First, this perspective clearly shows that the rentai and shūshi distinction was indeed vital for the preservation of kakari musubi. Second, it also renders support to an implicitly assumed claim that the no desu construction is the ModJ counterpart of the OJ kakari musubi (Onoe 1982; Funaki 1987; Schaffer 2002). This is because there exists a complementary distribution such that OJ loses its kakari musubi system, prompting ModJ to develop no desu, while MOk renews its kakari musubi system, and as a result never develops no desu. Third, the fact that ModJ and MOk have taken different paths due to the demise (ModJ) and renewal (NOk) of the rentai and shūshi distinction speaks for the essential character of the rentai and shūshi distinction in Okinawan and Japanese syntax. From the point of view of grammaticalization, the Okinawan case enriches crosslinguistically attested developmental pathways reported in Yap et al. (2004), and more recently elaborated in Yap and Matthews (in press), with added data. It demonstrates that Okinawan nominalized constructions exhibit all four functions often related to clausal type modifications: si for S-pronominal (headless relative clause) and complementizer; and kakari musubi for cleft/focus and stance marking. The development of stance marking functions from kakari musubi (cleft) substantiated Yap et al.’s claim that a cleft construction serves as a crucial link between a nominalizer and a stance marker. Their claim is especially noteworthy given the fact that as divergent as they are, both MOk clefts with the KP du and new rentai (nominalized verbal form) and ModJ clefts with nominalizer no have developed similar stance-marking functions. In a cross-linguistic perspective, it was shown that the OJ and OOk rentaikei shows versatility seen in Chantyal (Noonan 1997) and Lahu (Matisoff 1972). This
Nominalization in Okinawan 
paper argued that such versatility stemmed from its being essentially a semantic and functional analog of one word sentence, which is illocutionless, highly contextdependent and thetic in judgement.
Abbreviations comp cond cop gen hor imp inf ip kp mj modj mok mys
complementizer conditional copula genitive honorific imperative inference interjectional particle kakari particle Middle Japanese Modern Japanese Modern Okinawan Manyōshū
nmlz obj oj ook os prf pst q qt sp sub top
nominalizer object Old Japanese Old Okinawan Omorosōshi perfect past question particle quotative particle sentence-final particle subject topic
References Chamberlain, Basil Hall. 1895. Essay in Aid of a Grammar and Dictionary of the Luchuan Language. Supplement to Transactions of the Asiatic Society of Japan, Vol. 23 supplement. DeLancey, Scott. 1986. Relativization as nominalization in Tibetan and Newari. Presented at the 19th annual meeting of the International Conference on Sino-Tibetan Languages and Linguistics. Frellesvig, Bjarke. 2001. A common Korean and Japanese copula. Journal of East Asian Linguistics 1: 1–35. Funaki, Shuntaro. 1987. Kakari musubi. In Kokubunpo Kōza 3 (Seminars in Japanese Grammar 3), Akiho Yamaguchi (ed.), 278–306. Tokyo: Meiji Shoin. Handa, Ichiro. 1999. Ryūkyūgo Jiten (Dictionary of Ryūkyuan). Tokyo: Daigaku Shorin. Haraguchi, Yutaka. 1980. Juntai joshi no no teichaku – waka no zokugoyaku no bāi – (Establishment of nominalizer no – through the examination of Modern Japanese translations of waka). Kokugogaku 123: 47–57. Harris, Alice C. 2001. Focus and universal principles governing simplification of cleft structures. In Grammatical Relations in Change [Studies in Language Companion Series 56], Jan Terje Faarlund (ed.), 159–170. Amsterdam: John Benjamins. Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: CUP.
 Rumiko Shinzato Hashimoto, Shinkichi. 1969. Joshi, Jodōshi no Kenkyū (Studies on Particles and Auxiliary Verbs). Tokyo: Iwanami Shoten. Hattori, Shiro, Nakasone, Seizen & Hokama, Shuzen (eds). 1974. Ifa Fuyū Zenshū 3 (Complete Works of Ifa Fuyū 3). Tokyo: Heibonsha. Hattori, Shiro. 1959. Nihongo no Keitō (The Lineage of the Japanese Language). Tokyo: Iwanami Shoten. Hattori, Shiro. 1976. Ryūkyū hōgen to hondo hōgen (The Ryukyu dialects and the mainland dialects). In Okinawagaku no Reimei: Ifa Fuyū Seitan Hyakunen Kinenshi (The Dawning of Okinawan Studies: A Volume Commemorating the Hundredth Anniversary of the Birth of Ifa Fuyu), Ifa Fuyū Seitan Hyakunen Kinenkai (eds), 7–55. Tokyo: Okinawa Bunka Kyōkai. Heine, Bernd, & Reh, Mechthild. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Hendriks, Peter. 1998. Kakari particle and the merger of the predicative and attributive forms in the Japanese verbal system. Japanese and Korean Linguistics 7: 197–210. Hirayama, Teruo et al. (eds.). 1993. Gendai Nihongo Hōgen Daijiten (Dictionary of Japanese Dialects), Vol. 5, Tokyo: Meiji Shoin. Hokama, Shuzen (ed). 1995. Okinawa Kogo Dai-jiten (Dictionary of Old Okinawan Words). Tokyo: Kadokawa Shoten. Hokama, Shuzen & Saigō, Nobutsuna (eds). 1972. Omoro Sōshi (Book of Omoros). Tokyo: Iwanami Shoten. Hokama, Shuzen. 1977. Okinawa no Gengo to Rekishi (Okinawa’s Language and History). Tokyo: Iwanami Shoten. Horie, Kaoru. 1998. On the polyfunctionality of the Japanese particle no: From the perspectives of ontology and grammaticalization [Studies in Japanese Grammaticalization 6], Toshio Ohori (ed.), 169–192. Tokyo: Kurosio. Iwasaki, Shōichi. 1993. Functional transfer in the history of Japanese language. Japanese/Korean Linguistics 2: 20–32. Jorden, Eleanor Harz. 1962. Beginning Japanese. New Haven CT: Yale University Press. Kawabata, Yoshiaki. 1982. Dōshi katsuyō no shiteki tenkai (Historical development of verb conjugations). In Kōza Nihongogaku 2 Bunpōshi (Course in Japanese linguistics 2 History of Grammar), Kenji Morioka et al. (eds), 184–216. Tokyo: Meiji Shoin. Kobayashi, Takashi. 2002. Kakujoshi (Case particles). In Hōgen Chōsa Gaido Bukku (Dialect Research Guide Book), Takuichirō Onishi (ed.), 109–132. . Lehmann, Christian. 1995[1982]. Thoughts on Grammaticalization. Munich: Lincom. Mamiya, Atsushi. 1983. Omoro Sōshi no kakari musubi ni tsuite (On the kakari musubi of the Omoro Sōshi). Okinawa Bunka 61: 6–18. Matisoff, James. 1972. Lahu nominalization, relativization and genitivization. In Syntax and Semantics, Vol. 1, John Kimball (ed), 237–58. New York NY: Seminar Press. Matsumura, Akira (ed.). 1971. Nihon Bunpō Daijiten (Dictionary of Japanese Grammar). Tokyo: Meiji Shoin. Murasugi, Keiko. 1991. Noun Phrases in Japanese and English: A Study in Syntax, Learnability and Acquisition. Ph.D. dissertation, University of Connecticut. Nakajima, Fumio. 1987. Nihongo no Kōzō (The Structure of Japanese). Tokyo: Iwanami Shoten. Nishi, Yumiko. 2006. The emergence of Japanese complementizer no revisited. Japanese/Korean Linguistics 14: 127–138.
Nominalization in Okinawan 
Nohara, Mitsuyoshi. 1996. Okinawa no wakamono kotoba (Language of Okinawan Youth). Okinawa Bunka Kenkyū 22: 265–282. Nomura, Takashi. 1996. Ga-Shūshikei e (Change towards ga-finite form combination). Kokugo Kokubun 65(5): 524–541. Nomura, Takashi. 2002. Rentaikei ni yoru kakari musubi no tenkai (Genesis of kakari musubi from adnominal forms). Nihongogaku to Gengo Kyoiku (Japanese Linguistics and Language Education), Hiroto Ueda (ed.), 11–37. Tokyo: Tokyo Daigaku Shuppankai. Noonan, Michael. 1997. Versatile nominalization. In Essays on Language Function and Language Type. Dedicated to T. Givon, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Ono, Susumu. 1978. Nihongo no Bunpō wo Kangaeru (Thinking about Japanese Grammar). Tokyo: Iwanami. Ono, Susumu. 1993. Kakari Musubi no Kenkyū (A Study of Kakari Musubi). Tokyo: Iwanami. Ono, Toru. 1978. Nihongo no Sogenteki Kenkyū (An Etymological Study of the Japanese Language). Tokyo: Takayama Honten. Onoe, Keisuke. 1982. Bun no kihon kōsei – shiteki tenkai (Basic structure of sentences – historical development). Kōza Nihongogaku (Course in Japanese Linguistics), Kenji Morioka (ed.), 1–19. Tokyo: Meiji Shoin. Onoe, Keisuke. 2001 [1977]. Goretsu no imi to bun no imi (Word meanings and their sentence meaning). Bumpō to Imi I (Grammar and Meaning I). Tokyo: Kuroshio Shuppan. Quinn, Charles J. 2001. Why early Japanese rentaikei sentences didn’t mean ‘… koto’. In Cognitive-Functional Linguistics in an East Asian Context, Kaoru Horie & Shigeru Sato (eds), 303–331. Tokyo: Kurosio. Quinn, Charles J., Jr. 1997. On the origins of Japanese sentence particles ka and zo. Japanese/ Korean Linguistics 6: 61–89. Sakakura, Atsuyoshi. 1993. Nihongo Hyōgen no Nagare (The History of Japanese Expressions). Tokyo: Iwanami. Schaffer, Wolfram. 2002. Kakari musubi, noda-construction, and how grammaticalization theory meets formal grammar. Japanese/Korean Linguistics 10: 320–333. Serafim, Leon A. & Shinzato, Rumiko. 2000. Reconstructing the Proto-Japonic kakari musubi, *â•‚ka …â•‚(a)m-wo. Gengo Kenkyū 118: 81–118. Serafim, Leon A. & Shinzato, Rumiko. 2005. On the old Japanese kakari (focus) particle koso: Its origin and structure. Gengo Kenkyu 127: 1–49. Serafim, Leon A. & Shinzato, Rumiko. 2009: 116–130. Grammaticalization pathways for Japonic nominalizers: A view from the Western periphery. Japanese/Korean Linguistics 16. Shinzato, Rumiko. 1998. Kakari Musubi: Its functions and development. Japanese/Korean Linguistics 8: 203–216. Shinzato, Rumiko & Serafim, Leon A. 2003. Kakari musubi in comparative perspective: Old Japanese ka/ya and Okinawan â•‚ga/â•‚i. Japanese/Korean Linguistics 11: 189–202. Takeuchi, Lone. 1999. The Structure and History of Japanese: From Yamatokotoba to Nihongo. London: Longman. Takaesu, Yoriko. 2002. Uchinā Yamatoguchi wo megutte. Kokubungaku Kaishaku to Kanshō 67(7): 151–160. Traugott, Elizabeth Closs & Dasher, Richard B. 2002. Regularity in Semantic Change. Cambridge: CUP. Traugott, Elizabeth Closs. 1998. The rhetoric of counter-expectation in semantic change: A study in subjectification. In Historical Semantics and Cognition, Andreas Blank & Peter Koch (eds), 177–196.
 Rumiko Shinzato Uchima, Chokujin. 1990. Okinawa Gengo to Kyōdōtai: Uchi-Shakai no Ishiki to Kotoba (The Okinawan Language and Community: Insider Consciousness and Language). Tokyo: Shakai Hyōronsha. Uchima, Chokujin. 1994. Ryūkyūhōgen Joshi to Hyōgen no Kenkyū (A Study of the Particles and Expressions in Ryukyuan Dialects). Tokyo: Musashino Shoin. Ujiie, Yoko. 1992. nodesu bun no seiritsu to sono haikei – Nihongoshi to no taiwa (The establishment of the nodesu sentence, and its background: A dialogue with Japanese language history). Tsujimura Toshiki Kyōju Koki Kinen Nihongoshi no Shomondai (Tsujimura Toshiki Kyōju Koki Kinen: Nihongoshi no Shomondai), Tsujimura Toshiki Kyōju Koki Kinen Ronbunū Kankō Īnkai (eds), 554–572. Tokyo: Meiji Shoin. Whitman, John. 1997. Kakarimusubi from a comparative perspective. Japanese/Korean Linguistics 6: 161–178. Wrona, Janick. Forthcoming. The adnominal form in Old Japanese: Grammaticalization and degrammaticalization of an inflected form? . Yanagida, Seiji. 1985. Muromachi Jidai no Kokugo (Japanese Language of the Muromachi Period). Tokyo: Tōkyōdō. Yap, Foong Ha, Matthews, Stephen & Horie, Kaoru. 2004. From nominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins.
Versatility of nominalizations Where Japanese and Korean contrast Kaoru Horie
Nagoya University This paper presents a contrastive analysis of Japanese and Korean nominalized constructions, i.e. (I) Head Internal Relatives (HIRs) and (II) Sentence-final nominalized constructions. Japanese HIRs are shown to manifest a greater degree of cross-constructional extension than their Korean counterparts. Sentence-final nominalized constructions, which take on various pragmatic-semantic functions, also display a parallel cross-linguistic contrast in terms of interpretive versatility. This paper argues that these cross-linguistic contrasts are correlated with the differential degrees of functional extension of non-conclusive verb forms as nominal, and finally with the differential preferences toward noun-oriented structures between Japanese and Korean.
1.â•… Introduction Nominalization is known to serve various morpho-syntactic, semantic, pragmatic, discursive, and rhetorical functions. Indeed the existence of ‘versatile nominalization’ (Noonan 1997) has been attested abundantly in East and South East Asian languages (e.g. Matisoff 1991; Yap, Matthews, & Horie 2004). A similar phenomenon has recently been reported in a West African language Adioukrou (Niger-Congo), where the versatile subordinator/nominalizer èkè encodes relative clause, complement clause, and adverbial (conditional) clause (Horie, Pardeshi, & Kaul 2008). In Japanese and Korean, prime candidates for versatile nominalization are the so-called Head Internal Relatives (HIRs), structures known to involve nominalization cross-linguistically (Andrews 2007: 232). HIRs in Japanese and Korean have grammaticalized nominalizers no and kes respectively in head position, as in (1a, b). HIRs contrast with ‘regular’ relative clauses that have lexical heads (2a, b). HIRs in Japanese and Korean are structurally non-distinct from complement clauses, as shown in (1a′, 1b′):1 .â•… When a pair of Japanese and Korean sentences are given in examples (as in (1)), (a) and (b) refer to Japanese and Korean sentences respectively throughout this paper unless otherwise indicated.
 Kaoru Horie
Head Internal Relatives (HIRs) (1) a.
John-wa [doroboo-ga mise-kara dete kuru] no-o tukamae-ta. John-top2 ╛╛↜渀屮thief-nom shop-from come out nml-acc catch-pst
b. John-un [totwuk-i kakey-eyse nao-nun] kes-ul cap-ass-ta. John-top ╛╛↜渀屮thief-nom shop-from exit-att:prs nml-acc catch-pst-decl ‘John caught the thief who happened to come out of the shop.’
Externally Headed Relatives (2) a.
John-wa [mise-kara dete kuru] doroboo-o tukamae-ta. John-top ╛╛↜渀屮shop-from come out thief-acc catch-pst
b. John-un [kakey-eyse nao-nun] totwuk-ul cap-ass-ta. John-top ╛╛↜渀屮shop-from exit-att:pres thief-acc catch-pst-decl ‘John caught the thief who was coming out of the shop.’
Complement Clauses (1) a′. John-wa [doroboo-ga mise-kara dete kuru] no-o mi-ta. John-top ╛╛↜渀屮thief-nom shop-from come out nml-acc see-pst b′. John-un [totwuk-i kakey-eyse nao-nun] kes-ul po-ass-ta. John-top ╛╛↜渀屮thief-nom shop-from exit-att:prs nml-acc see-pst-decl ‘John saw the thief coming out of the shop.’
The fact that HIRs and complement clauses are identically coded (1a–1a′, 1b–1b′) is indicative of the versatility of the nominalized constructions involving no (Japanese) and kes (Korean) in the respective languages. While ‘versatile nominalization’ (Noonan 1997) is a useful descriptive notion for capturing an areally prominent grammatical phenomenon, it has not been employed for cross-linguistic comparisons. This study constitutes a first attempt to examine the differential degrees of versatility of nominalizations between two languages. The purposes of this paper are threefold. First, this paper analyzes the differential ‘versatility’ of HIRs and related nominal(ized) constructions between
.╅ The following abbreviations are used in the glosses (footnotes included). acc/Acc AdPast att conc conj cont dat decl
Accusative Adnominal Past Attributive Conclusive Conjunctive Continuous Dative Declarative
fut gen ger hon imp loc neg nml
Future Genitive Gerund Honorific Imperative Locative Negation Nominalizer
nom pol prs pst top sfp
Nominative Polite Present Past Topic Sentence Final Particle
Versatility of nominalizations: Where Japanese and Korean contrast 
Japanese and Korean by incorporating the functional-typological analysis of complex clauses proposed in Radical Construction Grammar (Croft 2001). Secondly, this paper then shifts its attention from HIRs to other nominal constructions in Japanese and Korean. Finally, it correlates the cross-linguistic differences to a more general cross-linguistic contrast between Japanese and Korean in terms of preference for noun-oriented structures. The organization of this paper is as follows. Section 2 revisits the original analyses of HIRs in Japanese and Korean in Horie (1993a, b) and casts them in view of the Radical Constructional grammar approach to complex clauses (Croft 2001). It thus highlights the differential ‘versatility’ of HIRs and related nominal(ized) constructions between Japanese and Korean. Section 3 presents a contrastive analysis of nominalized constructions occurring in sentence-final position. Section 4 correlates the cross-linguistic findings in Sections 2 and 3 to the differential degrees of versatility of nominalization between Japanese and Korean, and argues for a stronger preference for noun-oriented structures in Japanese than Korean. Section 5 presents the conclusion.
2.â•… H ead-internal relatives as a nominalized construction mediating between relative, complement, and adverbial clause Since the pioneering works on Japanese HIRs by S–Y. Kuroda (Kuroda 1974, 1975–76, 1976, 1976–77, 1992), HIRs in Japanese and Korean have received focused attention from linguists of varying theoretical persuasions (e.g. Cu 1983; Yang 1993; Jhang 1994; Hoshi 1995; Ohara 1996; Tsubomoto 1998; Nomura 2000, among many others). However, in spite of the close resemblance in the grammatical structures of Japanese and Korean, relatively few contrastive studies of HIRs have been produced, particularly within a functionalist framework (see Kim 2004 for a formal semantic approach). One of the few exceptions is a series of functional-typological analyses of Japanese and Korean HIRs by Horie (1993a, b). In what follows, I will revisit and recast major findings presented in Horie (1993a, b) in the light of Radical Construction Grammar (Croft 2001) and later developments in grammaticalization studies. The Radical Construction Grammatical framework is particularly relevant as it presents a more refined treatment of cross-constructional continuum originally noted by Horie (1993b). This will enable us to appreciate the close interaction between HIRs and other complex clauses, particularly adverbial clauses, than was possible at the time of these works.
2.1â•… Revisiting Horie’s (1993a, b) analyses of Japanese and Korean HIRs S–Y. Kuroda’s seminal works established that HIRs, which had received less attention than externally headed ‘regular’ relatives, are felicitous in Japanese. Kuroda also noted that Japanese HIRs are subject to some specific pragmatic-semantic constraint
 Kaoru Horie
(‘relevancy condition’) holding between events encoded by a matrix sentence and its co-occurring HIR (e.g. contemporaneity). Crucially, Kuroda noted that HIRs are subject to adverbial clause interpretations, a point that we will address under the rubric of ‘duality’ from a functional-typological viewpoint in this paper. Cognitive-functional linguists (e.g. Horie 1993a, 1993b, 1998; Ohara 1996; Tsubomoto 1998; Nomura 2000) have paid close attention to the ‘duality’ of Japanese HIRs which can profile both ‘entity’ (bold portion) and ‘event’ (bracketed portion), as in (3). This duality enables the nominalized construction in (3) to be compatible with both the HIR interpretation and the complement clause interpretation. (3) John-wa [doroboo-ga mise-kara dete kuru] no-o tukamaeta/mita. John-top ╛╛↜渀屮thief-nom shop-from come out nml-acc caught/saw ‘John caught the thief who (happened to be) coming out of the shop.’ (HIR interpretation) ‘John saw the thief coming out of the shop.’ (Complement interpretation)
Japanese HIRs show consistency in native speakers’ acceptability judgement. In fact, analogues of HIRs were already attested in Old Japanese (Kaiser 1991; Kuroda 1992; Horie 1993b), as in (4). The primary difference between Modern Japanese HIRs and their Old Japanese counterparts is the absence of nominalizer no, indicated by (Ø) in (4). This structure was possible because the attributive (glossed as ‘att’) predicate form in Old Japanese (in this case keru) served to nominalize the clause. (4) [Kasiwagi-no ki-ga ari keru ] (Ø)-wo ori-ni tamaeri keri. ╛╛↜渀屮oak-gen tree-nom exist pst: att -acc break-hon-pst: fin ‘(an honorable person) broke (a twig of) an oak tree which happened to be there.’ (Yamato Monogatari, 10th century; Yamaguchi 1992, p. 6)
However, how does this compare to the acceptability status of Korean HIRs? Primarily under the influence of Kuroda’s analysis of Japanese HIRs, Korean HIRs have been ‘identified’ by formally and functionally oriented linguists who have since tended to take their existence for granted (e.g. Yang 1993; Jhang 1994; Kim 2004). Horie (1993a), however, cast doubt on the very acceptability status of Korean HIRs through a contrastive study with their Japanese counterparts. Specifically, Horie paid close attention to the lexical/ grammatical status of the overt nominalizers occurring in head position of HIRs, i.e. no in Japanese and kes in Korean. While no has virtually no lexical meaning unlike other less semantically bleached nominalizers like mono (‘thing’), tokoro (‘place’) and koto (‘matter, fact’), kes still has some lexical (though fairly abstract) meaning: ‘thing, matter’. Horie (1993a) thus proposed classifying Japanese HIRs into three semantic subtypes (I–III) depending on whether no can be replaced by other semantically less bleached nominalizers (I) mono, (II) tokoro, or neither (III) (pp. 450–452), as in (5a–7a)(examples are slightly modified from the original ones for the purpose of the current paper):
Versatility of nominalizations: Where Japanese and Korean contrast 
Subtype I. No replaceable by mono (‘concrete thing’) when the former refers to a concrete product made from a source material referred to by a lexical noun occurring inside an HIR. (5) a. John-wa [namazakana-o hosi-ta] {no/mono}-o tabe-ta. John-top ╛╛↜渀屮raw fish-acc dry-pst nml/thing-acc east-pst ‘John ate fish dried from raw fish (lit. John ate the thing that (one) dried raw fish.’
Subtype II. No replaceable by tokoro (‘place’) when an entity referred to by the semantic head NP inside an HIR is engaged in an ongoing activity or in motion. (6) a.
John-wa [tomodati-ga hikoozyoo-ni tuita] {no/tokoro}-o kuruma-de John-top ╛╛↜渀屮friend-nom airport-to arrived ╛╛↜渀屮nml/place-acc car-by
hoteru-ni turete- itta. hotel-to take: ger- go:pst
‘John drove his friend, who had just arrived at the airport, to a hotel.’
Subtype III. No replaceable by neither when the semantic head NP inside an HIR refers to a static entity. (7) a.
John-wa [ringo-ga tukue-no ue ni oite aru] John-top ╛╛↜渀屮apple-nom desk-gen top-loc on be placed
no-o tabe-ta. nml-acc eat-pst
‘John ate an apple which happened to be placed on the desk.’
Some crucial structural differences were noted between Subtype (I) and Subtypes (II, III) in terms of the role of a lexical noun (‘bold portions’ in 5a, 6a, 7a) occurring inside the clause. It is fair to say that lexical nouns tomodati (‘friend’) (6a) and ringo (‘apple’) (7a) are semantic heads of HIRs as they function as direct objects of the matrix transitive verbs. However, it is less certain to call namazakana (‘raw fish’) (5a) a semantic head because it is not raw fish that was eaten, but the dried fish made from raw fish. Based on these structural differences, Subtype (I) is excluded from HIRs, which now consist of Subtypes (II) and (III). The function of a nominalizer no in HIRs (Subtypes II, III) was defined by Horie (1993a: 454) as in (8): (8) A nominalizer in HIRs should allow a matrix verb to access the semantically obligatory argument inside the clause while nominalizing the entire embedded clause as a syntactic argument. (partially modified)
Now, turning to Korean, an interesting contrast was observed between analogues to Subtype (I) (5b), on the one hand, and counterparts to Subtypes (II) and (III) (6b, 7b). It was found that, while the former structural type is widely accepted by native speakers, the latter structural types induce more variable acceptability judgments
 Kaoru Horie
(for instance, compare (6b) and (1b), which are structurally parallel but differ in terms of acceptability). This contrast is non-trivial from language-internal as well as comparative perspectives because (6b) and (7b), but not (5b), are candidates for HIRs in Korean. (5) b. John-un [nalsayngsen-ul malli-n] kes-ul mek-ess-ta. John-top ╛╛↜渀屮raw fish-acc dry-att:pst nml-acc eat-pst-decl ‘John ate fish dried from raw fish (lit. John ate the thing that (one) dried raw fish.’ (6) b. ?/OK John-un [chinkwu-ka konghang-ey tochakha-n] kes-ul John-top ╛╛↜渀屮friend-nom airport-to arrive-att:pst nml-acc
cha-lo hotheyl-ey teyli-ko ka-ss-ta. car-by hotel-to take-ger go-pst-decl
‘John drove his friend, who had (just) arrived at the airport, to a hotel.’ (7) b. ?/OK John-un [sakwa-ka chayksang wi-ey nohye John-top ╛╛↜渀屮apple-nom desk top-loc be put
iss-nun] kes-ul mek-ess-ta. exist-att:prs nml-acc eat-pst-decl
‘I ate an apple which happened to be placed on the desk.’
Horie thus concluded that Korean HIRs are not as fully grammaticalized as their Japanese counterparts. Furthermore, the differential degrees of grammaticalization of HIRs between Japanese and Korean were attributed to lexico-grammatical differences between the nominalizers no in Japanese and kes in Korean. Specifically, Horie suggested that no virtually has no lexical meaning and thus doesn’t interfere in the interpretive process of associating a clause-internal lexical NP (e.g. ringo in (6a)) with the matrix verb (e.g. totte tabeta in (6a)). Unlike no, kes still retains its lexical (if abstract) meaning and can stand in the way of such semantic association, thereby failing to fulfill the function (8) successfully performed by no. For instance, in (7b), the lexical meaning of kes, which is not yet completely bleached, may arguably cause one to wonder whether it is an apple (sakwa) or an unidentified thing (kes) that gets picked up and eaten. Such interference does not take place in (5b), an instance of non-HIR construction (see 5a) where the lexical meaning of kes (‘thing’) contributes to, rather than blocks, the successful interpretation of the nominalized clause. Horie (1998 a, b, 2008) presented additional comparative-typological studies of no and kes from the perspectives of functional extension and grammaticalization. The semantic peculiarity (‘duality’) of HIRs illustrated by example (3) is not an isolated phenomenon, but is observable with other types of constructions in other languages. For instance, the ‘predicative relative construction’ in French is known to yield an ‘event’ interpretation though it formally resembles a regular externally headed relative construction that typically profiles an ‘entity’ (e.g. 2a, b):
Versatility of nominalizations: Where Japanese and Korean contrast 
(9) J’ai vu [Marie qui pleaurait]. I have seen ╛╛↜Marie who was crying ‘I saw Mary crying’. (literally. ‘I saw Mary who was crying’.)
These cross-linguistic observations led Horie (1993b: 76) to propose the notion of ‘ontologically intermediate constructions’: (10) Ontologically intermediate constructions Ontologically intermediate constructions are those linguistic constructions (a) which show a mismatch between linguistic representation and ontological interpretation; and/or (b) which require access to two adjacent ontological levels (e.g. Directly Perceived Events and Indirectly Perceived Events) for a valid semantic interpretation.
The notion (10) was intended to capture the semantic peculiarity of the HIRs and other constructions exhibiting form-meaning ‘mismatch’ cross-linguistically. Section 2.2 recasts the observations and proposal presented in Horie (1993a, b) in view of the Radical Construction Grammar approach to the continuum of complex sentences (Croft 2001).
2.2â•… Continuum of complex sentences There has been growing interest in ‘constructions’ (Goldberg 1995, 2006) and formmeaning ‘mismatch’ phenomena (Francis & Michaelis 2004; Yuasa 2005) as well as ‘intersective constructional gradience’ (Aarts 2007; Bolinger 1961) observed across languages. Among various ‘constructional’ approaches, Radical Construction Grammar (RCG) (Croft 2001) is particularly relevant to the current analysis of ‘ontologically intermediate constructions’ such as Japanese and Korean HIRs. RCG offers a spatial representation of a continuum constituted by four types of complex sentences in traditional grammar, i.e. relative clause, complement, adverbial clause, and coordination based on cross-linguistic data, as shown in Figure 1. The continuum between each complex sentence type is corroborated by a construction, or a set of constructions, which has properties of two ‘pole’ complex sentence types. coordination
serial verbs paratactic clauses speech complements
complements
cosubordination
adverbial clauses
purpose clauses adjoined relative clause correlative clauses relative clauses internally headed relative clauses
Figure 1.╇ The continuum of complex sentence types (Croft 2001:╛322)
 Kaoru Horie
For our current concern, it is important to note that HIR (‘internally headed relative clause’ in Figure 1) is situated between relative clause and complement, reflecting its duality of profiling both ‘entity’ and ‘event’ noted in Section 2.1. Furthermore, it is also noteworthy that an HIR (1a, 1b), while structurally resembling a complement clause (1a′, 1b′), also shows affinity to an adverbial clause (cf. as-clause in English) that describes a co-occurring spatio-temporal situation, as in (11): (11) John-wa [doroboo-ga mise-kara dete kuru] John-top ╛╛↜渀屮thief-nom shop-from come out
{no/tokoro}-o tukamaeta/mita. ╛╛↜渀屮nml-acc caught/saw
‘John caught/saw the thief come out of the shop/as he came out of the shop.’
This suggests that the so-called head-internal relative construction formed by no-nominalization in Japanese is not incompatible with the ‘adverbial clause’ interpretation. The availability of adverbial clause properties to HIRs is reported in Tungusic languages in Malchukov (1996: 380), who states “(s)ynchronically Tungusic internal RCs can be regarded as a hybrid construction sharing features with both adverbial and relative clauses”. In fact, as pointed out by Cu (1983) and Horie (1997, 1998a, b, 2008), nonominalization readily yields ‘adverbial clause’ interpretations when it co-occurs with case particles ga (nominative), o (accusative), ni (dative/locative/allative), and de (instrumental/locative), and the combination of no and case particles have become grammaticalized as clause-linking conjunctions to varying degrees, e.g. semi-grammaticalized ‘antithetical’ conjunctions no-ga, no-o, and fully gramamticalized conjunctions noni (concessive/purposive), and node (causal). An interesting case in point is the semi-grammaticalized combination of no and the accusative case particle o, i.e. no-o, as shown in (12): (12) [Kesseki sita hoo-ga ii] no-o, muri-o ╛╛↜渀屮absence did side-nom good though impossible thing-acc site syusseki site i-ta no kamo sirenai. do:ger attendance do:ger exist-pst nml it may be ‘Though (he) should have been absent, (he) may have attended (the meeting) in spite of the adversative situation.’ (Cu 1983: 86; English translation by the current author)
In (12), the nominalized clause marked by no and followed by the accusative particle o does not function as a complement of any matrix clause. Instead, the clause serves as an antithetical adverbial clause that offers background/circumstantial information. As extensively discussed in Horie (1998a, 2008), kes-nominalization in Korean has not productively grammaticalized into clause-linking conjunctions.
Versatility of nominalizations: Where Japanese and Korean contrast 
The no-nominalized HIR construction is thus most aptly situated in the triangle region, as shown in Figure 2. The triangle region is the intersection of three complex sentence types, i.e. relative clause, complement clause, and adverbial clause, reflecting the interpretive flexibility/versatility of Japanese HIRs. coordination
serial verbs paratactic clauses speech complements
cosubordination
adverbial clauses
purpose clauses adjoined relative clause correlative clauses
complements
relative clauses internally headed relative clauses
Figure 2.╇ Constructional spatial representation of the Japanese HIRs
The kes-nominalized Korean HIRs, in contrast, fail to show such flexibility or versatility, as shown by the lower consistency in native speaker acceptability judgement (6b, 7b), indicating their lower degree of versatility and functional extension compared to their Japanese counterparts. This is graphically represented in Figure 3. coordination
serial verbs paratactic clauses speech complements
complements
cosubordination
adverbial clauses
purpose clauses adjoined relative clause correlative clauses relative clauses internally headed relative clauses
Figure 3.╇ Constructional spatial representation of the Korean HIRs
It is important to note here that such cross-constructional interpretive flexibility/ versatility is not limited to HIRs. Noun-Modifying Constructions, which include Externally Headed Relative Clauses (2a), also exhibit similar interpretive flexibility. Matsumoto (1997) drew attention to the importance of pragmatic factors that license Japanese Noun-Modifying Constructions (NMCs) which include, but are not limited
 Kaoru Horie
to ‘relative clauses’ and ‘complements with nominal heads’ (see also Comrie & Horie 1995), as in (13) and (14) (Yoshiko Matsumoto, personal communication): (13) [atama-no yoku-naru] hon ╛╛↜渀屮head-gen good-become book ‘the book (by reading which) one’s head gets better (one becomes smarter)’ (14) [hutor-anai] okasi-wa nai kasira ╛╛↜渀屮gain weight-neg sweets-top exist.not I wonder ‘I wonder if there aren’t any sweets (even though (one) eats them) which don’t make (one) gain weight.’
Recently, Wang, Horie, and Pardeshi (2009) presented a parallel corpus-based study of Japanese and Mandarin Chinese NMCs. They uncovered a consistent tendency, across different genres and authors, for more Japanese NMC tokens to be used in more functional varieties (i.e. ‘restricting’, ‘information adding’, and ‘complementing’ functions) than their Mandarin Chinese counterparts. Examples (15a, b) illustrate Japanese and Mandarin Chinese NMCs, while examples (15a′, b′) illustrate their corresponding non-NMC constructions (adverbial clause in Japanese, juxtaposition in Mandarin) with the same propositional content (examples (15a–b′) are from Wang, Horie and Pardeshi 2009). (15) NMCs a. [Watasi-no kotoba-o kii-ta] ozyoosan-wa ookata ╛╛↜渀屮I-gen words-acc hear-pst miss-top probably
K-o keibetsu-suru to demo tot-ta no-deshoo. (Kokoro) K-acc despise-do:nonpst that even take-pst nml-sfp
‘Miss, who heard my words, must think that I was despising K.’
b. [Ting-le wo hua de] xiaojie dagai ╛╛↜渀屮hear-pst my words gen miss probably
yiwei wo kanbuqi K ba. thought I despise K sfp
‘Miss, who heard my words, must think that I was despising K.’
(15) Non-NMC constructions a′. [Ozyoosan-wa watasi-no kotoba-o kii-te], ookata K-o ╛╛↜渀屮miss-top i-gen words-acc hear-conj probably K-acc
keibetsu-suru to demo tot-ta no-desyoo. (Kokoro) despise-do:nonpst that even think-pst nml-it will be
‘After Miss heard my words, (she) must think that I was despising K.’
b′. [Dagai shi xiaojie ting-le wo-de ╛╛↜渀屮maybe is miss hear-pst I-gen
hua], yiwei wo kanbuqi K ba. words thought I despised K sfp
‘Miss heard my words, (she) must think I was despising K.’
Versatility of nominalizations: Where Japanese and Korean contrast 
An NMC in Japanese (15a) can manifest ‘duality’ in terms of entity/event interpretations similar to an HIR (e.g. 3) and can be paraphrased by a more explicitly temporal/ situation clause-linkage construction (15a′). Crucially, there is no strong preference for the use of the latter construction in Japanese. In contrast, in Mandarin Chinese, there is a strong preference for a juxtaposed clause linkage construction (15b′) to an NMC (15b). This is another indication of cross-constructional versatility/functional flexibility of a nominal construction in Japanese parallel to HIRs
3.â•… F urther manifestation of constructional versatility: Sentence-final nominalizations in Japanese and Korean Japanese and Korean employ nominalizations not only in non-final position (typically Subject or Object position) as HIRs and complement clauses, but also in sentence-final position. The sentence-final positions in Japanese and Korean are typically occupied by various particles and suffixes that indicate the speaker’s subjective attitude and evaluation toward the propositional content conveyed, as well as her/his intersubjective assessment of/attention toward the addressee (cf. Horie & Taira 2002) It is thus very natural that sentence-final nominalizations formed with no in Japanese and kes in Korean take on various subjective or intersubjective meanings and functions (Traugott 2003; Horie 2007). Evans (2007) recently presented a cross-linguistic survey of what he calls ‘insubordination’, i.e. “the conventionalized main clause use of what, on prima facie grounds, appear to be formally subordinate clauses” (p. 367). ‘Insubordination’ is illustrated in (16), where a main clause infinitive is used to express ‘surprise’ in English. (16) To think that she should be so ruthless! (Quirk et al. 1985, quoted in Evans 2007: 404)
Evans noted that insubordination serves a variety of functions cross-linguistically as in (17): (17) a. indirection and interpersonal control (p. 387): “primarily imperatives and their milder forms such as hints, requests, but also permissives, warnings, and threats.” b. modal insubordination (p. 394): “to express various kinds of modal meaning, both epistemic- having to do with belief, truth, knowledge about the proposition and deontic, i.e. ‘concerned with action, by others and by the speaker himself ’ (Palmer 1986: 96) to bring about a state of affairs denoted by the proposition”
 Kaoru Horie
c. signaling presupposed material (p. 410): “to signal high levels of presupposed material in the insubordinate proposition, i.e. relatively specific presuppositions about the discourse context in which the sentence can occur. (…) Specific examples of this use of insubordination are: (a) negation, (b) focus constructions, (c) discourse contrast, (d) stipulated conditions before assenting to preceding assertions in interaction, (e) reiterations, (f) disagreement with assertions by the previous speaker.”
Evans’ notion of insubordination is highly suggestive for our analysis of sentencefinal nominalizations as the latter constructions at least partially fit the definition of ‘insubordination’. Nominalizations formed by no and kes exhibit versatility in sentence-final position as well, though in a very different way than they manifest constructional flexibility in sentence-medial position as HIRs and related constructions discussed in Section 2.2. They coalesce with copulas, as in no-da and kes-ita or they occur sentence-finally without copulas, as in n(o) and ke(s). Crucially, sentence-final nominalizations in Japanese and Korean develop a variety of pragmatic/semantic functions that can be characterized by employing Evans’ typology (17). The examples that follow in this subsection are taken from Horie, Kim, and Tamaji (2007) unless otherwise noted. I.
Signaling presupposed material: Evidential meaning
Cross-linguistically, nominalizations are known to serve as a source construction for evidentials. As observed by Aikhenvald (2004: 105), “Mood, modality, tense, person, nominalizations, and complement clauses can develop overtones similar to some semantic features of evidentials.” (emphasis added). Sentence-final nominalizations with copulas, i.e. no-da and kes-ita, can encode the presence of some evidence (linguistic or non-linguistic), which is presupposed and used by the speaker to make some inference/judgement, as in (18a, b). For instance, in (18b), the fact that the brushwood door was half open is taken to be nonchallengeable evidence by the speaker to conclude that someone came in the house. (18) a.
Yamada-san-ga ko-nai-na. Kitto yoozi-ga aru-nda. Yamada-Mr.-nom come-neg-sfp surely errand-nom be-noda ‘Mr. Yamada does not come. It must be that (he) has something to do’.
b. Salip-mwun-i pan-ccum yellyecye iss-ess-ta. brushwood-door-nom half-about open be-pst-decl
Tto again
tulli-n-ta. Cip-aney nwukwunka-ka wa iss-nun kes-ita. hear-prs-decl house-in someone-nom come exist-att:prs kes-ita
‘The brushwood door was half open. I can hear (sounds) again. It must be somebody came into the house.’ (Horie, Kim, & Tamaji 2007: 3)
As extensively discussed by Kim and Horie (2009), the evidential function of sentence-final nominalization is more conventionalized in Japanese and is more frequently attested than in Korean.
Versatility of nominalizations: Where Japanese and Korean contrast 
II. Modal insubordination: Deontic meaning vs. Epistemic meaning Sentence-final nominalizations in Japanese and Korean present a remarkable contrast in terms of the type of modality meaning they develop. Japanese sentence-final nominalization no-da can encode a deontic/interpersonal meaning of bringing one’s obligation to her/his attention (19a). Korean sentence-final nominalization kes-ita, in contrast, encodes an epistemic meaning of strong probability (19b). (19) a.
Asu kara siken desyoo? Ganbatte o-benkyoo suru no yo. tomorrow from exam be: fut hard hon-study do noda sfp ‘Your exam (week) will start tomorrow, right? You should work hard.’ (Izuhara 2007: 810, glosses and translation added)
b. Yuwel-i toy-myen pi-ka o–l kes-ita. June-nom become-if rain-nom come-att:fut kes-ita ‘When June rolls around, it will rain (I’m sure).’ (Martin 1992: 607, glosses added)
III. Indirectness and interpersonal control Sentence-final nominalizations in Japanese and Korean can both encode imperatives/ requests, as in (20a, b) (20) a.
Hayaku iku no. quickly go noda ‘Go quickly.’
b. Nayil-kkaci ceychwulha-l kes.3 tomorrow-until submit-att:fut kes ‘Submit (it) by tomorrow’.
.â•… An anonymous reviewer suggested comparing kes-nominalization with koto-nominalization. Indeed, this point was raised in my 1993 paper (Horie 1993a), as shown below: “Korean nominalizer kes typically performs a nominalizing function as a function word, as shown in (i): (i)
Mary-nun [John-i kil-ul kenne-n kes]-ul al-ass-ta. Mary-top John-nom street-Acc cross-AdPast nml-acc know-pst-decl ‘Mary learned that John crossed the street.’
In (i), kes nominalizes the entire embedded clause and turns it into the argument of the matrix verb. However, the Korean kes is not completely a function word; it derives from a lexical noun which means ‘concrete thing’ and ‘abstract thing,’ and it still retains these lexical meanings, even if it behaves like a function word in a variety of syntactic environments. In this respect, kes bears some resemblance to the Japanese functional morpheme koto, which also derives from a lexical noun meaning ‘abstract thing’.” (p. 455). It is noteworthy that Martin (1991) pointed out a possible historical/genetic connection between koto and kes.
 Kaoru Horie
However, the range of interpersonal control functions coded by sentence-final nominalization is very restricted in Korean. For instance, as pointed out by Horie, Kim, and Tamaji (2007: 5), noda can elicit the addressee’s immediate attention/response to a currently relevant event/state of affairs, as in (21a). In contrast, kes-ita is not felicitous in this context and a plain imperative verb form is employed instead, as in (21b). (21) (When explaining the rule of a game) a.
Yoku miru-nda-yo. well look-noda-sfp
b. Cal pwa. well look: imp ‘Please have a good look (and see how to play the game).’
A most conspicuous contrast emerges with ‘indirectness’ function. As noted by Horie, Kim, and Tamaji (2007: 5), noda can be used to preempt a potentially face-threatening act (Brown & Levinson 1978) aimed at an addressee, as a negative politeness strategy, as in (22a). The Korean counterpart kes-ita isn’t felicitous in this context and the bare conclusive verb final ending form is used, as in (22b). (22) a.
Onegai-ga aru-n desu. favor-nom exist-noda:pol
b. Pwuthak-i iss-supnita. favor-nom exist-pol ‘Can I ask a favor of you?’
4.â•… D ifferential degrees of preference for nominal structure between Japanese and Korean Sections 2 and 3 presented cross-linguistic evidence in favor of the greater versatility and functional extension of nominal(ized) constructions in Japanese compared to Korean. This section correlates the cross-linguistic contrast with the differing degrees of discreteness in verb-noun distinction between the two languages. Particular attention will be paid to the cross-linguistic contrasts in terms of (i) formmeaning correspondences manifested by non-conclusive (or non-final) verb forms, i.e. attributive and conjunctive verb forms, and (ii) the differential degrees of explicit distinction between non-conclusive and conclusive/final verb forms. It will be shown that there is a consistent tendency for Japanese to favor a nominal(ized) structure compared to Korean.
Versatility of nominalizations: Where Japanese and Korean contrast 
4.1â•… A ttributive and conjunctive verb forms in Japanese and Korean: Categorial and functional differences Languages have a variety of ‘non-conclusive’ (or non-final) verb forms like infinitives, gerunds, and participles, but their morpho-syntactic manifestations can vary a lot cross-linguistically (Koptejevskaja-Tamm 1993; Haspelmath & König 1995). Nonconclusive verb forms can thus pose an intriguing challenge to parts-of-speech assignment both language-internally as well as cross-linguistically (Hopper & Thompson 1984; Croft 1991, 2001). Particularly noteworthy is the cline of deverbalization (Givón 1990; Malchukov 2006) or “nouniness” squish (Ross 1973) along which various nonconclusive forms can be plotted. Recently, there has been renewed interest in such gradient phenomena (Aarts et al. 2004; Nikolaeva 2007). Non-conclusive verb forms in Japanese and Korean, which include so-called ‘attributive’ and ‘conjunctive’ forms, have similar inventories serving apparently similar syntactic functions such as noun modification and clause combining. However, upon closer examination, non-conclusive verb forms in Japanese and Korean exhibit intriguingly different form-meaning correspondences. Japanese attributive forms, besides serving to modify nouns (23), can form nominal clauses by themselves (24). (23) [eki-ni iku] miti ╛╛↜渀屮station-to go:att:prs way ‘a way of going to the station’ (lit. a going-to-the-station way’) (24) [yama-ni iku] -ni4 husawasii hukusooâ•… (archaic) ╛╛↜渀屮mountain-to go:att:prs to appropriate clothes ‘clothes appropriate for going to the mountains’
Korean attributive verb forms serve to modify nouns, and that is their exclusive function (25). (25) [yek-ey ka-nun ] kil ╛╛↜渀屮station-to go-att:prs way ‘a way of going to the station’ (lit. a going-to-the-station way’)
.╅ An anonymous reviewer pointed out that direct nominalization is fully acceptable with a purposive conjunction ni wa and that it contrasts with no-nominalization, which together with ni forms a purposive conjunction no ni. See Horie (1997a, b) for a functional linguistic analysis of environments favoring direct nominalization and no-nominalization.
 Kaoru Horie
The latter function, referred to as “direct nominalization” by Martin (1975), has been increasingly taken over by sentential nominalizers like no in Modern Japanese, but it continues to be attested even in Modern Japanese (24′). (24′) [yama-ni iku] no -ni husawasii hukusoo ╛╛↜渀屮mountain-to go:att:prs nml to appropriate clothes ‘clothes appropriate for going to the mountains’ (26) a. [Yama-ni iku] yori umi-ni iku hoo-ga ii. ╛╛↜渀屮mountain-to go:att:prs than sea-to go direction-nom good ‘It is better (I like it better) to go to the sea than to go to the mountains.’
In sharp contrast to this, Korean attributive forms lack this “direct nominalization” function altogether (26b): b. [San-ey ka-nun *(kes)] pota pata-ey ╛╛↜渀屮mountain-to go-att:prs ╛╛nml than sea-to
ka-nun phyen-i cohta. go-att:prs direction-nom good
‘It is better (I like it better) to go to the sea than to go to the mountains.’
Furthermore, attributive forms are virtually non-distinct from conclusive forms in Japanese (27a). (27) a. (Attributive) [eki-ni iku ] hito ╛╛↜渀屮station-to go:att person ‘a person going to the station (lit. a going-to-station person)’ (Conclusive) Eki-ni iku. station-to go:fin ‘(I) go to the station.’
This contrasts with the rigid formal distinction between attributive and conclusive forms in Korean (27b). (27) b. (Attributive) [yek-ey ka-nun] salam ╛╛↜渀屮station-to go-att:prs person ‘a person going to station (lit. a going-to-station person)’ (Conclusive) Yek-ey ka-n-ta. station-to go-prs-decl ‘(I) go to the station.’
The differential degrees of explicitness in distinguishing between attributive and conclusive verb forms in Japanese and Korean is presented in Table 1:
Versatility of nominalizations: Where Japanese and Korean contrast 
Table 1.╇ Distinction between Attributive and Conclusive Verb Forms in Japanese and Korean Tense (Mood)
Japanese
Past Present (non-past) Future (intention)
Korean
Attributive
Conclusive
Attributive
Conclusive
-ta -(r)u -(r)u
-ta -(r)u -(r)u
-n -nun -l
-(a/e)ss-ta -n-ta -keyss-ta
What is noteworthy from Table 1 is that while Japanese merges these two forms, Korean consistently distinguishes attributive and conclusive forms. In this connection, it is important to note the directionality of the merging. It is ‘attributive’ verb forms that replaced ‘conclusive’ (or ‘sentence-final’) verb forms, not the other way around. Iwasaki (2000: 243) noted that “(t)he distinction between Conclusive and Attributive forms gradually disappeared, and by the 14th century, the old Conclusive form of some verbs had been replaced by the Attributive form. In other words, the same (Attributive) form now functions both as the Conclusive and as the Attributive” (emphasis added). This seems to indicate that the functions related to attributive verb forms (e.g. modification, nominalization) were expanded to sentence-final predicate position, leading to the emergence of ‘new’ pragmatic/rhetorical functions and possibly at the cost of some functions originally served by conclusive verb forms (e.g. marking assertion). In fact, sentence-final use of attributive verb forms was “a way to suppress or background the force of the assertion” (Iwasaki 2000: 39). Another non-conclusive verb form, i.e. conjunctive verb form, also manifests an interesting cross-linguistic contrast. Conjunctive verb forms in Japanese consist of (i) a form immediately preceding the addressee politeness suffix -mas-, and (ii) V-te form. Conjunctive verb forms in Korean consist of (iii) a bare stem with the endings -a/-e(se) and (iv) a bare stem with the ending -ko. Conjunctive forms in Japanese and Korean serve to form adverbial clauses (28a, b). (28) a.
Nihon-ni iki tomodati-ni at-ta. Japan-to go:conj friend-dat meet-pst ‘(After) going to Japan, I met my friends.’
b. Ilpon-ey ka-se, chinkwu-lul manna-ss-ta. Japan-to go-conj friend-acc meet-pst-decl ‘(After) going to Japan, I met my friends.’
Additionally, one of the conjunctive verb forms in Japanese (i) serves to form a derived noun by itself (29). In contrast, Korean conjunctive forms lack this ‘nominal derivation’ function.
 Kaoru Horie
(29) Iki-wa 10 pun gurai desu ga, kaeri-wa go:conj-top 10 minutes about be:pol but, return:conj motto kakari-masu. more take-pol ‘Going takes about 10 minutes, but returning takes longer.’
In sum, attributive and conjunctive (“renyookei”) verb forms in Japanese, besides their ‘literal’ functions, can serve to form nominals by themselves, a property absent in Korean. Furthermore, attributive forms in Japanese, unlike their Korean counterparts, have merged with conclusive forms. The cross-linguistic contrasts presented in 4.1 suggest that the boundary between noun and verb is less discrete in Japanese than in Korean, and that even apparently verbal structures manifest nominal properties in Japanese.
4.2â•… Differential tendencies toward noun-oriented structure We conjecture that this potential propensity to nominal properties, or ‘nouniness’orientation in Japanese arguably leads to more ‘versatile’ nominalizations in Japanese relative to those in Korean, and accords with a stronger tendency for employing a nominal(ized) structure in Japanese where Korean employs a non-nominal(ized) or verbal structure. The fact that Japanese manifests a tendency toward nominal-oriented, stative/ existential predication has long been noted by cognitive and functional-discourse oriented Japanese linguists, most notably Ikegami (1981), Hinds (1986), Maynard (1997), and Kanaya (2003). However, there have been few studies that analyze the nominaloriented tendency in Japanese from a comparative viewpoint barring Kim (2003), a detailed quantitative contrastive study of Japanese and Korean. Based primarily on a corpus of Japanese novels and their Korean translations, Kim (2003) has convincingly demonstrated the contrastive tendencies in favor of nominal vs. verbal structure between Japanese and Korean. Kim started with nominal structures (indicated in italics) occurring in Subject (30a), Object (31a), Noun-modifying (32a), and Predicate (33a) positions in Japanese and examined whether these nominal structures are rendered into Korean as nominal structures, or, alternatively, as verbal structures (indicated in bold) (30b–33b). (30) a. [Morisita-no koto]-ga… ╛╛↜渀屮Morisita-gen thing-nom ‘Morisita’s matter…’ b. [Morisitha-ka cecilu-n il ]-i… ╛╛↜渀屮Morisita-nom do-att:pst thing-nom ‘The thing Morisita did…’ (ibid: 20, glosses and translation added)
Versatility of nominalizations: Where Japanese and Korean contrast 
(31) a. [Omosiroi katati]-o siteiru yo. ╛╛↜渀屮interesting shape-acc do: cont sfp ‘(It) has an interesting shape, you know.’ b. Caymiisskey sayngkye-ss-e. interestingly be made-pst-decl ‘(It) is formed in an interesting way.’(ibid: 30, glosses and translation added) (32) a. [Ame-no] hi toka-wa… ╛╛↜渀屮rain-gen day such-top ‘As for the days of rain… b. [Pi o-nun] nal-un… ╛╛↜渀屮rain come-att:prs day-top ‘As for the rain falling days…’ (ibid: 40) (33 = 21) (When explaining the rule of a game) (33) a.
Yoku miru-nda-yo. well look-noda-sfp ‘(It is that) you look closely.’
b. Cal pwa. well look: imp ‘Look closely.’
Kim’s finding is striking in that in each of these four positions there is a consistent tendency for (some) Japanese nominal structures to be rendered into verbal structures in Korean, as shown in (34) (based on Table 14 in Kim 2003, p. 67): (34) Subject position: Object position: Noun Modifying position : Predicate Position (with copula): (without copula):
Japanese (100%) vs. Korean (90.4%) Japanese (100%) vs. Korean (92.8%) Japanese (100%) vs. Korean (81.6%) Japanese (100%) vs. Korean (60.4%) Japanese (100%) vs. Korean (64%)
Particularly noteworthy is the fact that about 40% of the tokens of Japanese sentencefinal nominalization in predicate position are rendered into Korean as verbal structures. This observation accords with the finding in Section 3. Kim made the following cross-linguistic generalization: “When representing approximately the same meaning in the same linguistic setting, the Japanese language exhibits relative preference for a noun-oriented structure, while the Korean language, on the contrary, prefers a verboriented structure” (p. 3, emphasis added). Where these differential patterns of preference, i.e. preference for noun-oriented structure in Japanese vs. preference for verb-oriented structure in Korean, originates is an intriguing research question that is beyond the scope of this paper. It is probably
 Kaoru Horie
safe to say at this stage that use of nominalization in various intra-sentential positions, both sentence-medial and sentence-final, is a preferred rhetorical strategy in Japanese. In this connection, it is not insignificant to note some rhetorical functions of sentence-final nominalization in Japanese which have been identified by previous researchers. For instance, Maynard (1997: 115) states that sentence-final nominalization serves the function of “objectifying and stativizing the event through nominalization”. A similar observation was offered by Hinds (1986), who argued that Japanese has a tendency “to express events as states rather than actions” (p. 58) by “end(ing) a sentence, not with a verb, but with a nominal” (p. 62), as shown in (35) (from Hinds, 62–63, glosses and emphasis added, translations partially modified): (35) K: A, honto. Soide sokode umare-ta wake? oh really well there be born-pst reason ‘(lit.) Oh really. Well, (is it that) you were born there?’ A: Soko-de umare-ta wake. there be born-pst reason ‘(It’s that) I was born there.’ K: Soo. Sokode umare-te sokode sodat-ta wake? oh there be born-ger there grow up-pst reason ‘(Is it that) you were born there and grew up there?’ A: Sodat-ta wake. be born-pst reason ‘(It’s that) I grew up there.’
This short conversational exchange is characterized by the constant presence of a sentence-final nominalizer wake meaning ‘reason’ in both interactants’ turns. In Korean, while sentence-final nominalization is available as a means of expression, the extent to which it is actually employed is more restricted than in Japanese. It is thus improbable that sentence-final nominalization is employed in a situation parallel to (35) in Korean. What is crucially needed at the next stage is a systematic cross-linguistic inquiry into interactional, rhetorical, and cognitive motivations for the use and non-use of nominalization in Japanese and Korean.
5.â•… Conclusion This paper started with a comparative study of Japanese and Korean Head Internal Relatives (HIRs), which occur in sentence-medial (non-final) position. Japanese HIRs manifest a greater degree of cross-constructional extension than their Korean counterparts, and are subject to multiple complex clause interpretations such as complement clause and adverbial clause interpretations. Sentence-final nominalized constructions,
Versatility of nominalizations: Where Japanese and Korean contrast 
which take on various pragmatic-semantic functions, also display a parallel crosslinguistic contrast in terms of interpretive versatility. Finally, these cross-linguistic contrasts are correlated with the differential degrees of functional extension of nominal non-conclusive verb forms, and finally with the differential preferences toward noun-oriented structures.
Acknowledgements Thanks are due to two anonymous reviewers for very helpful suggestions and constructive criticism, Andrew Barke, Juyeon Ryu, Spring Ryan, Ahran Kim, Joungmin Kim and Karen Grunow-Hårsta for stylistic suggestions, and particularly to Foong Ha Yap and Janick Wrona for constant encouragement, editorial guidance, and friendship.
References Aarts, Bas, David Denison, Evelien Keizer & Gergana Popova (eds). 2004. Fuzzy Grammar. Oxford: OUP. Aarts, Bas. 2007. Syntactic Gradience. The Nature of Grammatical Ind eterminacy. Oxford: OUP. Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: OUP. Andrews, Avery D. 2007. Noun phrase structure. In Language Typology and Syntactic Description, Vol. II: Complex Constructions, Timothy Shopen (ed.), 206–236. Cambridge: CUP. Bolinger, Dwight L. 1961. Generality, Gradience and the All-or-None. The Hague: Mouton. Brown, Penelope & Levinson, Stephen C. 1978. Politeness. Cambridge: CUP Croft, William. 1991. Syntactic Categories and Grammatical Relations. The Cognitive Organization of Information. Chicago IL: University of Chicago Press. Croft, William. 2001. Radical Construction Grammar. Oxford: OUP. Cu, Le Van. 1983. No ni yoru bun umekomi no koozoo to hyoogen no kinoo (Structure and Expressive Function of Sentence Embedding with No). Tokyo: Kurosio. Evans, Nocholas. 2007. Insubordination and its uses. In Finiteness, Irina Nikolaeva (ed), 366–431. Oxford: OUP. Goldberg, Adele. 1995. Constructions. A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Goldberg, Adele. 2006. Constructions at Work. The Nature of Generalization in Language. Oxford: OUP. Francis, Elaine J. & Michaelis, Laura A. (eds). 2004. Mismatch: Form-Function Incongruity and the Architecture of Grammar. Stanford CA: CSLI. Givón, T. 1990. Syntax. Vol. II. Amsterdam: John Benjamins. Haspelmath, Martin & König, Ekkehard. (eds). 1995. Converbs in Cross-Linguistic Perspective. Berlin: Mouton de Gruyter. Hinds, John. 1986. Situation vs. Person Focus. Tokyo: Kurosio. Hopper, Paul J. & Thompson, Sandra A. 1984. The discourse basis for lexical categories in Universal Grammar. Language 60: 703–752.
 Kaoru Horie Horie, Kaoru. 1993a. Internally headed relative clauses in Korean and Japanese: where do the differences come from? In Harvard Studies in Korean Linguistics 5, Susumu Kuno, Ik-Hwan Lee, John Whitman, Joan Mailing, Yong-Se Kang & Young-Joo Kim (eds), 449–458. Seoul: Hanshin. Horie, Kaoru. 1993b. A Cross-linguistic Study of Perception and Cognition Verb Complements: A Cognitive Perspective. Ph.D. dissertation, University of Southern California. Horie, Kaoru. 1997a. Three types of nominalization in Modern Japanese: no, koto and zero. Linguistics 35–5: 879–894. Horie, Kaoru. 1997b. Reanalysis as a means of “recycling” conventionalized expressions: A case study from Japanese. In Proceedings of the 16th International Congress of Linguists. (CDRom Version Paper No. 0400). Oxford: Elsevier. Horie, Kaoru. 1998a. Functional duality of case-marking particles in Japanese and its implications for grammaticalization: a contrastive study with Korean. In Japanese/Korean Linguistics 8, David Silva (ed.), 147–159. Stanford CA: CSLI. Horie, Kaoru. 1998b. On the polyfunctionality of the Japanese particle no: From the perspectives of ontology and grammaticalization. In Studies in Japanese Grammaticalization: Cognitive and Discourse Perspectives, Ohori, Toshio (ed), 169–192. Tokyo: Kurosio. Horie, Kaoru. 1999. From core to periphery: a study on the directionality of syntactic change in Japanese. In Cognition and Function in Language, Barbara Fox, Dan Jurafsky & Laura A. Michaelis (eds), 1–14. Stanford CA: CSLI. Horie, Kaoru. 2007. Subjectification and intersubjectification in Japanese: A comparativetypological perspective. Journal of Historical Pragmatics 8(2): 311–323. Horie, Kaoru. 2008. The grammaticalization of nominalizers in Japanese and Korean: A constrastive study. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Linguistics 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 169–187. Amsterdam: John Benjamins. Horie, Kaoru & Kaori Taira. 2002. Where Korean and Japanese differ: modality vs. discourse modality. In Japanese/Korean Linguistics 10, Noriko Akatsuka & Susan Strauss (eds), 178–191. Stanford CA: CSLI. Horie, Kaoru, Kim, Joung-Min & Tamaji, Mizuho. 2007. Where Japanese contrasts with Korean and Mandarin Chinese: Intersubjectivity, modality, and the differential pragmaticsemantic foundations across languages. Studies in Pragmatics 9: 1–16. (The Pragmatics Society of Japan). Horie, Kaoru, Pardeshi, Prashant & Kaul, Guy. 2008. Transparency vs. economy. How does Adioukrou resolve the conflict? In Asymmetric Events [Converging Evidence in Language and Communication 11], Barbara L. Tomaszczyk (ed.), Asymmetric Events, 193–206. Amsterdam: John Benjamins. Hoshi, Koji. 1995. Structural and Interpretive Aspects of Head-Internal and Head-External Relatives. Ph.D. dissertation, University of Rochester. Ikegami, Yoshihiko. 1981. Suru to naru no gengogaku (The Linguistics of Doing and Becoming). Tokyo: Taishukan. Iwasaki, Shoichi. 2000. Suppressed assertion and the functions of the final-attributive in prose and poetry of Heian Japanese. In Textual Parameters in Older Languages [Current Issues in Linguistic Theory 195], Susan C. Herring, Pieter van Reenen & Lene Schøsler (eds), 237–272. Amsterdam: John Benjamins. Izuhara, Shoji (ed). 2007. Nihongo ruigi hyoogen tukaiwake ziten (A Dictionary of Japanese Synonymous Expressions and their Usage). Tokyo: Kenkyusha.
Versatility of nominalizations: Where Japanese and Korean contrast 
Jhang, Sea-Eun. 1994. Headed Nominalizations in Korean: Relative Clauses, Clefts, and Comparatives. Ph.D. dissertation, Simon Fraser University. Kanaya, Takehiro. 2003. Nihongo bunpoo no nazo o toku. Aru nihongo to suru eigo. (Solving Mysteries in Japanese Grammar. “Existing” Japanese and “Doing” English). Tokyo: Chikuma Shobo. Kaiser, Stefan. 1991. Cirumnominal Relative Clause in Classical Japanese. An Historical Study. Göttingen: Otto Harrassowitz. Kim, Eun-Ae. 2003. Nihongo no meisi sikoo koozo to kankokugo on doosi sikoo koozoo (The nominal-oriented structure of Japanese and the verbal-oriented structure of Korean). Chosen Gakuhoo 188: 1–83. (Journal of the Academic Association of Koreanology in Japan). Kim, Joung-Min & Kaoru Horie. 2009. Intersubjectification and the textual functions of Japanese noda and Korean kes-ita. In Japanese/Korean Linguistics 16, Yukinori Takubo, Tomohide Kinuhata, Szymon Grzelak & Kayo Nagai (eds), 279–288. Stanford CA: CSLI. Kim, Min-Joo. 2004. Event Structure and the Internally Headed Relative Clause Construction in Korean and Japanese. Ph.D. dissertation, University of Massachusetts. Koptejevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. Kuroda, Shigeyuki. 1974. Pivot-independent relativization in Japanese. Papers in Japanese Linguistics 3: 59–93. (Reprinted in Kuroda 1992, 114–145). Kuroda, Shigeyuki. 1975–76. Pivot-independent relativization in Japanese. Papers in Japanese Linguistics 4: 85–96. (Reprinted in Kuroda 1992, 146–157). Kuroda, Shigeyuki. 1976. Headless relative clauses and the relevancy condition. BLS 2: 269–279. Kuroda, Shigeyuki. 1976–77. Pivot-independent relativization in Japanese. Papers in Japanese Linguistics 6: 59–93. (Reprinted in Kuroda 1992, 157–174). Kuroda, Shigeyuki. 1992. Japanese Syntax and Semantics. Dordrecht: Kluwer. Malchukov, Andrej L. 1996. Internal relative clauses in Tungusic languages in a synchronic and a diachronic perspective. Sprachtypologie und Universalienforschung 49(4): 358–382. Malchukov, Andrej L. 2006. Constraining nominalization: Function/form competition. Linguistics 44(5): 973–1008. Martin, Samuel E. 1975. A Reference Grammar of Japanese. New Haven CT: Yale University Press. Martin, Samuel E. 1991. Recent research on the relationships of Japanese and Korean. In Sprung from Some Common Source, Sydney Lamb & Douglas E. Mitchell (eds), 269–292. Stanford CA: Stanford University Press. Martin, Samuel E. 1992. A Reference Grammar of Korean. Tokyo: Tuttle. Matisoff, James A. 1991. Areal and universal dimensions of grammaticalization in Lahu. In Approaches to Grammaticalization, Vol. 2: Types of Grammatical Markers, Elizabeth C. Traugott, & Bernd Heine (eds), 383–452. Amsterdam: John Benjamins. Matsumoto, Yoshiko. 1997. Noun-Modifying Constructions in Japanese: A Frame-Semantic Approach [Studies in Language Companion Series 35]. Amsterdam: John Benjamins. Nikolaeva, Irina. (ed.) 2007. Finiteness. Oxford: OUP. Nomura, Masuhiro. 2000. The Internally-Headed Relative Clause Construction in Japanese: A Cognitive Grammar Approach. Ph.D. dissertation, University of California, San Diego. Maynard, Senko K. 1997. Japanese Communication. Honolulu HI: University of Hawaii Press. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givon. Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins.
 Kaoru Horie Ohara, Kyoko. 1996. A Constructional Approach to Japanese Internally Headed Relativization. Ph.D. dissertation, University of California, Berkeley. Palmer, Frank Robert. 1986. Mood and Modality. Cambridge: CUP. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1985. A Comprehensive Grammar of the English Language. London: Longman. Ross, John R. 1973. Nouniness. In Three Dimensions of Linguistic Theory, Osamu Fujimura (ed.), 137–257. Tokyo: TEC. Traugott, Elizabeth C. 2003. From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: CUP. Tsubomoto, Atsuro. 1998. Bun renketu no katati to imi to goyooron (Form, meaning, and pragmatics of sentence linkage). In Modality to Hatuwa Kooi (Modality and Speech Act), Noriko Akastuka & Atsuro Tsubomoto (eds), 100–193. Tokyo: Kenkyusha. Wang, Luming, Horie, Kaoru & Pardeshi, Prashant. 2009. Toward a functional typology of noun modifying constructions in Japanese and Chinese: A corpus-based account. Studies in Language Sciences 8, Shunji Inagaki, Makiko Hirakawa, Setsuko Arita, Yahiro Hirakawa, Hiromi Morikawa, Mineharu Nakayama, Hidetosi Sirai & Jessika Tsubakita (eds), 213–228. Tokyo: Kurosio. Yamaguchi, Gyoji. 1992. Kodaigo no zyntai koozoo (Adnominal structure of the Old Japanese). Kokugo Kokubun (Japanese Language & Japanese Literature) 693: 1–16. Yang, Byong-Seon. 1993. Clause structure and NP Accessibility Hierarchy of Korean internally headed relative clauses in Role and Reference Grammar. In Harvard Studies in Korean Linguistics 5, Susumu Kuno, Ik-Hwan Lee, John Whitman, Joan Mailing, Young-Se Kang & Young-Joo Kim (eds), 593–602. Seoul: Hanshin. Yap, Foong Ha, Matthews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implications for unidirectinality from a crosslinguistic perspective. In Up and Down the Cline. The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon. (eds.), 137–168. Amsterdam: John Benjamins. Yuasa, Etsuyo. 2005. Modularity in Language: Constructional and Categorial Mismatch in Syntax and Semantics. Berlin: Walter de Gruyter.
part v
Austronesian languages
The functions of -an and =ay in Kavalan* Fuhui Hsieh
Tatung University This paper reassesses the functions of -an and =ay constructions in Kavalan, a highly endangered Formosan language spoken in eastern Taiwan. The two types of constructions have previously been identified as nominalization and relativization constructions. In the present analysis, Croft’s theory of conceptual space for parts of speech is used to show that both -an and =ay in Kavalan have a wider range of distribution and functions than has been identified in earlier studies. In spite of differences in morphology, -an and =ay share something in common with regard to their functions and syntactic behavior. In other words, they are not two distinct operations at the syntactic level, nor do they display single, clear-cut categorical differences in terms of semantics and pragmatics.
1.â•… Introduction The main purpose of this paper is to reassess the functions of -an and =ay, the so-called nominalization and relativization constructions, respectively, in Kavalan, a highly endangered Formosan language spoken in eastern Taiwan. Kavalan displays a predicate/verb-initial word order, as in (1):1 (1) a.
m-lizaq=iku tu sunis-ku af-happy=1sg.nom obl child-1sg.gen ‘I am happy for my child’ or ‘I like my child’
*I would like to thank our Kavalan informants: Abas (Pan Tien-li 潘天利), Ngengi (Lin A-fen 林阿份), Pilaw (Lin Ai-yu 林愛玉), Api (Zhu A-bi 朱阿比), Imui (Pan Jin-mei 潘金妹), Buya (Xie Zong-xiu 謝宗修), and Raciang (Pan Jin-ying 潘金英), for their astute intuition on the language and their patience. My sincere thanks also go to the editors, Foong Ha Yap, Karen Grunow-Hårsta and Janick Wrona, three anonymous reviewers, and Elizabeth Zeitoun for their very valuable comments on the pre-final draft of this paper. They are not responsible for any remaining errors and omissions, however. .â•… The glosses used in this paper are mostly based on the Leipzig Glossing Rules (2004 version) from http://www.eva.mpg.de/lingua/files/morpheme.html. Abbreviations used in this study are listed below: a.
special orthographic symbols: ’ glottal stop R voiced uvular fricative
 Fuhui Hsieh
b. pukun-an-ku ya lazat unay2 hit-lf-1sg.gen nom person that ‘I hit that person.’ c.
kebalan3 lazat unay Kavalan person that ‘That person is a Kavalan.’
However, unlike many other Formosan languages, Kavalan makes only a two-way distinction in its focus system between actor focus (AF) and locative focus (LF). In its linguistic development, Kavalan shows focus attrition of locative focus (LF) and patient focus (PF) (cf. Shibatani 2005).4 The PF suffix -en, which is common in Formosan languages, disappeared in this language, and the LF suffix -an has lost its function in denoting a location nominative. With the exception of very few examples, however, the nominative argument licensed by the LF -an denotes a patient, rather than a location. Because of such a mismatch between form and function, scholars have been showing divergence in labeling the Kavalan -an suffix. Some (e.g. Chang 1997, 2000; Chang & Lee 2002; among others) label it as PF in terms of function, while some others (e.g. Huang & Sung 2008; Hsieh 2007) label it as LF in terms of form. In an AF clause, the verb is usually marked with the prefix m- or the infix , as in (1a), and the Nominative NP, usually marked by ya or a, is usually an actor or an experiencer of psychological verbs, while the Patient argument always takes the oblique case
b.
acc af cau clf.hum epl exist gen ipl lf lnk loc
Accusative case Agent Focus Causative Prefix human classifier exclusive plural Existential Marker Genitive Case inclusive plural Locative Focus Linker Locative Case
ncm neg nmz nom obl pfv pl pn poss rel sg
Non-common Name Marker Negation Nominalization Nominative Case Oblique Case Perfective plural Proper Noun/Personal Name Possessive Relativization singular
.â•… Note that Kavalan, like Mandarin Chinese, does not specify ‘NUMBER’, ‘GENDER’, and ‘DEFINITENESS’ in its nouns, nor does it have definite articles. The specific interpretation relies on the contextual information. .â•… The word Kavalan, a lexical nominalization with -an nominalizer, means ‘people living in the plains.’ These people called themselves Kavalan to distinguish themselves from the other aboriginal people in the mountain areas, e.g. the Atayals. .â•… For a discussion on focus attrition, see Shibatani (2005).
The functions of -an and =ay in Kavalan 
marker tu.5 In LF clauses, the verb is marked with the -an suffix, as in (1b), and the Nominative NP usually denotes Patient, while the Agent is marked by the genitive case. In tackling with the issue of nominalization and relativization in a particular language, one has to confront two cross-linguistic problems. First, what cross-linguistic criteria are used to delineate parts of speech? Second, what cross-linguistic criteria are used to define nominalization and relativization? While defining parts of speech is comparatively rather easy for Indo-European languages, it has been a challenging task for Formosan languages, since “there is not much morphological difference in shape between a noun and a verb in Austronesian languages in general” (Li 2000: 64; also see Li 2002). Many nouns can be used as verbs without any change in morphology, and vice versa. In other words, telling a noun from a verb cannot depend simply upon their morphological forms. Moreover, as noted by many scholars, the morphemes that occur in indicative verb forms are also used in nominalization in many Austronesian languages (cf. English 1986 and Ramos 1971 for Tagalog; Yamada 1966 for Itbayaten; Rau 2002 for Yami; Huang 2002 for Mayrinax Atayal). In Kavalan, for example, the LF verbal suffix -an can also be identified as a nominalizer, as shown in (2b): (2) a.
supaR-an-ku ti-utay know-lf-1sg.gen ncm-pn ‘I know Utay.’
b. kiya=ma a supaR-an-ku little=only nom know-nmz-1sg.gen ‘What I know is just a little.’
In (2a), -an serves as a verbal marker to the stem supaR ‘know’, yielding a verbal reading ‘to know’, while -an in (2b) is a nominalizer, yielding a nominal reading for supaR-an ‘what I know’. The presence of nominative case a reinforces the nominal status of supaR-an. This intriguing phenomenon makes the analysis of the Kavalan nominalization particularly difficult, as will be further discussed in Section 3.1. Many cross-linguistic studies take a functional approach to analyze both problems. From a functional point of view, nouns are to do reference, namely to refer to an object or entity in the world, while adjectives are to do modification and verbs to do predication. In a similar vein, nominalization is defined as any process that transforms a verb, adjective, or adverb into a noun to do reference, and relativization as the process which transforms a linguistic form to do modification.
.â•… In the literature, there has been a hot debate on the grammatical status of the case marker tu (whether it is accusative or oblique), and this issue is, of course, critical regarding whether Kavalan should be seen as an accusative or ergative language. Liao’s (2002) decisive paper, The interpretation of tu and Kavalan ergativity, appears to have ceased the debate.
 Fuhui Hsieh
Such a functional approach provides a convenient but oversimplified account of the issue. In Kavalan, for example, many nouns are also used to do predication, as shown in (1c). To solve this cross-linguistic difficulty, Croft (2001: 92) in his Radical Construction Grammar proposes the theory of conceptual space for parts of speech. Figure 1 below provides a thumbnail sketch of his theory.
REFERENCE
MODIFICATION
PREDICATION
OBJECTS
unmarked noun
object modifier
object predication
PROPERTIES
property reference
unmarked adjective
property predication
ACTIONS
action reference
action modifier
unmarked verb
Figure 1.╇ Conceptual space for parts of speech (Croft 2001: 92)
Figure 1 shows that a prototypical ‘noun’ is used to do object reference;6 nevertheless, it is not uncommon that a noun is used to do object modification, as the word ‘conference’ in the phrase ‘conference room’, and it is not uncommon either that a noun is used to do object predication. The latter is prevalent in Kavalan, where we often find a noun used as a predicate, especially in equational constructions, as in (1c). Next, a prototypical ‘adjective’ is used to do property modification, such as color terms red, white and black; however, the adjective red can be used as a predicate as in I want my hair red, and it can sometimes also be used as a referential expression, as in Red is the color I like best. Finally, an unmarked verb is aimed to do action predication, but in many cases the action verbs, such as run, fly and walk, can do action modification, as in the relative clause The girl walking in the garden is my sister. Many languages, like English, employ overt nominalization of action words when they are used as referring expressions, namely action reference, as in Walking is a good form of exercise. Croft’s theory of conceptual space for parts of speech will be employed as the base for the discussions in this paper. Now, let us proceed to review nominalization and relativization phenomena in Kavalan. In addition to the aforementioned -an form, there is another marker, =ay, which will be extensively discussed here. In the literature, the -an form is regarded
.â•… The term “object” here is not used in the sense of syntactic object (as opposed to syntactic subject). It is used in the sense of “ontological entity” a la Lyons (1972).
The functions of -an and =ay in Kavalan 
as nominalized construction, mainly as lexical nominalization, while the =ay form is used as a relative clause marker, as shown in (3): (3) yau tanian m-Ruziq=ay tu klisiw-ku sunis ’nay exist where af-steal=rel obl money-1sg.gen child that ‘Where is that child who stole my money?’ (Li 1996: 93–4)
According to previous studies (Chang & Lee 2002; Hsin 1996; Li 1996; Lee 1997), the differences between these two forms lie not only in morphology, but also in syntax and pragmatics. Syntactically, the -an form cannot take another noun phrase as its complement, while the =ay form can. In terms of function, -an turns its host into an argument, while =ay turns its host into a modifier. In other words, -an (nominalization) and =ay (relativization) are treated as two distinct, categorical linguistic operations in Kavalan. However, I will demonstrate with concrete examples in this paper that both -an and =ay have a wider range of distribution and functions than has been identified in earlier studies. Moreover, in spite of the difference in morphology, they share something in common with regard to their functions and syntactic behavior. In other words, claiming that the -an form is nominalization and the =ay form relativization is an oversimplified account of the functions of -an and =ay. Detailed arguments and demonstrations in this paper will provide a clearer and more comprehensive picture about these two forms.
2.â•… The =ay form I identify four patterns of the =ay form in terms of their functions and syntactic behavior, Action Modifier, Property Modifier, Action Reference, and Adverbial Subordination. When the =ay form serves as an Action Modifier, it does two things simultaneously, namely, it nominalizes and relativizes an action word to modify the head noun, as in (4) Pattern 1. When the =ay form serves as a Property Modifier, it is a lexical attributive modifier, as in (5) Pattern 2. Finally, when the =ay form serves as an Action Reference, it nominalizes the verb and is used as a referring expression, as in (6) Pattern 3. Semantically, Pattern 3 can be viewed as a headless relative clause; morphosyntactically, it behaves like a lexical nominal. From this nominal, the =ay form is further used non-referentially to serve as an adverbial subordinate clause, as in (7) Pattern 4.
(4) Pattern 1
m-RaRiw=ti sunis [q<m>an=ay tu biabas-ku] af-run=pfv child eat=rel obl guava-1sg.gen ‘The child who ate my guava ran away.’
 Fuhui Hsieh
(5) Pattern 2
m-lizaq=iku tu [Raya=ay baut] af-like=1sg.nom obl ╛╛╛big=rel fish ‘I like big fish.’
(6) Pattern 3
mai=pama mawtu [s<em>inap=ay-ku] neg=yet af.come sweep=rel-1sg.gen ‘My servant has not come yet.’
(7) Pattern 4
Raya dames utuz a yau big in.particular earthquake lnk that [m-suRaw=ay ci-aliung tu m-qaytun] ╛╛╛af-fall=rel ncm-pn obl af-car ‘There was a strong earthquake when Aliung fell over while riding a motor.’
Each pattern will be discussed in detail in the following sub-sections.
2.1â•… Pattern 1: Action modifier In Pattern 1, the enclitic =ay goes with a clause. The hosts that =ay can encliticize to are mostly action verbs, which carry focus markers and can take an NP complement. The enclitic =ay usually attaches to the main verb in the clause, as in (8a) below, though it can be put at the end of the whole embedded clause as well, as in (8b). (8) a.
m-RaRiw=ti sunis [q<m>an=ay tu biabas-ku] af-run=pfv child eat=rel obl guava-1sg.gen ‘The child who ate my guava ran away.’
b. m-RaRiw=ti sunis [q<m>an tu biabas-ku ]=ay af-run=pfv child [eat obl guava-1sg.gen]=rel ‘The child who ate my guava ran away.’
This type of =ay clause serves as a clausal modifier and provides an action background information for the head noun, something regarding what the head noun has done in an action. In (8a), q<m>an=ay tu biabas-ku ‘(the one) who ate my guava’ is used to modify the head noun sunis ‘child’ and support background information about the child who ran away in the main clause. In this regard, the =ay marker in Pattern 1 behaves like a canonical relativizer, which transforms the encliticized clause into a clausal modifier. Moreover, this type of =ay construction displays a rather flexible word order, occurring either in pre-nominal or post-nominal position (cf. Payne 1997), as in (9) and (10).
The functions of -an and =ay in Kavalan 
(9)
(10)
a.
m-RaRiw=ti tazungan unay [m-Ruziq=ay tu klisiw-ku] -run.away= girl that [-steal= money=1.] ‘That girl who stole my money ran away.’
b.
m-RaRiw=ti [m-Ruziq=ay tu klisiw-ku] tazungan unay -run.away= [-steal=rel obl money=1.] girl that ‘That girl who stole my money ran away.’
a.
ala-an-ku biabas [t<m>ibuk=ay] bring--1. guava [fall=] ‘I brought back the guavas that fell down (from the tree).’
b.
ala-an-ku [t<m>ibuk=ay] biabas bring--1. [fall=] guava ‘I brought back the guavas that fell down (from the tree).’
2.2â•… Pattern 2: Property modifier Despite its similarity in function with Pattern 1, Pattern 2 reveals something different in syntax. First, the syntactic structure of Pattern 2 is not clausal, but lexical; the modifier, i.e. the =ay encliticized form, and the modified head noun together form a noun phrase, which can take a case marker in accordance with the syntactic relation this NP assumes in the clause. Moreover, unlike the =ay clause in Pattern 1, which can occur either pre-nominally or post-nominally, the =ay encliticized form in Pattern 2 is fixed, occurring before the modified head noun only. Second, the intrinsic characteristics of the hosts that =ay can encliticize with are different. In Pattern 2, the hosts are stative predicates, such as terms denoting color, number and dimension, which do not carry any focus marker. Third, the semantic relation between the modifier and the modified is different, too. The =ay encliticized form in Pattern 2 does not provide an action background information for the head noun; rather, it is used to do property modification, either as attributive modifier or genitive modifier. To begin with, I will use the examples in (11) to demonstrate that the syntactic structure of the =ay encliticized form in Pattern 2 is essentially lexical and that this encliticized form behaves like an attributive adjective. (11) a.
[Raya=ay baut]noun phrase big=rel fish ‘big fish’
 Fuhui Hsieh
*b. {[baut Raya]=ay}*noun phrase ╇╛╛fish big=rel c. [baut unay]topic [Raya=ay]comment ╛╛╛fish that big=rel ‘That fish is a big-sized one.’ (Lit.: ‘That fish belongs to those which are big.’)
As seen in (11a) above, the =ay modifier in Pattern 2 can only precede the modified head noun, in this case baut ‘fish’. Unlike Pattern 1, the word order of the =ay marker here is not flexible and cannot be moved out of the noun phrase it modifies, hence the ungrammaticality of (11b). When the =ay form follows the noun, this word order is acceptable only when analyzed as a topic + comment clause as in (11c), where baut unay ‘that fish’ serves as the topic and the =ay form as a headless relative clause conveying the meaning ‘those which are big-sized ones’. When the =ay construction is put into another sentence, as in (12), a clearer picture obtains. In (12a) Raya=ay baut takes the oblique case marker tu and serves as a patient argument of the verb m-lizaq ‘to like’, while baut Raya=ay is not an acceptable word order for an argument here, as in (12b). (12) a.
m-lizaq=iku tu Raya=ay baut af-like=1sg.nom obl big=rel fish ‘I like big fish.’
*b. m-lizaq=iku tu baut Raya=ay af-like=1sg.nom obl fish big=rel Intended meaning: ‘I like big fish.’
Moreover, this =ay form cannot take another tu marked argument as its complement, as in (13a and 13c), nor can it take a genitive argument as its agent, as in (13a and 13b). (13) *a.
Raya=ay-ku tu baut big=rel-1sg.gen obl fish
*b. Raya=ay-ku baut big=rel-1sg.gen fish *c.
Raya=ay tu baut fish=rel obl fish
Paradigmatically, the =ay form in Pattern 2 behaves in an entirely different way from that in Pattern 1; syntagmatically, it is a lexical modifier in a noun phrase. As noted earlier, Pattern 2 =ay can be further sub-classified into two types: attributive modifier, as in (14), in which the =ay encliticized form provides attribute properties for the head noun, and genitive modifier, as in (15), which shows possession relation between the modifier and the head noun.
The functions of -an and =ay in Kavalan 
(14) a.
mazmun=ay sunis clf.hum.many=rel child ‘many children’
b. suqaw=ay lazat bad=rel person ‘gangsters; bad men’ c.
tengen=ay buqes black=rel hair ‘black hair’
(15) a.
tiana=ay sunis zau who=rel child this ‘Whose child is this? (Lit.: This is whose child?)’
b. zaku=ay sunis zau 1sg.poss=rel child this ‘This is my child.’
2.3â•… Pattern 3: Action reference In Pattern 3, the =ay marker encliticizes to both action verbs and stative predicates and the resulting form functions to do reference, referring to any entity involved in the activity or any entity carrying the property. Semantically, this =ay encliticized form can be viewed as a headless relative clause as identified in Payne (1997). According to the parameter and definition proposed by Payne, headless relative clauses are “those clauses which themselves refer to the noun that they modify, usually when the head noun is non-specific” (Payne ibid: 328). Yet, morphosyntactically, it behaves like a lexical nominal, as this =ay encliticized form cannot take another NP as its complement, nor does it modify another NP. Evidence that it forms a nominal expression can be seen in its ability to take a genitive case marker, e.g. -ku in (6) above. Sentences in (16) and (17) are additional examples. (16) a. m-susup aizipna masang, iwaliw tangi aizipna si-klisiw =ay af-poor 3sg.nom before, instead now 3sg.nom have-money=rel ‘He was poor before; (unexpectedly) now he has become a person that has money.’ b. t<m>aqsi=ay study=rel ‘those who study’ (17) m-lizaq=iku tu kitut=ay af-like=1sg.nom obl small=rel ‘I like anything that is small in size.’
 Fuhui Hsieh
We may consider Pattern 3 =ay as an agent nominalization, since the derived form refers to an agent involved in doing an activity or carrying a property. However, Pattern 3 =ay differs from the pa-…-an form (the other agent nominalization in Kavalan, which we will discuss in detail in Section 3.3) in that the pa-…-an form denotes a habitual agent, whereas the Pattern 3 =ay form refers to a non-habitual agent. By habitual agent I mean a person who engages in the activity by profession, while the non-habitual agent refers to anyone who is occasionally involved in the activity denoted by the verb. The subtle difference between these two types of agents can be observed by the following pairs: students vs. (those) who study, cooks vs. (those) who cook, and singers vs. (those) who sing. We may therefore say that Pattern 3 in general is not as highly lexicalized as the pa-…-an form. Nonetheless, a few of the =ay forms in Pattern 3 are found to have acquired lexical status due to their high frequency in discourse, as shown in (18) and (19) below. (18) yau=ti si-kubu=ay exist=pfv wear-hat=rel ‘Here comes a police officer.’ (< ‘Here comes one who wears a hat.’) (19) a.
seminap=ay sweep=rel ‘servant; sweeper’ (< ‘one who sweeps’)
b. mai=pama mawtu s<em>inap=ay-ku neg=yet af.come sweep=rel-1sg.gen ‘My servant has not come yet.’ c.
s<em>inap=ay s<em>inap=ay-ku sweep=rel sweep=rel-1sg.gen ‘My servant is the one who sweeps (here).’
In the Kavalan villages during the Japanese occupation period, only police officers had the privilege to wear hats. The Kavalan therefore used the term si-kubu=ay ‘one who wears a hat’ to refer to police officers. It was so frequently used that the term acquired lexical status. It now continues to be used to refer to the police, even though everyone now is free to wear a hat. The other example is seminap=ay ‘one who sweeps’. Since the servant is frequently hired to sweep and clean the house for others, this term has also acquired lexical status, and applies to servants in general. Nonetheless, it is worth noting that these examples are ambiguous; si-kubu=ay can refer to either those who wear a hat or police officers, and seminap=ay those who sweep or servants. Disambiguation is often dependent on context.
2.4â•… Pattern 4: Adverbial subordinate clauses As pointed out in Yap, Grunow-Hårsta and Wrona this volume, the nominalizer is easily reinterpreted as a subordinator, as the nominalization construction is frequently
The functions of -an and =ay in Kavalan 
used in topic position to set up the stage (e.g. temporal or suppositional background) for the upcoming clause. In other words, nominalization constructions are apt to be extended non-referentially to serve as adverbial subordinate clauses. This development is also attested in Kavalan, as in (20) below. (20) a.
Raya dames utuz a yau big in.particular earthquake lnk that
m-suRaw=ay ci-aliung tu m-qaytun af-fall=rel ncm-pn obl af-car
‘There was a strong earthquake when Aliung fell over while riding a motor.’
b. wama zana iRuR=ay zana iRuR=ay ma tamun-ta only 3.poss river=rel 3.poss river=rel only dish-1ipl.gen
zana lazing=ay wama zana lazing=ay tamun-ta 3.poss sea=rel only 3.poss sea=rel dish-1ipl.gen
‘If/When we have only those from the river, those from the river will be our dishes. If/When we have those from the sea only, those from the sea will be our dishes.’
In both (20a) and (20b) above, =ay serves as a subordinator; it marks a subordinate clause which sets up the temporal frame for the main clause.
2.5â•… Interim summary It should now have become clear that Kavalan =ay contributes to more than just relativization. In terms of function, the =ay form not only does modification (Pattern 1 and 2), but also does reference (Pattern 3). Moreover, it is extended non-referentially and functions as an adverbial subordinate clause (Pattern 4). Even in modification, the =ay form engages in division of labor, often with the aid of syntax: Pattern 1, a canonical relative clause, provides action modification, while Pattern 2, a lexical attributive modifier, functions to do property modification. The distinction in syntax may lead to entirely different sentence readings. By adopting Croft’s (2001) theory of conceptual maps for parts of speech as a framework, I have identified the distribution and functions of the Kavalan=ay forms as in Figure 2.7
.â•… The adverbial function does not appear in Croft’s original conceptual maps, but is added here to encompass the four patterns of the Kavalan =ay form.
 Fuhui Hsieh
REFERENCE
MODIFICATION
Adverbial
Pattern 2 adjective-like modifier (property modification) Pattern 1 relative clause (action modification)
Pattern 4 adverbial subordinate clauses
OBJECTS PROPERTIES
ACTIONS
Pattern 3 headless relative clause (action nominals)
Figure 2.╇ Semantic map for the =ay form in Kavalan
These four different patterns are not unrelated. As many studies show that relative clauses are secondary derivations from nominalization constructions (e.g. this is the case for Old Chinese nominalizer zhe (Yap & Wang this volume), and it is also the case for many Tibeto-Burman languages (Genetti this volume)). Ross (1995: 742ff) reconstructs PAn (Proto-Austronesian) *-ay as a ‘projective’ form for locative focus (or locative voice) verbs, and notes that it is well attested across Formosan languages.8 Although =ay is not found for the locative verbs in the synchronic Kavalan data, Ross’ (ibid) reconstruction might shed some light on the historical origin of the Kavalan =ay. One may very well hypothesize that the Kavalan =ay first emerged as a locative noun, developed into a light noun, was then grammaticized as a nominalizer, then, gradually gaining its “projective” readings, evolved as a projective marker on LF verbs and eventually developed into a LF verb form.9 Of course, this hypothesis needs to be further supported by comparative evidence. From this nominal use, the Kavalan =ay further obtains its secondary derivation of relativization; thus, we may have Pattern 3 [VO=ay] > Pattern 1 N [VO=ay]. As for Pattern 2 (lexical modifier), it can be viewed as a sub-type of Pattern 1 (clausal modifier), as Pattern 2 are lexical terms that have developed via grammaticalization from Pattern 1. Note that from the nominal Pattern 3 (headless relative clause), the =ay form is also found to be used non-referentially to serve as adverbial subordinate clauses (Pattern 4).
.â•… The term “projective” refers to the “finite verb form used to express intention, possibility, and exhortation” (Ross 1995: 742). .â•… The marker -ay is used as a LF verb suffix in Puyuma, another Formosan language spoken in southeastern Taiwan (cf. Stacy Teng 2008).
The functions of -an and =ay in Kavalan 
3.â•… The -an form10 As mentioned earlier, the Kavalan -an suffix can serve either as a nominalizer or as a verbal suffix, by which the duality of functions frequently poses an analytical problem in analyzing the Kavalan nominalization. I will therefore begin this section by clarifying this problem before I go any further. Next, I will show with concrete examples that the -an form is not a lexical nominalization, for the -an marked form can take a genitive argument as its agent and a tu marked argument as its complement. Then, I will discuss two types of the -an form. The first type, behaving like a headless relative clause, is to do reference. The second type, like the English gerund, is used to refer to the activity denoted by the whole -an clause. Finally, I will discuss the pa-…-an form, which is mainly used to do reference, viz. the so-called agent nominalization. I will then compare the difference between the pa-…-an form and the Pattern 3 =ay, which is also used to do reference.
3.1â•… One analytical problem Let me begin with a frequently cited example, as in (21): (21) nengi sanu-an na sunis a yau good educate-nmz gen child lnk that ‘That child’s education is good.’
The -an form in (21) is analyzed as a lexical nominalization by previous studies (cf. Chang & Lee 2002). As shown in the translation, sanu-an is analyzed as a lexical nominal conveying the meaning ‘education’, and na as the genitive case, which can be roughly translated as ‘of ’ in English.11 In other words, the relation between the derived noun and the source verb was treated as patterning with those attested in English pairs, e.g. educatev à educationn and instructv à instructionn.
.â•… This -an form can go with many different prefixes and together form clausal nominals. Apart from the qena-…-an form identified by Chang and Lee (2002), there are ni-…-an, sa-…-an, qa-…-an, and <en>…-an. The prefix ni- and the infix <en> are perfective markers, while the prefix sa- mainly conveys purposive meaning along with many other usages (cf. Jiang 2006). Although the prefix qena- is attested to be derived from the compounding of the stative prefix qa- and the perfective infix <en> and mainly goes with stative verbs, the prefix qa- alone conveys irrealis reading in the -an construction. .â•… In Kavalan, the genitive case marker and the 3rd person genitive pronoun share the same form, na, in morphology. This morpheme frequently poses problems in analyzing naturally spoken data. See Appendix for the personal pronominal system and the case marking system in Kavalan.
 Fuhui Hsieh
I here propose an alternative analysis; that is, to treat the -an suffix in (21) as a locative focus marker and the na morpheme as the 3rd person bound pronoun, which denotes the agent of the verb sanu ‘to educate; to instruct’, as in (22a) below. (22) a. nengi sanu-an-na sunis a yau good instruct-lf-3sg.gen child lnk that ‘It is easy (good) for someone to instruct the child (i.e. to tell the child to do some work).’ b. nengi sanu-an-su sunis a yau good instruct-lf-2sg.gen child lnk that ‘It is easy (good) for you to instruct that child.’ c.
nengi sanu-an-na ti-utay good instruct-lf-3sg.gen ncm-pn ‘It is easy (good) for one to instruct Utay.’
*d. nengi sanu-an ni utay good instruct-lf gen pn *‘Utay’s education is good.’
Obviously, the divergence lies in the grammatical status of the morpheme na in these two analyses. If we analyze it as a genitive case marker, the -an form will be treated as a lexical nominalization and the whole sentence will render a reading as in (21). When we analyze it as an agent, i.e. 3rd person genitive pronoun, however, the -an form will be analyzed as a clausal nominalization reading as in (22a). That the na form in (22a) is the 3rd person bound pronoun and that the -an form is indeed a clausal nominalization can be evidenced by the other two syntactically parallel sentences in (22b) and (22c). If the analysis in (21) were correct, we could not explain the occurrence of the second person genitive -su in the nominalization construction in (22b). Moreover, if the analysis of the sentence in (21) as lexical nominalization were to be sustained, the genitive case marker for the proper name Utay should be ni, instead of na (compare 22c and 22d). We may use the existential negation construction as another syntactic test. In Kavalan, the negated noun phrase in the existential construction should take the oblique marker tu following the negator mai, as shown in (23a). If sanu-an were a lexical nominalization, it should be able to take the oblique marker tu in the negation construction, but this is not acceptable as seen in (23b). (23) a.
mai tu klisiw na lazat ’nay neg obl money gen person that ‘That person does not have money. (Lit. ‘There is no money of that person.’)
The functions of -an and =ay in Kavalan 
*b. mai tu sanu-an na lazat ’nay neg obl instruct-nmz gen person that
Having thus settled the analytical dispute, I will now proceed to examine the distribution and functions of Kavalan -an in the following sub-sections.
3.2╅ Two types Let me begin with the example in (24) below. (24) niz-an-ta q<m>an tu [ni-kiala-an na tama-na all-lf-1ipl.gen eat obl ╛╛╛pfv-pick-nmz gen father-3sg.gen
tu biabas] obl guava
‘We ate all of the guavas that his father picked.’
It is clear that -an is not used as a lexical nominalization here, as the -an marked form ni-kiala-an ‘pfv-pick-nmz’ in (24) not only takes a tu marked noun tu biabas ‘guavas’ as its complement, but also is followed by a genitive marked agent, i.e. na tama-na ‘by his father’. Moreover, the whole -an marked clause in (24) modifies and refers to the embedded tu marked argument, i.e. the guavas (that his father picked), which in turn functions as the patient of the matrix verb q<m>an ‘eat’. Note that case marking inside the -an clause does not follow the same principles that operate in main clauses; the patient argument biabas takes an oblique case marker tu instead of a nominative.12 This tu marked argument in the -an clause cannot be analyzed as an argument of the matrix clause, since it cannot be moved to the position right after the matrix verb. It is the -an clause as a whole that receives case marking on the basis of its grammatical role in the matrix clause. Note the oblique marker tu preceding the whole -an clause, indicating the -an clause as a whole functions as the Patient of the matrix verb qman ‘eat’, viz. the guavas that his father picked were something we ate. In this regards the -an construction is quite similar to a headless relative clause with the embedded tu marked argument as a default head noun (Payne 1997: 328; see also Cole et al. 1982: 116).13
.â•… In LF clauses, the patient argument takes the nominative case marker ya. .â•… Cole et al. (1982: 116) treat such a syntactic structure as a headless relative clause, since “the relativized NP appears in its normal position within the relative clauses.” However, Dryer (2005: 366) treats it as an internally headed relative clause. I don’t want to go into the dispute on the terminology here. I follow Payne’s and Cole et al.’s terminology, simply because the Kavalan -an very often goes without the tu marked argument, which is obviously a “headless” clause.
 Fuhui Hsieh
This is of course reminiscent of the =ay form in Pattern 1 (the RCs, or relative clauses). For comparison, I juxtapose these two constructions in (25). (25) a.
pukun-an-ku sunis ’nay [q<m>an=ay tu qawpiR] hit-lf-1sg.gen child that eat obl sweet.potato ‘I hit the child who ate sweet potatoes.’
b. m-liuq=ti [sa-paluma-an-ku tu sinsuli] af-rotten=pfv ╛╛╛sa-plant-nmz-1sg.gen obl plum ‘(The seedlings of) the plums that I am going to plant are rotten.’
As revealed in (25), we can see that there are subtle differences between these two types of relative clauses. The first one is the occurrence of an external head noun in the =ay clause, which is absent from the -an clause. This makes their respective functions different; the =ay clause is used to do modification, whereas the -an clause is used mainly to do reference. Next, it is worth noting that the relativized NP in the =ay clause is usually the subject (or actor) argument in AF clauses, whereas in the -an clause it is always the embedded tu marked argument that gets relativized. This tu marked argument in -an clauses is indeed the patient/theme argument in the clause. The S/O alignment in relativization is another piece of syntactic evidence of Kavalan ergativity, in addition to the identification of the oblique case marker tu (Liao 2002). What is intriguing is that the relativized patient argument is marked by tu, an oblique marker marking theme argument in an intransitive or extended intransitive clause, i.e. AF-clause. This particular marking makes -an appear to relativize an internal argument.14 Nonetheless, an -an clause often goes without a tu marked argument. In this case, the whole -an clause refers to any entity that the verb acts on, as shown in (26) below. (26) a.
lasam ni-imet-an-su af.dirty pfv-hold-nmz-2sg.gen ‘What you hold is dirty.’
b. m-patay=ti ni-paluma-an-ku af-die=pfv pfv-plant-nmz-1sg.gen ‘What I planted is dead.’
It is of particular interest to note a very exceptional use of the -an clause, which is used to denote the location where the event takes place, as shown in (27), taken from naturally spoken data.15
.╅ I would like to thank one of the reviewers for pointing this out to me. .╅ The data is taken from the NTU Corpus of Formosan Languages at http://www.corpus. linguitics.ntu.edu.tw (cf. Sung, Su, Hsieh & Lin 2008).
The functions of -an and =ay in Kavalan 
(27)
Pear-buya 1.
yau baqian ’nay usiq._ exist elder.male that one
2. ...(2.4)matiw .go
ta
Location
ni-paluma-an-na tu sinsuli._ -plant--3. plum
‘There was an old man. (He) went to the place where he grew plums.’
Entity
(28) m-niz=ita q<m>an tu ni-paluma-an-na -all=1. eat -plant-nmz-3.
tu
sinsuli plum
‘We all ate the plums that he planted.’
In (27) the -an clause does not refer to an entity but rather denotes the location where the event occurs, i.e. the place where the old man grew plums; note that the locative case marker ta precedes the -an clause. In the absence of the locative case marker ta, the same -an clause in (28) refers to the tu marked argument, i.e. the plums that the old man grew. We can easily see that the contextual clues, i.e. case marking and the semantics of the matrix verb, may help determine the exact reading. At present, example (27) is the only example I have found; further study may be needed to disclose this particular function. The second type of -an clauses is used to refer to the activity that the whole -an clause denotes, as in (29). (29) a.
nengi=ti [kiala-an na tama-na tu biabas] good=pfv ╛╛╛pick-nmz gen father-3sg.gen obl guava His father’s picking guava is good.’
b. qaqaytisan [zaqis-an-na tu paRin] dangerous ╛╛╛climb-nmz-3sg.gen obl tree ‘His tree climbing is dangerous.’ *c.
tayta-an-ku [kiala-an na tama-na tu biabas] see-lf-1sg.gen ╛╛╛pick-nmz gen father-3sg.gen obl guava Intended meaning: ‘I saw his father’s picking guavas.’
*d. qapaR-an-ku [zaqis-an-na tu paRin] catch-lf-1sg.gen ╛╛╛climb-nmz-3sg.gen obl tree Intended meaning: ‘I caught his the climbing tree.’
 Fuhui Hsieh
As revealed in (29), the -an construction is on a par with the English gerund. Sentence (29a) says the activity (i.e. his father’s picking guava) is good, and (29b) depicts that his tree-climbing is dangerous. The difference in the syntactic behavior between these two types of -an clauses is that the headless relative clause is embedded and plays a semantic role, mostly patient, in the matrix clause. Moreover, the headless relative clause takes a case marker in accordance with the semantic role it assumes in the sentence. However, the gerundive use mostly occurs as topic/subject arguments in adjectival-predication constructions;16 moreover, the gerundive -an clause cannot be used as a complement clause in a finite clause. Therefore, it is unacceptable to say ‘I saw his father’s picking guavas’ or ‘I caught his climbing the tree’, as is seen in the the unacceptability of (29c–d).
3.3â•… Lexical -an nominals: Agent nominalization As observed by Chang and Lee (2002: 363), in Kavalan, an agentive nominalization is formed by attaching the affixal complex pa-…-an to a source verb, as in (30) below. (30) a.
pa-Ribaut-an cau-catch.fish-nmz ‘fisherman’
b. pa-taqsi-an cau-study-nmz ‘student’ c.
pa-klawklaway-an cau-work-nmz ‘worker’
While this is true, it is not the only way to form an agentive nominalization in Kavalan. Another possible way to derive an agentive noun is by encliticizing =ay to the source verb as shown by the examples in (31) and (32), the latter reproduced from (19) earlier: (31) qay-Ruziq=ay aisu qay-steal=rel 2sg.nom ‘You are the one who stole (my stuff)’ or ‘You are a thief ’.
.╅ An adjectival-predication construction is a construction where an adjectival verb serves as the main predicate, which does not take any patient argument as its complement. Like other Austronesian languages, Kavalan, known to generally lack a copula, simply juxtaposes topic/ subject arguments and adjectival predicates.
The functions of -an and =ay in Kavalan 
(32) a.
seminap=ay sweep=rel ‘servant; sweeper’ (< ‘one who sweeps’)
b. seminap=ay-ku sweep=rel-1sg.gen ‘my servant’
As described in the previous section, the =ay derived noun is usually used when the head noun is a non-habitual agent, someone who does the activity mentioned only occasionally or just once in a while. The pa-…-an derived agentive noun makes reference to a habitual agent, someone who does the activity by profession. This contrast is highlighted in the examples in (33). (33) a.
salekiaw ‘dance’
pa-salekiaw-an salekiaw=ay
‘dancer’ ‘one who dances’
b.
satezay ‘sing’
pa-satezay-an satezay=ay
‘singer’ ‘one who sings’
c.
sa’may ‘cook (rice)’
pa-sa’may-an sa’may=ay
‘servant;cook’ ‘one who cooks’
d.
sudad ‘write’
pa-sudad-an sudad=ay
‘office-worker’ ‘one who writes’
A related issue I would like to reiterate here is that the noun-verb boundary in the pa-…-an derived nominal construction is not a clear-cut line. While English (and other languages as well) uses an unmarked noun (object-reference), Kavalan uses an action nominal (action-reference) to encode the concept ‘teacher; one who teaches’, as in (34) and (35). (34) a.
pa-tud-an tu pataqsian cau-teach-nmz obl student ‘teacher’
b. pa-tud-an=iku tu pataqsian cau-teach-nmz=1sg.nom obl student ‘I am a teacher. (Lit.: I teach students.)’ (35) [pa-tud-an timaiku] mai mawtu taqsian ╛╛╛cau-teach-nmz 1sg.acc neg af.come school ‘My teacher did not come to school.’ Lit. ‘The one who teaches me did not come to school.’
 Fuhui Hsieh
Syntactically, the nominal patudan ‘teacher’ in (34) takes a tu marked argument as its patient and in combination with this patient argument refers to the concept ‘teacher’. Note that the syntactic structure in (34b) is exactly the same as that of an AF-clause, viz. a predicate followed by a bound nominative argument and a tu patient argument. The same description applies in (35). Based on these linguistic data, we may conjecture that the Kavalan language tends to use action reference, viz. event or activity, to code the concept of entity.17
3.4â•… Interim summary It should now be very clear that the -an form is essentially not lexical nominalization. One of its main functions, headless relative clause, is to do reference and, at the same time, modification (modifying the argument embedded in the clause). The other function, the gerundive usage, is to do action reference. The former mostly occurs in an embedded clause, whereas the latter in an adjectival-predication structure. The distribution and functions of -an are summarized in the semantic map in Figure 3 below. REFERENCE
MODIFICATION
OBJECTS PROPERTIES
-an clause (headless relative clause: entity, location)
ACTIONS
-an clause (gerund) pa-…-an lexical nominalization
Figure 3.╇ Semantic map of the Kavalan -an
Comparing the semantic maps of =ay in Figure 2 with -an in Figure 3, we can see that both have their distinct functions, yet at the same time share some functions in common. The =ay form displays three distinct patterns in terms of their syntactic behavior and functions, i.e. Action Modifier in Pattern 1 (headed or ordinary relative .â•… This phenomenon is not uncommon in other Formosan languages. For example, in Saisiyat, where nominalization processes are rather productive, many of what Lyons (1977: 442–7) calls ‘first-order nouns’ or what Wierzbicka (2000: 291) names ‘prototypical nouns’ in Saisiyat are derived from nominalization, e.g. kama-manra:an kama-af.walk ‘man’, min-koring-an min-hit-lf ‘woman’, ka-ngesel-an ka-have.running.nose-lf ‘nose’, and ha-hila ca-sun.rise ‘the sun’. Moreover, the Saisiyat ma’iyaeh ‘man’ might track its root back to the verb ’i’iyaeh ‘be alive’ (cf. Yeh 2003).
The functions of -an and =ay in Kavalan 
clause), Property Modifier in Pattern 2 (adjective-like modifier), Action Reference in Pattern 3 (headless relative clause), and subordination in Pattern 4 (subordinate clauses). The -an form has two main functions: the headless relative clause is used to do both modification and reference, while the gerundive usage is used to denote an activity. In terms of functions, both -an and =ay constructions can be used as relative clauses to do modification and reference. Syntactically, the -an clause does not have an external head noun to modify, while the =ay construction, except the headless relative clause in Pattern 3, always has an external head noun. Moreover, the relativized NP in the =ay clause is usually the subject (or actor) argument in AF clauses, whereas in the -an clause it is always the embedded tu marked argument that gets relativized. Interestingly enough, this S/O alignment in relativization further evidences syntactic ergativity in Kavalan. As for the agent nominalization in Kavalan, it can be manifested by either the affixal complex pa-…-an or the Pattern 3 =ay strategy. The difference between the pa-…-an derived agentive noun and the =ay derived agentive noun is that the former denotes a habitual agent while the latter refers to someone who does the action occasionally or just once in a while. Nonetheless, it is of particular interest to note that the pa-…-an derived lexical nominal, like a predicate, can take another patient argument as its patient and in combination with this patient denotes the agent. We may thus conclude that relativization and nominalization appear not to be two unrelated linguistic mechanisms in Kavalan, especially as many studies (e.g. Shibatani 2009; Yap & Wang this volume; Genetti this volume) reveal that relative clauses are secondary derivations from nominalization constructions. Moreover, many studies in this volume (e.g. Shinzato this volume; Yap & Wang this volume; Yap, Grunow-Hårsta & Wrona this volume) indicate that headless relative clauses are nominalization. On the basis of just the Kavalan data and on purely logical grounds, we might posit the following relations among the various functions of =ay and -an. lexical source unknown
-an nominal
lexical source unknown
locative noun =ay nominal
gerund
RCs
attributive modifier genitive modifier
Adverbial subordinate clause Figure 4.╇ Relations among the various uses of the Kavalan =ay and -an
 Fuhui Hsieh
4.â•… Conclusion I hope I have made it clear that simply labeling =ay and -an constructions as respective relativization and nominalization in Kavalan indeed is not a sufficient way of depicting their functions. In this paper, I have demonstrated with concrete examples that =ay and -an in Kavalan are not two distinct operations in syntax, nor do they display single, clear-cut categorical differences in terms of semantic and pragmatic functions. I have also shown that the Kavalan language tends to use action reference, viz. event or activity, to code the concept of entity.
Appendices Table A.╇ Personal pronominal system in Kavalan (cf. Chang 2000: 84) Number
Singular
Plural
Bound
Free Form
Person
Nom
Gen
Nom
Acc
Loc
Poss
1st
=iku
-ku
aiku
timaiku
timaikuan tamaiku
zaku
2nd
=isu
-su
aisu
timaisu
timaisuan tamaisu
zasu
3rd
---
-na
aizipna
timaizipana
tamaizipana
zana
1st.Inc.
=ita
-ta
aita
timaita
timaitaan tamaita
zaita
1st.Exc.
=imi
-niq
aimi
timaimi
timaimian tamaimi
zaimi
2nd
=imu
-numi
aimu
timaimu
timaimuan tamaimu
zaimu
3rd
---
-na
qaniau
qaniau
qaniauan
zana
Table B.╇ Case marking system in Kavalan (cf. Chang 2000: 68) Nominative
Oblique
Genitive
Locative
Common noun
ya/a
tu
na
ta…an (in/on/at); sa- (to); maq- (from)
Personal Name & Proper Name
ya/a
tu
ni
…an
The functions of -an and =ay in Kavalan 
References Chang, Yung-li. 1997. Voice, Case and Agreement in Seediq and Kavalan. Ph.D. dissertation, National Tsing Hua University. Chang, Henry Yung-li. 2000. Gemalanyu Cankaoyufa (A Kavalan Reference Grammar) [Series on Formosan Languages 12]. Taipei: Yuanliou. Chang, Henry Yung-li & Lee, Amy Pei-jung. 2002. Nominalization in Kavalan. Language and Linguistics 3(2): 349–367. Cole, Peter, Harbert, Wayne & Hermon, Gabriella. 1982. Headless relative clauses in Quechua. International Journal of American Linguistics 48(2): 113–124. Croft, William. 2001. Radical Construction Grammar. Oxford: OUP. Dryer, Matthew S. 2005. Order of relative clause and noun. In The World Atlas of Language Structure, Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds), 366–367. Oxford: OUP. English, Leo Lames. 1986. Tagalog-English Dictionary. Manila: National Book Store. Genetti, Carol. This volume. Nominalization in Tibeto-Burman languages of the Himalayan areas: A typological perspective. In Yap, Grunow-Hårsta & Wrona (eds). Hsieh, Fuhui. 2007. Language of Emotion and Thinking in Kavalan and Saisiyat. Ph.D. dissertation, National Taiwan University. Hsin, Aili. 1996. Noun phrase structure and focus marking in Kavalan. Tsing Hua Journal of Chinese Studies, New Series 26(3): 323–364. Huang, Lillian M. 2002. Nominalization in Mayrinax Atayal. Language and Linguistics 3(2): 197–225. Huang, Shuping & Sung, Li-May. 2008. The undergoer ma- in Kavalan. Oceanic Linguistics 47(1): 159–184. Jiang, Haowen. 2006. Spatial Conceptualizations in Kavalan. MA thesis, National Taiwan University. Lee, Amy Pei-jung. 1997. The Case-marking and Focus Systems in Kavalan. MA thesis, National Tsing Hua University. Li, Paul Jen-kuei. 1996. Yilanxian Nandao Minzu Yu Yuyan (The Formosan Tribes and Languages in I-Lan). I-Lan: I-Lan County Government. Li, Paul Jen-kuei. 2000. Nominalization in Pazih. Paper presented at the Workshop on Nominalization in Formosan Languages, Academia Sinica, Taiwan, October 21–22. Li, Paul Jen-kuei. 2002. Nominalization in Pazih. Language and Linguistics 3(2): 227–239. Liao, Hsiu-chuan. 2002. The interpretation of tu and Kavalan ergativity. Oceanic Linguistics 41(1): 140–158. Lyons, John. 1977. Semantics. Vol. 1 & 2. Cambridge: CUP. Lyons, John. 1972. Structural Semantics: An Analysis of Part of the Vocabulary of Plato. Oxford: Blackwell. Payne, Thomas. 1997. Describing Morphosyntax: A Guide for Field Linguists. Cambridge: CUP. Ramos, Terisita V. 1971. Tagalog Dictionary. Honolulu HI: University of Hawai’i Press. Rau, D. Victoria. 2002. Nominalization in Yami. Language and Linguistics 3(2): 165–195. Ross, Malcom D. 1995. Reconstructing Proto-Austronesian verbal morphology: Evidence from Taiwan. In Austronesian Studies Relating to Taiwan [Symposium Series of the Institute of History and Philology, Academia Sinica, Number 3], Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho & Chiu-yu Tseng (eds), 727–791. Taipei: Academia Sinica.
 Fuhui Hsieh Shibatani, Masayoshi. 2005. The attrition of the Austronesian focus system. Paper presented at Taiwan-Japan Joint Workshop on Austronesian Languages, June 23–24. Taipei: National Taiwan University. Shibatani, Masayoshi. 2009. Elements of complex structures, where recursion isn’t: The case of relativization. In Syntactic Complexity: Diachrony, Acquisition, Neuro-cognition, Evolution [Typological Studies in Language 85], T. Givon & Masayoshi Shibatani (eds), 163–198. Amsterdam: John Benjamins. Shinzato, Rumiko. This volume. Nominalization in Okinawan: From a diachronic and comparative perspective. In Yap, Grunow-Hårsta & Wrona (eds). Starosta, Stanley, Pawley, Andrew & Reid, Lawrence. 1982. The evolution of focus in Austronesian languages. In Papers from the Third International Conference on Austronesian Linguistics, Vol. 2: Tracking the Travelers, Amran Halim, Lois Carrington & Stephen A. Wurm (eds), 145–170. Canberra: Pacific Linguistics. Sung, Li-may, Su, Lily I-wen, Hsieh, Fuhui & Lin, Zhemin. 2008. Developing an online corpus of Formosan Languages. Taiwan Journal of Linguistics 6(2): 79–118. Teng, Stacy Fang-ching. 2008. A Grammar of Puyuma, an Austronesian Language. Canberra: Pacific Linguistics. Wierzbicka, Anna. 2000. Lexical prototypes as a universal basis for cross-linguistic identification of ‘parts of speech’. In Approaches to the Typology of Word Classe, Petra M. Vogel & Bernard Comrie (eds), 285–317. Berlin: Mouton de Gruyter. Yamada, Yukihiro. 1966. A Preliminary Itbayaten Vocabulary. Quezon City: Institute of Asian Studies, University of the Philippines. Yap, Foong Ha & Wang, Jiao. This volume. From light noun to nominalizer and more: On the grammaticalization of↜ Zhe and Suo in Old and Middle Chinese. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha, Grunow-Hårsta, Karen & Wrona, Janick. This volume. Nominalization strategies in Asian languages. In Yap, Grunow-Hårsta & Wrona (eds). Yeh, Marie M. 2003. A Syntactic and Semantic Study of Saisiyat Verbs. Ph.D. dissertation, National Taiwan Normal University.
Clausal nominalization in Budai Rukai Li-May Sung
National Taiwan University This paper aims to explore the morphosyntactic characteristics of the two nominalizers -anә and -Ø occurring pervasively in both clausal and lexical nominalizations in Budai Rukai, one of the Formosan languages spoken in the southern part of Taiwan. It shows that relativization plays a role in the emergence of nominalization constructions under investigation. Budai nominalization constructions also exhibit both nominal and verbal properties. Their distinctive nominality is signaled by the presence of case markers and genitive subjects. Yet their internal structure resembles in nearly all respects that of a sentence. Retention of adverbial phrases, oblique-marked object argument, and various verbal categories including tense, aspect, voice and valency morphemes is clear evidence of the verbal nature of such a nominalization process. The result of this work matches various cross-linguistic observations attested in Koptjevskaja-Tamm (2003), Malchukov (2006), Comrie and Thompson (2007), and Noonan (2007).
1.â•… Introduction Ever since Lees (1960), Chomsky (1970), and Comrie (1976), a major issue that has occupied the linguistic literature for more than four decades is the status of deverbal nominals and gerunds in (1a–c), and, more particularly, their relation to the corresÂ� ponding verbal predicate in (1d) and the degree of retention or loss of nominal/verbal properties during the nominalization processes. (1) a. John’s construction impressed us. (deverbal nominal: result nominal) b. John’s construction of sailing boats for ten years impressed us. (deverbal nominal: event nominal) c. John’s constructing sailing boats impressed us. (gerund) d. John constructs sailing boats. (verbal predicate)
Studies on deverbal nominals, as seen in the examples above, often distinguish between two types: result (1a) and event nominals (1b). As has been discussed extensively in the literature, couched in different frameworks, a certain amount of consensus has
 Li-May Sung
emerged that example (1a) has all the characteristics of a noun, whereas example (1b) is more verb-like, referring to an event or a process associated with an argument structure. Gerunds in (1c) are also commonly compared to those of another type, as seen in (1c′):
(1) c′. John’s constructing of sailing boats impressed us.
Though both types possess transparent morphological -ing affixation and appear to be of nominal category, the syntactic behavior in (1c) is parallel to that of sentential syntax requiring clause-like case marking on the object arguments, while (1c′) is characteristic of nominal syntax, in the sense that both arguments take possessive (genitive) forms. Though much of the work in the literature is concerned with the English construction, what has become apparent is that these puzzling properties have counterparts in a variety of languages, and thus inquiry into the problem of such ‘mixed-categories’ has expanded considerably (cf. Koptjevskaja-Tamm (1993), Malchukov (2004), and many studies in this volume). Budai Rukai, a Formosan language spoken in southern Taiwan, exhibits interesting data (2) that are analogous to those of English (1):1 (2) a.
ma-a2aw ka ta-suTa-suTav-anә2 stat.fin-big nom nfut-red-cure-nmz ‘The hospital is big.’
.╅ Rukai, according to Li (1977), includes six major dialects: Tanan, Budai, Labuan, Maga, Tona, and Mantauran. Geographically, Budai is spoken in Wutai Township in the south of Pingtung County in southern Taiwan; it is linguistically closer to Tanan and Labuan, and these three form a subgroup, while Tona, Maga and Mantauran are claimed in Li (1977) to form another subgroup. However, the position of Mantauran within the Rukai family remains controversial. More recently, two alternative hypotheses have been postulated that view Mantauran as being the first offshoot or being closer to the Budai-Tanan-Labuan subgroup (see Zeitoun 2007: 5 and fn 6). .╅ The glosses of the data taken from other authors will stay as they are in the original sources. Abbreviations used in this paper are as follows: 1 2 3 act caus comp fin fut gen
first person second person third person active voice marker causative complementizer finite marker future genitive
imp loc neg nfin nfut nmz nom obl pass
imperative Locative negative non-finite marker non-future nominalizer nominative oblique passive voice marker
pfv pl recp red rel sg stat
perfective plural reciprocal reduplication relative singular stative marker
Clausal nominalization in Budai Rukai 
b. masamali-aku ku ta-suTav-anә ki saTabu ki TaiTai surprise-1sg.nom obl nfut-cure-nmz gen saTabu obl TaiTai ‘I am surprised at Salabu’s curing Lailai.’ c.
wa-suTaw ki TaiTai ka saTabu fin-cure obl TaiTai nom saTabu ‘Salabu cured Lailai’
The same morpheme -anә in Budai, as shown in (2a–b), functions either as a derivational (lexical) nominalizer in the example of ta-suTa-suTav-anә ‘hospital’, or as a clausal nominalizing element in the example of ta-suTav-anә ki saTabu ki TaiTai ‘Salabu’s curing Lailai’. The distinction between (2a) and (2b) correlates with an interpretational distinction between nouns simply denoting an entity and nouns expressing an event. The examples in Budai appear to correspond to the result nominal (1a) and the gerund (1c) in English and there are no corresponding event nominals as those of English in (1b). In addition to -anә, another nominalization (3) exists in Budai, which will be referred to as -Ø (zero) nominalization: no overt nominalization affix is employed, and the verb is manifested in an infinitival form. Unlike -anә, -Ø nominalization, to the best of my knowledge, is used only in clausal nominalization.
(3) Nominalized Infinitive with -Ø
masamali-aku ku Ø-suTaw-Ø ki saTabu ki TaiTai surprise-1sg.nom obl nfin-cure-nmz gen saTabu obl TaiTai ‘I am surprised at Salabu’s curing Lailai.’
These nominalizations are considerably productive in Budai and may appear in a wide range of syntactic structures. The primary concern in this work is to focus on examining the morphosyntactic characteristics of clausal nominalizations involving the two elements -anә and -Ø. This paper is organized as follows. With a brief description of Budai grammar as background knowledge provided in Section 2, Section 3 begins with an exploration of the -anә and -Ø nominalized constructions and shows that relativization plays a role in the emergence of nominalization constructions under investigation. Their distinctive nominality is signaled by the embeddedness under a case marker/demonstrative and the presence of genitive subject. Section 4 investigates the internal structures of both -anә and -Ø nominalizations and argues that, despite clear evidence of nominal features, the internal structures resemble in nearly all respects that of a sentence. Of particular relevance for the current discussion is that the retention of various verbal categories is also suggestive of the verbal nature of such a nominalization process. Issues regarding word order and tense/aspect specifications pertaining to nominalization will be elaborated upon as well. Section 5 subsequently
 Li-May Sung
turns to the lexicalization of these nominalizing elements and concludes with a discussion of the correlation between clausal nominalization and lexical nominalization, highlighting upon the role of relativization as the key mechanism in deriving the resultant structures. Section 6 concludes this paper.
2.â•… A grammatical sketch of Budai Before delving into a detailed investigation of the nominalization, a brief introduction of the grammatical properties of Budai is in order, with particular reference to those that are relevant to the issues addressed in this paper. Like many other Austronesian languages, Budai is a copula-less and basically predicate-initial language; the word order may be either X(O)S or XS(O), much less rigid than in English. This is illustrated in (4–8).
(4) Nouns as predicates
a.
ŋu2a2әkay [SUBJECT ka/ku saTabu] Rukai nom saTabu ‘Salabu is a Rukai.’
b. kai ka ŋu2a2әkay [SUBJECT ka/ku saTabu] neg ka Rukai nom saTabu ‘Salabu is not a Rukai.’
(5) Weak quantifiers as predicates
a. takaTa [SUBJECT ka/ku taTagi ki Taŋpaw] many nom male friends gen Taŋpaw ‘Langpaw has many male friends.’ (Lit: ‘Langpaw’s male friends are many.’) b. kai takaTa [SUBJECT ka/ku taTagi ki Taŋpaw] neg many nom male friends gen Taŋpaw ‘Langpaw does not have many male friends.’
(6) Adjectives as predicates
a.
duduli [SUBJECT ka laymay] red nom clothes ‘The clothes are red.’
b. kai duduli [SUBJECT ka laymay] neg red nom clothes ‘The clothes are not red.’
Clausal nominalization in Budai Rukai 
(7) Stative verbs as predicates
a.
ma-dalamә ki/*ka/*ku TaiTai [SUBJECT ka/ku saTabu] stat.fin-like obl TaiTai nom saTabu ‘Salabu likes Lailai.’
b. kai ma-dalamә ki/*ka/*ku TaiTai [SUBJECT ka/ku saTabu] neg stat.fin-like obl TaiTai nom saTabu ‘Salabu does not like Lailai.’
(8) Action verbs as predicates
a.
Ø-wa-Tumay ki/*ka/*ku TaiTai [SUBJECT ka/ku saTabu] act-fin-beat obl TaiTai nom saTabu ‘Salabu beat Lailai.’
b. kai Ø-wa-Tumay ki/*ka/*ku TaiTai [SUBJECT ka/ku saTabu] neg act-fin-beat obl TaiTai nom saTabu ‘Salabu does not beat Lailai.’
As examples (4a)–(8a) show, noun phrases, weak quantifiers, adjectives, statives, and verbs all may function as predicates. When these examples are negated, the negator kai is employed for all types of predicates, as illustrated in (4b)–(8b), with one co-occurrence restriction for nominal predicates (4b) where the case marker ka is obligatory; ka is the only case marker, not ku or ki, which co-exists with the negator kai. Nominal case markers in Budai precede the noun phrase, and in most cases, they are obligatory.3 They may appear on full noun phrases and names. As Table 1 shows, Budai makes a three-way case distinction between Nominative, marking a subject argument, Genitive, marking a non-subject agent/actor, and Oblique, marking, among other things, a non-subject argument.4
.â•… Kuo (1979) uses the terminology ‘article’ or ‘determiner’ instead of ‘case-marker’ for ku, ka and ki in Budai. In this paper I am not in a position to argue which terminology would best describe these elements; I have simply followed the recent Formosan literature by using the terminology ‘case-marker’ for the sake of convenience. .â•… In a ki- passive construction, the non-subject agent/actor does not bear the genitive case-marker but rather the oblique. This is illustrated in example (10). And in a nominalized complement clause as illustrated below, the passive patient subject would receive genitive marking (-li): (i)
masamali ka saTabu ku ta-ki-2aә2aŋәT-anә-li surprise nom saTabu obl nfut-pass-cheat-nmz-1sg.gen ‘Salabu is surprised at my being cheated.’
 Li-May Sung
Table 1.╇ Nominal case system in Budai (cf. Li 1997; Chen 1999; Chen 2008; Zeitoun 2000) Nominative ku ka ki
ku (−visible, +distance ± animate) ka (+visible, −distance ± animate)
Oblique ku (−visible, +distance, −human, ± generic) ka (+visible, −distance, −human) ki (+specific, +human) (+generic, −human)
Genitive
ki (± animate)
With overt nominal case marking, the word order of arguments is relatively free, being either VOS or VSO. Where two arguments in a sentence are both marked by the same case markers, word order is then crucial in distinguishing objects from subjects. The subject would preferably be clause-final, with the object preceding it; that is, the resulting order is VOS.5 In addition, case markers or demonstratives such as kai ‘this [+visible, −distance]’, kuini/kui ‘that [+visible, −distance]’, kavai ‘that [+visible, +distance]’ or ku(i)2a ‘that [−visible, −distance]’ have all been grammaticalized into a new kind of markers that indicate nominal complementation, subordination, or modification associated with relative clauses (cf. Kuo 1979; Chen & Sung 2005: Footnote 8). This is a typical historical development, commonly observed in a great number of languages, including the Formosan family to varying degrees (cf. Noonan 2007; Tang 2008). These noun phrase markers, such as a case marker or demonstrative, are also called as “substantivizers” in Yap, Grunow-Hårsta and Wrona (this volume), in which they provide a comprehensive cross-linguistic comparison of diverse nominalization strategies, and also look at the relationships between nominalization and a variety of substantivization processes in the language samples presented in this volume. Pronouns in Budai can be roughly divided into two sets, namely free pronouns and clitic pronouns, as listed in Table 2. For free pronouns, there are topic and oblique pronouns; nominative personal pronouns do not exist in Budai. ku used in topic pronouns is a topic marker, which appears obligatorily with the pronominal stems. The clitic pronouns are usually attached to the first verbal element in the clause, but if there is a preverbal element such as a negator, they are forced to encliticise to the negator. Clitic pronouns can be either nominative or genitive, although the difference is only evident in the first person singular: -aku (nominative) and -li (genitive). In the other first and second persons, four pronouns (-ta, -nay, -su, -numi) are ambiguous, in that they can either be nominative or genitive. In the third person, only the genitive exists as an overt clitic pronoun.
.╅ Or the subject could be topicalized, yielding a SVO order.
Clausal nominalization in Budai Rukai 
Table 2.╇ Pronominal system of Budai (adapted from Chen 1999: 10) Person
Plurality
1
Singular Plural
2
Singular Plural Singular Plural
3
Visible/ Inclusive
+ Inclusive + Exculsive ± Visible ± Visible
Free
Bound
Topic
Nom.
Oblique
Nominative
Genitive
ku aku ku ta ku nay ku su ku numi ku ini ku ini
Ø Ø Ø Ø Ø Ø Ø
nakuanә mitaanә nayanә musuanә numianә inianә linianә
-(C)aku, -naw -ta -nay -su -numi -Ø -Ø
-li -(i)ta -nay -su -numi -ini -lini
Rukai, as discussed in Li (1973), Li (1977), Kuo (1979), Starosta (1995), Chen (1999), Zeitoun (2007), and Chen (2008), manifests a clear morphological voice distinction between active and passive sentences, unlike most Formosan languages (e.g. Kavalan (Hsieh, this volume) or Saisiyat (Yeh, this volume)), which exhibit instead a distinct ‘focus/voice’ marking system. This is illustrated in (9–10).
(9) Active construction in Budai
Ø-wa-Tumay ki/*ka TaiTai [SUBJECT ka/*ki saTabu] act-fin-hit obl TaiTai nom saTabu ‘Salabu hit Lailai.’
[Agent subject]
(10) Passive construction in Budai6 ki-a-Tumay ki/*ka saTabu [SUBJECT ka/*ki TaiTai] pass-fin-hit obl saTabu nom TaiTai ‘Lailai is hit by Salabu.’
[Patient subject]
.â•… Note that crosslinguistically and diachronically, GET verbs sometimes develop into passive markers (e.g. English (i)). Budai Rukai could also be a case of a verbalizer ki- ‘get’ developing into a passive marker as shown in (ii) and (iii). See Zeitoun and Teng (2009) for the arguments of the existence of *ki-N ‘get/obtain’ in PAn, in which *ki- is believed to have been reinterpreted as a passive marker in Rukai, Paiwan and Puyuma.
(i)
Mary got hit [by John].
(ii)
[ta-ki-tai-tai-anә] kai asasanә nfut-get-red-taro-nmz this now ‘Right now is the time of harvesting taros.’
(iii)
ki-a-Tumay ki saTabu ku TaiTai pass-fin-beat obl saTabu nom TaiTai ‘Lailai is hit by Salabu.’ (Lit: ‘Lailai got hit by Salabu.’)
 Li-May Sung
The affixes Ø- and ki- attached to a verb like Tumay ‘hit’ indicate active and passive voices respectively.7 In an active construction such as (9), an agent or an actor is .â•… I assume, in this paper, that a phonologically null element Ø, but not the morpheme w- in wa- (a view taken by Zeitoun et al. (1996), Chen (1999) and Zeitoun (2000)), is in fact the morphological marker for active voice. The prefix wa- as a portmanteau, not just -a-, should be treated as a verb class marker for non-future finite tense. This claim is supported by the following reasons. When examining the predicates more carefully, several verb classes can be distinguished in terms of their finiteness, as illustrated below.
(i) Verb class
Active finite form
Active non-finite form (take future as an illustration here)
I
wa-Tumay ‘hit’
Ti-Ø-Tumay ‘will hit’
II
ma-rimuru ‘forget’
Ti-Ø-ka-rimuru ‘will forget’
III
tu-a-daanә ‘do/build a house/houses’ ki-a-ragay ‘glad, happy’
Ti-Ø-tu-daanә ‘will build a house’ Ti-Ø-kiragay ‘will be glad, happy’
IV
Ø-duduli ‘red’ Ø-paθagili ‘start’
Ti-(a)-Ø-duduli ‘will become red’ Ti-Ø-paθagili ‘will start’
Take the type I verb class as an example. The passive voice ki- can show up in imperative constructions like (ii-b) while in an active imperative sentence like (ii-a), wa- in wa-2aTað ‘dance’ is omitted as a whole. If the morphological distinction between active and passive is w- and ki- respectively, there would be no way to explain why the whole morpheme wa- is omitted in imperative constructions. (ii) Imperative construction a. Active Ø-2aTað-a! act-dance-imp ‘Dance!’ b. Passive ki-papanav-a ki tina-su! pass-take.a.bath-imp obl mother-your ‘(Go) to be bathed by your mother!’
(Zeitoun 2000: 49)
Second, in a future active sentence, the future prefix Ti- can co-occur with the passive voice ki- (without -a-), as illustrated in (iii). As (iii-a) shows, for a verb like wa-kәla ‘come’, the whole morpheme wa-, not just w-, is omitted in a future sentence when Ti- is prefixed to the stem of the verb. (iii)
Sentences with future tense a. Active
Ti-Ø-kәla luiya ku saTabu fut-act-come tomorrow nom saTabu ‘Salabu will come tomorrow.’
Clausal nominalization in Budai Rukai 
marked as the subject, while in a passive construction such as (10), the patient or the theme is marked as the subject. With regard to the tense system, Budai exhibits a twoway contrast in which future a-, (T)i-, or Tika-(a)- is distinguished from non-future, rather than a three-way one, in which past contrasts with present and future, as in English. In addition, the past and present interpretation in Budai usually depends on the occurrence of temporal adjuncts, aspectual affixes (e.g. perfective marker -ŋa), and verbal reduplication indicating progressive.
3.â•… Nominalization, relativization and its nominality Before we examine the extent of retention of verbal categories inside nominalization in Budai, let us first consider various forms of clausal nominalization and their general characteristics.8 Nominalized clauses with -anә or -Ø are considerably pervasive in Budai; furthermore, they correlate to relativization in interesting ways. Examples (11) and (12) are two instances of ordinary relative clauses, in which the finite clause stands in appositive relation to the modified head noun, lasu ‘man’. The case marker ku or
b.
Passive
Ti-ki-Tumað-aku ki ina fut-pass-hit-1.sg.nom obl mother ‘I will be hit by (my) mother.’
Third, in a serial verb construction (iv–b), ki- can be used in the second verb Tumay ‘hit’ to indicate a (non-finite) passive voice, whereas the whole morpheme wa- is omitted on Tumay ‘hit’ in (iv–a) in which it still indicates an active voice.
(iv)
Serial verb construction a.
Active
ma-Tas-aku Ø-Tumay ki lavavalakә stat.fin-hate-1.sg.nom act-hit obl children ‘I hate to hit children.’
Passive
b.
ma-Tas-aku ki-Tumay ki ina stat.fin-hate-1.sg.nom pass-hit obl mother ‘I hate to be hit by (my) mother.’
To conclude, we argue that the active voice marker in Budai is a null element, and wa- should be analyzed as a non-future finite verb class marker. .╅ Various forms of nominalized clauses discussed in this section are also attested in the Mantauran dialect of Rukai. See Zeitoun (2002) for details.
 Li-May Sung
ka preceding the lexical head is reanalyzed as the relativizer.9 When the relativized argument is the subject, either an agent/actor (11) or a patient (12), the relative clause is essentially identical to the ordinary predicative clause counterpart (9) (in this case, Ø-wa-Tumay ki TaiTai) or (10) (in this case, ki-a-Tumay ki saTabu), and no nominalization is involved. (11) Active agent subject relativized ŋu-a-bәrә [SUBJECT ku Ø-wa-Tumay ki TaiTai ku lasu] go-fin-escape nom act-fin-hit obl TaiTai rel man ‘The man that hit Lailai ran away.’ (12) Passive patient subject relativized ŋu-a-bәrә [SUBJECT ku ki-a-Tumay ki saTabu ku lasu] go-fin-escape nom pass-fin-hit obl saTabu rel man ‘The man that was hit by Salabu ran away.’
When it is the non-subject argument, i.e. the internal object, of (9) that gets relativized, this is where -Ø (zero) nominalization comes into play, as illustrated in (13): (13) Active patient object relativized a.
ŋu-a-bәrә [SUBJECT ku Ø-Tumay-Ø ki saTabu ku lasu] go-fin-escape nom act-hit-nmz gen saTabu rel man ‘The man that Salabu hit ran away.’
b. ŋu-a-bәrә [SUBJECT ku2a Ø-Tumay-Ø-li ku2a lasu] go-fin-escape distal.that act-hit-nmz-1sg.gen rel man ‘The man that I hit ran away.’
The distal demonstrative ku2a in (13b) accompanying the head noun lasu has undergone a similar re-analysis and gets reinterpreted as a relativizer.10 Two notable
.â•… Budai allows both pre-nominal and post-nominal relative clauses. In the post-nominal clauses, it is the case marker accompanying the finite clause now re-analyzed as the relativizer. As noted in Yap, Grunow-Hårsta and Wrona (this volume), there is a strong tendency among Austronesian languages to recruit noun phrase markers (e.g. case markers and demonstratives) as relativizers, while Korean, Japanese, Sino-Tibetan languages are more inclined to recruit light nouns as nominalizers and relativizers (see also Yap and Wang (this volume)). In other words, different languages may make use of rather diverse nominalization strategies, with each language possibly having undergone different grammaticalization pathways. .â•… Note further that in identifying the intended referent, proximity concord is strongly enforced between the subject-marking case marker/demonstrative and the relativizer preceding the highlighted head noun. The proximity concord can be roughly represented as follow:
Clausal nominalization in Budai Rukai 
features distinguishing (13) from (9) include the use of “bare” root form Tumay and the presence of genitive case assigned to the agent/actor ki saTabu/-li. The same observation is attested in the -anә type of nominalization as well, which will be investigated below. When the agent/actor is presented in a genitive form, the genitive has a subjectlike role, different from the modifier-like interpretation as in laymay ki saTabu ‘Salabu’s clothes’. It is of particular interest to note that the non-subject agent/actor ki saTabu in the passive construction (10) cannot be relativized, suggesting that saTabu here at some syntactic point has been ‘suppressed’ first to an adjunct position, where further relativization was not applicable. Internal adjuncts, including location and instrument, are also relativizable, with which the -anә type of nominalization plays a role. The nominalizer -anә, together with a location head ta- or an instrument head sa-, form a double nominalization (in the sense that they are circumfixal) as illustrated in (14) and (15). (14) [ta-ka-kәcәT-anә] kai tayhuku nfut-stat.nfin-cold-nmz this Taipei ‘Taipei is a place that is cold.’ (15) ka muni, ma-Tas-aku arakai nom muni, stat.fin-hate-1sg.nom use [ka sa-pa-ka-mwa-mwal-anә ki laymay] obl sa-caus-stat.nfin-red-dry-nmz obl clothes ‘Muni doesn’t like to use the thing (machine) that makes clothes dry.’
The above -anә clauses clearly involve a headless-strategy towards relativizing or pseudo-clefting location/instrument expressions.11 The prefix ta-, in addition to
(i) (ii) (iii) (iv)
[ku (case marker) [relative clause….ku (REL) N]] [ka (case marker) [relative clause….ka (REL) N]] [kai (demonstrative) [relative clause….kai (REL) N]] [kuini (demonstrative) [relative clause…. kuini (REL) N]]
Also interestingly, the use of the distal demonstratives ku2a in (13b) indicates additionally the relative time of the event occurred. This semantic extension from an event-referring to a temporal-referring function is also attested in Abui (a Papuan language) demonstratives. See Yap, Grunow-Hårsta and Wrona (this volume) and Kratochvìl (this volume) for the discussion of the relationship between nominalizers and demonstratives. .â•… A headless relative clause can be analyzed as one with an empty head e which is coindexed with the trace t inside the relative clause. This could be partially represented as:
(i)
[DP/NP ku/ka [S …. ti ….] REL [N ei]]
 Li-May Sung
indicating a location where the action/state associated takes place, can also convey certain adverbial readings, such as a manner result (16–18) and a temporal (19). (16)
ŋi-a-caәmә [ka ta-Tumað-anә ki saTabu ki muni] ŋi-fin-sick nom nfut-hit-nmz gen saTabu obl muni ‘Muni got hurt (sick) badly as a result of Salabu’s beating.’ (Lit: ‘The result of Salabu’s beating Muni is serious.’)
(17) Q: ma-tu-manә kai Tibaŋә? stat.fin-do-what this window ‘What happened to this window?’ A: [ta-ki-kupa-anә]! nfut-pass-steal-nmz ‘(It) got stolen! (Unbelievable!)’ (18) ma-Tas-aku kanә [ka lacәŋә [ka ta-aga-anә stat.fin-hate-1sg.nom nfin.eat obl vegetable rel nfut-cook-nmz ki TaiTai]] gen TaiTai ‘I hate to eat the vegetables that Lailai has cooked (because they did not look good in the way that Lailai cooked them).’ (19) a.
[ta-ki-tai-tai-anә] kai asasanә nfut-get-red-taro-nmz this now ‘Right now is the time of harvesting taros.’
b. [lu ka ta-ki-tai-tai-anә], ŋu-a-daTanә madu maganaәTә if/when obl nfut-get-red-taro-nmz go-fin-field have.to all ‘At the time of harvesting taros, all (of us) have to go to the field.’
With an anti-causative interpretation, where the extraneous agent/actor is pragmatically less significant and the semantic patient is much more prominent, the prefix niwould be employed, as seen in (20a). The nominalized clause ni-…-anә, occasionally accompanied by the passive marker ki-, is often found depicting the so-called “spontaneous events” along the line of Kemmer (1993), as evidenced by the examples given in (20b–e). These spontaneous events are known to designate in particular the change of state of an entity, including physiological, physiochemical or natural changes, in which the role of agent/actor is linguistically insignificant and thus unexpressed. (20) a.
saŋu-a-әsai [ka ni-sinav-anә ka laymay] smell-fin-fragrant nom ni-wash-nmz rel clothes ‘The clothes that got washed smell good.’
Clausal nominalization in Budai Rukai 
b. i-a-kai inu [ku ni-kupa-anә ku aðaðamә] (Kuo 1979: 32)12 loc-fin-this where nom ni-steal-nmz rel birds ‘Where is the stolen bird?’ c.
[ni-(ki-)kupa-anә] kai aðaðamә ni-(pass-)steal-nmz this bird ‘This bird is a stolen one.’
d. [ni-(ki-)muәlәb-anә] ka TawTaw2u ni-(pass-)open-nmz nom door ‘The door is open.’ e.
[ni-(ki-)kanә-anә] kui kaaŋә ni-(pass-)eat-nmz that fish ‘That fish was what was eaten.’
(Kuo 1979: 19)
(Kuo 1979: 18)
(Kuo 1979: 19)
The -anә nominalization, without co-occurring with any prefixes, may be taken to denote a property ascribed to the intended referent. Examples in (21) illustrate this usage of nominalization, featuring a copular-type pseudocleft construction (in this case, a headless relative strategy), albeit without overt copula-support.13 (21) a.
sa-paysu-anә ka saTabu14 possess-money-nmz nom saTabu ‘Salabu is one who possesses (lots of) money.’
b. sa-daan-anә ka saTabu possess-house-nmz nom saTabu ‘Salabu is one who possesses (lots of) houses.’ c.
sa-awTu-anә ka saTabu possess-head-nmz nom saTabu ‘Salabu is one who is quite smart.’
.â•… Data taken from Kuo (1979) have been re-glossed in this paper, but the English translations stay as they are in the original source. Kuo analyzed ni- as expressing ‘in the state of ’ and ni-, especially when not co-occurring with the passive marker ki-, is characterized to describe a situation when “the state of the subject is imaginable as one achievable without intervention of an agent” (Kuo 1979: 18). .â•… We will not discuss details of various possible interpretations (e.g. predicational, specificational, definitional or identificational etc.) associated with such copular-type pseudoclefts, as addressed in the literature (cf. Dikken 2006). We will leave this for future investigation. .â•… The morpheme sa- here is a verbalizer, functioning differently from the instrumental marker sa-. However, it is quite probable that this verbalizer sa- with its interpretation of ‘possess’ or ‘with’ is an extension from the instrument origin sa-. This is a topic for further research.
 Li-May Sung
d. sa-la-manә-man-anә ka saTabu possess-pl-red-things-nmz nom saTabu ‘Salabu is one who possesses lots of (precious) things.’
The -anә nominalized structure also appears in conditional/temporal clauses headed by the lu ‘if/when’ subordinator, where the property of quality of (in)animate entities is perceived through sensory perceptions by human being in a generic or impersonal sense. This is exemplified in (22). (22) a.
kai si-silu, [lu 2әl-anә] ŋia ma-kuciŋaTә this red-glass.bead if/when see-nmz seem stat.fin-expensive ‘This necklace seems expensive when looking at it.’
b. [lu uŋul-anә], ŋia bava kai acilay if/when drink-nmz, seem wine this water ‘The water seems like beer when drinking it.’ c.
[lu sabali-anә], saŋu-a-mia kai swantu if/when smell-nmz smell-fin-stink this garlic ‘This garlic smells stinky when smelling it.’
d. [lu 2әәl-anә kai Tulay], arua-kaa2aa2aw if/when see-nmz this kid getting.more-big ‘This kid looks getting bigger.’
(Zeitoun 2000: 54)15
One commonality shared by the foregoing examples of ni-…-anә and -anә is that they both feature “low saliency” of event agent/actor, as suggested in Kemmer (1993), usually leaving the semantic patient or undergoer as the sole argument in the whole event structure. This is reminiscent of the valency-decreasing function of Proto-Oceanic *ma- (Evans & Ross 2001) and many of *ma- reflexes attested in some Formosan languages such as Kavalan, Amis and Paiwan (Huang & Sung 2008). The nominalized clauses, either -Ø type or -anә type, also appear in an NP locus position, where they function as noun complements selected by predicates such as masamali ‘be surprised’, malisi ‘angry; not good’, makuro ‘feel pity on’, and kiasәasә ‘grateful’, etc. (cf. Kuo 1979).16 As revealed in (23), nominal case markers like ku/ka or demonstratives kai/kuini have to be present and the use of a factive complementizer like alaka would be illicit. When a factive complementizer is employed, the verb appears in a finite form as in (24), rather a nominalized one.
.â•… English glossing and translation are mine. .â•… The complement clause with the nominalizer -anә is termed ‘Deverbal NP’ and the one with the nominalizer -Ø ‘Action Nominal’ in Kuo (1979).
Clausal nominalization in Budai Rukai 
(23) a.
masamali-aku [ku/ka/kai/*alaka Ø-Tumay-Ø-inii surprise-1sg.nom obl/obl/this/*comp act-hit-nmz-3sg.gen
ki/*ka/*ku saTabui ki TaiTai/inianә] gen saTabu obl TaiTai/3sg.obl
‘I am surprised at Salabu’s (frequently) hitting Lailai/her.’
b. masamali-aku [ku/ka/kai/*alaka ta-Tumað-anә surprise-1sg.nom obl/obl/this/*comp nfut-hit-nmz
ki/*ka/*ku saTabu ki TaiTai/inianә] gen saTabu obl TaiTai/3sg.obl
‘I am surprised at (the bad consequence resulting from) Salabu’s hitting Lailai/her.’ c. masamali-aku [ka (ta-ka-)θarir-anә ki muni] surprise-1sg.nom obl (nfut-stat.nfin-)beautiful-nmz gen muni ‘I am surprised at Muni’s beauty.’ (Lit: ‘I am surprised at how beautiful Muni has been right now after she grows up.’) (24) masamali-aku alaka/*ku/*ka/*kai Ø-wa-Tumay ki TaiTai surprise-1sg.nom comp/*obl/*obl/this act-fin-hit obl TaiTai ka saTabu nom saTabu ‘I am surprised that Salabu hit Lailai.’
From a cross-linguistic point of view, all of the nominalized data under discussion so far in Budai resemble closely well-known gerundive constructions such as the English POSS-ing gerunds, the Korean nominalization (Yoon 1996; Yap & Matthews 2008), the Italian nominalized infinitives (Zucchi 1993), the Hebrew gerunds (Siloni 1997), the Tagalog/Austronesian exclamatives and temporal adjuncts (Kaufman, this volume) and many other examples which have been thoroughly examined in the literature. Overt morphosyntactic cues for nominal features of gerundives lie in the inclusion of nominal categories such as case, determiner, number, singular/plural, etc., and the relation that a nominalized predicate has with its arguments. In some cases, both the subject and the object may bear a genitival relation to the nominalized predicate as exemplified in English (25a), by means of the possessive ’s marker and the preposition of, while in other situations, for example Uzbek (25b), only the subject is genitivized (-niŋ) and the object stays in its usual object coding (-ni). (25) English vs. Uzbek a.
(Noonan 2007: 70)
Algernon’s shooting of the aardvark drew international attention.
b. X ftin bu fdam-niŋ ˇJ oˇJ a-ni oğirla-š-i-ni istadi woman this man-gen chicken-obj steal-nzr-3sg-obj want.3sg ‘The woman wanted the man to steal the chicken.’
 Li-May Sung
Nominalized data in Budai also display these two similar hallmarks of nominality, i.e. embeddedness under a case marker/demonstrative and the presence of genitive subject. The first is signaled by the canonical argument position that the nominalized clause occupies, in conjunction with a nominal case marker ka/ku or a deitic demonstrative kai/kuini. For instance, the nominalized clause can appear in the position of a subject (13) and (16), object (15) and (18), complement (23), and object of a preposition (26) as illustrated below; and the presence of case marker/demonstrative clinches the identity as a nominal. (26) a.
Ti-ki-paysu-aku pakai [ku a-laŋað-anә(-li) fut-get-money-1sg.nom for obl fut-buy-nmz(-1sg.gen) ku daanә]17,18 obl house ‘I will (want to) earn money in order to buy a house.’
.â•… Li (1973: 111) mentions that there are four prepositions found in the Tanan dialect of Rukai. In Budai, several forms are also found to be preposition-like, including tualai ‘from’, i- as in i-bәlәŋә ‘loc-top; in the sky’ and y-a-kai ‘loc-fin-this; exist, have, at’ (cf. Zeitoun et al. 1999), pi- indicating causative of location (cf. Blust 2003) as in pi-cakәna ‘caus-ground; on the ground’, and pakai ‘for’. Budai shares with Puyuma, Amis, and Thao in having an invariable maker i for location nouns (cf. Ross & Stacy 2005: 756). .â•… pakai ‘for’ in Budai is presumably made up of a causative pa- and a demonstrative kai. It has the following characteristics of behaving as a preposition. First, in most cases, pakai does not occur in a predicate position, as shown in (i). Second, examples like (ii) cannot be analyzed as a serial verb construction consisting of two ‘verbs’, since the whole constituent [pakai ….] can be optionally ‘preposed’ to the sentence-initial position; such a movement is impossible for a true serial verb construction as shown in (iii). Third, the element that follows pakai is nominal in nature, as in (iv). All these facts tend to suggest that [pakai …] is a prepositional phrase having the structure of P-NP. (i) *pakai [subject ku a-laŋað-anә ku daanә] for obl fut-buy-nmz obl house (ii) a.
Ti-ki-paysu-aku [pakai [ku a-laŋað-anә ku daanә]] fut-get-money-1.sg.nom for obl fut-buy-nmz obl house ‘I will earn money in order to buy a house.’
b.
[pakai [ku a-laŋað-anә ku daanә]], Ti-ki-paysu-aku
(iii) a.
ma-Tas-aku Tumay ki lavavalakә stat.fin-hate-1.sg.nom hit obl kids ‘I hate to hit kids.’
b. *[Tumay ki lavavalakә], maTas-aku
(iv)
Ti-ki-paysu-aku pakai inianә/ki TaiTai fut-get-money-1.sg.nom for 3sg.obl/obl TaiTai ‘I will earn money for her/Lailai.’
Clausal nominalization in Budai Rukai 
b. Ti-ki-paysu-aku pakai (ku a-bað-anә) [ku laŋay-Ø fut-get-money-1sg.nom for (obl fut-give-nmz) obl buy-nmz
ki TaiTai ku daanә] gen TaiTai rel house
‘I will (want to) earn money (for paying) for the house that Lailai bought.’
The second nominality is clearly distinguished by the appearance of the agent/actor subject in genitive case ki saTabu (or -ini) seen in (23), in contrast to its finite predicate counterpart in nominative case ka saTabu in (24). Additionally, note that the patient object argument ki TaiTai or inianә retains its usual oblique case in (23) and (24). All foregoing Budai examples fit perfectly into the typological observation by Malchukov (2006), in that subjects are more likely to be assimilated to nominal syntax than direct objects are. That is, subject arguments are more inclined to receive genitive encoding in the process of nominalization, while object arguments may retain their sentential marking, in accordance with Comrie’s (1976) generalization. Before we proceed to the issue of verbality, one remark is worth a brief discussion. Despite the fact that Rukai is notably characterized as belonging to a disparate subgroup, due to its peculiarity of dichotomic active/passive voice system, Budai nominalization data under discussion here reflect more or less the “four-voice” contrasts of PAn *-um-, *-en, *-an, and *Si- (Wolff 1973), for actor voice (AF), patient voice (PF), locative voice (LF) and instrument voice (IF) respectively.19 A functionally similar prefix *Sa- is also reconstructable (Ross 1995; Blust 1999); however, the division of function between *Si- and *Sa- is still unclear. Paiwan is a good example, as it characterizes this four-way verbal distinction of focus/voice (i.e. AF, OF/PF, LF and IF): (27) Paiwan (southern Taiwan) a.
(Ferrell 1982: 31; cited in Ross & Teng 2005: 741)
q<m>alup a tsautsau tua vavuy i (tua) gadu tua vuluq hunt tpc man obl pig loc (obl) mountain obl spear ‘The man hunts the pigs in the mountains with a spear.’
b. qalup-en nua tsautsau a vavuy i (tua) gadu tua vuluq hunt-of gen man tpc pig loc (obl) mountain obl spear ‘The man hunts the pigs in the mountains with a spear.’
.╅ Starosta, Pawley, Reid (1982) have argued that the focus/voice system of Philippinetype languages evolved from a PAn system of nominalization, in which certain derived nouns were reinterpreted as verbs in main clauses. This is known as the verbs-from-nominalizations hypothesis.
 Li-May Sung
c.
qalup-an nua tsautsau tua vavuy a gadu tua vuluq hunt-lf gen man obl pig tpc mountain obl spear ‘The man hunts the pigs in the mountains with a spear.’
d. si-qalup nua tsautsau tua vavuy i (tua) gadu a vuluq if-hunt gen man obl pig loc (obl) mountain tpc spear ‘The man hunts the pigs in the mountains with a spear.’
Budai has apparently lost synchronic traces of voice markings of *-um-, *-en which most Formosan languages still cling to, as evidenced by Paiwan (27). The finite stative marker ma- in ma-dalamә ‘like’ is one of few relics showing a link of a *ma- reflex (cf. Evans & Ross 2001) in non-nominalizing contexts.20 The prefix ni- in ni-…-anә nominalization appears to be a possible reflex of the aspectual marker *-in-, assuming a portmanteau analysis of *-in- simultaneously marking a PV voice and perfectivity (cf. Starosta, Pawley & Reid 1982; Ross 1995; Hsieh, this volume, and Yeh, this volume). Note that the perfective marker -ŋa is more pervasively used in both non-nominalized and nominalized structures in modern Budai. The morpheme sa- in sa-…-anә has retained its instrumental function of *Si-/*Sa- in the process of nominalization.21 The nominalizing morpheme -anә, a reflex of *-an, has lost its origin as a location voice and has evolved to encompass almost all participants (agentless, non-subject undergoers or adjuncts) involved in relativization, subordination and nominalized complementation. The prefix ta- in ta-…-anә mainly denotes a location and various adverbial interpretations (e.g. a result manner or temporal). The occurrence of ta- as a location indicator is not as widespread as the other locative marker i- across Formosan languages, synchronically speaking. ta- typically occurs in nominalizing contexts in Budai, while there is no such a restriction for i- as in i-bәlәŋә ‘loc-top; in the sky’, y-a-kai ‘loc-fin-this; exist, have, at’ (cf. Zeitoun et al. 1999), and pi-cakәna ‘caus.loc-ground; on the ground’ (cf. Blust 2003). The use of ta- as a locative marker is also attested in Kavalan (Hsieh, this volume), as illustrated in the naturalistic excerpts (28). Although no proposal has been made to connect these two uses of the prefix ta- in Budai and Kavalan, further investigation into other Formosan languages in the future might provide a coherent explanation.
.â•… Corresponding to ma-, the morpheme ka- also marking stativity appears in non-finite constructions. See Blust (1999) and Zeitoun and Huang (2000) for the discussion of the affixal ka-. .â•… The morpheme sa- as a voice marker in many Formosan languages is often characterized to encode either an instrumental or a benefactive semantic subject. sa- in Budai nominalized clauses is restricted to the former function; for the benefactive interpretation, the prepositionlike pakai ‘for’ is more prevailing.
Clausal nominalization in Budai Rukai 
(28) a. 1. yau baqian ’nay usiq._ exist elder.male that one
(NTU Corpus: pear_buya)22
2. …(2.4) matiw ta ni-paluma-an-na tu sinsuli._ af.go loc pfv-plant-nmz-3sg.gen obl plum ‘There was an old man. (He) went to the place where he grew plums.’ b. 16. … matiw,\ â•… AF.go
(NTU Corpus: map_imuy)
17. … pasazi ta-lepaw-an ni utay.\ â•… toward.here loc-house-loc gen pn ‘(He) came towards Utay’s house.’ c. 51. … yau=ti ya sunis a yau,_ â•… exist=pfv nom child lnk that
(NTU Corpus: pear_api)
52. …(1.1) m-espes,\ af-stroke 53. … ta nunuq-na ta==,_ â•… loc shank.bone-3sg.gen loc ‘The child stroked the bone of his shank.’
In the following section we shall draw attention to the verbal-like properties of the nominalized constructions.
4.â•… Nominalization and its verbality 4.1â•… Evidence of a verbal category Up to this point, we have investigated the morphosyntactic characteristics of the -Ø and -anә nominalizations and their distinctive nominality. We shall now consider the internal structure of these nominalized constructions. Despite the above distributional evidence for nominal features, the internal structure of both nominalizations resembles in nearly all respects that of a sentence. If we look inside these nominalizations, they manifest a number of syntactic properties that are typical of verbs rather than of nouns. One of the most distinguishing feature is the retention of clause-like case marking on their object argument as shown in (23), reproduced below as (29), where the
.╅ The data have been taken from the NTU Corpus of Formosan Languages at . See Sung, Su, Hsieh and Lin (2008) for an introduction of the corpus.
 Li-May Sung
nominalized verb Tumay-Ø or ta-Tumað-anә still assigns its usual oblique case to the direct object ki TaiTai/inianә as its finite counterpart does in (30): (29) a.
masamali-aku [ku/ka/kai/*alaka Ø-Tumay-Ø-ini surprise-1sg.nom obl/obl/this/*comp act-hit-nmz-3sg.gen
ki/*ka/*ku saTabu ki TaiTai/inianә] gen saTabu obl TaiTai/3sg.obl
‘I am surprised at Salabu’s (frequently) hitting Lailai/her.’
b. masamali-aku [ku/ka/kai/*alaka ta-Tumað-anә surprise-1sg.nom obl/obl/this/*comp nfut-hit-nmz
ki/*ka/*ku saTabu ki TaiTai/inianә] gen saTabu obl TaiTai/3sg.obl
‘I am surprised at (the bad consequence resulting from) Salabu’s hitting Lailai/her.’ (30) Ø-wa-Tumay ki TaiTai/inianә ka/ku saTabu act-fin-beat obl TaiTai/3sg.obl nom saTabu ‘Salabu beat Lailai/her.’
In addition, both nominalized clauses in (29) are not semantically interpreted as a referential location entity or a derived nominal expression; rather, they denote the events of hitting. This is often referred to as action nominalization or event nominalization in the literature. Moreover, these nominalized constructions also extend to signal an additional propositional content presupposed by the speaker. In other words, they are packaged particularly with certain subjective speaker attitudes toward the action event. For example, with zero-nominalization, (29a) signals a habitual implication of the hitting event as perceived by the speaker.23 Example (29b), with -anә nominalization on the other hand, presupposes that Lailai was hit by Salabu, and emphasizes the speaker’s strong reaction to the serious consequence that the hitting has resulted in. As noted earlier in Section 3, a similar sense of speaker perspective or evaluation is implied to a certain degree in the nominalization examples of temporal adjuncts lu …-anә expressing ‘smell (stinky), look (expensive), taste like (something)’ and many pseudo-cleft types of construals such as ta-…-anә ‘a place/time for V-ing’ or ni-…-anә ‘something that got V-en’. Interestingly, the following “stand-alone” or “non-embedded” nominalizations also convey strong speaker attitude:24
.â•… This habitual interpretation is sometimes accompanied by morphological reduplication. See also Footnote 32. .â•… Stand-alone (unembedded) nominalizations could be closely related to cleft constructions, possibly involving a syntactic process of copula ellipsis, as addressed in the literature (see Yap, Grunow-Hårsta and Wrona (this volume). Possible implications of such analysis, in particular for a copula-less language, Budai, is a topic for future research.
Clausal nominalization in Budai Rukai 
(31) Q: a-su Tumay ki saTabu? why-2sg hit obl saTabu ‘Why did you hit Salabu? A:
lisi-anә-li (inianә)! angry-nmz-1sg.gen (3sg.obl) ‘I am (very) angry (at him)!’ (Lit: ‘My anger (at him)!’)
(32) Q: ani kamani ka muni ka mu-a-cu-cubungu? how.come.3sg KAMANI nom muni rel go-fin-red-front ‘How come it is Muni who went on stage?’ A: θarir-anә-ini! beautiful-nmz-3sg.gen ‘She is beautiful!’ (Lit: ‘Her beauty!’)
Yap, Grunow-Hårsta and Wrona (this volume) refer to these constructions as ‘stance constructions’. This sort of “speaker perspective nature”, pertaining to the form of nominalizations, is widely attested across Altaic, Sinitic, Tibeto-Burman, and Austronesian language families, as observed in Watters (2008), Yap and Matthews (2008), Kaufman (this volume), Potsdam (this volume), and many papers presented in this volume. Clearly, this intimate association cannot be accidental and robust crosslinguistic parallels point to the same historical track of recruiting nominalization as a shared cognitive strategy to “package our thoughts and viewpoints”, as concluded in Yap and Matthews (2008). Turning back to the verbal nature of nominalization, another commonly deployed diagnostic lies in its ability to license adverbial phrases inside the nominalized construction. Consider the following example of POSS-ing gerunds in English: (33) English [His having called me up sooner] surprised me.
In (33), a manner adverbial such as sooner can appear inside a gerundive construction. Similar examples from other languages like (34)–(36) point to the same distribution as well: the presence of adverbials such as ppalli ‘quickly’ in Korean (34), improvvisamente ‘suddenly’ in Italian (35) and le’at ‘slowly’ in Hebrew (36) is felicitous in nominalized constructions. (34) Korean: -(u)m/-ki nominalization
(Yoon 1991)25
[Chelswu-ka chayk-ul ppalli ilk-ess-um]-i pwunmyengha-ta Chelswu-nom book-acc quickly read-pst-nml-nom evident-decl ‘It is evident that Chelswu read the book quickly.’
.╅ Data from other authors stay as they are in the original sources.
 Li-May Sung
(35) Italian: nominalized infinitive
(Zucchi 1993: 54)
il suo scrivere quella lettera improvvisamente his write (inf) that letter suddenly (36) Hebrew: gerund
(Siloni 1997: 162 (7a))
be-nos’e-nu le’at, hiclaxnu li-mco ‘et ha-derex in-driving-1pl slowly, succeeded (we) to-find acc the-way ‘Driving slowly, we succeeded to find the way.’
In this respect, a similar observation holds for Budai, as shown in (37): (37) a.
masamali-aku ku ta-Tumað-anә ki saTabu ki TaiTai surprise-1sg.nom obl nfut-hit-nmz gen saTabu obl TaiTai
saŋutuTulu kuiya three.times yesterday
‘I am surprised at Salabu’s hitting Lailai three times yesterday.’
b. masamali-aku ku Ø-Tumay-Ø ki saTabu ki TaiTai surprise-1sg.nom obl act-hit-nmz gen saTabu obl TaiTai
saŋutuTulu kuiya three.times yesterday
‘I am surprised at Salabu’s hitting Lailai three times yesterday.’
Adverbial phrases such as the temporal adverbs kuiya ‘yesterday’ and frequency adverbs like saŋutuTulu ‘three times’ are not only permissible within these nominalized clauses with -anә and -Ø, they are also allowed to occur freely between the nominalized verb and the subject agent/actor, or between the subject agent/actor and the object patient, as shown in (38).26 (38) a. masamali-aku ku ta-Tumað-anә saŋutuTulu (kuiya) ki saTabu ki TaiTai b. masamali-aku ku ta-Tumað-anә ki saTabu saŋutuTulu (kuiya) ki TaiTai
.â•… It should be noted that not all types of adverbs can be used as diagnostics. For instance, manner adverbs are not an appropriate test in nominalized clauses for Budai, and perhaps are not an appropriate one for most of the Formosan languages, since manner adverbs usually appear as the matrix verbal predicate in the so-called “serial verb construction”, which consists of a string of two verbs followed by the matrix subject, as shown in (i). (i)
maridalә TayTay ka saTabu kuiya fast run nom saTabu yesterday ‘Salabu ran very fast yesterday.’ (Lit: ‘Salabu was very fast yesterday while running.’)
As the literal translation provided in (i), maridalә ‘fast’ is the matrix predicate and it takes the remainder of the clause as its complement. Related issues are addressed in Chang (2006) for Kavalan and Holmer (2006) for Seediq.
Clausal nominalization in Budai Rukai 
Under the standard assumption that adverbs typically modify verbs and not nouns, the presence of adverbs inside these constructions supports the claim that the internal structure of the nominalization in Budai is, in fact, verbal.
4.2â•… Evidence of other verbal categories Next, we shall examine whether or not there is any evidence for any other syntactic categories than the verb itself that might be included in the process of nominalization. Languages vary considerably regarding the extent to which typical verbal categories (e.g. tense, aspect, voice, etc.) can be retained along the typological view of Koptjevskaja-Tamm (2003), Malchukov (2006), Comrie and Thompson (2007), and Noonan (2007). Squamish, as given in Noonan (2007), is an extreme case where nominalized clauses retain all of the verbal inflection, clitics, and particles found in main clauses: (39) Squamish a.
(Noonan 2007: 71)
Na č-n wa c’↜aq′-an-umi fact declar-1sg prog hit-trans-2sg.obj ‘I was hitting you.’
b. Č-n Sč-iws kwi n-s-na declar-1sg tired-body art 1sg.poss-nom-fact
wa c’↜aq′-an-umi prog hit-trans-2sg.obj
‘I’m tired of hitting you.’
In the following section, we will explore this research question relating to Budai nominalization. Budai is well-known to exhibit copious verbal affixations, including verbalizer, voice, aspect, tense and so on. Whenever two prefixes co-occur, the temporal prefix always precedes the voice marker, which subsequently precedes the verb stem. This is exemplified in (40). When (40) is negated, two affixal orders are allowed as seen in (41). This suggests that a negator behaves clearly like an auxiliary verb which can host a tense category and a pronominal clitic. (40) Ti-ki-Tumað-aku ki ina fut-pass-hit-1sg.nom obl mother ‘I will be hit by (my) mother.’ (41) a.
Ti-kai-naku ki-Tumay ki ina fut-neg-1sg.nom pass-hit obl mother
b. kai-naku Ti-ki-Tumay ki ina neg-1sg.nom fut-pass-hit obl mother ‘I will not be hit by (my) mother.’
 Li-May Sung
The relevant affixal order is hence roughly depicted as: (42) (neg) > (pronominal clitic) > future.tense > (neg) > (pronominal clitic) > voice > v > pronominal clitic
We shall now look at whether these verbal affixes are retained within the nominalization constructions. Example (43) below reveals that a voice prefix, such as ki-, as well as future tense prefixes such as a-, (T)i-, or Tika-(a)- are present inside the nominalized clauses marked by -anә, just as they are frequently found in main clauses in (40)–(41). This shows that the nominalizer -anә takes its scope not only over a verb stem, but also over verbal categories, such as passive voice and (future) tense. (43) a.
masamali ka saTabu ku ta-ki-2aә2aŋәT-anә-li surprise nom saTabu obl nfut-pass-cheat-nmz-1sg.gen ‘Salabu is surprised at my being cheated.’
b. Ti-kisamula ki-paysu ka saTabu pakai ku fut-work.hard get-money nom saTabu for obl
a-ki-ka-dalam-anә nakuanә fut-pass-stat.nfin-like-nmz 1sg.obl
‘Salabu will (want to) work hard to earn money in order to be liked by me.’
In addition, certain valency-changing morphemes, including reciprocal, reflexive and causative, are also retained within the -anә nominalization, as illustrated in (44). (44) a.
masamali-aku ka ta-ma-la-Tumað-anә surprise-1sg.nom obl nfut-recp-red-hit-nmz
ki saTabu si muni gen saTabu and muni
‘I am surprised at Salabu and Muni’s hitting each other.’
b. masamali-aku ka ta-ŋi-pa-pacað-anә-ini surprise-1sg.nom obl nfut-self-caus-die-nmz-3sg.gen ‘I am surprised at his killing himself.↜’ c.
masamali-aku ki saTabu ka ta-pa-Tumað-anә surprise-1sg.nom gen saTabu obl nfut-caus-hit-nmz
ki muni ki TaiTai obl muni obl TaiTai
‘I am surprised at Salabu’s making Muni hit Lailai.’
As for the nominalizer -Ø, an active example like (23a), reproduced as (45), contains the active voice Ø-,27 which leads us to the conclusion that the zero nominalization has
.â•… See Footnote 7 for arguments proposing a null element Ø-, rather than w-, as an active voice marker.
Clausal nominalization in Budai Rukai 
scope over at least a voice category. The passive voice ki-, however, is incompatible in the construction containing the nominalizer -Ø, as seen in (46). This incompatibility could be due to the fact that the nominalizer -Ø is strictly restricted to a non-passive interpretation. Furthermore, a future tense affix like Ti- is not compatible inside the -Ø nominalized clause, as in (47). This indicates that the nominalizer -Ø has no scope either over a (future) tense. (45) masamali-aku [ku/ka/kai/*alaka Ø-Tumay-Ø-ini surprise-1sg.nom obl/obl/this/*comp act-hit-nmz-3sg.gen ki/*ka/*ku saTabu ki TaiTai/inianә] gen saTabu obl TaiTai/3sg.obl ‘I am surprised at Salabu’s (frequently) hitting Lailai/her.’ (46) *masamali-aku ku ki-Tumay-Ø-ini ki TaiTai surprise-1sg.nom obl pass-hit-nmz-3sg. gen TaiTai ‘I am surprised at Lailai’s being hit.’ (47) *masamali-aku ku Ti-Ø-Tumay-Ø ki saTabu ki TaiTai surprise-1sg.nom obl fut-act-hit-nmz gen saTabu obl TaiTai ‘I am surprised at Salabu’s being going to hit Lailai.’
Example (45) also illustrates one additional point regarding the issue of word order. As noted earlier, Budai is basically a predicate-initial language, like many other Austronesian languages. With overt nominal case marking, the word order of arguments is relatively free, being either VOS or VSO. The word order seen in (45), however, is rather fixed, with ki saTabu interpreted as the active agent subject. The -anә data (23b), now represented as (48), conforms to this VS(A)O order restriction in a similar way. Switching the order between saTabu and TaiTai results in ungrammaticality on the reading ‘I am surprised at Salabu’s hitting Lailai.’ Likewise, in a passive construction such as (49), the patient subject TaiTai has to be placed before the oblique agent saTabu, displaying again a strict VS[Obl(A)] word order. (48) masamali-aku [ku/ka/kai/*alaka ta-Tumað-anә surprise-1sg.nom obl/obl/this/*comp nfut-hit-nmz ki/*ka/*ku saTabu ki TaiTai/inianә] gen saTabu obl TaiTai/3sg.obl ‘I am surprised at (the bad consequence resulting from) Salabu’s hitting Lailai/her.’ (49) masamali-aku ku/ka/kai/*alaka ta-ki-Tumað-anә surprise-1sg.nom obl/obl/this/*comp nfut-pass-hit-nmz ki TaiTai ki saTabu gen TaiTai obl saTabu ‘I am surprised at Lailai’s being hit by Salabu.’
 Li-May Sung
All the above examples have come to suggest that grammatical subjects in Budai, whether they are agent or patient, are very salient in a syntactic sense, and must precede grammatical objects or oblique agent at some underlying level. Consider now other syntactic categories, such as negation or aspect. As shown in examples (50)–(51), a verbal negator kai is ruled out in both nominalized clauses. Only a nominal negator kai ka, external to the nominalized construction, can be felicitous. (50) Negation in -anә nominalization a.
malisi-aku ka/ku kai ka ta-kәla-anә ki saTabu angry-1sg.nom obl neg ka nfut-come-nmz gen saTabu
b. *malisi-aku ka/ku kai ta-kәla-anә ki saTabu c. *malisi-aku ka/ku ta-kai-kәla-anә ki saTabu ‘I am angry at Salabu’s not coming’ (51) Negation in -Ø nominalization a.
malisi-aku ka/ku kai ka kәla-Ø ki saTabu angry-1sg.nom obl neg ka come-nmz gen saTabu
b. *malisi-aku ka/ku kai kәla-Ø ‘I am angry at Salabu’s not coming’
ki
saTabu
Then what about aspect? A progressive aspect via a morphological reduplication of the verb base is permissible within both nominalized constructions of -anә and -Ø, as shown in (52), while a perfective aspect -ŋa has to occur outside the scope of the nominalization, as illustrated in (53). (52) Progressive aspect in -anә and -Ø nominalizations a.
kialaTa-laTa-aku ku ta-sәna-sәnað-anә listen-red-1sg.gen obl nfut-red-sing-nmz
ki aðaðamә28 gen bird
‘I am listening to the singing sound of the bird.’
b. butulu ku kanә-kanә-Ø-li pork nom red-eat-nmz-1sg.gen ‘The thing that I am eating is pork.’
.╅ The interlinear glosses have been slightly modified.
(Kuo 1979: 72)
Clausal nominalization in Budai Rukai 
(53) Perfective aspect in -anә and -Ø nominalizations a.
malisi ka saTabu ka ta-kan-anә-ŋa-li angry nom saTabu obl nfut-eat-nmz-pfv-1sg.gen ‘Salabu is angry at my having eaten already.’
b. laŋay-li ku/kuini kanә-Ø-ŋa-su29 buy-1sg.gen nom/that eat-nmz-pfv-2.sg.gen ‘The thing that you have eaten is what I bought.’
Our findings regarding the scope of the -anә and -Ø nominalizations over different categories other than the verb stem are summarized in Table 3, quite in support of the hierarchy of verbal categories proposed in Bybee (1985) and Malchukov (2006). Retention of these categories in nominalization constructions is also clear evidence of the verbal nature of such a nominalization process.
Table 3.╇ Scope of the nominalizers in Budai
verb stem verb-valence
causative reciprocal reflexive progressive (inner) aspect active voice passive voice tense perfective (outer) aspect (verbal) negation
-anә
-Ø
ü yes ü yes ü yes ü yes ü yes ü yes ü yes ü yes -------
ü yes ü yes ü yes ü yes ü yes ü yes -------------
In addition, the fact that -anә nominalization excludes the perfective affix -ŋa but retains the progressive one (via lexical reduplication) implies the different nature that these two aspectual markers may possess, suggesting that they might reside in structurally different slots; that is, the perfective -ŋa may occupy a relative outer aspect position, whereas the progressive aspect deploying a lexical strategy stays close to the
.â•… Since the nominalizer -Ø here is phonologically null, it is difficult to judge whether the perfective aspect -ŋa occurs outside or inside the scope of -Ø nominalization. Here I simply draw the same conclusion from the nominalizer -anә for the nominalizer -Ø.
 Li-May Sung
verb, and thus appears in an inner slot. Also worth noting is that the inclusion of the tense feature within the -anә nominalization poses a challenge to the general claim that syntactic nominalizations containing a verbal phrase would lack tense specifications, as has often been observed in similar constructions in other languages such as English gerunds, Hebrew gerunds, or Italian nominalized infinitives (cf. Siloni 1997). For example, POSS-ing gerunds in English systematically exclude all tensed verbs or tensed auxiliaries, as shown in (54) (cf. Pullum 1991: 768): (54) Lack of tense feature in POSS-ing gerunds in English
a. b. c. d.
*my talkeding to your sister *John’s sanging the song *his woulding have walked the dog *her canning play tennis quite well
Either from a cross-linguistic or a typological perspective, Budai is definitely not the only language allowing tense specifications in the nominalized constructions. For instance, the nominalized clauses in Cuzco Quechua, as pointed out in Lefebvre and Muysken (1988: 18), contain tense features as well. This is illustrated in (55): (55) Cuzco Quechua [+Main Tense] (main clauses)
-Ø- -rqa- -sqa- various
present, unmarked simple past sudden discovery future
[−Main Tense] (nominalized clauses)30 -na- -sqa- -y- -q-
finite, unrealized finite, realized non-finite agentive
Many similar examples showing retention of tense features in the process of nominalization are attested in a wide variety of languages documented in Koptjevskaja-Tamm (2003) and Malchukov (2006). Fijian provides another example of this sort, with tense and aspect markers totally retained. (56) Fijian
(Dixon 1988: 131; cited in Malchukov 2006: 991)
au tadra-a [a o-mu aa/na la’o mai] 1sg dream-tr art class-your past/fut come here ‘I dreamt that you had/will come.’
.â•… Nominalizations in Quechua have the following characteristics as generalized in Lefebvre and Muysken (1988). One is that case marking is possible on nominalized verbs, but never on non-nominalized ones. In addition to converting the [−N] feature of a verbal root into a [+N] feature, the nominalizing affixes also specify some sort of tense, in which -na-, for example, encodes future tense and -sqa- encodes past tense. The tense expressed by the nominalizing affixes is relative to the tense expressed by the main verb. This is why Lefebvre and Muysken label them [−Main Tense].
Clausal nominalization in Budai Rukai 
Additionally, Yap and Matthews (2008) and data in this volume also show ample evidence that there exists a potential close link between nominalizers, tense/aspect markers, and mood particles when addressing the nominalization phenomena across Asian and Austronesian languages. On the basis of foregoing robust crosslinguistic parallels, the case of Budai nominalized constructions retaining tense features is certainly not unusual.
5.â•… Clausal nominalization vs. lexical nominalization We have thus far seen that both -anә and -Ø nominalizations are highly pervasive phenomena found in Budai, displaying partly nominal and partly verbal syntax on a par with gerundive constructions detected in a number of languages. Nominalization in Budai is not just visible on the clausal level. Rather, the nominalizer -anә, together with many affixes, are also found to derive many lexical nouns, as listed in Table 4. Table 4.╇ Lexical nominalization Affixes
Base
Derived meaning(s)
Examples
Predicative forms
ta-…-anә
V
place for V-ing (reduplication is obligatory)
wa-suTaw ‘cure’ wa-buTu ‘teach’ m-ubanaw ‘take a shower’ wa-aga ‘cook’
N
place/container for N
V
a certain time/day for V-ing
ta-suTa-suTav-anә ‘medical clinic; hospital’ (Note: not ta-suLav-anә) ta-buTu-buTu-anә ‘the place where people teach something; classroom; church’ ta-ubana-banav-anә ‘bathroom’ ta-aga-aga-anә ‘kitchen’ ta-hana-anә ‘vase (container for flowers)’ ta-sabik-anә ‘container for betel nuts’ ta-laymað-anә ‘bedroom closet’ ta-cәkә-cәkәl-anә ‘familyhood; marriage’ ta-isa2asa2a-anә ‘resting day; Sunday’ ta-ki-tu-lalak-anә ‘birthday’
hana ‘flowers’ sabiki ‘betel nuts’ laymay ‘clothing’ cәkәlә ‘village’
wa-isa2a ‘take a rest’ tu-a-lalakә ‘give birth;do-fin-kid’ (Continued)
 Li-May Sung
Table 4.╇ Lexical nominalization╇ (Continued) Affixes
Base
Derived meaning(s)
Examples
Predicative forms
ka-…-anә
N
Genuinely locally made; related to Budai cultural tradition
ka-bava-anә ‘true wine; self-brewed rice wine’ ka-kipiŋ-anә ‘traditional clothing’ ka-diTuŋ-anә ‘true pottery’ ka-silu-anә ‘true glass beads’ ka-kaaŋ-anә ‘true fish (freshwater fish (usually from rivers, not bred on farms for commercial purposes))’ ka-daanә ‘true house, the one inherited by the eldest son’
bava ‘wine’ kipiŋ ‘clothing’ diTuŋu ‘pottery’ silu ‘glass beads’ kaaŋә ‘fish’ daanә ‘house’
sa-…-anә
V
materials for V-ing; something to V with
sa-suTav-anә ‘something used for curing; medicine’ sa-buTu-anә ‘materials for teaching’ sa-ubanav-anә ‘things (water) for washing sb’s face or taking a bath’ sa-aga-anә ‘things used for cooking, such as water, pepper, or soy sauce’ sa-uŋul-anә ‘soup’
wa-suTaw ‘cure’ wa-buTu ‘teach’ m-ubanaw ‘take a shower’ wa-aga ‘cook’ wa-uŋulu ‘drink’
sa-…-anә
V
tools for V-ing (reduplication is strongly preferred)
sa-paca-pacas-anә ‘pen; pencil; chalk’ sa-supi-supiT-anә ‘eraser’ sa-aga-aga-anә ‘kitchen utensil’
wa-pacasә ‘write’ wa-supiTә ‘remove’ wa-aga ‘cook’
sa-ŋu…. -anә
V
Part of sth. that has been left behind from V-ing
saŋu-suTav-anә ‘medical balm left from curing’ saŋu-aga-anә ‘things left behind from cooking’ saŋu-ubanav-anә ‘water left from washing sb’s face’ saŋu-uŋul-anә ‘wine or water left from drinking
wa-suTaw ‘cure’ wa-aga ‘cook’ m-ubanaw ‘take a shower’ wa-uŋulu ‘drink’
saŋua…. -anә
V
strong smell of
saŋua-sa-suTav-anә ‘strong smell of medicine’ saŋua-mia-anә ‘smell stinky’
wa-suTaw ‘cure’ mia ‘stinky smell’ (Continued)
Clausal nominalization in Budai Rukai 
Table 4.╇ (Continued) Affixes
Base
Derived meaning(s)
Examples
Predicative forms
kala-….. -anә
V/N
season
kala-2aTaŋә2aŋ-anә ‘summer’ kala-kәcәT-anә ‘winter’ kala-bәcәŋ-anә ‘harvest season for millets’ kala-rigarig-anә ‘spring’ kala-lidar-anә ‘fall’ kala-әdal-anә ‘rainy season’ kala-tukunuð-anә ‘harvest season for (Ficus pumila var. awkeotsang) plants’31
ma-2aTaŋә2aŋә ‘hot’ ma-kәcәTә ‘cold’ bәcәŋә ‘millet’ wa-rigarigi ‘to sprout’ ma-lidarә ‘wither’ wa-әdalә ‘rain’ tukunui ‘(Ficus pumila var. awkeotsang) plants’
…-anә
V
abstract concept; dimension; color; value
θarir-anә ‘goodness; beauty’ Tigil-anә ‘smartness’ әTәŋ-anә ‘length’ caәm-anә ‘(a certain type of) disease’
ma-θariri ‘good; pretty’ ma-Tigili ‘smart’ ma-әTәŋә ‘long’ ma-caәmә ‘sick’
Note: Shaded areas indicate data taken from Zeitoun (2000) and double-checked with my language consultants; the English translation is mine. Note: The prefix tara- derives agentive nouns (e.g. tara-sәna-sәnay ‘singer’, tara-paca-pacas ‘writer’, tara-iduT-iduTi ‘dancer’), not involving the nominalizing element -anә.31
As revealed in Table 4, the same mechanism of relativization still plays a central role in shaping the resultant forms. These -anә affixed or circumfixed nouns involve relativizing different types of undergoer participants, and are used for either entity or property references, such as place terms ta-suTa-suTav-anә ‘hospital’, instrumental terms sa-paca-pacas-anә ‘pen; pencil; chalk’ and value/dimension terms θarir-anә ‘goodness’ etc. These derived nominals do not refer to an event or a process expressing the actions of ‘healing’ or ‘writing’, as we have demonstrated in Sections 3 and 4. Moreover, they may be further modified by a relative clause as shown in (57), just as regular non-derived nouns would be in (58). (57) Relative clause modification in derived nouns affixed with -anә muaka2ulu ku ma-a2aw ku ta-suTa-suTav-anә fall.down nom stat.fin-big rel nfut-red-cure-nmz ‘The big hospital fell down.’
.╅ The fruit of this kind of plants is used to make jelly, a popular Taiwanese-style snack.
 Li-May Sung
(58) Relative clause modification in non-derived nouns ma-pituTu ku ma-a2aw ku tawpuŋu stat.fin-fat nom stat.fin-big rel dog ‘The big dog is fat.’
It is also worth noting that some of the root stems that -anә attaches to in Table 4 are unarguably nominal in nature, such as ta-hana-anә ‘vase (container for flowers)’, ka-silu-anә ‘true glass beads’, kala-bәcәŋ-anә ‘harvest season for millets’, etc. From these examples, it can be seen that the derivational function of -anә has widened significantly, having been extended from applying mainly to verb roots to now encompassing any possible lexical roots. Additionally, these lexically derived nominals display a certain degree of morphological irregularity. When ta-…-anә is formed with a verbal root, e.g. wa-buTu ‘teach’, the resultant form (ta-buTu-buTu-anә), accompanied obligatorily by a morphological reduplication, yields the meaning of ‘a place where people teach something; classroom; church’. With a nominal root, some ta-…-anә examples also require reduplication (e.g. ta-cәkә-cәkәl-anә ‘familyhood; marriage’), whereas some obviously do not (e.g ta-hana-anә ‘vase; container for flowers’). As for examples of abstract concept/state nominalization … -anә (e.g. caәm-anә ‘disease’) or material nominalization sa-…-anә (e.g. sa-suTav-anә ‘medicine’), morphological reduplication is not detected either, and in contrast, tool nominalization sa-…-anә (e.g. sa-supi-supiTanә ‘eraser’) would strongly favor reduplication. This lexical idiosyncrasy makes -anә derived nouns differ from the corresponding clausal nominalization instances of -anә, in that clausal nominalization generates rather predictable morphological forms.32 Clausal and lexical nominalizations still display a distinct nature, to a certain degree. Also, as pointed out in Genetti (this volume), clausal nominalization and lexical/ derivational nominalization differ essentially in terms of the domains to which they apply (lexical root versus clause) and in terms of the syntactic category of the resultant structures (single word versus noun phrase). Nonetheless, they are not two unrelated processes, historically and diachronically, as observed across the language families in the present volume.
6.â•… Concluding remarks To summarize our findings so far regarding Budai nominalization, two types of nominalization have been distinguished, i.e. clausal and lexical nominalizations, both
.╅ Morphological reduplication can also be used in clausal nominalization construction. In this case, it may yield either a progressive or habitual interpretation of the event, rather than a result nominal interpretation.
Clausal nominalization in Budai Rukai 
extensively employing the -anә nominalizer, together with various prefixes ta-, sa-, ni- etc. While clausal and lexical nominalizations are closely correlated, each manifests a distinct nature. The mechanism of relativization plays a key role in shaping the resultant structures in question. The intimate relationship between nominalization and relativization is fairly robust across different language families discussed in the literature and represented in this volume (see in particular Yap, Grunow-Hårsta and Wrona for a comprehensive review). Many forms of nominalized data in Budai also reflect synchronic traces of “four-voice” contrasts of PAn *-um-, *-en, *-an, and *Si-, despite the fact that Budai features notably a dichotomic active/passive voice system. We see evidence of Budai nominalizer -anә, a reflex of location voice *-an, being grammaticalized to encompass almost all (undergoer) participants that get relativized. Figure 1 provides a sketch of this grammaticalization pathway of Budai -anә and its relation to the various nominalized constructions we have examined in this work. wa-/-a-/ma-/ki-a
ordinary relativization
nominalizer -Ø
PAn *-an (location voice)
nominalizer -anә (agentless, undergoers, adjuncts) ta-…-anә sa-…-anә ni-…-anә (lu)…-anә
nominal complement ka/ku [VP-anә /VP-Ø] relative clause [VP-anә /VP-Ø] ka/kurelN pseudo-cleft clause [VP-anә /VP-Ø] ka/kurelN (headed) [VP-anә /VP-Ø] (headless)
Figure 1.╇ Grammaticalization pathway of Budai -anә and its relation with various nominalized constructions
Of particular relevance for our current discussion is that many of the -anә and -Ø nominalized clauses we have examined exhibit a mixture of both nominal and verbal properties. Such mixed properties are precisely analogous to those of English gerundives and their counterparts from a variety of languages. Both -anә and -Ø constructions are clearly nominal, accepting case markers/demonstratives and genitive subjects. Yet their internal structure resembles in nearly all respects that of a sentence. Oblique-marked object arguments are permissible, modification is adverbial, and various verbal categories are retained, all of which are indicative of verbal syntax, rather than nominal syntax. As has been typologically addressed in Koptjevskaja-Tamm (2003), Malchukov (2006), Comrie and Thompson (2007) and Noonan (2007), nominalized clauses vary considerably with respect to the extent in which verbal categories can be retained. Budai has been shown in this paper to exemplify the retention of verbal categories, including (active/passive) voice, aspect,
 Li-May Sung
tense and valency morphemes during both -anә and -Ø nominalization processes. Nonetheless, these two nominalized clauses differ slightly in regards to the inclusion of tense features. The -Ø nominalization is distinguished from the -anә nominalization in that the outer layer which hosts the tense feature is not retained. This is again consistent with the otherwise well-attested tendency for tense to be lost prior to aspect, and meshes perfectly with the generalization of hierarchical verbal categories proposed in the typological literature of Bybee (1985) and Malchukov (2006). Finally, we have also demonstrated that among verbal arguments, the object in nominalization in Budai retains its sentential oblique form, while the subject, no matter whether it is agent/actor subject or patient subject, receives possessive (genitive) encoding. That is, S genitivizes prior to O, in accordance with Comrie’s (1976) generalization (cf. Malchukov 2006). We are seeing evidence of nominalization constructions exhibiting both nominal and verbal properties, with retention of oblique form for the object suggestive of verbal predication, while use of genitive subjects suggests nominal marking for the whole clause, since the object is the internal argument while the subject is the external argument.
Acknowledgments My sincere thanks go to the reviewers, book editors and in particular Foong Ha Yap, who has been an invaluable source of advice and has contributed considerably to my thoughts on the nature of nominalization through the revising process. I also owe a debt of gratitude to Peter Cole, Gabriella Hermon, James Yoon, and Shuanfan Huang for their comments on an earlier draft of this article. Parts of the material included in this study were presented at the sixth AFLA conference held in Amsterdam, and I am indebted to the people in attendance for their suggestions and comments. Thanks are also due to my Budai consultants, SaTabu Ka2әsәŋanә (Chi-chuan Ke), TaiTai Ka2әsәŋanә (Chia-ying Pa), VaTyalanә Ka2әsәŋanә (Chao-hsien Wang), and Balәngә Kasәpәlanә (Chu-hua Ke), for providing the data on which this paper is based. If not otherwise mentioned, the data used in this paper come from my own field work. Errors and oversights are of course the sole province of mine. This work was supported by a two-year project entitled The Austronesians: language, gene, culture, and archaeology, National Taiwan University.
References Blust, Robert. 1999. Notes on Pazeh phonology and morphology. Oceanic Linguistics 38: 321–365. Blust, Robert. 2003. Three notes on early Austronesian morphology. Oceanic Linguistics 42: 438–478. Bybee, Joan. 1985. Morphology: A Study of the Relation between Meaning and Form [Typological Studies in Language 9]. Amsterdam: John Benjamins.
Clausal nominalization in Budai Rukai 
Chang, Henry Yungli. 2006. Adverbial modifiers as matrix verbs in Kavalan. In Clause Structure and Adjuncts in Austronesian Languages, Hans-Martin Gärtner, Paul Law & Joachim Sabel (eds), 43–82. Berlin: Mouton de Gruyter. Chen, Cheng-fu. 1999. Wh-words as Interrogatives and Indefinites in Rukai. MA thesis, National Taiwan University. Chen, Cheng-fu. 2008. Aspect and Tense in Rukai: Interpretation and Interaction. Ph.D. dissertation, University of Texas at Austin. Chen, Cheng-fu & Sung, Li-May. 2005. Interrogatives as polarity items in Kucapungan Rukai. Concentric: Studies in Linguistics 31(1): 95–117. Chomsky, Noam. 1970. Remarks on nominalizations. In Readings in English Transformational Grammar, Roderick A. Jacobs & Peter S. Rosenbaum (eds). Waltham MA: Ginn. Comrie, Bernard. 1976. The syntax of action nominals: A cross-linguistic study. Lingua 40: 177–201. Comrie, Bernard & Thompson, Sandra A. 2007. Lexical nominalizations. In Language Typology and Syntactic Description, Vol. 3, Timothy Shopen (ed.), 334–381. Cambridge: CUP. Dikken, Marcel. 2006. Specificational copular sentences and pseudoclefts. In The Blackwell Companion to Syntax, Vol. IV, Martin Everaert & Henk van Riemsdijk (eds), 292–409. Oxford: Blackwell. Dixon, Robert Malcolm Ward. 1988. A Grammar of Boumaa Fijian. Chicago IL: University of Chicago Press. Evans, Bethwyn & Ross, Malcolm. 2001. The history of Proto-Oceanic *ma-. Oceanic Linguistics 40: 269–290. Ferrell, Raleigh J. 1982. Paiwan Dictionary [Pacific Linguistics C-73]. Canberra: Australian National University. Genetti, Carol. This volume. Nominalization in Tibeto-Burman languages of the Himalayan area: A typological perspective. In Yap, Grunow-Hårsta & Wrona (eds). Holmer, Arthur. 2006. Seediq – Adverbial heads in a Formosan language. In Clause Structure and Adjuncts in Austronesian Languages, Hans-Martin Gärtner, Paul Law & Joachim Sabel (eds), 83–123. Berlin: Mouton de Gruyter. Hsieh, Fuhui. This volume. The functions of -an and =ay in Kavalan. In Yap, Grunow-Hårsta & Wrona (eds). Huang, Shuping & Sung, Li-May. 2008. The undergoer focus ma- in Kavalan. Oceanic Linguistics 47(1): 159–184. Kemmer, Suzanne. 1993. The Middle Voice [Typological Studies in Language 23]. Amsterdam: John Benjamins. Kaufman, Daniel. This volume. Exclamatives and temporal nominalizations in Austronesian. In Yap, Grunow-Hårsta & Wrona (eds). Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. Koptjevskaja-Tamm, Maria. 2003. Action nominal constructions in the languages of Europe. In Noun Phrase Structure in the Languages of Europe, Frans Plank (ed.), 723–761. Berlin: Mouton de Gruyter. Kratochvìl, František. This volume. Discourse-structuring functions of Abui demonstratives. In Yap, Grunow-Hårsta & Wrona (eds). Kuo, John Ching-hua. 1979. Rukai Complementation. MA thesis, Fu Jen Catholic University, Taiwan. Lees, Robert. 1960. The Grammar of English Nominals. The Hague: Mouton. Lefebvre, Claire & Muysken, Pieter. 1988. Mixed Categories: Nominalizations in Quechua. Dordrecht: Kluwer.
 Li-May Sung Li, Paul Jen-kuei. 1973. Rukai Structure [Special Publication No. 64]. Taipei: Institute of History and Philology, Academia Sinica. Li, Paul Jen-kuei. 1977. The internal relationships of Rukai. Bulletin of the Institute of History and Philology 48: 1–92. Taipei: Academia Sinica. Li, Paul Jen-kuei. 1997. A syntactic typology of Formosan languages – case markers on nouns and pronouns. In Chinese Languages and Linguistics IV: Typological Studies of Languages in China [Symposium series of the Institute of History and Philology 2], Chiu-yu Tseng (ed.), 343–378. Taipei: Academia Sinica. Malchukov, Andrej L. 2004. Nominalization, Verbalization: Constraining a Typology of Transcategorial Operations [Lincom Studies in Language Typology 8]. Munich: Lincom. Malchukov, Andrej L. 2006. Constraining nominalization: Function/form competition. Linguistics 44(5): 973–1008. Noonan, Michael. 2007. Complementation. In Language Typology and Syntactic Description, Timothy Shopen, (ed.), 52–154. Cambridge: CUP. Potsdam, Eric. This volume. Expressing exclamatives in Malagasy. In Yap, Grunow-Hårsta & Wrona (eds). Pullum, Geoffrey K. 1991. English nominal gerund phrases as noun phrases with verb-phrase heads. Linguistics 29: 763–799. Ross, Malcom D. 1995. Reconstructing Proto-Austronesian verbal morphology: Evidence from Taiwan. In Austronesian Studies Relating to Taiwan [Symposium Series of the Institute of History and Philology, Academia Sinica, Number 3], Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho & Chiu-yu Tseng (eds), 727–791. Taipei: Academia Sinica. Ross, Malcolm & Teng, Stacy Fang-ching. 2005. Formosan languages and linguistic typology. Language and Linguistics 6(4): 739–781. Siloni, Tal. 1997. Noun Phrases and Nominalizations: The Syntax of DPs. Dordrecht: Kluwer. Starosta, Stanley. 1995. A grammatical subgrouping of Formosan languages. In Papers from the International Symposium on Austronesian Studies Relating to Taiwan, Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho, & Chiu-yu Tseng (eds), 683–726. Taipei: Academia Sinica. Starosta, Stanley, Pawley, Andrew & Reid, Lawrence. 1982. The evolution of focus in Austronesian. In Papers from the Third International Conference on Austronesian Linguistics 2: Tracking the Travellers [Pacific Linguistics C-75], Amran Halim, Lois Carrington & Stephen A. Wurm (eds), 145–170. Canberra: Australian National University. Sung, Li-May, Su, Lily I-wen, Hsieh, Fuhui & Lin, Zhemin. 2008. Developing an online corpus of Formosan languages. Taiwan Journal of Linguistics 6(2): 79–118. Tang, Jane Chih-Chen. 2008. Functional extension vs. grammaticalization: A typological study of modification markers in Formosan nominals. Language and Linguistics 9(4): 917–966. Watters, David. 2008. Nominalization in the Kiranti and Central Himalayish languages of Nepal. Linguistics of the Tibeto-Burman Area, 31(2): 1–43. Wolff, John. 1973. Verbal inflection in Proto-Austronesian. In Parangal kay Cecilio Lopez, Andrew Gonzalez (ed.), 71–91. Quezon City: Linguistic Society of the Philippines. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], López-Couso, María José & Elena Seoane (eds), in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins.
Clausal nominalization in Budai Rukai 
Yap, Foong Ha & Wang, Jiao. This volume. From light noun to nominalizer and more: On the grammaticalization of zhe and suo in Old and Middle Chinese. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha, Grunow-Hårsta, Karen & Wrona, Janick. This volume. Nominalization strategies in Asian languages. In Yap, Grunow-Hårsta & Wrona (eds). Yeh, Marie Meili. This volume. Nominalization in Saisiyat. In Yap, Grunow-Hårsta & Wrona (eds). Yoon, James Hye Suk. 1991. Korean nominalizations, lexicalism, and morphosyntactic interface. Ms, University of Illinois, Urbana-Champaign. Yoon, James Hye Suk. 1996. Nominal gerund phrases in English as phrasal zero derivations. Linguistics 34: 329–356. Zeitoun, Elizabeth. 2000. A Reference Grammar of Rukai [Series on Formosan Languages 8]. Taipei: Yuanliou Publishing Company. (In Chinese). Zeitoun, Elizabeth. 2002. Nominalization in Mantauran (Rukai). Language and Linguistics 3(2): 241–282. Zeitoun, Elizabeth. 2007. A Grammar of Mantauran (Rukai) [Language and Linguistics Monographs series A4-2]. Taipei: Institute of Linguistics, Academia Sinica. Zeitoun, Elizabeth & Huang, Lillian M. 2000. Concerning ka-, an overlooked marker of verbal derivation in Formosan languages. Oceanic Linguistics 39(2): 391–414. Zeitoun, Elizabeth, Lillian M. Huang, Marie M. Yeh & Anna H. Chang. 1999. Existential, possessive, and locative constructions in Formosan languages. Oceanic Linguistics 38(1): 1–42. Zeitoun, Elizabeth, Huang, Lillian M., Yeh, Marie M., Chang, Anna H. & Wu, Joy J. 1996. The temporal, aspectual, and modal systems of some Formosan languages: A typological perspective. Oceanic Linguistics 35(1): 21–56. Zeitoun, Elizabeth & Teng, Stacy. 2009. From ki-N ‘get N’ in Formosan languages to ki-V ‘get’ V-ed’ (passive) in Rukai, Paiwan and Puyuma. In Discovering History Through Language: Papers in Honour of Malcolm D. Ross [Pacific Linguistics PL605], Bethwyn Evans (ed.), 479–500. Canberra: Australian National University. Zucchi, Alessandro. 1993. The Language of Propositions and Events: Issues in the Syntax and the Semantics of Nominalization. Dordrecht: Kluwer.
Nominalization in Saisiyat* Marie Mei-li Yeh
National Hsinchu University of Education Studies on Formosan languages (Huang 2002; Li 2002; Rau 2002; Zeitoun 2002) have pointed out that lexical and syntactic nominalization are indistinct morphologically. In Saisiyat, not only are lexical nominalization deriving argument nouns and syntactic nominalization forming relative or subordinate clauses morphologically identical, the nominalizers are also found to function as tense or aspect markers. In terms of categoriality, (Hopper & Thompson 1984, 1985), such phenomena indicate a cline from verbs in a special temporal/ aspectual frame to verbs in subordinate clauses serving background function and then to derived nouns. In light of this, the relationship between marking tense/aspect and these two types of nominalization can be captured by a decategorialization process from verb to noun triggered by discourse function.
1.â•… Introduction Saisiyat is an Austronesian language spoken by the aboriginal people inhabiting the northern Taiwan mountain areas. According to Li (1985, 1997) and Blust (1999), it belongs to the Northwestern subgroup of the Austronesian language family. Lexical nominalization, the process of deriving lexical nouns from verbs or adjectives, is a very productive process in Saisiyat. According to Yeh (2003: 164–166), many semantically prototypical nouns are derived from lexical nominalization.1 As shown in (1), nouns
*The data presented in this paper are drawn from the author’s own field notes. I would like to express my greatest gratitude to all the informants I have consulted. An earlier version of this paper was presented at the Workshop on Nominalization in Formosan Languages, Academia Sinica, Oct. 21–22, 2000. I have benefited from discussions with the participants, especially Jane Tang. I am also grateful to Henry Yungli Chang and Foong Ha Yap for their insightful and constructive comments and discussions that contributed a lot to the development of this paper. Thanks are also due to the two anonymous reviewers for their helpful comments. .â•… Hopper and Thompson (1985: 152) hold that “certain ‘prototypical’ percepts of thinglike entities will be coded in a grammatical form identifiable as Noun, while prototypical percept of action or events will be coded as grammatical Verbs.” These prototypical nouns should correspond to the first order nouns in Lyons (1977: 442–447), which are common
 Marie Mei-li Yeh
referring to natural objects such as ‘sun’, ‘rain’, or plant species are derived from verbs, as are nouns referring to body parts, as in (2).2 (1) a.
ha~hila: ‘the sun’3 nmlz~sun.rise
b. ’a~’oral ‘rain’ nmlz ~to rain c.
ta~toba’ ‘fish poison’ (plant species, Derris poison) nmlz~to poison.fish
(2) a.
ka-’otet-an nmlz-fart-lv
‘hip’
b. ka-ngesel-an ‘nose’ nmlz-have.running.nose-lv c.
rotol-an ‘joint’ join< nmlz>lv
These two sets of examples illustrate major strategies for forming derived nouns in Saisiyat – first, by way of Ca- reduplication, with C representing a copy from the first consonant of the stem and a, the fixed vowel /a/,4 and second, by adding ka- or -into a verb marked by Locative Voice (LV) marker -an.5 Notice that without any other and concrete nouns, readily re-identified and can be referred to as individuals. According to Croft (1990: 245), words denoting human beings, animate entities or physical objects such as woman, dog, car, house and flower are prototypical nouns. .â•… One of the reviewer points out that it is common for the nouns in (1) to be derived from verbs since they refer either to reoccurring events in the nature or to the usage of the plant. However, in another Formosan language Atayal the root tuba’ functions as a noun denoting ‘fish poison.’ Compared with Atayal, forming nouns from verbs appear to be characteristic of Saisiyat. .â•… Leipzig glossing is followed in general, with the following abbreviation added:
av Actor Voice incpt Inceptive iv Instrumental Voice lig ligature lv Locative Voice pv Patient Voice prsn Personal Name q Question
For typographic convenience, ae, oe, e,:, ’, S, and ng are used to represent low front unrounded vowel /æ /, mid front rounded vowel /œ/, mid central vowel /ә/, long vowel, glottal stop /ô/, voiceless alveolar fricative /s/ (with s representing voiceless dental fricative) and velar nasal /ŋ/ respectively. .â•… The term is coined by Blust (1998). According to Blust (1998, 2003), at least two functions can be reconstructed for Pan *Ca- reduplication – to form a numeral in counting humans and to derive instrumental nouns. .â•… Most Formosan languages show a Philippine-type voice system (or Focus system in some studies) in which four categories – Actor, Patient (or Undergoer), Locative, and Instrumental
Nominalization in Saisiyat 
markers, the adding of ka- yields an instrumental noun whereas -in- infixation alone results in a derived object noun as shown in (3) and (4). (3) a.
ka-ka:at nmlz-to write
b. ka-ko:as nmlz-to comb (4) a.
ka:at write< nmlz>
‘pen’ ‘comb’ ‘book, letter’
b. ’omaS ‘salted pork; pickled vegetable’ to salt/pickle< nmlz>
These are examples of lexicalized nouns derived from verbs via nominalization. Therefore, Ca- reduplication, the prefix ka- as well as the infix -in- can be analyzed as nominalizers in Saisiyat. However, the three morphological strategies also play a role in the verbal paradigm of Saisiyat (Yeh 1995), with Ca- and ka- marking futurity and -in- perfectivity, as shown in (5).6 In other words, the affixes for deriving lexical nouns also function as a kind of tense or aspect marker.7 (5) a.
hini’ malat rim’an ka~kilmaeh ka kahoey this knife tomorrow fut.iv~chop acc wood ‘This knife is what is going to be used to chop wood tomorrow.’
b. yako nak hini’ m-patawaw, ka-’angang-en=ay ni ’oya’ 1sg.nom like this av-do fut-scold-pv=q gen mother ‘Will I be scolded by Mother, doing it this way?’ c.
baki’ ma’rem haseb=ila tiyamsong grandfather sleep.av five=incpt hour ‘Grandfather has slept for five hours.’
To complicate the picture further, these affixes also function in relative or subordinate clauses, which are regarded as a kind of syntactic nominalization, as will be discussed in Voice – are distinguished. They fall into two large groups – Actor Voice (AV) and Non-Actor Voice (NAV) based on the case realization of the Actor (Ogawa & Asai 1935: 30) – in AV construction, the Actor is marked as Nominative whereas in NAV construction, it is marked as Genitive. .â•… Ca- reduplication and -in- are reconstructed as the Pan future/durative and perfective aspect markers respectively (Wolff 1973; Ross 1995, 2002). .â•… The same phenomenon is found in other world languages. Shibatani and Awadh (2009) reports a productive lexical nominalization process to derive nouns from perfective verbs. According to Genetti (this volume), in some Tibeto-Burman languages, nominalized verbs deriving lexical verbs or lexical adjectives also function in a larger system of tense-aspect marking.
 Marie Mei-li Yeh
the next section.8 In other words, the future or perfective verb form, lexical nominalization and syntactic nominalization are morphologically identical in Saisiyat. Therefore, two issues are addressed in this paper: 1. Whether lexical nominalization exists in Saisiyat or not? 2. What is the relationship between future or perfective verb form and syntactic as well as lexical nominalization? The rest of this paper is organized as follows: section two is an introduction on nominalization in Saisiyat, including lexical and syntactic nominalization, section three deals with the relationship between futurity or perfectivity and nominalization, and section four concludes the discussion.
2.â•… Nominalization in Saisiyat 2.1â•… Types of nominalization In Chomsky (1970), two types of derived nominals – gerundive and derived nominals – are distinguished on the basis of their productivity, their relation with the proposition and their internal structure. While gerundive nominals can be formed freely from propositions, hold a regular semantic relation with the proposition, and does not have the internal structure of a noun phrase, derive nominals cannot be formed freely from propositions, bear a varied and idiosyncratic semantic relation with the proposition, and have the internal structure of a noun phrase. In addition, Chomsky (1970) also postulates the existence of a third kind, a mixed type, which bears the internal structure of a noun phrase, and yet cannot be modified by adjectives. In Comrie and Thompson (2007) nominalization process for forming nouns from lexical verbs and adjectives are distinguished form the process for forming noun phrases from predicates and propositions, including action nominals (noun phrase containing a derived noun and one or more reflexes of a proposition or a predicate) and clausal nominalization (nominalization with no lexically derived noun). The three are different syntactically – while derived nominals of the first type, lexical nominalization, behave syntactically like nouns, action nominals display syntactic characteristics of nouns and verbs, and in a nominalized clause the head, instead of showing any nominal properties, often possesses verbal characteristics.
.╅ According to Genetti (this volume), clausal nominalization in Tibeto-Burman languages also functions in different kinds of embedded structures.
Nominalization in Saisiyat 
Shibatani and Awadh (2009: 22) also point out the importance in making distinction between lexical nominalization and grammatical nominalization. First of all, lexical nominalization forms nouns belonging to the lexical category of noun, whereas grammatical nominalization creates nominal expressions that do not have lexical status. In addition, lexical nominalizations identify the referent and classify them as belonging to a particular class of entity while grammatical nominalizations function to characterize an entity referred to in terms of an event without specifying the identification or names. For example, the expression what I bought yesterday characterizes an entity by referring to an event of the speaker’s buying it yesterday, with the identity unidentified – it could be a book, a newspaper, an umbrella or anything that could be bought. In this paper, the process for deriving nouns from lexical verbs or adjectives will be referred to as lexical nominalization, and the process for deriving noun phrases from predicates and propositions, as syntactic nominalization. According to Chomsky (1970), in English the distinction between different types of nominalization is also manifested in their morphology (-ing versus -ity, -ion, -al, etc.) In Saisiyat, however, argument nouns derived from lexical nominalization are identical in form to the verbs in clausal nominalization in that both involve the voice (or focus markers, following the terminology of Austronesian linguistic tradition). In other words, different types of nominalization do not appear to be morphologically distinct. In what follows, it will be shown that they are different in terms of syntactic behaviors although they share the same forms.
2.2â•… Lexical nominalization forming names of an argument Lexical nominalization, according to Comrie and Thompson (2007: 334), may derive nouns referring to the action/state and or the arguments of the verbs/adjectives. Action/ state nouns in Saisiyat such as ka-Spi’ ‘dream’, kin-Sekela’ ‘knowledge’ and kak-’aeyaeh ‘illness’ are formed by adding the prefix ka-, kin- or ka(k)- to action verb ‘iSpi’ ‘to dream’, and state verbs Sekela’ ‘to know’ and ’aeyaeh ‘sick, ill.’ Lexical nominalization deriving nouns referring to arguments will be discussed in this paper since voice markers are involved to mark the argument and therefore this type of nominalization manifest the fact that lexical nominalizations are in the same form as syntactic nominalization (cf. Li 2002; Huang 2002; Raw 2002, and Zeitoun 2002). As shown in the introductory section, adding ka- to verb roots yields instrumental nouns. Besides, Ca- reduplication is also a productive way to derive instrumental nouns (Yeh 2009).9 Blust (1998) points out that the meaning of instrument can sometimes be opaque. For example, the nouns referring to natural phenomena such as sun and rain in (1) and those referring to body .â•… The use of Ca- reduplication in the formation of instrumental nouns is also attested in Pazeh, Puyuma, and Thao (Blust 1998), and Atayal (Huang 2002: 219).
 Marie Mei-li Yeh
parts in (2) are regarded as representing opaque instrumental meanings, and therefore are called ‘non-canonical’ instrumental nouns. In contrast, canonical instrumental nouns refer to such derived nouns as comb, broom/brush, sieve, chopsticks, and lids, whose instrumental meanings are transparent. The function of Saisiyat Ca- in forming instrumental nouns are attested not only in the reflexes of pan words such as sa-sapoeh ‘broom’ (from sapoeh ‘sweep’) and ha-ha:ep ‘sift’ (from ha:ep ‘sieve’) but also in nouns referring to new inventions, as shown in the following examples from Yeh (2009). (6) a.
ka~kaskes nmlz~cold
b. ha~hiyop ka bokes nmlz~blow acc hair c.
pa~pori’ nmlz~apply (balms)
d. pa~potoy nmlz~wrap
‘air conditioner’ ‘hair dryer’ ‘plaster’ ‘bandage’
For the formation of other kinds of argument nouns, the voice affixes as well as the prefix ka- are involved. For example, agentive nouns can be formed by adding ka- to verbs marked with ma-, a kind of Actor Voice (AV) marker (Ross 1995). (7) a.
ka-ma-’omalop10 nmlz-av-hunt
b. ka-ma-kiSka:at nmlz-av-read c.
ka-ma-marma’ nmlz-av-steal
d. ka-ma-hayap nmlz-av-fly
‘hunter’ ‘student’ ‘thief ’ ‘plane’
Besides agentive nominalization, patient and location nouns can be formed by adding ka- to verbs marked with Patient Voice (PV) marker -en and Location Voice (LV) marker -an respectively. (8) a.
ka-si’ael-en nmlz-eat-pv
b. ka-ra’oe:-en nmlz-drink- pv c.
ka-i:ba:-en nmlz-wear- pv
‘food’ ‘drink’ ‘clothes’
.╅ Why sometimes the verbs appear in Actor Voice form remains a puzzle to be solved.
(9) a.
Nominalization in Saisiyat 
ka-patol-an nmlz-sing-lv
b. ka-si’ael-an nmlz-eat-lv c.
ka-talek-an nmlz-cook-lv
‘the place for singing’ ‘the place for eating; restaurant’ ‘the place for cooking; kitchen’
Table 1 below summarizes the mechanisms for deriving lexical nouns referring to arguments in Saisiyat. The examples illustrate how different nominalization processes can be applied to the same verb stem to yield different nouns.11 Table 1.╇ Formation of argument nouns in Saisiyat Type
Form
Example
Agentive
ka-ma-V
ka-ma-’omalop ‘hunter’ ka-ma-ka:at ‘writer’ ka-’alop-en ‘game’ ka-ka:at-en ‘the thing to be written, homework’ ’-in-alop ‘game been hunted’ k-in-a:at ‘what is written, book, letter, word’ ka-’alop-an ‘hunting area’ ka-ka:at-an ‘place for writing, desk’ ka-ka:at ‘pen’ ’a-’alop ‘hunting instrument’
ka-V-en Patient
V-in-
Locative
ka-V-an
Instrument
ka-V Ca~V
Notice that the formation of patient and instrumental nouns both involve two nominalizers – namely, ka- or -in- for patient nouns, and ka- or Ca- for instrumental nouns. It is reported in Yeh (2003: 26–27) that future instrumental voice (IV) Ca- and ka- are allomorphs, with ka- being the variant of Ca- occurring when the verb stems contain two identical/similar syllables or pa-.12 As shown below, the same condition holds of the formation of instrumental nouns. (10) a.
ka-pahiyo’ ‘flash lighter’ nmlz-light.up
b. ka-bo:bo: ‘fan’ nmlz-to fan
.╅ As shown in the gloss, the nouns show different degree of lexicalization. Some of them require context to render a specific reference. .╅ The motivation for ka- to occur with verb stems containing two identical/similar syllables is to avoid more than two identical/similar syllables in one word. For words initiated with pa-, the motivation is to avoid mistaking the reduplicated part as the causative morpheme.
 Marie Mei-li Yeh
With regard to -in- and ka- marked patient nouns, the difference lies in their temporal/ aspectual implication, as shown below. (11) a. ka-tawbon-en ‘something to be pounded’ b. tawbon ‘something pounded; e.g. rice cake’
While the ka- marked patient noun, which refers to ‘things to be V-en’, is irrealis, the noun derived from -in-, indicating ‘things that have been V-ed’, is realis.13 In other words, derived nouns seem to retain temporal or aspectual distinction.
2.3â•… Functions of nominalization Comrie and Thompson (2007: 377–379) point out that in addition to subjects and objects, nominalization also functions in subordinate and relative clauses. Hopper and Thompson (1984), in discussing the discourse basis for the lexical category noun (N) and verb (V), points out that property of being a prototypical instance of the grammatical category N or V is imposed on linguistic forms by discourse – while prototypical N’s introduce a participant, prototypical V’s report an actual event of the discourse. In typical environments, N and V display the widest degree of morpho-syntactic contrast, whereas in non-typical environments, the contrast is usually neutralized. For example, background materials are subsidiary to foregrounded events, and therefore are often marked as dependent. The V in many types of dependent clauses, not being typical instances of prototypical V tends to be low in categoriality, and therefore may appear in a nominalized form. Among them, the V in relative clauses and purpose clauses in some languages are reported to be nominalization. As shown in the following examples, the verbs in Saisiyat purpose clauses are marked by ka-. Consequently, verbs in purpose clauses are in the same form as lexical nouns derived from nominalization. (12) a.
yao ’am14 mari’ ka yasam ka~kilmaeh ka kahoey15 1sg.nom fut take acc axe purp~chop acc wood ‘I am going to take an axe to chop wood.’
.â•… For realis and irrealis as the basic temporal distinction in most Formosan languages, we refer you to Zeitoun et al. (1996). .â•… Notice that future in Actor Voice constructions is marked by ‘am, which is argued to be derived from the verb denoting “desire” in Yeh (2006). .â•… One of the reviewer points out the possibility of analyzing this construction as SVC (serial verb construction). However, this construction does observe the AV-only constraint (the second verb of SVC is restricted to be marked as AV) in Formosan SVC construction. For details, we refer the readers to Chang (2011).
Nominalization in Saisiyat 
b. ’oya’ talek ka ’aelaw no korkoring ka-si’ael-en mother cook acc fish dat child purp-eat-pv ‘Mother cooked fish for the children to eat.’ c.
yami ma-’az’azem ’am paskayzaeh ka taew’an 1pl.excl.nom av-think fut make acc house
ka-rowasek-an nonak purp-live-lv self
‘We are thinking that we will build a house for ourselves to live in.’
An actor voice (AV) purpose clause, though not in the same form as agentive nominalization, is also marked by ka-, as shown in (13).16 (13) a.
’oya’ talek ka kashaw ’ima rikrika: ka-pa-rae’oe: ’iSo’on mother cook acc soup rel hot purp-cau-drink 2sg.acc ‘Mother cooked hot soup so as to let you drink.’
b. yako talek ka pazay ka-pa-si’ael 1sg.nom cook acc rice purp-cau-eat ‘I cooked rice so as to feast the guest.’
ka ’aeh’aehael acc guest
Relativization and nominalization, according to Comrie and Thompson (2007: 379), are indistinct in some languages. This phenomenon has been reported to be attested in many Formosan languages (Li 2002 on Pazeh, Huang 2002 on Mayrinax Atayal, Raw on Yami, Zeitoun 2002 on Mantauran Rukai). Shibatani (2008) proposes to analyze relativization in Indonesian languages (e.g. Sasak) as nominalization, remarking that it is “just one pattern of noun modification involving a nominalized clause as a modifier.” Comrie & Thompson (2007: 338) points out that in some languages, agentive nominalization can also be used to modify another noun. For instance, in the following Mandarin Chinese example, agentive nominalization appears to be used as a modifier. To be more specific, relativization is identical to agentive nominalization in form. Mandarin (14) a.
(Comrie & Thompson 2007: 338)17
chao-fan de ren cook-rice nmlz person ‘a/the person who cooks’
.â•… Notice that the marker in purpose clauses is pa- instead of ma-. From the meaning and case marking of the subordinate clause, pa- could be the causative prefix. However, there is another possibility that pa- is the root form of ma- (cf. Ross 1995). According to Huang (2001), future marker for AV in Mayrinax Atayal is pa-. Whether the pa- in Saisiyat AV purpose clauses and the future pa- in Mayrinax Atayal are the same deserves further study. .â•… This is the transcription in Comrie and Thompson (2007), only with the omission of tone marking. As a native speaker, I feel that ‘cook’ should be chou-fan, rather than chao-fan ‘fry rice.’
 Marie Mei-li Yeh
b. chao-fan de cook-rice nmlz ‘a cook’
In Saisiyat, however, not only agentive nominalization (15), but also the other types of argument nominalization can function as a noun modifier as shown in (16). Agentive relative clause (15) a.
[ka-ma-talek ’ini ’apih] ka18 ma’iyaeh rae’iw=ila nmlz-av-cook for prsn lig man leave=incpt ‘The man who used to cooked for ‘apih has left.’
b. tatini’ sarara’ ka [ka-ma-rae’oe: ka pinobae:aeh] kabina:o’ old.man like acc nmlz-av-drink acc wine lady ‘The old man likes the lady who drinks/is drinking wine.’ (16) a.
yako sarara’ [ni ’oya’ talek] ka ’aelaw 1sg.nom like gen mother cook lig fish ‘I like the fish cooked by Mother.’
b. [sa~si’ael ni ’obay] ka sasoway ’inaro: Ca~eat gen prsn lig chopsticks long ‘The chopsticks ‘obay uses for eating are long.’ c.
[ka-hiwa:-en rim’an] ka waliSan rae’iw ila ka-cut-pv tomorrow lig boar run.away incpt ‘The boar that is going to be killed tomorrow has run away.’
d. hini’ [mayhael mita’ ka-kakrowaSek-an] ka taewan this future 1pl.gen ka-live-lv lig house ‘This is the house where we are going to live in the future.’
As shown in the following table, lexical nouns derived from verbs referring to arguments and verbs in relative or purposive clauses are in the same form as future, habitual or perfective verb forms.
.╅ The ligature ka functions to indicate a kind of modification relationship between the modifier and the head noun. As shown in the following example, it can occur after the demonstrative, numeral, or adjective optionally. (i) a.
hini’ (ka) ngiyaw this lig cat
‘this cat’
b.
to:o’ (ka) ngiyaw three lig cat
‘three cat’
c.
’ima ngangilah-an (ka) tatpo’ ‘red hat’ rel red-lv lig hat
Nominalization in Saisiyat 
Table 2.╇ Lexical nominalization, syntactic nominalization and tense-aspect
Agent Patient Location Instrument
Lexical nominalization
Syntactic nominalization
Temporal/aspectual
ka-ma-V ka-V-en V-inka-V-an Ca~V ka-V
ka-pa-V ka-V-en V-inka-V-an Ca~V ka-V
Habitual, Future Future Perfective Future Future Future
2.4â•… Lexical nominalization vs. syntactic nominalization Given that lexical nominalizations are identical in form to syntactic nominalizations and the future or perfective verb forms, it is tempting to argue that the so-called lexical nominalizations are something like headless relative clauses and that there is in fact no such process as lexical nominalization in Saisiyat. However, the deverbal lexical nouns exhibit syntactic properties of nouns, including demonstrative, case, number and possessive marking.19 First of all, as shown below, the derived lexical nouns can be preceded by a demonstrative or a numeral. (17) a.
hini’ ka-i:ba:-en ma’an rim’an ka-i:ba:-en this nmlz-wear-pv 1sg.gen tomorrow nmlz-wear-pv ‘This dress is what I am going to wear tomorrow.’
b. hiza’ ka-ma-manra:an ka-ma-’omalop ka waliSan that nmlz-av-walk nmlz-av-hunt acc boar ‘That man always hunts boars.’ c.
hini’ ’aehae’ ka-i:ba:-en ’okik niSo ka-i:ba:-en this one nmlz-wear-pv neg 2sg.gen nmlz-wear-pv ‘This dress is not for you to wear.’
Secondly, they can co-occur with such case markers as the accusative case marker ka (18a–b), the genitive case marker noka (18c), and the locative case marker ray (18d). (18) a.
mayhael yako ’am matol pak-bazae’ ka ka-ma-kiSka:at future 1sg.nom fut sing.av cau-hear acc nmlz-av-read ‘In the future I will sing for my classmates to listen to.’
.╅ According to Hopper and Thompson (1984: 718), the grammatical attributes of N include determiners, case, number, and gender affix, as well as classifiers, modifiers and morphemes signaling nounhood.
 Marie Mei-li Yeh
b. ’oya’ paskayzaeh ka ka-i:ba:-en no mother make acc nmlz-wear-pv dat
korkoring ka-i:ba:-en child nmlz-wear-pv
‘Mother made clothes for the children to wear.’
c.
yako noka ka~ko:as ko:as ka bokeS 1sg.nom gen nmlz~comb comb acc hair ‘I comb my hair with a comb.’
d. korkoring ma’ saboeh ray ka-kiSka:at-an ’am kiSka:at child also all loc nmlz-read-lv prog read ‘The children are all studying at school.’
Thirdly, they can appear in a possessive construction. (19) a.
hini’ ’an noka baboy a ka-si’ael-en this poss pig lig nmlz-eat-pv ‘This is pigs’ food.’
b. sia ’anso’a=ay ka-ma-manra:an 3sg.nom 2sg.poss=q mnlz-av-walk ‘Is he your husband?’ (Lit: ‘Is he your man?’)
Semantic and phonological evidence also appears to show that they are lexical nouns. In terms of semantics, some derived nouns such as ka-ma-manra:an ‘man’ (< walk) have taken on special and unpredictable meanings, indicating that they are more lexicalized. With regard to phonological evidence, the pronunciation of ka-i:ba:-en ‘clothes’ seems to show that the verb stem and the affix could be fused. The first consonant of the verb i:ba:en ‘to wear’ was a flap, which was lost and has caused a compensatory lengthening of the following vowel or a smooth onset sounding like /h/ (Li 1978; Tsuchida 1964). However, for the words referring to clothes, i.e. nouns derived from lexical nominalization, the pronunciation with the smooth onset [kahiba:en] is not accepted. Instead, ka- and the first vowel undergo coalescence to [kayba:en] in fast speech; only in future forms or purpose function is the pronunciation with smooth onset accepted. On the other hand, the verbs in syntactic nominalization such as purpose clauses retain such verbal properties as voice and transitivity.20 As shown in the following examples, verbs in purpose clauses still retain voice distinction.
.╅ According to Comrie and Thompson (2007: 345), verbal properties include tense, aspect, voice, transitivity and negation.
(20) a.
Nominalization in Saisiyat 
’oya’ talek ka pazay no korkoring ka-si’ael-en mother cook acc rice dat child nmlz-eat-pv ‘Mother cooked rice for the child to eat.’
b. korkoring hawaeh ’ini baki’ ka-kaS’abo’-an child open dat grandpa nmlz-enter-lv ‘The child opened (the door) for Grandpa to enter.’
Besides, transitivity is retained, as evidenced by the fact that the object is marked as accusative case in the following AV purpose clauses. (21) a.
’oya’ talek ka kashaw ’ima rikrika: ka-pa-ra’oe: ’iSo’on mother cook acc soup rel hot purp-cau-drink 2sg.acc ‘Mother cooked hot soup so as to let you drink.’
b. yako talek ka pazay ka-pa-si’ael ka ’aeh’aehael 1sg.nom cook acc rice purp-cau-eat acc relative ‘I cooked rice so as to feast the relatives.’
Therefore, although lexical nominalization and syntactic nominalization appear to be morphologically identical, they are distinct syntactically. This can explain why the two can co-occur in the same sentence, as in the following examples. (22) a.
hini’ ka-i:ba:-en ma’an rim’an ka-i:ba:-en this nmlz-wear-pv 1sg.gen tomorrow fut-wear-pv ‘This dress is what I am going to wear tomorrow.’
b. yako ’am mari’ ka ha~hiwa: ha~hiwa: ka ’ayam 1sg.nom fut take acc nmlz~cut purp~cut acc meat ‘I am going to take a knife to cut the meat.’
In (22a), the first ka-i:ba:-en is a derived noun referring to clothes, whereas the second one is a syntactic nominalization indicating purpose. In a similar vein, the first ha-hiwa: in (22b) is a derived noun referring to the instrument for cutting, whereas the second ha-hiwa: is a syntactic nominalization denoting the purpose of the event.
3.â•… F uturity/perfectivity, lexical nominalization and syntactic nominalization Now let us proceed to the relationship between lexical nominalization, syntactic nominalization (purpose subordination and modification) and the future or perfective verb form. For ease of discussion, the functions of Ca-, ka- and -in- are summarized in Table 3 below.
 Marie Mei-li Yeh
Table 3.╇ Functions of Saisiyat nominalizers Verb form
Argument nouns
Syntactic nominalization
Ca-/ka-
IV future
instrumental nouns
ka-
PV, LV future
-in-
perfective
agent, patient, location nouns patient nouns
relative or purpose clause modifying instrumental nouns relative or purpose clause modifying agent, patient & location nouns relative clause modifying patient nouns
As shown in Table 3, in terms of temporal or aspectual marking, ka- and Camanifest the same function. They both mark futurity – while Ca- marks futurity of instrument voice (IV), ka- marks futurity of patient voice (PV) or locative voice (LV). The difference lies in the fact that there are overt voice markers in PV and LV constructions, whereas no overt voice marking appears in IV constructions. In the following, the relationship between Ca- and ka- will be addressed, followed with the discussion on the relationship between futurity and syntactic as well as lexical nominalization.
3.1â•… Functions and semantic development of Ca- and kaComparing Saisiyat with other Formosan languages, we find that ka- appears to correspond functionally to Ca- reduplication. For example, as pointed out in Zeitoun et al. (1996: 37) and Huang (2001: 56–57), non-actor voice (NAV) irrealis in Mayrinax Atayal is marked by Ca- reduplication, as shown below. Note that as in Saisiyat, verbs in IV constructions bear no extra surface voice marker. Mayrinax Atayal (23) a.
(Zeitoun et al. 1996: 37)
ta~tuting-un=cu nku’ ’ulaqi’ fut~beat-pv=1sg.nom gen child ‘The child will beat me.’
b. ha~hihip-an ni’ yumin ’i’ limuy fut~kiss-lv gen Yumin nom Limuy ‘Yumin will kiss Limuy.’ c.
ba~Ø-baiq nku’ nabakis ku’ xuil ’i’ limuy21 fut~iv-give gen old.man nom dog acc Limuy ‘The old man will give the dog to Limuy.’
.╅ Note that IV marker si- is multifunctional: besides instrumental and beneficiary arguments, it also marks the theme argument of transportation verbs and is thus dubbed under different terms such as accessory focus (Starosta, Pawley & Reid 1982) and circumstantial focus (Huang 1993; Ross 1995).
Nominalization in Saisiyat 
Nanwang Puyuma, Central Amis (Zeitoun et al. 1996: 43) and Pazeh (Li 2000: 94) also make use of Ca- reduplication to mark futurity or irrealis. Not only is the correspondence found in the expression of futurity or irrealis, but it is also found in nominalizations. For example, in Mayranax Atayal, patient, location and instrumental nouns are formed by Ca~V-un, Ca~V-an, and Ca~V respectively, while in Saisiyat they are formed by ka-V-en, ka-V-an, and ka-V respectively. Take a look at the table below. Table 4.╇ Mayrinax Atayal Ca- vs. Saisiyat kaNouns
Mayrinax Atayal (Huang 2002)
Saisiyat
Form
Example
Form
Example
Patient
Ca~V-un
ka-V-en
Locative
Ca~V-an
Instrumental
Ca~V
na~niq-un ‘food’ Ca~eat-pv na-niq-an ‘restaurant’ Ca~eat-lv pa~patiq ‘pen’ Ca~write
ka-si’ael-en ‘food’ ka-eat-pv ka-si’ael-an ‘restaurant’ ka-eat-lv ha~hiwa: ‘a saw’ Ca~to saw ka-bobo: ‘a fan’ ka-to fan
ka-V-an Ca~V ka-V
In Paiwan and Pazeh, Ca- is also found to co-occur with patient voice (PV) suffix -en and locative voice (LV) suffix -an to derive object and location nouns, as shown in the examples below. Paiwan (24) a.
ka~kan-en Ca~eat-pv
(Chang 2000: 77; Tseng 2003: 59) ‘food; something to be eaten’
b. ka~kesa-an ‘kitchen’ Ca~cook-lv Pazeh (25) a.
(Lin 2000: 75–76, Li & Tsuchida 2001: 96)
ka~kita-en ‘something to be looked at’ Ca~look-pv
b. da~depex-an ‘school’ Ca~read-lv
Thus, it seems clear that Saisiyat ka- is a functional equivalent of Ca- in other Formosan languages, the function of which as a futurity or irrealis marker (Zeitoun et al. 1996) or as a derivational affix to derive instrumental nouns (Blust 1998) has been acknowledged across Formosan languages. Based on the functional identity shared by IV ka- and the ka- in other voice constructions and the fact the morpheme ka- appears to be the functional equivalent of Ca- in other Formosan languages, we will take the two as arising from the same form. Now what is the relationship between the three functions – expressing temporal or aspectual information, marking verbs in background or subordinate clauses as well
 Marie Mei-li Yeh
as deriving instrumental nouns – manifested by Ca- or ka-? In Yeh (2003), it is proposed that the use of Ca- is extended from the function to mark futurity motivated by the goal-orientation property, and the function of deriving instrumental nouns is a further extension motivated by a metonymic principle naming an instrument by its function or purpose. The relationship between the three functions of Ca- can be represented as the figure below. Future verb form
GOAL
Purpose or relative clause
METONYM
Instrumental nouns
Figure 1.╇ Semantic extension of Ca-/ka-
Under this analysis comes the question: If the ka- in other voice constructions is a variant of Ca-, which is the future IV marker, then how can we account for its co-occurrence with the other voice markers? It is postulated that the function of ka- as a purpose marker has been grammaticalized and therefore can be extended to other voice constructions. The following dialogue seems to show that the function of ka- is to denote purpose. (26) A: yao ’am talek ka ralom 1sg.nom fut cook acc water ‘I am going to boil some water.’ B: ’am-powa’ fut -do ‘What for?’ A: ka-pa-pranaw ka korkoring purp-cau-bathe acc child ‘For the child to take a bath.’
The purposive function of ka- also appears to be revealed in the coinage of words referring to new innovations. For example, for a new word referring to a coffee mug, which profiles the function of the cup, our Saisiyat informants resort to ka-. In contrast, to create a word to name a paper cup, which features in the material, the perfective marker -in- is used. Compare: (27) a.
ka-rae’oe: ka kafei talobon purp-drink acc coffee cup ‘cup for drinking coffee; coffee cup’
b. noka zinowas paskayzaeh talobon gen paper make cup ‘cup made from paper; paper cup’
However, if what has been grammaticalized is the purposive function of ka-, then how can we explain the fact that it can also mark futurity in other voice constructions?
Nominalization in Saisiyat 
Hopper and Traugott (1993: 1) point out that the English auxiliary for immediate future ‘be going to’ is developed from the motion verb ‘go’ in a specific context – purposive directional construction, which induces an inference from purposive to futurity. Consider: English
(Hopper & Traugott 1993)
(28) a. I am going to London to marry Bill. (Purposive directional construction) b. I am going to marry Bill. (Immediate future)
Therefore, it is possible that the future function of ka- in other voice constructions is developed from the purposive function since purpose is a goal projected in front of us. As for lexical nominalization, as is the case in instrumental voice (IV) constructions, it is developed from the purposive function by the metonymic principle naming something on the basis of its purpose, with the meaning of argument (agent/actor, patient or location) marked by the voice affixes. Take the expression ka-si’ael-en for example. In the following sentence, it is subordinated to another event to express the purpose of a foregrounding event. (29) ’oya’ talek ka ’aelaw no korkoring ka-si’ael-en mother cook acc fish dat child purp-eat-pv ‘Mother cooked fish for the child to eat it.’
However, when used alone to predicate the subject as shown below, we get the meaning ‘X is for the purpose of being V-ed by someone’, which is irrealis because purpose is usually the goal projected ahead, so the meaning ‘X is what is going to be V-ed by someone’ comes about. If the agent, which is marked as genitive case, is absent or is interpreted as the possessor instead of the agent, we get ‘X is something to be eaten’ or by metonymic extension, ‘X is someone’s food.’ (30)
hini’ ma’an ka-si’ael-en22 this 1sg.gen ka-eat-pv ‘This is for the purpose of my eating.’ ‘This is what is going to be eaten by me.’ ‘This is my food.’
.╅ Note that the case marking in a purpose clause and that in the predicate is also different: in a purpose clause, the agent is marked by dative case markers, but when functioning as the predicate, it is marked by genitive case markers. Compare: (ii) a.
’oya’ talek ka tatimae’ ’iniman ka-si’ael-en mother cook acc dishes 1sg.dat purp-eat-pv ‘Mother cooked dishes for me to eat.’
b.
hini’ ma’an ka-si’ael-en this 1sg.gen nmlz-eat-pv ‘This is for me to eat.’
 Marie Mei-li Yeh
Applying this development to LV and AV, we get the following developments. (31) a. X ka-V-an
‘X is a location for the purpose of V-ing’ → X will be the location where people V → X is the location to V
b. X ka-ma-V ‘X is a person for the purpose of V-ing’ → X is a person who Vs by profession; X is a V-er
Note that informants usually translate ka-ma-V as someone specialized in V-ing. If kais taken to denote a purpose or a goal, with ma-, the AV affix highlighting the person who performs the action or event, then ka-ma-V can be interpreted as ‘someone who is for the purpose of V-ing’. As a result, the implication ‘someone who specializes in V-ing’ or ‘someone who Vs by profession’ comes naturally.23 (32) a.
ka-ma-tortoroe’ nmlz-av-teach
b. ka-ma-matawaw nmlz-av-work
‘teacher’ ‘worker’
To sum up, the semantic extension of Ca-/ka- can be represented by the following figure. Ca-
Purpose
Nominalization
ka-
Purpose
Nominalization
Future IV
Grammaticalization Purposive to future Future
AV, PV and LV Metonymic extension
Metonymic extension Nominalization
Figure 2.╇ Semantic extension of Ca-/ka-
Notice that Ca- reduplication does not serve as pure instrumental voice marking. Rather, it marks both instrumental voice and future. It is this portmanteau property
.â•… Many agentive nouns in Mandarin Chinese also come from verbs by the same metonymic extension. For example, the word jian-jun ‘general’ comes from the verb phrase ‘command the army’, and jia-shi ‘driver’ from ‘drive’ with both jia meaning ‘drive’ and shi meaning ‘sail’. Other examples include: zhang-gui (manage-reception desk) ‘cashier’, shou-wei (to guard-protect) ‘guard’. I owe Prof. Dylan Tsai (2002, p.c.) for directing my attention to this phenomenon in Mandarin Chinese.
Nominalization in Saisiyat 
that makes possible its use in purposive construction. This function in the variant kahas undergone grammaticalization and extended to be used in the other type of constructions. Under the mechanism of metaphoric extension from purposive to future, ka- is extended to mark future in patient voice and locative voice construction.
3.2â•… Functions and semantic development of -inNow, let us turn to the other nominalizer -in-, which also manifests at least three functions – marking verbs in a special temporal or aspectual frame, marking verbs in syntactic nominalization, and deriving patient nouns from verbs. Given that -in- is polyfunctional, it is possible for -in- marked verbs to be ambiguous, as shown below. (33) a.
ma’an ka:at ha’ino’= ila 1sg.gen write where=incpt ‘Where is my book?’
b. hini’ ni hi:a’ ka:at? this gen who write ‘By whom is this (book) written?’
In the first example (33a), V-in- refers to an entity ‘book’ or ‘something that has been written’, but in the second one (33b) it denotes an event.24 In the following discussion, the functions of -in- will be introduced first, and then efforts will be made to figure out the relationship between different usages manifested by -in-.
3.2.1â•… Functions of -inFirst, as mentioned in Section 1, the infix -in- functions as a nominalizer to derive patient nouns, as shown in (34). (34) a.
tnawbon pound.rice
‘rice cake (made from pounding rice)’
b. po-bae:aeh ‘beer (made by brewing over charcoals)’ put-charcoal c.
’alingo’ take.picture
‘picture’
.╅ The case is somewhat like that of -en in English. On the one hand, the English -en serves an aspectual/voice marker, that is, an inflectional affix; on the other hand, it can be used as a derivational affix to derive adjectives from verbs. English (iii) a. The money was stolen by the banker. b. The stolen money was found yesterday.
(Haspelmath 1994: 151)
 Marie Mei-li Yeh
As noted by many linguists (e.g. Blust 1998: 52), -in- also functions as an aspectual marker indicating perfectivity. Consider:25 (35) a.
’obay ra’oe:=ila ka ralom ’obay drink=incpt acc water ‘’obay has drunk water.’
b. yako kahi:a’ moSa’ ila hiza 1sg.nom yesterday go.av to there ‘I went there yesterday.’
As shown in (36), -in- does not co-occur with any overt patient voice (PV) marker. Therefore, it is analyzed as displaying a portmanteau function in PV constructions (c.f. Huang 1995: 49 & 151, Blust 1998: 52).26 (36) a.
kayba:en bahi’=ila clothes wash=incpt ‘The clothes have been washed.’
b. tatimae’ talek=ila ni ’oya’ dish cook=incpt gen mother ‘The dishes have already been cooked by Mother.’
In the following sentences, -in- does not imply a perfective meaning. Instead, it marks the verbs in the subordinate clause indicating ‘in the state of ’. (37) a.
baki’ maywawa:a’ si’ael grandpa lie.av eat ‘Grandpa ate lying.’
b. yami si’ael ka pazay ma-sa:eng 1pl.excl.nom eat acc rice av-sit ‘We ate rice sitting.’
In addition to marking the verbs in subordinate clauses, -in- can also mark the verbs in a relative clause. Consider: (38) a.
ni ’oya’ talek ’alaw kayzaeh si’ael-en gen mother cook fish good eat-pv ‘The fish cooked by Mother is delicious.’
.â•… As shown in the two examples, ila in Saisiyat is multifunctional. It can function as a motion verb, a case marker indicating destination/goal, or an aspect marker indicating the inception of a state. For the grammaticalization of ila, we refer you to Lu & Su (2003). When co-occurring with V-in, a sense of relevant to present is rendered. .â•… Note that in Huang’s later work (2001: 56–57), it is analyzed as the perfective aspect marker.
Nominalization in Saisiyat 
b. tibabih ka hiza’ b otoe’ katin untie acc that tie cow ‘Untie that cow which has been tied.’
Consequently, as in the case of Ca- or ka-, three major functions are manifested by -in-: a special verb form, syntactic nominalization and lexical nominalization. Only that in the case of -in-, the verb form is the perfective instead of the future verb form. Now, what is the relationship between the perfective verb form and syntactical as well as lexical nominalizations? How can we establish the relationship between different functions of -in-? As mentioned earlier, it is possible for -in- to mark perfective or to mark perfective and patient voice simultaneously. Linguists appear to hold different views as to whether -in- functions primarily as an aspectual marker or as a portmanteau indicating voice and aspect. For the portmanteau analysis, Starosta, Pawley and Reid (1982) argues that pan *-in- was a perfective PV marker denoting ‘the N affected by V-ing’, which was in complementary distribution with the PV *-un denoting ‘the N to be V-ed’. As it developed into an aspectual marker denoting perfectivity, its use began extending to the other voice constructions. Ross (1995: 756) also treats V-in- as a portmanteau, holding that PV perfective form meaning ‘something V-ed’ in contrast to the PV neutral V-en meaning ‘something to be V-ed’.27 For the aspect marker analysis, Blust (1998: 52) argues that pan *-in- “functioned primarily to signal past tense or perfective aspect”, and Huang (2001: 56), based on the fact that -in- can co-occur with other voice markers, also argues that the analysis as an aspect marker is more plausible. It is noteworthy that the use of an identical form to code perfective aspect and passive voice is not unique of Formosan or Austronesian languages. In English, past participle -en also has two uses – perfective aspect (39a) and passive voice (39b).28 English29 (39) a. He’d stolen the flowers from our garden. b. Would it make a difference if it could be maintained that Laura had died before the Tongue was stolen?
Haspelmath (1994), observing the orientation as well as tense/aspect asymmetry between active and passive participles across languages, proposes that passive
.╅ Note that the term neutral refers to the finite indicative form that is not marked by any tense or aspect (Ross 1995: 742). .╅ According to Payne (1997: 227), English -ee marker in patient nominalizations comes from the French past participle. .╅ The two examples are from The LDOCE (Longman Dictionary of Contemporary English) On Line (http://www.ldoceonline.com/).
 Marie Mei-li Yeh
participles in English and many other languages evolve from resultative, which is semantically more natural. This argument is in line with the proposal in Bybee et al. (1994) that past or perfective developed from resultative.30 In the following discussion, the possibility of -in- being resultative will be examined.
3.2.2â•… Infix -in- as resultative Bybee et al. (1994) points out that perfective and resultative are similar to completive, anterior, and simple past conceptually – they all describe a situation that is completed prior to a certain referent time.31 On the basis of crosslinguistic evidence, they propose that the path of development is from resultative or completive to anterior and then to perfective or simple past. Yeh (1991) and Yeh (1995) observe that besides perfectivity, -in- exhibits other similar functions such as designating relative past or a resultant state. Then, it is legitimate for one to raise the possibility that the perfective use of -inarises from resultative. Fist of all, Yeh (1991: 81) remarks that the Saisiyat infix -in- can indicate relative past by marking the first verb in a sequence of events, as shown in (40). Note that this function corresponds to anterior in Bybee et al. (1994: 54). (40) a.
yako potngor ray taew’an korkoring ma’rem 1sg.nom arrive.av loc house child sleep.av ‘When I arrived home, the child has already fallen asleep.’
b. yako So: rim’an si’ael=ila 1sg.nom when tomorrow eat = incpt
yako ’am m-wa:i’=ila 1sg.nom fut av-come=incpt
‘Tomorrow I will come after eating.’
Additionally, Yeh (1995: 379) observes that -in- can also denote a state, including the resultant state of a previous event, as shown below. (41) a.
katesnenan hawaeh door open ‘The door is open.’
.â•… Slobin (1994) points out that parallel development can be found in child language and that crosslinguistic researches show that “results” are particularly salient to 1- and 2-year-olds. .â•… The following definitions of these aspects from Bybee et al. (1994: 54) are adopted in this paper. Anterior: situations occur prior to reference time and is relevant to the situation at reference time Resultative: a state exist as result of past action Perfective: situations viewed as bounded temporally
Nominalization in Saisiyat 
b. ma’an katesnenan Si-kon-hawaeh 1sg.gen door IV-push-open ‘I pushed the door open.’
Summing up what is mentioned above, the infix -in- can mark perfective aspect, patient voice,32 anterior, and resultative. Following Bybee et al. (1994) the relationship between these functions can be established as Figure 3:
RESULTATIVE
ANTERIOR
PERFECTIVE
PATIENT OREINTED Figure 3.╇ Extension of -in- from resultative to perfective
According to Bybee et al. (1994: 69 & 86), both the change from resultative to anterior and that from anterior to perfective or past involve a generalization of meaning or a loss of specificity triggered by discourse function. Anterior sense evolves when the discourse function of resultative – setting the stage for a subsequent action – is extended from situations where actions produce states to situations where actions precede other actions.33 For past or perfective, overuse of anterior on the part of the speaker to frame the events as highly relevant to current time weakens the force of the current relevance; eventually only past or perfective is inferred by the hearer. As for Patient Voice marking, Bybee et al. (1994: 54) remarks that resultative is similar to passive since it often takes patient as the subject. Haspelmath (1994: 164) also points out that resultative is oriented toward the patient of events. The difference lies in the fact that resultative can apply to intransitive verbs without a change of subject, as in He is gone. Now, let us turn to the function of -in- in nominalizations. Note that both the definition of resultative as “a present state exists as the result of a previous action” and
.╅ Notice that in this paper, following what is generally accepted in Austronesian linguistics, patient voice is used instead of passive voice. It is pointed out in many studies (e.g. Shibatani 1988) that patient focus (patient voice) constructions in Austronesian languages are not the same as passive voice in accusative-type languages. Nevertheless, what concerns here should be patient-oriented and therefore the difference can be overlooked in this aspect. .╅ Slobin (1994) discusses the discourse origins of a similar development in present perfect.
 Marie Mei-li Yeh
the function in discourse to set the stage for subsequent actions (Bybee et al. 1994: 69) imply the possibility of occurrence with another action. As the meaning of “resultant state’ is generalized to state in general, it becomes possible for the resultative -in- to assume the function of marking the verb in a subordinate clause, denoting ‘in the state of ’. By extending the background function further from setting the stage for an event to grounding an NP34 motivated by the semantic extension taking property as a kind of state, the use of -in- to mark the verbs in a relative clause comes about. As for the function of -in- as a nominalizer to derive patient nouns, it can be regarded as a grammaticalization of headless relative clause by the metonymic extension using the property of an entity to refer to the entity. Figure 4 below summarizes the semantic development of -in- from resultative to nominalizations.
Resultative
Subordination (in the state of)
Semantic generalization
Syntactic nominalization
Property as state Modification (relative clause) Metonymy
Lexical nominalization
Derived nouns Figure 4.╇ Semantic extension of -in-
4.â•… Conclusion The semantic developments presented above for Ca-/ka- and for -in- indicate a direction from main verbs (future or perfective verb forms) to subordination (purpose clauses) or modification (relative clauses) to lexical nominalization. Comrie (1976), based on a crosslinguistic study on action nominals, concludes that different types of nominalization exist as a continuum. Syntactic and lexical nominalizations in Saisiyat also appear to exist as a continuum along the axis of decategorialization. According to Hopper and Thompson (1984: 709), categoriality is the property of being a prototypical instance of the grammatical category N or V. It is imposed on linguistic forms by discourse – while prototypical N’s introduce a participant, prototypical V’s report an
.â•… Fox and Thompson (1990: 300) remarks that: “relative clauses both provide one kind of grounding and interact strongly with other grounding devices.”
Nominalization in Saisiyat 
actual event of the discourse. In typical environments, N and V display the widest degree of morphosyntactic contrast, whereas in non-typical environments, the contrast is usually neutralized. Take for example the V in a purpose clause such as ‘People study in order to learn’ (Hopper & Thompson 1985: 175). It is low in categoriality since it reports no event, and therefore, in most languages it lacks tense/aspect and person morphology. As syntactic and lexical nominalizations in Saisiyat appear to exhibit different degrees of categoriality – with syntactic nominalization exhibiting more, and lexical nominalization less properties of V’s – they can be said to represent different stages of decategorialization, as shown in the following figure. Decategorialization: Verb
Syntactic nominalization Lexical nominalization, Nouns
Discourse function: Event-reporting background
participant-introducing
Figure 5.╇ Relationships between decategorialization and discourse function
To be specific, nominalization in Saisiyat can be taken as a kind of decategorialization triggered by discourse functions with different types of nominalization representing its different stages. Syntactic nominalization – including purpose and relative clauses functioning as background, are low in categoriality and therefore lie in the middle stage of the continuum, and lexical nominalization, with its derived nouns functioning to introduce an argument or a participant in the discourse, represents the cases that are pushed further towards the end of decategorialization. What has been advanced here is an alternative analysis from the perspective of categoriality where lexical nominalizations are regarded as syntactic nominalizations lexicalized. Shibatani and Awadh (2009) have proposed a similar analysis of lexical nominalization in Soqotri, a South Arabian language of Yemen. However, the semantic development proposed here seems to run in the opposite direction from the development of focus morphology from lexical nominalization to voice marking proposed by Starosta, Pawley and Reid (1982) and Ross (1995). Their nominalization-to-voice analysis is based on two facts. First of all, the Actor in a NAV construction is marked by genitive case markers. Secondly, in Formosan and Philippine languages these voice affixes are also used to form nouns (Ross 1995). For example, the neutral forms of different voice forms in Paiwan can be used to form nominalizations: k-em-an ‘eater, someone who eats,’ kan-en ‘food, something to be eaten,’ kan-an ‘place where one eats,’ and si-kan ‘eating utensil, something to eat with’ (Ross 1995: 752). However, in Saisiyat, it is not the neutral voice forms that are involved in nominalization. Rather, what appears in Saisiyat nominalizations are the future and perfective verb forms. Thus, the voice as nominal deriving analysis does not seem to be applicable to Saisiyat.
 Marie Mei-li Yeh
References Blust, Robert. A. 1998. Ca- reduplication and Proto-Austronesian grammar. Oceanic Linguistics 37(1): 29–64. Blust, Robert. A. 1999. Subgrouping, circularity and extinction: Some issues in Austronesian comparative linguistics. In Selected Papers from the Eighth International Conference on Austronesian Linguistics [Symposium Series of the Institute of Linguistics (Preparatory Office), Academia Sinica, No. 1], Elizabeth Zeitoun & Paul Jen-kuei Li (eds), 31–94. Taipei: Academia Sinica. Blust, Robert. A. 2003. Three notes on early Austronesian morphology. Oceanic Linguistics 42(2): 438–478. Bybee, Joan, Perkins, Revere & Pagliuca, William. 1994. The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago IL: University of Chicago Press. Chomsky, Noam. 1970. Remarks on nominalization. In Readings in English Transformational Grammar, Roderick. A. Jacobs & Peter S. Rosenbaum (eds), 184–221. Waltham MA: Ginn and Co. Chang, Henry Yungli. 2010. On the syntax of adverbial verbs in Formosan languages. In Austronesian and Theoretical Linguistics [Linguistik Aktuell/Linguistics Today 167], Raphael Â�Mercado, Eric Potsdam & Lisa de Mena Travis (eds), 183–208. Amsterdam: John Benjamins. Chang, Hsiou-chuan. 2000. Paiwan Reference Grammar. Taipei: Yuan Liu. (In Chinese). Comrie, Bernard. 1976. The syntax of action nominals: A cross-linguistic study. Lingua 40: 177–201. Comrie, Bernard & Thompson, Sandra A. 2007. Lexical nominalization. In Language Typology and Syntactic Description, Vol. III: Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 334–381. Cambridge: CUP. Fox, Barbara A. & Thompson, Sandra A. 1990. A discourse explanation of the grammar of relative clauses in English conversation. Language 66: 297–316. Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1991. Grammaticalization: A Conceptual Framework. Chicago IL: Chicago University Press. Haspelmath, Martin. 1994. Passive participles across languages. In Voice: Form and Function [Typological Studies in Language 27], Barbara A. Fox & Paul J. Hopper (eds), 151–177. Amsterdam: John Benjamins. Hopper, Paul & Thompson, Sandra A. 1984. Lexical categories in universal grammar. Language 60(4): 703–752. Hopper, Paul & Thompson, Sandra A. 1985. The iconicity of the universal categories ‘noun’ and ‘verbs’. In Iconicity in Syntax [Typological Studies in Language 6], John Haiman (ed.), 151–186. Amsterdam: John Benjamins. Hopper, Paul & Traugott, Elizabeth C. 1993. Grammaticalization. Cambridge: CUP. Huang, Lillian M. 1993. A Study of Atayal Syntax. Taipei: Crane. Huang, Lillian M. 1995. A Study of Mayrinax Atayal Syntax. Taipei: Crane. Huang, Lillian M. 2001. Focus system of Mayrinax Atayal: A syntactic, semantic and pragmatic perspective. Journal of Taiwan Normal University Humanities and Social Science 46(1–2): 51–69. Huang, Lillian M. 2002. Nominalization in Mayrinax Atayal. Nominalization in Formosan Languages [Special Issue of Language and Linguistics 3(2)], 197–226. Taipei: Academia Sinica.
Nominalization in Saisiyat 
Li, Paul J.-K. 1978. A comparative vocabulary of Saisiyat dialects. Bulletin of the Institute of History and Philology 49(2): 133–199. Taipei: Academia Sinica. Li, Paul J.-K. 1985. Linguistic criteria for classifying the Ataylic dialect groups. Bulletin of the Institute of History and Philology 56(4): 699–718. Taipei: Academia Sinica. Li, Paul J.-K. 1997. The Ethnic Group and Migration of Formosan People. Taipei: Chang-ming Culture Company. (In Chinese). Li, Paul J.-K. 2002. Nominalization in Pazih. Nominalization in Formosan Languages. [Special Issue of Language and Linguistics 3(2)], 227–240. Taipei: Academia Sinica. Li Paul Jen-kuei & Shigeru, Tsuchida. 2001. Pazeh Dictionary [Language and Linguistics Monograph Series Number A2]. Taipei: Institute of Linguistics, Preparatory Office, Academia Sinica. Lin, Yin-jin. 2000. The Pazeh Language. Taipei: Yuanliu. (In Chinese). Lu, Louis Wei-lun & Su, Lily I-wen. 2003. From main verb, verbal aspect to current relevant state: a case study on the grammaticalization of Saisiyat polysemous ila. Paper presented at The 2nd Workshop of Formosan Languages: on the Notion of Verb. Institute of Linguistics, Preparatory Office, Academia Sinica, November 1–2. Lyons, John. 1977. Semantics. Vol. 1 & 2. Cambridge: CUP. Ogawa, Naoyoshi & Asia, Erin. 1935. Myths and Traditions of the Formosan Native Tribes. Taipei: Taihoku Imperial University. Payne, Thomas E. 1997. Describing Morphosyntax: A Guide for Field Linguistics. Cambridge: CUP. Rau, Victoria D. 2002. Nominalization in Yami. Nominalization in Formosan Languages. [Special Issue of Language and Linguistics 3(2)], 165–196. Taipei: Academia Sinica. Ross, Malcom D. 1995. Reconstructing Proto-Austronesian verbal morphology: Evidence from Taiwan. In Austronesian Studies Relating to Taiwan [Symposium Series of the Institute of History and Philology, Academia Sinica, Number 3], Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho & Chiu-yu Tseng (eds), 727–791. Taipei: Academia Sinica. Ross, Malcom D. 2002. The history and transitivity of Western Austronesian voice and voice marking. In The History and Typology of Western Austronesian Voice Systems, Fay Wouk & Malcolm Ross (eds), 17–62, Canberra: Pacific Linguistics. Shibatani, Masayoshi. 2008. Relativization in Sasak and Sumbawa, Eastern Indonesia. Language and Linguistics 9(4): 865–916. Shibatani, Masayoshi & Khaled Awadh bin Makhashen. 2009. Nominalization in Soqotri, a South Arabian language of Yemen. In The Linguistics of Endangered Languages – Contributions to Morphology and Morpho-Syntax, W. Leo Wetzels (ed.), 9–31. Leiden: Brill. Slobin, Dan I. 1994. Talking perfectly: Discourse origins of the present perfect. In Perfective on Grammaticalization, William Pacgliuca (ed.), 119–133. Amsterdam: John Benjamins. Starosta, Stanley, Pawley, Andrew K. & Reid. Lawrence A. 1982. The evolution of focus in Austronesian. In Papers from the Third International Conference on Austronesian Linguistics, Vol. 2: Tracking the Travelers, Amran Halim, Lois Carrington & Stephen A. Wurm (eds), 145–170. Canberra: Pacific Linguistics C-75. Tseng, Meylysa. 2003. Reduplication as Affixation in Paiwan. MA thesis, National Chung Cheng University, Taiwan. Tsuchida, Shigeru. 1964. Preliminary report on Saisiyat: phonology. Gengo Kenkyu 46: 42–52. Wolff, John U. 1973. Verbal inflection in Proto-Austronesian. In Parangal kay Cecilio Lopez, Andrew Gonzalez (ed.), 71–91. Quezon City: Linguistic Society of the Philippines.
 Marie Mei-li Yeh Yeh, Marie M. 1991. Saisiyat Structure. MA thesis, Tsing Hua University, Hsinchu, Taiwan. Yeh, Marie M. 1995. A preliminary report on tense and aspect in Saisiyat. In Papers in Formosan Languages, 369–384. Taipei: Education Research Committee, Ministry of Education. (In Chinese). Yeh, Marie M. 2003. A Syntactic and Semantic Study of Saisiyat Verbs. Ph.D. dissertation, National Taiwan Normal University. Yeh, Marie M. 2006. The grammaticalization of ’am in Saisiyat. In Streams Converging Into An Ocean: Festschrift in Honor of Professor Paul Jen-huei Ki on His 70th Birthday, [Language and Linguistics Monograph Series Number W-5], Henry Y. Chang, Lillian M. Huang & Dah-an Ho (eds), 393–413. Taipei: Institute of Linguistics, Academia Sinica. Yeh, Marie M. 2009. Ca-reduplication in Formosan languages. Grazer Linguistische Studien 71: 135–156. Zeitoun, Elizabeth. 2002. Nominalization in Mantauran (Rukai). Nominalization in Formosan Languages [Special Issue of Language and Linguistics 3(2)], 241–282. Taipei: Academia Sinica. Zeitoun, Elizabeth, Huang, Lillian M., Yeh Marie M., Chang, Anna H. & Wu, Joy J. 1996. The temporal, aspectual and modal systems of some Formosan languages: A typological perspective. Oceanic Linguistic 35: 21–52.
Rise and fall of referentiality Articles in Philippine languages* Naonori Nagaya Rice University
In this paper, we look into expansion and spread of non-referential uses in articles in Philippine languages with special attention to two competing articles, ang and yung, in Tagalog. There are four claims we make in this paper. First, the functions of ang-marked nominals have been expanding from referential to non-referential uses. Second, yung is emerging as an article for referential nominals, taking over and expanding the referential uses of ang. Third, there is an emerging semantic distinction between ang and yung in spoken discourse: ang is for non-referential uses and yung for referential ones. Lastly, in other Philippine languages, as well, there is evidence of an existing article obtaining non-referential uses and a newer article emerging for referential nominals.
1.â•… Articles and referentiality The prototypical function of nominals, namely nouns and nominalized elements, is to refer to an entity, and that of articles is to ground nominals into a discourse, specifying a type of reference such as definiteness or specificity (cf. Hopper & Thompson 1984; Croft 1991, 2001). However, it often happens that the function of nominals goes beyond reference and expands to non-referential functions, and that articles also accommodate the functional change of nominals. For instance, the English indefinite article a/an was once employed only for already identified particulars, as in a (certain)
*I thank Laura Rabalais, Masayoshi Shibatani, Hiroto Uchihara, Foong Ha Yap and three anonymous reviewers for their valuable and insightful comments on drafts of the paper. I am also grateful to Farah Cunanan, Kathleen Diaz, Pi Ramos, Mark Felix Albert Santiago and Gilbert Tinio for sharing their precious native speakers’ intuition of Tagalog with me, to Rossana Santa Sy and Maria Christina Sy for Partido Bikol and to Lorie Sibolboro for Ilokano. Of course, responsibility for any errors is purely my own. The research presented here was partially presented as Nagaya (2007b, c, ms.) and supported in part by the National Science Foundation grant for the project “Austronesian voice systems: an eastern Indonesian perspective” (BCS-0617198) headed by Masayoshi Shibatani.
 Naonori Nagaya
man, but later comes to indicate non-specific, unidentified entities, as in we called a taxi (Hopper & Martin 1987). Kaufman (this volume), Potsdam (this volume), and Yap (this volume) also point out that in Austronesian languages nominals can be employed for a variety of non-referential uses such as exclamatives. With regards to this topic, articles in Tagalog and other Philippine languages are worthy of attention. They introduce various elements as referential expressions. In (1), for example, the Tagalog article ang marks a noun, an adjective and a clause as a referential expression.1 See (1a), (1b), and (1c) respectively. (1) a.
Dating ang=aso. arrive nom=dog ‘The dog arrived.’
b. Dating ang=ma-ganda. arrive nom=adj-beautiful ‘The beautiful one arrived.’ c.
Dating ang=ka~kanta. arrive nom=asp~sing ‘The one who will sing arrived.’
In contrast, the same article can be employed to create non-referential expressions as well. It leads a sentential adverb of some kind in (2) and an exclamatory expression in (3). Here we can see the polysemy of ang-phrases from referential to non-referential uses. (2) Ang=alam=ko, mag-a~aral si=Icel sa=Japan. nom=know=1sg.gen af-asp~study p.nom=Icel dat=Japan ‘As far as I know (lit. what I know is), Icel will study in Japan.’
.╅ Abbreviations used in this paper are: adj af art asp caus dat exc
adjective actor focus article aspect marker causative dative exclusive
gen grn inc lf lk neg nom
genitive gerund inclusive locative focus linker negation nominative
p pf q rl sg sp
personal name and kinship term patient focus question marker realis singular spontaneous
1-first person, 2-second person, 3-third person, “< >”-infix, “=”-cliticization, and “~”-reduplication. The diagraph ng represents a velar nasal except that the article ng is pronounced as [naŋ] and the plural marker mga as [maŋa].
Rise and fall of referentiality 
(3) Ang=ganda=mo! nom=beautiful=2sg.gen ‘How beautiful you are! (lit. ‘Your beauty.’)’
The aim of this paper is to look into such expansion and spread of non-referential uses in nominals and articles of Philippine languages with special reference to two competing articles ang and yung in Tagalog. There are four main claims we argue in this paper. First, the functions of ang-phrases, i.e. nominals marked by ang, have been expanding from referential to non-referential uses, obtaining more discourse-related and subjective meanings (cf. Traugott 1989). In the literature, it has long been believed that ang-phrases are always referential (Schachter & Otanes 1972; Schachter 1976, 1977, to name a few). By examining ang-phrases construction by construction, however, we argue that they can be used either referentially or non-referentially depending on the particular construction in which they are found (Nagaya 2007b, c, ms.). Second, yung is emerging as an article for referential nominals. The article yung is grammaticalized from the distal demonstrative iyon and is often considered as a replacement for ang in conversations (Reid 1978; Himmelmann 2005b). However, we show that the difference in style is not the end of the story: yung does not have all of the non-referential functions that ang has, but displays nominalizing functions that ang does not demonstrate. Yung is not just a simple renewal form of the functions of ang, but also an addition to the article system in Tagalog. Third, there are two correlated semantic changes going on in the Tagalog article system: the primary function of ang is shifting from referential to non-referential uses, whereas yung is taking over the referential uses of ang. In the spoken corpus we examined, the referential uses of ang appear less frequently than its non-referential uses, and yung is almost always used referentially (Nagaya 2007b, c, ms.). The semantic contrast being formed between ang and yung is in terms of referentiality. Lastly, the same contrast between existing and emerging articles is observed in other Philippine languages as well. We can look at two cases from two Philippine languages: ang and su in Partido Bikol and ti and diay in Ilokano. In these languages, the existing articles ang and ti are acquiring non-referential uses, while the emerging ones su and diay are mainly used for introducing referential nominals. Both elicited examples and naturally-occurring data are examined in this paper, but we mainly focus on elicited examples in order to compare Tagalog with other languages. See Nagaya (2007b, c, ms) for the details of the actual usage of Tagalog ang and yung and their referentiality in a spoken discourse. This paper is organized as follows: we examine the usage of Tagalog ang and yung in Sections 2 and 3 respectively, observing that ang and yung have both referential and non-referential uses. In Section 4, based on Nagaya (2007 b, c, ms.), we make a quantitative analysis of the Tagalog spoken corpus to discuss how often ang and yung are
 Naonori Nagaya
used referentially and how often they are not. In Section 5, we extend our analyses to other Philippine languages. Finally, this paper is concluded in Section 6.
2.â•… Fall of referentiality: The usage of Tagalog ang Tagalog and other Philippine languages are typical head-initial languages. Predicates usually occupy the clause-initial position, and nominals are introduced either by an article or by a linker.2 As shown in Table 1, Tagalog articles carry a variety of nominal information such as case, number, noun class and deictic meaning (Schachter & Otanes 1972; Himmelmann 2004, 2005b, ms.; Reid & Liao 2004). There are three different case forms distinguished by articles: nominative, genitive and dative.3 The genitive case is used for marking not just a possessor but also an agent or patient which is not the primary grammatical argument; the dative case is for a recipient, a goal, a location and other peripheral semantic roles. The usage of the nominative case is discussed in relation to ang in this section. Personal name articles are for personal names and kinship terms (e.g. Kath, Marfeal, nanay ‘mother’ and tatay ‘father’), whereas common nouns are for others, including nominalized constituents. Number distinction is only made in personal name articles. Table 1.╇ Tagalog article system
Personal name
Singular Plural
Common noun Distal Medial Proximal
NOM
GEN
DAT
si sina ang yung yang itong
ni nina ng [naŋ] nung niyang nitong
kay kina sa – – –
.â•… Tagalog articles are also referred to as “construction markers” (Reid 1978), “case markers” (Kroeger 1993), “phrase markers” (Ross 2002), “determiners” (Reid 2002; Reid & Liao 2004), “nominal specifiers” (Reid 2006) and so on. .â•… There is quite a bit of controversy surrounding the case-marking typology of Tagalog. Some insist that Tagalog has an ergative-absolutive case-marking pattern (Cena 1977; Payne 1982; De Guzman 1988; Reid & Liao 2004), and others reject the analysis (Kroeger 1993; Ross 2002; Himmelmann 2005a, b). In this paper, by “nominative case” we simply mean the primary grammatical case, and do not intend to support any specific analysis of the Tagalog case-marking system.
Rise and fall of referentiality 
In this section we are concerned with ang, whose function is to introduce common nouns in the nominative case. From the beginning of Philippine linguistics, angphrases have been analyzed as always referential, that is, to have a definite, indefinite or generic reference (Bloomfield 1917; Blake 1925; Schachter & Otanes 1972; McFarland 1976, 1978; Schachter 1976, 1977; Himmelmann 1991, 2004, 2005b among others). For example, Schachter and Otanes (1972: 96) say “[a]n ang-plus-unmarked-noun topic [read “ang-phrase,” NN] normally expresses either a DEFINITE meaning, ‘an identified member of the class’ (corresponding to ‘the’ plus noun in English) or a generic meaning,” Cena (1977) claims “[t]he topic [“ang-phrase,” NN] is absolutely referential,” Kroeger (1993: 14) notes “[t]he nominative argument [“ang-phrase,” NN] is normally interpreted as being definite […], but can also be generic […],” and Himmelmann (1991: 15) concludes “ang-phrases are always referential, but they may involve all kinds of referentiality (definite, indefinite, generic […]).”. Although there is no immediate evidence that the original meaning of ang is referential, its reconstructed form suggests that this is the case. Reid (1978, 2002) argues that Tagalog “ang [aŋ] NP” is historically derived from
(4) *ôi
na-ŋ
np
in Proto-Philippines: *ôi is the reconstructed form of the article for nominative common nouns,*na ‘that’ is a demonstrative, and *-ŋ is the linking element which relates the demonstrative to the following NP. Since demonstratives usually have a specific referent in the physical surrounding environment, it is natural to assume that the original function of ang was referential. Indeed, it is well-known that definite articles are almost always derived from demonstratives (e.g. þæs > the in English) (Greenberg 1978; Diessel 1999). In sum, it is believed that ang-phrases are always referential in modern Tagalog and the reconstructed form suggests that its original function was referential too. In this section, however, by carefully examining the usage of ang in a variety of constructions, we argue that the hypothesis that ang-phrases are always referential is not true in modern Tagalog; but rather ang-phrases can be either referential or non-referential depending on the particular construction in which they are used. Before turning to the body of our analysis, we have to make clear what we mean by “referential” and “non-referential.” Referentiality is defined by many linguists in many different ways, and there is much debate over the definitions (see Lyons 1999; Payne 1997, for example). Facing the controversy, we choose to stick to what is standard in the Philippine linguistics: we use the term “referential” for nouns or nominalized constituents that have a definite, indefinite or generic reference. Non-referential nominals, on the contrary, have none of these. In this definition many kinds of nominals can be non-referential, but the most important one is a property-denoting nominal. For example, the verb-predicate clause in (5) indicates
 Naonori Nagaya
a situation in which the participant referred to by the noun estudyante ‘student’ carried out the action of running described by the verb. The subject noun estudyante ‘student’ is referential, because it has a definite reference. In contrast, the copular clause in (6) means that the subject Lyndie has the property of being a student. Here, the noun estudyante ‘student’ does not refer to a specific individual, but is used to denote a property. Therefore it is non-referential. (5) Ta~takbo ang=estudyante. asp~run nom=student ‘The student is running.’ (6) Estudyante si=Lyndie. student p.nom=Lyndie ‘Lyndie is a student.’
In order to examine the referentiality of ang-phrases in this definition, we take a close look at ang-phrases construction by construction. There are two major reasons why we take the construction-based approach to the referentiality of ang-phrases. First, it often happens that a given nominal is referential in one construction but non-referential in another. For example, the noun estudyante ‘student’ is referential when it is used as the primary grammatical argument of a verb-predicate clause as in (5), but non-referential when it appears as the complement of a predicational copular clause as in (6). When we analyze the referentiality of a given nominal, it is necessary to take into consideration the syntactic construction where the nominal is used. Second, unlike articles in other languages (e.g. a/an and the in English), ang has the syntactic function of marking the nominative case. Thus, its distribution in sentences is limited to several positions. We cannot analyze the referentiality of ang-phrases without looking at their syntactic function. In modern Tagalog, the distribution of ang-phrases is restricted to the syntactic positions listed in (7) (Nagaya 2007 b, c, ms.). In the rest of this section, we discuss each of these uses, arguing that ang-phrases can be used either referentially or non-referentially.
(7) Distribution of ang-phrases:
a. b. c. d. e. f. g.
Unattached nominal (Section 2.1) Primary grammatical argument of a verb-predicate clause (Section 2.2) Copular subject (Section 2.3) Copular complement (Section 2.3) Complement clause (Section 2.4) Parenthetical (Section 2.5) Exclamative (Section 2.6)
Rise and fall of referentiality 
2.1â•… Unattached nominals Ang can introduce a nominal which is not subcategorized by any predicate and can be used on its own right: for example, the titles of books, songs and narratives as in (8), (9) and (10) (henceforth, the relevant ang-phrases and their free translations are put in boldface). (8) ang=Biblia nom=bible ‘the Bible’ (9) Wikipedia, ang=ma-laya=ng ensiklopedya Wikipedia nom=adj-free=lk encyclopedia ‘Wikipedia, the free encyclopedia’ (10) ang=ulol na unggoy at ang=ma-runong na pagong nom=foolish lk monkey and nom=adj-clever lk turtle ‘the foolish monkey and the clever turtle’
(10) is adopted from the first line of one of the narrative texts collected in Bloomfield (1917: 16–17) (also cited in Himmelmann 2004: 1486–1488). It refers to the two main characters of the narrative. As they refer to a specific individual, ang-marked unattached nominals are referential.
2.2â•… Verb-predicate clauses Tagalog and other Philippine languages are well-known for their voice/transitivitymarking verbal morphology, often referred to as the focus system. In this system, the primary participant of an event is realized in the nominative case or as the nominative pronoun form, and its semantic role is marked on the verb by one of the focus affixes (m-/-um- for Actor Focus, -in for Patient Focus, -an for Locative Focus, and i- for Circumstantial Focus).4 The article ang is employed to introduce such a primary grammatical argument. Let us consider (11), (12) and (13). In (11), the agent participant counts as the primary grammatical argument, and is marked by ang. Accordingly, the verb takes the Actor Focus marker -um-. In contrast, the semantic role of the primary grammatical argument is a patient in (12) and a recipient in (13). Thus, the verb is marked by -Ø in (12) and by -an in (13) respectively (the Patient Focus marker is -Ø in realis mood and -in in irrealis mood). Again, the primary grammatical argument is headed by ang. (11) Dating ang=bata sa=simbahan. arrive nom=child dat=church ‘The child arrived at the church.’
.╅ See Nagaya (2009) for the details of this voice system.
 Naonori Nagaya
(12) Kain-Ø=ko ang=buko pie na gawa sa=Batangas. eat-pf=1sg.gen nom=coconut pie lk made dat=Batangas ‘I ate the coconut pie made in Batangas.’ (13) Bigy-an ni=Kim ng=pera ang=kaibigan=niya. give-lf p.gen=Kim gen=money nom=friend=her ‘Kim gave her friend money.’
As has been analyzed by Schachter and Otanes (1972) and Schachter (1976, 1977), to name a few, the ang-phrase of a verb-predicate clause is always referential, and definite in most cases. It can also have an indefinite reading, when used with an indefinite numeral such as isa ‘one’ and ilan ‘several.’ Compare (11) and (14). (14) Dating ang=isa=ng bata sa=simbahan. arrive nom=one=lk child dat=church ‘A child arrived at the church.’
2.3â•… Copular clauses Tagalog copular clauses are composed of two juxtaposed nominals, that is, a copular complement and a copular subject.5 In the canonical order, a copular complement occupies the clause-initial position, followed by a copular subject, complying with the predicate-initial constituent order of this language, as in (15).6 (15) Copular complement + copular subject
.â•… In the original manuscript, the term “equational clause” was used instead of the “copular clause.” This is partially because there is no copulative element found in Tagalog copular clauses, and partially because it is common to refer to sentence type (15) as “equational” in Austronesian linguistics (cf. Himmelmann 2005a). However, two reviewers rejected the use of “equational” for this sentence pattern; a Tagalog “equational” clause can be an adjectivepredicate or prepositional phrase-predicate clause, which is not “equational” in a usual sense. Thus, we use the “copular clause” rather than the “equational clause” in this paper. Also, the term “copular complement” is chosen over “copular predicate” because in our analysis copular complements are predicative (i.e. denoting a property or action) only in predicational copular clauses. The terms “copula subject” and “copula complement” are originally introduced by Curnow (2000) and Dixon (2001) for their typology of copular clauses. .â•… In this paper, Tagalog copular clauses are translated in such a way that a copular subject is followed by a copular complement in English free translations. This is intended to be faithful to the structure of Tagalog copular clauses in (15), although the resulting English translations might sound a bit unnatural in specificational and pseudocleft clauses (cf. Schachter & Otanes 1972: 530).
Rise and fall of referentiality 
In our analysis, four kinds of copular clauses are distinguished in terms of semantics and pragmatics: predicational clauses, specificational clauses, pseudocleft clauses and identificational clauses.7 Formally speaking, both copular subject and complement are marked in the nominative case, except for copular complements of predicational clauses. In the rest of this subsection, we analyze the referentiality of ang-phrases in each kind of copular clause. Our conclusions are presented in Table 2 in advance. Table 2.╇ Four kinds of copular clauses Clause type
Copular complement
Copular subject
Predicational clause Specificational clause Pseudocleft clause Identificational clause
Non-referential Referential Referential Referential
Referential Non-referential Non-referential Referential
2.3.1â•… Predicational clauses Predicational clauses simply predicate a property of a specific individual. Structurally, a specific individual is realized as a copular subject and is marked in the nominative case, whereas a property is represented as a copular complement and is not marked by any article. For example, (16) is a predicational clause, meaning that the copular subject ang babae ‘the woman’ has the property of being the lead in the movie. (16) Bida ang=babae sa=pelikula. lead nom=woman dat=movie ‘The woman is the lead in the movie.’
A copular complement of this type of copular clause can be an adjective phrase or a prepositional phrase as well. Predicational clause (17), for example, has the adjective matangkad ‘tall’ as its copular complement. (17) Ma-tangkad ang=babae. adj-tall nom=woman ‘The woman is tall.’
.╅ See Higgins (1979), Decklerk (1988), Nishiyama (2003) and Mikkelsen (2004, 2005) for detailed studies on copular clauses. Although there is not enough space here for details, it is necessary to note that as is often the case with copular clauses in other languages, some copular clauses in Tagalog are ambiguous in terms of the four kinds listed in Table 2.
 Naonori Nagaya
When they are used as the copular subject of a predicational clause, ang-phrases are always referential, referring to a specific individual about which a copular complement tells something.
2.3.2â•… Specificational clauses The function of a specificational clause is to single out a specific individual who has a given property. Syntactically, a property is realized as a copular subject and a specific individual as a copular complement. For instance, (18) is a specificational clause, identifying who has the property of being the lead in the movie. It does not add any new information to the copular subject, unlike predicational clauses, but simply identifies who it is. (18) Ang=babae ang=bida sa=pelikula. nom=woman nom=lead dat=movie ‘The lead in the movie is the woman (not any others).’
In other words, the function of a specificational clause is to specify a value for a variable: (18) specifies the value “the woman” for the variable “X is the lead in the movie.” As Higgins (1979) and Declerck (1988) observe, specifying a value for a variable in specificational clauses is similar to providing an answer to a question: “Who is the lead in the movie? – The woman.”. Predicational and specificational clauses look much the same, but differ in many ways. Compare the communicative ends of (16) and (18). The predicational clause in (16) is uttered to inform the hearer of a property which he or she did not know before, while in (18) the point of the utterance is to inform the hearer of who it is that has a certain property that he or she already knows of. In (18), the hearer already knew that someone was the lead in the movie, but did not know who it was. Although they seem to share the same truth-conditional meaning, they have totally different communicative functions. Semantically, both predicational and specificational copular clauses have a specific individual and a property as their semantic components, but these semantic components are aligned in different positions. A property is expressed as a copular complement in predicational clauses but as a copular subject in specificational clauses; a specific individual is realized as a copular subject in the former but as a copular complement in the latter (Table 3). On their appearance, specificational clauses are inverted predicational clauses.8
.â•… For this very reason, specificational clauses have more recently been termed “inverse” copular clauses (Moro 1997).
Rise and fall of referentiality 
Table 3.╇ Predicational and specificational clauses Clause type
Copular complement
Copular subject
Predicational clause Specificational clause
Property Specific individual
Specific individual Property
The semantic contrast between predicational and specificational clauses becomes clear when they are negated. Compare (19a) and (19b), which are the negated versions of (16) and (18). The negated predicational clause in (19a) means that the copular subject does not have the property of being the lead in the movie. It implies that the woman might have a different property. She might be the supporting actress, for example. On the other hand, the negated specificational clause in (19b) indicates that the property of being the lead is not assigned to the woman but someone else. The existence of someone who has the property of being the lead in the movie is implied. (19) a.
Hindi bida ang=babae sa=pelikula. neg lead nom=woman dat=movie ‘The woman is not the lead in the movie (but something else).’
b. Hindi ang=babae ang=bida sa=pelikula. neg nom=woman nom=lead dat=movie ‘The lead in the movie is not the woman (but someone else).’
Another interesting contrast between predicational and specificational clauses is in information structure.9 In a typical interpretation, predicational clauses are used to put pragmatic focus on a property, but specificational clauses on a specific individual. To illustrate, look at question-answer pair (20). In (20Q) the speaker is asking who the lead in the movie is. To answer this question, it is pragmatically correct to use the specificational clause in (18), but not the predicational clause in (16). Since the constituent in the answer that corresponds to the question word in the question is focused on (Halliday 1967), what is on focus in (20A) is the specific individual ang babae ‘the woman’ rather than the property (ang) bida sa pelikula ‘the lead in the movie.’ Therefore, specificational clause (18), which puts focus on the specific individual, is appropriate here. (20) Q: Sino ang=bida sa=pelikula? who.nom nom=lead dat=movie ‘Who is the lead in the movie?’ A: *Bida ang babae sa pelikula. (=16) Ang=babae ang bida sa pelikula. (=18)
.╅ See Kaufman (2005) and Nagaya (2005, 2007a) for the details of information structure in Tagalog.
 Naonori Nagaya
In the literature of Philippine linguistics, the specificational clause has been referred to as the “identificational sentence” (Ramos 1971: 77–79, 110–114) or the “definitized predicate” sentence (Schachter & Otanes 1972: 529–531), and both copular subject and complement have been treated to have a definite reference. However, we analyze the copular subject of a specificational clause as non-referential. This is because in our semantic analysis of specificational clauses, their copular subject is not used to refer to a specific individual, but to denote a property (or to introduce a variable). Semantically, it corresponds to the copular complement of a predicational clause. Both the copular complement of a predicational clause and the copular subject of a specificational clause denote a property and therefore should be equally treated as non-referential. This analysis is borne out by pronominalization phenomena (Mikkelsen 2004, 2005). A unique trait in Tagalog pronominalization is that the third personal pronoun siya can be used to replace a referential nominal, whether it is animate or inanimate (Nagaya 2007b, c, ms.). Here we examine this siya-pronominalization particularly in two syntactic contexts, left-dislocation and question-answer pairs (Mikkelsen 2004, 2005). To begin with, let us confirm that referential nominals can leave siya as its resumptive pronoun in left-dislocation. Look at (21) and (22), which are left-dislocated versions of (11) and (12) respectively. The primary grammatical argument of a verbpredicate clause, either animate or inanimate, is referential and thus can be referred to by siya. Note that leaving a resumptive pronoun is optional in this language (Nagaya 2005, 2007a). (21) Ang=bata, dating (=siya) sa=simbahan. nom=child arrive=3sg.nom dat=church ‘The child, (s/he) arrived at the church.’ (22) Ang=buko pie na gawa sa=Batangas, nom=coconut pie lk made dat=Batangas kain-Ø=ko (=siya) kahapon. eat-pf=1sg.gen=3sg.nom yesterday ‘The coconut pie made in Batangas, I ate (it) yesterday.’
The same is true of question-answer pairs (23) and (24). The primary grammatical argument in the question can be referred to by siya in the answer. (23) Q: Dating=na=ba ang=bata sa=simbahan? arrive=already=q nom=child dat=church ‘Did the child already arrive at the church?’ A: Oo, dating=na=siya sa=simbahan. yes arrive=already=3sg.nom dat=church ‘Yes, s/he already arrived at the church.’
Rise and fall of referentiality 
(24) Q: Kain-Ø=mo=na=ba eat-pf=2sg.gen=already=q
ang=buko pie na gawa sa=Batangas? nom=coconut pie lk made dat=Batangas
‘Did you already eat the coconut pie made in Batangas?’
A: Oo, kain-Ø=ko=na=siya kahapon. yes eat-pf=1sg.gen=already=3sg.nom yesterday ‘Yes, I already ate it yesterday.’
However, siya cannot replace a non-referential nominal like a predicate nominal. For instance, the predicate nominal estudyante ‘student’ in (6) cannot leave a resumptive pronoun in left-dislocation or be pronominalized in a question-answer pair. See (25) and (26). It is therefore concluded that the availability of siya hinges on the referentiality of its antecedent. (25) *Ang=estudyante, siya si=Lyndie. nom=student, 3sg.nom p.nom=Lyndie Intended for ‘Lyndie is a student.’ (26) Q: Estudyante=ba si=Lyndie? student=q p.nom=Lyndie? ‘Is Lyndie a student?’ A: *Oo, siya si=Lyndie yes 3sg.nom p.nom=Lyndie Intended for ‘Yes, Lyndie is (a student).’
Now, let us consider the predicational clause in (27), whose copular subject is clearly referential. (27) ascribes the property of being a doctor to the specific woman to whom the copular subject is referring. (27) Doktora ang=babae=ng iyon. doctor nom=woman=lk that ‘That woman is a doctor.’
When left-dislocated, the copular subject in (27) can leave the resumptive pronoun siya as in (28). (28) Ang=babae=ng iyon, doktora (=siya). nom=woman=lk that doctor=3sg.nom ‘That woman, (she) is a doctor.’
(29) is a question-answer pair made from (27). (29Q) is a predicational clause whose copular complement is an interrogative pronoun. This content question is asking what property can be predicated of the copular subject. Then (29A) is stating that the property in question is “being a doctor.” Notice that the copular subject ang babaeng
 Naonori Nagaya
iyon ‘that woman’ in (29Q) is referred to by siya in (29A). The same goes for the polar question in (30). (29) Q: Ano ang=babae=ng iyon? what nom=woman=lk that ‘What is that woman?’ A: Doktora=siya. doctor=3sg.nom ‘She is a doctor.’ (30) Q: Doktora=ba ang=babae=ng iyon? doctor=q nom=woman=lk that ‘Is that woman a doctor?’ A: Oo, doktora=siya. yes doctor=3sg.nom ‘Yes, she is a doctor.’
Thus, the copular subject of a predicational clause can be referred to by siya both in left-dislocation and in question-answer pairs. This is not surprising at all; the copular subject of this type of copular clause refers to a specific individual, and the specific individual can be pronominalized. However, this is not the case with the copular subject of a specificational clause. Let us use (31) for illustration. (31) Si=Kath ang=pinaka-maganda=ng babae sa=Cainta. p.nom=Kath nom=most-beautiful=lk girl dat=Cainta ‘The most beautiful girl in Cainta is Kath (not any others).’
(31) is a specificational clause, meaning that it is Kath (not any others) that has the property of being the most beautiful girl in Cainta. It provides the value “Kath” for the variable “X is the most beautiful girl in Cainta.” Crucially, the copular subject of this specificational clause cannot be referred to by siya in left-dislocation (32) or in question-answer pairs (33) and (34). (32) Ang=pinaka-maganda=ng babae sa=Cainta, si=Kath (=*siya). nom=most-beautiful=lk girl dat=Cainta p.nom=Kath=3sg.nom ‘The most beautiful girl in Cainta, (*she) is Kath.’ (33) Q: Sino ang=pinaka-maganda=ng babae sa=Cainta? who.nom nom=most-beautiful=lk girl dat=Cainta ‘Who is the most beautiful girl in Cainta?’ A: Si=Kath(=*siya). p.nom=Kath=3sg.nom ‘(*She) is Kath.’
Rise and fall of referentiality 
(34) Q: Si=Kath=ba ang=pinaka-maganda=ng babae sa=Cainta? p.nom=Kath=q nom=most-beautiful=lk girl dat=Cainta ‘Is the most beautiful girl in Cainta Kath?’ A: Oo, si=Kath(=*siya). yes p.nom=Kath=3sg.nom ‘Yes, (*she) is Kath.’
Note that the copular clause si Kath siya itself is grammatical. In (35), for instance, the copular clause si Kath siya is used to state that the person the speakers are looking at is identical to their friend Kath. We call this type of copular clause an identificational clause (see Section 2.3.4 below). (35) (While they were chatting in a coffee shop, speakers A and B saw a girl walking on the street. The girl looked like their friend, Kath, but they were not sure.) A: Tingn-an=mo yung=babae. look-lf=2sg.gen nom=woman ‘Look at the woman.’
Parang si=Kath=siya. seem p.nom=Kath=3sg.nom ‘It seems she is Kath.’
B: Oo=nga, baka si=Kath=siya. yes=indeed maybe p.nom=Kath=3sg.nom ‘Yeah, she might be Kath.’
Also, the ang-phrase pinaka-magandang babae sa Cainta ‘the most beautiful girl in Cainta’ can be referential in other constructions. For example, when it is used as the primary grammatical argument of a verb-predicate clause, it has a definite reference as in (36Q). There is no problem in pronominalization here. See (36A). As noted earlier, a referential nominal in one context may be non-referential in another. (36) Q: alis=na=ba ang=pinaka-maganda=ng babae sa=Cainta? leave=already=q nom=most-beautiful=lk girl dat=Cainta ‘Did the most beautiful girl in Cainta leave already?’ A: Oo, alis=na=siya. yes leave=already=3sg.nom ‘Yes, she already left.’
To summarize, the copular subject of a predicational clause can be referred to by the personal pronoun siya, but that of a specificational clause cannot. The different results of the pronominalization test between the copular subject of a predicational clause and that of a specificational clause show the difference in referentiality between the two. The former can be pronominalized by siya because it has a referent; the latter cannot
 Naonori Nagaya
be because it is not referential. The copular subject of a specificational clause does not refer to a specific individual but denotes a property. In contrast, the copular complement of a specificational clause, which represents a specific individual, is referential. Indeed, the copular complement of the specificational clause in (18) can be pronominalized in left-dislocation (37) and in question-answer pair (38). (37) Ang=babae, siya ang=bida sa=pelikula. nom=woman 3sg.nom nom=lead dat=movie ‘The woman, the lead in the movie is her.’ (38) Q: Ang=babae=ba ang=bida sa=pelikula? nom=woman=q nom=lead dat=movie ‘Is the lead in the movie the woman?’ A: Oo, siya ang=bida sa=pelikula. yes 3sg.nom nom=lead dat=movie ‘Yes, the lead in the movie is her.’
To conclude, the copular subject of a specificational clause is non-referential, whereas its copular complement is referential.
2.3.3â•… Pseudocleft clauses Pseudocleft clauses, or simply pseudoclefts, can be analyzed in the same manner as specificational clauses. What we call pseudoclefts is a subtype of specificational clauses whose copular subject is a nominalized clause (or headless relative clause).10 The pseudocleft clause in (39), for example, picks out a specific individual who is involved in the action of arriving at the church. In other words, it provides the value “the child” for the variable “X arrived at the church.” (39) Ang=bata ang=dating sa=simbahan. nom=child nom=arrive dat=church ‘The one who arrived at the church is the child (not any others).’
It is crucial to note that pseudoclefts are to be distinguished from the superficially similar example in (40). It is a straightforward predicational clause, the copular subject of which happens to be a nominalized clause.
.â•… Lambrecht (2001: 467) defines a cleft construction as “a complex sentence structure consisting of a matrix clause headed by a copula and a relative or relative-like clause whose relativized argument is coindexed with the predicative argument of the copula. Taken together, the matrix and the relative express a logically simple proposition, which can also be expressed in the form of a single clause without a change in truth conditions.” Except for the fact that Tagalog does not have a copula, this definition matches what we call pseudoclefts.
Rise and fall of referentiality 
(40) Ma-tangkad ang=dating sa=simbahan. adj-tall nom=arrive dat=church ‘The one who arrived at the church is tall.’
In order to elucidate the characteristics of pseudoclefts, it is useful to compare pseudocleft (39) with verb-predicate clause (11), repeated here as (41). (39) and (41) represent truth-conditionally the same situation but differ in many ways. Semantically, the verbpredicate clause in (41) describes the action in which the ang-phrase is involved, while the pseudocleft in (39) tells us who it is that is involved in the action. The inverted predication we have already looked at between predicational and specificational clauses is observed between verb-predicate and pseudocleft clauses. (41) Dating ang=bata sa=simbahan. arrive nom=child dat=church ‘The child arrived at the church.’
When they are negated, verb-predicate and pseudocleft clauses imply different things. Compare the negated verb-predicate clause in (42a) and the negated pseudocleft in (42b). Only in the latter construction is it implied that there is someone who arrived at the church. (42) a.
Hindi dating ang=bata sa=simbahan. neg arrive nom=child dat=church ‘The child did not arrive at the church.’
b. Hindi ang=bata ang=dating sa=simbahan. neg nom=child nom=arrive dat=church ‘The one who arrived at the church is not the child (but someone else).’
In terms of information structure, similarly to specificational clauses, pseudoclefts are employed to put pragmatic focus exclusively on a specific individual, but verbpredicate clauses on an action. For this reason, narrowly focused arguments such as question words and exhaustively identified individuals (‘only,’ Rizzi 1997) must appear as a copular complement of a pseudocleft clause (Nagaya 2005, 2007a). For example, it is not allowed to use the question word sino ‘who’ to form a content question in a verb-predicate clause (43a). The pseudocleft is required instead as in (43b). (43) a.
*Dating sino sa=simbahan? arrive who.nom dat=church Intended for ‘Who arrived at the church?’
b. Sino ang=dating sa=simbahan? who.nom nom=arrive dat=church ‘Who arrived at the church?’ (lit. ‘Who is the one who arrived at the church?’)
 Naonori Nagaya
The same contrast can be observed regarding exhaustive identification. The examples in (44) show that it is necessary to choose a pseudocleft clause (44b) in order to exhaustively identify the child as the one who arrived at the church. Verb-predicate clause (44a) is grammatical but does not have the reading of exhaustive identification. (44) a. Dating=lang ang=bata sa=simbahan. arrive=only nom=child dat=church *‘Only the child arrived at the church.’ ‘The child only arrived at the church (i.e. s/he did not do anything).’ b. Ang=bata=lang ang=dating sa=simbahan. nom=child=only nom=arrive dat=church ‘The one who arrived at the church is only the child.’
Since the pseudocleft clause is a subtype of specificational clauses, its copular subject is analyzed as non-referential, while its copular complement is referential. The copular subject of this type of clause denotes an action (or variable), while its copular complement refers to a specific individual (or value), specifying who it is that is involved in the action. Again, this analysis is supported by the pronominalization test. Let us consider the referentiality of the copular subject of a pseudocleft by examining the pseudocleft in (45). (45) Si=Marfeal ang=nag-u~ukulele. p.nom=Marfeal nom=af.rl-asp~play.ukulele ‘The one who is playing the ukulele is Marfeal (not any others).’
The copular subject ang nag-uukulele ‘the one who is playing the ukulele’ cannot be referred to by siya in left-dislocation (46) or in question-answer pairs (47) and (48). (46) Ang=nag-u~ukulele, si=Marfeal (=*siya). nom=af.rl-asp~play.ukulele p.nom=Marfeal=3sg.nom ‘The one who is playing the ukulele, (*he) is Marfeal.’ (47) Q: Sino ang=nag-u~ukulele? who.nom nom=af.rl-asp~play.ukulele ‘Who is the one who is playing the ukulele?’ A: Si=Marfeal(=*siya). p.nom=Marfeal=3sg.nom ‘(*He) is Marfeal.’ (48) Q: Si=Marfeal=ba ang=nag-u~ukulele? p.nom=Marfeal=q nom=af.rl-asp~play.ukulele ‘Is the one who is playing the ukulele Marfeal?’ A: Oo, si=Marfeal(=*siya). yes p.nom=Marfeal=3sg.nom ‘Yes, (*he) is Marfeal.’
Rise and fall of referentiality 
On the other hand, the copular complement of a pseudocleft can be pronominalized, which shows that it is referential. See (49) and (50). (49) Ang=bata, siya ang=dating sa=simbahan. nom-child 3sg.nom nom=arrive dat=church ‘The child, the one who arrived at the church is him/her.’ (50) Q: Ang=bata=ba ang=dating sa=simbahan? nom=child=q nom=arrive dat=church ‘Is the one who arrived at the church the child?’ A: Oo, siya ang=dating sa=simbahan. yes 3sg.nom nom=arrive dat=church ‘Yes, the one who arrived at the church is him/her.’
To conclude, the copular subject of a pseudocleft clause is non-referential, while its copular complement is referential.
2.3.4â•… Identificational clauses The function of identificational clauses is to declare that the referent of a copular subject is identical to that of a copular complement. For example, (51) states that ang babaeng iyon ‘that woman,’ about whom the speaker is talking, is the same person as Osang. The reversed order is possible, like (52), and does not change the meaning. It is clear from its function that both copular subject and complement are referential in this type of copular clause. (51) Si=Osang ang=babae=ng iyon. p.nom=Osang nom=woman=lk that ‘That woman is Osang.’ (52) Ang=babae=ng iyon si=Osang. nom=woman=lk that p.nom=Osang ‘Osang is that woman.’
2.4â•… Complement clauses Ang also introduces an infinitival complement clause in impersonal adjective-predicate clauses as in (53) and (54). In both cases, the ang-phrase represents a closed proposition, not a specific individual, and thus it is non-referential.11
.â•… The linker na/-ng is also employed for introducing a complement clause, and it is more common than marking a complement clause with ang. Himmelmann (1997) argues that the linker also developed from a demonstrative. See also Diessel (1999: 130–132) and Reid and Liao (2004: 484).
 Naonori Nagaya
(53) Ma-hirap ang=mag-mahal ng=syota ng=iba. adj-hard nom=af-love gen=girlfriend gen=another ‘It is hard to love the girlfriend of another (person).’ (54) Ma-saya ang=maki-pag-kwentuhan. adj-fun nom=af-grn-chat ‘It is fun to chat.’
2.5â•… Parentheticals Parentheticals are defined here as expressions used to situate the following utterance in a discourse in terms of epistemic attitude or quotation (Nagaya 2007b, c, ms, cf. “stance markers” in Yap this volume). They usually appear in the sentenceinitial position,12 and work as sentential adverbs of some kind. Ang can be used for marking such a parenthetical. In (55), for instance, ang introduces totoo ‘truth,’ which expresses the speaker’s epistemic attitude toward the following utterance (cf. “epistemic parenthetical” in Thompson & Mulac 1991).13 (2) is also an example of this type of parentheticals. (55) Ang=totoo, may gusto=siya sa=akin. nom=truth have crush=3sg.nom dat=3sg.dat ‘In fact (lit. the truth is), s/he has a crush on me.’
Unlike other uses of ang, ang-parentheticals are not required constituents of a clause. (55) is still grammatical even if the ang parenthetical is omitted. See (56). (56) May gusto=siya sa=akin. Nhave crush=3sg.nom dat=3sg.dat ‘S/he has a crush on me.’
.╅ In constructed examples, ang-marked parentheticals can appear in the sentence-final position as well (see Nagaya 2005, 2007a for the sentence-initial/final position in Tagalog). However, no parenthetical appears in the sentence-final position in the corpus examined in Section 4. .╅ Foong Ha Yap (p.c.) points out that this use of ang is reminiscent of epistemic uses of the definiteness marker -nya in Colloquial Indonesian (Englebretson 2003) and in Malay (Yap this volume).
Rise and fall of referentiality 
In other cases, parentheticals have textual or discourse-related meanings. In (57), the ang-phrase ang sabi niya ‘his/her statement’ or ‘what s/he said’ is employed to indicate that the utterance to follow is a direct quotation. This type of parenthetical is one of the major markers for quotations in Tagalog, which Schachter and Otanes (1972: 170) call the “quotation tag.” See also LaPolla and Poa (2005). (57) Ang=sabi=niya, may gusto=siya sa=akin. nom=statement=3sg.gen have crush=3sg.nom dat=3sg.dat ‘What s/he said was, s/he has a crush on me.’
(58) is cited from the Tagalog spoken corpus (see Section 4). The parenthetical ang ginawa ko ‘what I did’ frames the following utterance as a reaction to the rumor the speaker heard. To be more precise, the parenthetical indicates the textual relationship between the two events: it means that the speaker pursued the girl because he heard the rumor. To understand the parenthetical in (58), it is necessary to access a more complex discourse environment than a single sentence. (58) (The speaker has a crush on his classmate, but she has a boyfriend.) tapos may na-dinig=ako na, then have sp.rl-hear=1sg.nom lk kapag ni-ligaw-an=ko=daw=siya, if rl-court-lf=1sg.gen=hearsay=3sg.nom bi~break-in=daw=niya yung=boyfriend=niya. asp~break-pf=hearsay=3sg.gen nom=boyfriend=3sg.gen ‘Then I heard something like if I court her, she would break up with her boyfriend.’ di ang=gawa-Ø=ko, well nom=do-pf=3sg.gen subuk-an=ko=nga, ni-ligaw-an=ko=nga. try-lf=1sg.gen=indeed rl-court-lf=1sg.gen=indeed ‘Well, what I did was, I did try (courting her), I did court (her).’
These parentheticals are not employed to refer to a specific individual but to convey a subjective, textual, or discourse-related meaning. Therefore, they are non-referential.
2.6â•… Exclamatives There are several ways of forming an exclamative in Tagalog (see Blake 1925: 169–172 for example), but the most productive one is the construction “ang + adjectival
 Naonori Nagaya
root.”,14,15 (59) is an ordinary adjective-predicate clause, in which the proposition that Kim is beautiful is analytically expressed by the subject and the adjectival predicate. The ang-marked exclamative in (60), in contrast, expresses the same proposition as (59), but the proposition is packed into the single nominalized expression marked by ang, yielding an exclamatory meaning. Notice that the subject in (60) is marked in the genitive case rather than in the nominative case, which explicitly shows that (60) is nominalized. Another example is given in (61), which is a nominalized version of (40). See also (3). (59) Ma-ganda si=Kim. adj-beautiful p.nom=Kim ‘Kim is beautiful.’ (60) Ang=ganda ni=Kim! nom=beautiful p.gen=Kim ‘How beautiful Kim is!’ (lit. ‘Kim’s beauty.’) (61)
Ang=tangkad ng=dating sa=simbahan! nom=tall gen=arrive dat=church ‘How tall the one who arrived at the church is!’ (lit. ‘The tallness of the one who arrived at the church.’)
Ang-exclamatives are non-referential. They are not employed to refer to a specific individual, but to express the speaker’s feeling at the time of utterance.
.â•… Exclamatives created by nominalization are not uncommon at all. Kaufman (this volume), Potsdam (this volume) and Wrona (this volume) provide examples of nominalized exclamatives from typologically different languages. Bendette McFarland (p.c.) suggests that it might be possible to analyze ang with an exclamatory meaning as a contracted form of ano-ng ‘what-LK,’ which is another exclamativeforming strategy. We do not take this possibility for the following reasons: (i) there are semantic differences between ang-exclamatives and anong-exclamatives, (ii) the CVCV-reduplication for intensification is available to ang-exclamatives but not to anong-exclamatives (for example, ang ganda-ganda mo but *anong ganda-ganda mo), and (iii) there exist nominalized exclamatives created by an article in other Philippine languages as well (see Section 5 and Kaufman this volume). .â•… Another way of forming an exclamative in Tagalog is the prefix ka- (Bloomfield 1917; Blake 1925, Schachter & Otanes 1972), whose cognates are also used for forming an exclamative in other Philippine languages (Kaufman this volume). Interestingly, the exclamative with the prefix ka- is not used in the Tagalog spoken corpus (Section 4), while the ang-exclamative occurs frequently. Thus, we might be able to speculate that the exclamative with ka- is disappearing in modern Tagalog because of the emergence of the exclamative with ang.
Rise and fall of referentiality 
2.7â•… Development from referential to non-referential uses In this section, we have argued that ang-phrases can be either referential or nonreferential according to the construction where they appear. Our discussions can be summarized as in (62) and (63). (62) Referential uses of ang-phrases: a. b. c. d.
Unattached nominal Primary grammatical argument of a verb-predicate clause Copular subject of predicational and identificational clauses Copular complement of identificational, specificational, and pseudocleft clauses
(63) Non-referential uses of ang-phrases: a. Copular subject of specificational and pseudocleft clauses (ï…€ property/action/variable) b. Complement clause (ï…€ closed proposition) c. Parenthetical (ï…€ the speaker’s epistemic attitude/quotation/textual relation) d. Exclamative (ï…€ the speaker’s feeling)
At the beginning of this section, we pointed out that the original function of ang should be referential according to its reconstructed form (Reid 1978). If this assumption is correct, our analyses summarized in (62) and (63) suggest that ang has been grammaticalizing from referential to non-referential uses. In other words, the article ang has been acquiring non-referential functions. Of course, further research is definitely needed to substantiate our hypothesis, and it still remains uncertain in what order ang-phrases have acquired these non-referential uses. Nevertheless, it is crucial to mention that the non-referential uses of ang-phrases display at least two properties characteristic of grammaticalization. First, ang-phrases have obtained more discourse-related and more subjective meanings at the emergence of non-referential uses (cf. “stance markers” in Yap this volume). In referential uses, ang-phrases simply refer to a specific individual. In non-referential uses, in contrast, their meaning is more abstract, textual, and subjective. Instead of referring to a specific individual, the copular subject of specificational and pseudocleft clauses is involved in the manipulation of a property, action, or variable. The complement clause marked by ang represents a closed proposition. Ang-parentheticals indicate the speaker’s epistemic attitude toward a proposition or a textual meaning, while ang-exclamatives express the speaker’s intense feeling at the moment of utterance. Both are deeply anchored in the speaker’s experience, and cannot be interpreted properly if separated from the existence
 Naonori Nagaya
of the speaker. In this sense, they are highly subjective. This higher degree of subjectivity is often associated with grammaticalization (Traugott 1989).16 Second, ang-phrases display de-categorization or a loss of categoriality in nonreferential uses (Heine, Claudi & Hünnemeyer 1991; Heine & Kuteva 2007). Angphrases are clearly nominal in referential uses, but lose their noun-like properties in non-referential uses. The copular subject of specificational and pseudocleft clauses cannot be referred to by the third person pronoun siya. The complement clause marked by ang is not nominal but infinitival. Ang-parentheticals function as sentential adverbs, and they are not required in a clause. Ang-exclamatives can serve as stand-alone sentences indicating a propositional meaning on their own right. Thus, ang-marked nominals show reduced categoriality in non-referential uses. As the discussion above shows, although our hypothesis on the development of ang-phrases needs more research for verification, the non-referential uses of angphrases demonstrate the semantic changes and de-categorization that are characteristic of grammaticalization.
3.â•… Rise of referentiality: The usage of Tagalog yung Modern Tagalog has a set of articles that are grammaticalized from the demonstratives: iyon ‘that (distal)’ + -ng > yung, iyan ‘that (medial)’ + -ng > yang, and ito ‘this’ + -ng > itong (see Table 1 again; -ng is a linking element). See also McFarland (1976), Reid (1978), and Himmelmann (2005b). Among these demonstrative-derived articles, yung is currently the most often used and is considered as nearly a replacement for ang in conversations (Reid 1978; Himmelmann 2005b). Yung is not a demonstrative any longer but an article in modern Tagalog. Reid (1978: 58) observes that yung is “usable as a definite article without demonstrative significance.” As a sign of grammaticalization, yung shows phonological reduction: iyon [ijón] + -ng [ŋ] → yung [juŋ]. Morphologically, while the demonstrative iyon is
.â•… Concerning the nature of semantic changes that accompany grammaticalization, Traugott (1989: 34–35) proposes three semantic-pragmatic tendencies.
i. Tendency I: Meanings based in the external described situation > meanings based in the internal (evaluative/perceptual/cognitive) described situation. ii. Tendency II: Meanings based in the external or internal described situation > meanings based in the textual and metalinguistic situation. iii. Tendency III: Meanings tend to become increasingly based in the speaker’s subjective belief state/attitude toward the proposition.
Rise and fall of referentiality 
a free word, yung is a proclitic: it cannot be used independently and forms a single phonological unit with a constituent that accompanies it. It is clear that the original meaning of yung is deictic and thus referential, because it is derived from the distal demonstrative iyon. In this section, we point out that, similarly to ang, yung has been developing from deictic to referential uses and from referential to non-referential uses. It is also emphasized that yung is not just a renewal or replacement for ang, but displays additional functions which ang does not have.
3.1â•… Referential and non-referential uses of yung Similarly to ang, yung has both referential and non-referential functions. See (64)–(70) for referential uses and (71)–(73) for non-referential uses. (64)
(Unattached nominal) Yung=Biblia. nom=bible ‘The Bible.’ (e.g. as an answer to a question)
(65)
(Primary grammatical argument of a verb-predicate clause) Dating yung=bata sa=simbahan. arrive nom=child dat=church ‘The child arrived at the church.’
(66)
(Copular subject of a predicational clause) Bida yung=babae sa=pelikula. lead nom=woman dat=movie ‘The woman is the lead in the movie.’
(67)
(Copular subject of an identificational clause) Si=Osang yung=babae=ng iyon. p.nom=Osang nom=woman=lk that ‘That woman is Osang.’
(68)
(Copular complement of an identificational clause) Yung=babae=ng iyon si=Osang. nom=woman=lk that p.nom=Osang ‘Osang is that woman.’
(69)
(Copular complement of a specificational clause) Yung=babae ang=bida sa=pelikula. nom=woman nom=lead dat=movie ‘The lead in the movie is the woman.’
(70)
(Copular complement of a pseudocleft clause) Yung=bata ang=dating sa=simbahan. nom=child nom=arrive dat=church ‘The one who arrived at the church is the child.’
 Naonori Nagaya
(71)
(Copular subject of a specificational clause) Si=Piolo yung=bida sa=pelikula. p.nom=Piolo nom=lead dat=movie ‘The lead in the movie is Piolo.’
(72)
(Copular subject of a pseudocleft clause) Si=Lorie yung=dating sa=simbahan. p.nom=Lorie nom=arrive dat=church ‘The one who arrived at the church is Lorie.’
(73)
(Complement clause) Ma-hirap yung=mag-mahal ng=syota ng=iba. adj-hard nom=af-love gen=girlfriend gen=another ‘It is hard to love the girlfriend of another (person).’
However, it is not true that yung has exactly the same distribution as ang: parentheticals with yung are not preferred, and exclamatives cannot be formed with yung. See (74) and (75). (74) ?Yung=totoo, may gusto=siya sa=akin. nom=true have crush=3sg.nom dat=3sg.dat ‘In fact (lit. the truth is), s/he has a crush on me.’ (cf. 55) (75) *Yung=ganda ni=Kim! nom=beautiful p.gen=Kim Intended for ‘How beautiful Kim is!’ (cf. 60)
Thus, yung has been grammaticalized to such an extent that it can be used either referentially or non-referentially. However, its grammaticalization has not yet gone so far as ang. Yung does not have all of the non-referential uses that ang has developed.
3.2â•… Carried-over and expanded functions of yung Although this has received little attention in prior studies, the article yung has several functions which ang does not: one is carried over from its demonstrative meaning, and another is an expanded nominalizing function that only yung displays. First, yung may still convey a spatial deictic meaning. In (76), for instance, the yung-phrase is used deictically, referring to a specific ball away from both speaker and hearer. This interpretation is not possible when ang is used instead of yung. (76) Kun-in=mo yung=bola! get-pf=2sg.gen nom=ball ‘Get the ball (which is far from the speaker and hearer)!’
Second, yung can nominalize constituents that ang and other articles cannot. For example, yung introduces a personal name already marked by the personal name
Rise and fall of referentiality 
Â� article si in (77), yielding the interpretation that John Rey, to whom the si-phrase is referring, is away from the speaker and hearer.17 (77) Na-kausap=mo=na=ba yung=si=John Rey? sp.rl-talk.to=2sg.gen=already=q nom=p.nom=John Rey? ‘Did you already talk to John Rey?’
In (78), yung leads the temporal noun kagabi ‘last night’ to refer to the message the speaker sent the night before. Ang is not possible in these contexts. (78)
(The speaker is saying that s/he sent a message to the hearer by mistake.) Wrong send yung=kagabi. wrong send nom=last.night ‘The message I sent to you last night was the one sent to a wrong address.’18
Moreover, yung can nominalize a full clause. In (79), the yung-marked clause yung nakabikini ka ‘you were wearing a bikini’ works as a referential expression, meaning the picture (drawing, video, etc.) where the hearer was wearing a bikini. (79) Pa-kita=mo sa=akin yung=naka-bikini=ka. caus-see=2sg.gen dat=1sg.dat nom=wear.bikini=2sg.nom ‘Show me the picture (drawing, video etc.) where you were wearing a bikini.’
(80) is a pseudocleft in which the copular subject is left-dislocated. The copular complement is the yung-marked complex sentence yung kapag nag-birthday ako maraming regalo ‘when I have a birthday, there are a lot of gifts,’ which refers to the specific event the speaker likes the most. Again, the use of ang in these environments results in ungrammatical sentences. .â•… One of the reviewers, who is a native speaker of Tagalog, reports that (77) has an alternative reading in which John Rey is socially distant (or distanced) from the speaker, and that one cannot use (77) if John Rey is the speaker’s brother, for example, regardless of his location at the time of utterance. .â•… We have to note, however, that the interpretation that the yung-phrase refers to the message is due to pragmatic reasons rather than syntactic ones. The more precise translation for yung kagabi is ‘something pertaining to the last night’ and the interpretation in question is obtained because the copular complement in (78) is wrong send ‘the message sent to a wrong address.’ Thus, in other contexts the same yung-phrase can mean a different thing. (i)
Ma-saya yung=kagabi. adj-fun nom=last.night ‘What happened last night (e.g. party) was fun.’
(ii)
Su~suut-in=ko=ulit yung=kagabi. asp~wear-pf=1sg.gen=again nom=last.night ‘I will wear again the one I wore last night (e.g. jacket).’
 Naonori Nagaya
(80) Ang=pinaka-gusto=ko, nom=most-like=1sg.gen yung=kapag nag-birthday=ako marami=ng regalo. nom=when af.rl-have.birthday=1sg.nom many=lk gift ‘What I like the most is the situation that when I have my birthday, there are a lot of gifts.’
3.3â•… Summary In this section, we have showed that yung has both referential and non-referential uses like ang, but lacks some non-referential uses that ang has, and possesses some referential uses that ang does not. Therefore, yung is not a mere replacement for ang but is also an addition to the Tagalog article system.19
4.â•… Evolution from referential to non-referential uses We have so far observed that both ang and yung have developed referential and nonreferential uses but with different focal points. Ang has a variety of non-referential uses, even being able to create parentheticals and exclamatives. In contrast, the main function of yung is referential. It still retains its deictic implication in some contexts and has an expanded nominalizing function but it lacks the function of forming an exclamative. The next question to ask is, then, how frequently are ang and yung used referentially? Also, how often are they employed non-referentially? In order to answer these questions, we have to look at the actual usage of these articles.
4.1╅ Corpus-based approach to ang and yung Nagaya (2007b, c, ms) has conducted a quantitative investigation into referential and nonreferential uses of ang and yung in a spoken corpus. A spoken corpus rather than a written one is examined, because, as noted earlier, yung is usually used in informal speech, like conversations, and is avoided in formal style (Reid 1978; Himmelmann 2005b). The corpus examined, the Tagalog spoken corpus, was collected by asking native speakers to do three distinct tasks: have conversations, tell narratives, and relate pear stories. Native speakers of Tagalog had conversations with their close friends (conversations), produced a story about a specific topic given to them (narratives), and retold .╅ As Foong Ha Yap (p.c.) notes, there are some functional overlaps between Tagalog yung and Malay yang (see Yap this volume). But they are not etymologically directly related. According to Scott Paauw (p.c.), the etymology of Malay yang has been much discussed, with no conclusive answer as yet. One of the plausible hypotheses is that yang developed from ia (third person singular pronoun) + nan/nang (Old Malay equivalent to yang).
Rise and fall of referentiality 
the Pear film (Chafe 1980) (pear stories). The total duration of time is approximately 303 minutes, and the total number of words is 45,708 (see Table 4). All participants who have contributed to this corpus are native speakers of Tagalog born and raised in Manila or other Tagalog-speaking provinces such as Bulacan, Rizal and Quezon. Table 4.╇ Tagalog spoken corpus Corpus Conversations Narratives Pear stories Total
# of Texts
Total # of words
Total duration of time
4 34 24 62
18,965 17,084 9,659 45,708
115 minutes 111 minutes 77 minutes 303 minutes
In Table 5 the frequency of occurrence of each use of ang and yung in the Tagalog spoken corpus is given. As seen from the table, ang is often used non-referentially. Interestingly, the most frequent use of ang in our spoken corpus is to form an exclamative: 157 examples of ang (36.4%) are employed for this purpose. In contrast, the use of ang as the primary grammatical argument of a verb-predicate clause is relatively infrequent. It consists of 78 examples (18.1%), which is as frequent as the ang-marked parentheticals. In terms of referentiality, 112 examples of ang (26.0%) are used referentially, while 319 examples (74.0%) are employed non-referentially. See Table 6. Table 5.╇ Referential and non-referential uses of ang and yung Referentiality
ang
Uses
yung
#
%
#
%
6 78 28 0 0 0 0
1.4 18.1 6.5 0.0 0.0 0.0 0.0
205 756 234 17 1 21 4
14.5 53.5 16.6 1.2 0.1 1.5 0.3
Referential
Unattached nominal Primary argument of a verb-predicate clause Subject of a predicational clause Subject of an identificational clause Complement of an identificational clause Complement of a specificational clause Complement of a pseudocleft clause
Non-referential
Subject of a specificational clause Subject of a pseudocleft clause Complement clause Parenthetical Exclamative
40 50 0 72 157
9.3 11.6 0.0 16.7 36.4
97 78 0 0 0
6.9 5.5 0.0 0.0 0.0
Total
431
100.0
1413
100.0
 Naonori Nagaya
Table 6.╇ Referentiality of ang and yung ang
Referentiality
yung
#
%
#
%
Referential uses
112
26.0
1238
87.6
Non-referential uses
319
74.0
175
12.4
Total
431
100.0
1413
100.0
On the other hand, yung is almost always used referentially. 1,238 examples of yung (87.6%) are used referentially, while 175 examples (12.4%) are employed nonreferentially. The most frequent use of yung is to introduce the primary grammatical argument of a verb-predicate clause (756 examples, 53.5%), followed by marking the copular subject of a predicational clause (234 examples, 16.6%). Its non-referential uses are not impossible but rarely appear in the spoken corpus. The use of forming a parenthetical or exclamative, which constitutes a significant fraction of the total uses of ang, is not found for yung. In conclusion, ang frequently appears in non-referential uses, but yung is often employed referentially in the spoken corpus.
4.2â•… Renewal and expansion in the Tagalog article system The results revealed in the previous sections suggest that two semantic changes are going on in the Tagalog article system. Both ang and yung have been grammaticalizing from deictic to referential uses and from referential to non-referential uses. But the developments of the two articles have proceeded at different paces and to different stages. Ang-phrases are now becoming less referential, obtaining discourse-related and subjective meanings. Its primary usage has been moving from referential to nonreferential, subjective and discourse-related uses. The original demonstrative meaning of ang is completely bleached out in modern Tagalog, and its frequent uses are nonreferential ones in spoken discourse. The widely-accepted hypothesis that ang-phrases are always referential can no longer be upheld. Keeping in step with the de-referentialization of ang, yung is now taking over the referential uses of ang, as hinted already in Reid (1978). This process of grammaticalization has made progress to such an extent that in spoken discourse yung is the default article for marking referential nominals. It still maintains its original deictic meaning, and has acquired an expanded nominalizing function that ang does not have. Although yung has non-referential uses as well, they are rarely observed. Taken together, there is an emerging semantic distinction between ang and yung in spoken discourse: ang is for non-referential uses, and yung for referential uses.
Rise and fall of referentiality 
5.â•… Development of non-referentiality across Philippine languages Traditionally, not only Tagalog ang but also its equivalents in other Philippine languages have been believed to be referential. Reid and Liao (2004: 469), among others, observe “[i]n all Philippine languages, Nominative phrases [“ang-phrases and their equivalents,” NN] typically have a definite interpretation, that is, the speaker assumes that the addressee knows the general reference of the actant which is the head of the phrase.” In this section, we explore the possibility that our analyses of Tagalog articles can be applied to article systems in other Philippine languages, by looking at cases from two other Philippine languages: ang and su in Bikol (Partido dialect, Central Philippines) and ti and diay in Ilokano (Northern Philippines). We point out that nominals marked by these articles can be either referential or non-referential, and that in these languages, too, there is evidence of an existing article obtaining non-referential uses and a newer article emerging for referential nominals.
5.1â•… Bikol ang and su The Partido dialect of Bikol has two articles for common nouns in the nominative case, ang and su. Both are considered to mark a definite nominal, but su is more specific and more emphatic than ang (McFarland 1978).20 Indeed, ang and su can be used almost interchangeably as in (81)–(88). When both articles are available, su allows for a distal deictic reading, but ang does not. (81)
(Primary grammatical argument of a verb-predicate clause) Nag-abot ang/su=aki sa=simbahan. af.rl-arrive nom=child dat=church ‘The child arrived at the church.’
(82)
(Copular subject of a predicational clause) Bida ang/su=bayi sa=pelikula. lead nom=woman dat=movie ‘The woman is the lead in the movie.’
.â•… Ang [aŋ] in this dialect is equivalent to an [an] in the dialect of Bikol on which McFarland (1978) works. Pangasinan (Northern Philippines), Matigsalug Manuvu (South Mindanao) and other Manobo languages have su as a demonstrative (Ricardo Ma. Duran Nolasco p.c., and Mark Felix Albert Santiago p.c.). However, su is not a demonstrative but an article in Partido Bikol. Demonstratives in this language are ito ‘that (distal),’ iyan/an ‘that (medial),’ and ini ‘this.’
 Naonori Nagaya
(83)
(Copular subject of an identificational clause) Si=Osang ang/su=bayi=ng itu. p.nom=Osang nom=woman=lk that ‘That woman is Osang.’
(84)
(Copular complement of an identificational clause) Ang/Su=bayi=ng itu si=Osang. nom=woman=lk that p.nom=Osang ‘Osang is that woman.’
(85)
(Copular complement of a specificational clause) Ang/Su=bayi ang=bida se=pelikula. nom=woman nom=lead dat=movie ‘The lead in the movie is the woman.’
(86)
(Copular complement of a pseudocleft clause) Ang/Su=aki ang=nag-abot sa=simbahan. nom=child nom=af.rl-arrive dat=church ‘The one who arrived at the church is the child.’
(87)
(Copular subject of a specificational clause) Si=Piolo ang/su=bida sa=pelikula. p.nom=Piolo nom=lead dat=movie ‘The lead in the movie is Piolo.’
(88)
(Copular subject of a pseudocleft clause) Si=Lorie ang/su=nag-abot sa=simbahan. p.nom=Lorie nom=af.rl-arrive dat=church ‘The one who arrived at the church is Lorie.’
Interestingly, ang is the only available article to introduce a complement clause and a parenthetical. See (89) and (90). Su cannot be used for these types of non-referential expressions. (89) Ma-sakit ang/*su=ma-moot sa=ilusyo ning=iba. adj-hard nom=af-love dat=girlfriend gen=another ‘It is hard to love the girlfriend of another (person).’ (90) Ang/*Su=tutuo, mai gusto=siya sakuya. nom=true have crush=3sg.nom 3sg.dat ‘In fact (lit. the truth is), s/he has a crush on me.’
It is not possible, however, to use ang or su to form an exclamative in this language. They are not yet grammaticalized to accommodate a subjective meaning like in exclamatives. For this purpose, another article ung is employed like (91). This article is used only for forming an exclamative in this dialect of Bikol.
Rise and fall of referentiality 
(91) Ung=gayun=mo. art=beautiful=2sg.gen ‘How beautiful you are! (lit. ‘Your beauty.’)’
In summary, the two articles in Partido Bikol, ang and su, show the same grammaticalization patterns from referential to non-referential uses as ang and yung in Tagalog. Ang can be used even for the non-referential uses that are discourse-related and subjective, but su cannot. Neither of them can create an exclamative.
5.2â•… Ilokano ti and diay There are two articles for core arguments in Ilokano, ti and diay. Ti is the older article employed mainly in written discourse, whereas diay recently developed from the distal demonstrative pronoun daydiay and is often used in informal speech (Reid 1978; Rubino 1997). Ti and diay can be used either referentially or non-referentially as in (92)–(98). However, the consultant reports that diay is preferred for referential uses (92)–(96) and ti for non-referential uses (97)–(98). (92)
(Primary grammatical argument of a verb-predicate clause) Sangpet ?ti/diay=ubing idiay=simbaan. arrive nom=child there=church ‘The child arrived at the church.’
(93)
(Copular subject of a predicational clause) Bida ?ti/diay=babai idiay=pelikula. lead nom=woman there=movie ‘The woman is the lead in the movie.’
(94)
(Copular subject of an identificational clause)21 Ni=Osang ?ti/diay=babai nga daydiay. p.nom=Osang nom=woman lk that ‘That woman is Osang.’
(95)
(Copular complement of a specificational clause) ?Ti/Diay=babai ti=bida idiay=pelikula. nom=woman nom=lead there=movie ‘The lead in the movie is the woman.’
.╅ The distal demonstrative daydiay rather than ti or diay is used for making the copular complement of an identificational clause. (i)
Daydiay nga babai ni=Osang. that lk woman p.nom=Osang ‘Osang is that woman.’
 Naonori Nagaya
(96)
(Copular complement of a pseudocleft clause) ?Ti/Diay=ubing ti=sangpet idiay=simbaan. nom=child nom=arrive there=church ‘The one who arrived at the church is the child.’
(97)
(Copular subject of a specificational clause) Ni=Piolo ti/?diay=bida idiay=pelikula. p.nom=Piolo nom=lead there=movie ‘The lead in the movie is Piolo.’
(98)
(Copular subject of a pseudocleft clause) Ni=Lorie ti/?diay=sangpet idiay=simbaan. p.nom=Lorie nom=arrive there=church ‘The one who arrived at the church is Lorie.’
Crucially, diay cannot create a complement clause (99), a parenthetical (100) or an exclamative (101). Ti is the only available article to mark these types of non-referential uses. (99) Na-rigat ti/*diay=ag-ayat ti=nobya ti=sabali (a tao). adj-hard nom=af-love nom=girlfriend nom=another lk person ‘It is hard to love the girlfriend of another.’ (100) Ti/*Diay=kuskusto, adda kayat=na kaniak. nom=truth have crush=3sg.nom 1sg.dat ‘In fact (lit. the truth is), s/he had a crush on me.’ (101) Ti/*Diay=gago! nom=stupid ‘How stupid it is!’
Thus, in Ilokano, the existing article ti has developed a wide variety of non-referential functions, while the newly-grammaticalized article diay only covers referential functions and a few non-referential functions.
5.3â•… Summary In this section we have suggested that our analyses of the Tagalog articles can be applied to the articles in Partido Bikol and Ilokano. Nominals marked by the nominative articles in these languages are often analyzed as referential (or definite). However, they are referential in some constructions but non-referential in other constructions. Also, we have confirmed that in these languages as well, the older articles can be used for different kinds of non-referential uses, while the emerging ones are mainly for referential nominals and have not obtained several non-referential uses.
Rise and fall of referentiality 
The development of the articles in these languages is not equal or homogeneous, however. Tagalog ang and Ilokano ti have obtained the function of creating an exclamative, whereas Partido Bikol ang has not acquired this function. Tagalog yung can be used even for complement clauses and parentheticals, but this is not the case with Partido Bikol su or Ilokano diay. From these observations, one may speculate that the semantic changes that are going on in the Tagalog articles system, namely, (i) the de-referentialization of the older article and (ii) the takeover of referential uses by the new article, are taking place in other Philippine languages too. Needless to say, further research and more data from other Philippine languages are necessary to determine whether this hypothesis is valid or not.
6.â•… Conclusions This paper has examined the referential and non-referential uses of articles and nominals in Philippine languages and proposed the hypothesis that their development has been proceeding from referential to non-referential uses. Tagalog ang and yung have their origin in demonstratives, the former being grammaticalized from *na in Proto Philippines and the latter from iyon in Tagalog. The older article ang has lost its original deictic meaning, and its functions have been shifting from referential uses to non-referential uses. Although it has long been believed that ang-phrases are always referential, our careful examination of ang-phrases in various constructions has revealed that they can be either referential or non-referential. They have even acquired the function of expressing discourse-related and subjective meanings as parentheticals and exclamatives. Moreover, it was observed that in the spoken corpus the referential uses of ang occur less frequently than its non-referential uses. In contrast, the newly-grammaticalized article yung is taking over the referential functions that ang is losing. In the Tagalog spoken corpus, yung almost always marks referential nominals. At the same time, it still keeps its original deictic meaning, unlike ang, and has obtained an additional nominalizing function which ang does not demonstrate. Taken together, there is an emerging semantic difference between ang and yung: ang is preferred for non-referential uses, and yung for referential uses. Lastly, by examining ang and su in Partido Bikol and ti and diay in Ilokano, this paper has also suggested that nominals marked by these articles, which have been believed to be always referential, can be either referential or non-referential, and that in these languages as well we see evidence of an existing article acquiring non-referential uses and a newer one emerging for referential nominals.
 Naonori Nagaya
References Blake, Frank R. 1925. A Grammar of the Tagalog Language. New Haven CT: American Oriental Society. Bloomfield, Leonard. 1917. Tagalog Texts with Grammatical Analysis. Urbana IL: University of Illinois. Cena, Resty M. 1977. Patient primacy in Tagalog. Presented at the LSA Annual Meeting, Chicago. Chafe, Wallace L. 1980. The Pear Stories: Cognitive, Cultural, and Linguistic Aspects of Narrative Production. Norwood NJ: Ablex. Croft, William. 1991. Syntactic Categories and Grammatical Relations: The Cognitive Organization of Information. Chicago IL: University of Chicago Press. Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: OUP. Curnow, Timothy Jowan. 2000. Towards a cross-linguistic typology of copula constructions. In Proceedings of the 1999 Conference of the Australian Linguistic Society, John Henderson (ed.), 〈http://www.linguistics.uwa.edu.au/research/als99/proceedings〉. Declerck, Renaat. 1988. Studies on Copular Sentences, Clefts, and Pseudo-clefts. Leuven: Leuven University Press. De Guzman, Videa P. 1988. Ergative analysis for Philippine languages: An analysis. In Studies in Austronesian Linguistics, Richard McGinn (ed.), 323–345. Athens OH: Center for Southeast Asia Studies, Center for International Studies, Ohio University. Diessel, Holger. 1999. Demonstratives: Form, Function and Grammaticalization [Typological Studies in Language 42]. Amsterdam: John Benjamins. Dixon, Robert Malcolm Ward. 2002. Copula clauses in Australian languages: A typological perspective. Anthropological Linguistics 44: 1–36. Englebretson, Robert. 2003. Searching for Structure: The Problem of Complementation in Colloquial Indonesian Conversation [Studies in Discourse and Grammar 13]. Amsterdam: John Benjamins. Greenberg, Joseph H. 1978. How does a language acquire gender markers? In Universals of Human Language, Vol. 3, Word Structure, Joseph H. Greenberg, Charles A. Ferguson & Edith Moravcsik (eds), 47–82. Stanford CA: Stanford University Press. Halliday, Michael Alexander Kirkwood. 1967. Notes on transitivity and theme in English. Part 2. Journal of Linguistics 3: 177–274. Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1991. Grammaticalization: A Conceptual Framework. Chicago IL: University of Chicago Press. Heine, Bernd & Kuteva, Tania. 2007. The Genesis of Grammar. Oxford: OUP. Higgins, Francis Roger. 1979. The Pseudo-cleft Construction in English. New York NY: Garland. Himmelmann, Nikolaus P. 1991. The Philippine Challenge to Universal Grammar. Köln: Institut für Sprachwissenschaft. Himmelmann, Nikolaus P. 1997. Deiktikon, Artikel, Nominalphrase: Zur Emergenz syntaktischer Struktur. Tübingen: Niemeyer. Himmelmann, Nikolaus P. 2004. Tagalog (Austronesian). In Morphology: A Handbook on Inflection and Word Formation, Vol. 2, Geert Booij, Christian Lehmann & Joachim Mugdan (eds), 1473–1490. Berlin: de Gruyter. Himmelmann, Nikolaus P. 2005a. The Austronesian languages of Asia and Madagascar: Typological characteristics. In The Austronesian Languages of Asia and Madagascar, Alexander Adelaar & Nikolaus P. Himmelmann (eds), 110–181. London: Routledge.
Rise and fall of referentiality 
Himmelmann, Nikolaus P. 2005b. Tagalog. In The Austronesian Languages of Asia and Madagascar, Alexander Adelaar & Nikolaus P. Himmelmann (eds), 350–376. London: Routledge. Himmelmann, Nikolaus P. Ms. Notes on noun phrase structure in Tagalog. In Special issue of Australian Journal of Linguistics, Michael Ewing & Simon Musgrave (eds). Hopper, Paul J. & Martin, Janice. 1987. Structuralism and diachrony: The development of the indefinite article in English. In Papers from the 7th International Conference on Historical Linguistics [Current Issues in Linguistic Theory 48], Anna Giacalone Ramat, Onofrio Carruba & Guiliano Bernini (eds), 295–304. Amsterdam: John Benjamins. Hopper, Paul J. & Thompson, Sandra A. 1984. The discourse basis for lexical categories in universal grammar. Language 60: 703–752. Kaufman, Daniel. 2005. Aspects of pragmatic focus in Tagalog. In The Many Faces of Austronesian Voice Systems: Some New Empirical Studies, I Wayan Arka & Malcom Ross, (eds), 175–196. Canberra: Pacific Linguistics. Kaufman, Daniel. This volume. Exclamatives and temporal nominalizations in Austronesian. In Yap, Grunow-Hårsta & Wrona (eds). Kroeger, Paul. 1993. Phrase Structure and Grammatical Relations in Tagalog. Stanford CA: CSLI. Lambrecht, Knud. 2001. A framework for the analysis of cleft constructions. Linguistics 39: 463–516. LaPolla, Randy J. & Poa, Dory. 2005. Direct and indirect speech in Tagalog. Workshop on Direct and Indirect Speech, Research Centre for Linguistic Typology, La Trobe University, June 8, 2005. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. McFarland, Curtis D. 1976. A Provisional Classification of Tagalog Verbs [Study of Languages and Cultures of Asia and Africa Monograph Series No.8]. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies. McFarland, Curtis D. 1978. Definite objects and subject selection in Philippine languages. In Studies in Philippine Linguistics, Vol. 2, Casilda Edrial-Luzares & Austin Hale (eds), 139–182. Manila: Linguistic Society of the Philippines. Mikkelsen, Line. 2004. Specifying Who: On the Structure, Meaning, and Use of Specificational Copular Clauses. Ph.D. dissertation, University of California Santa Cruz. Mikkelsen, Line. 2005. Copular Clauses: Specification, Predication, and Equation [Linguistik Aktuell/Linguistics Today 85]. Amsterdam: John Benjamins. Moro, Andrea. 1997. The Raising of Predicates: Predicative Noun Phrases and the Theory of Clause Structure. Cambridge: CUP. Nagaya, Naonori. 2005. Information structure and constituent order in Tagalog. Paper presented at the 2005 International Course and Conference on Role and Reference Grammar, Academia Sinica, Taipei, June 26–30. Nagaya, Naonori. 2007a. Information structure and constituent order in Tagalog. Language and Linguistics 8: 343–372. Nagaya, Naonori. 2007b. Rethinking referentiality of Tagalog ang-phrases. Paper presented at the 2007 annual conference of the Australian Linguistic Society, Adelaide, September 26–28. Nagaya, Naonori. 2007c. Referentiality of Tagalog ang-phrases revisited: referential, definite, specific, or none of the above? Paper presented at the Linguistic Colloquium, Rice University, Houston, November 8. Nagaya, Naonori. 2009. The middle voice in Tagalog. Journal of the Southeast Asian Linguistics Society 1: 159–188.
 Naonori Nagaya Nagaya, Naonori. ms. Rethinking the referentiality of Tagalog ang-phrases. Rice University, Houston, Texas. Nishiyama, Yuji. 2003. Nihongo Meishiku no Imiron to Goyooron [Semantics and Pragmatics of Noun Phrases in Japanese]. Tokyo: Hituzi Syobo. Payne, Thomas E. 1982. Role and reference related subject properties and ergativity in Yup’ik Eskimo and Tagalog. Studies in Language 6: 75–106. Payne, Thomas E. 1997. Describing Morphosyntax: A Guide for Field Linguists. Cambridge: CUP. Potsdam, Eric. This volume. Expressing exclamatives in Malagasy. In Yap, Grunow-Hårsta & Wrona (eds). Ramos, Teresita V. 1971. Tagalog Structures. Honolulu HI: University of Hawaii Press. Reid, Lawrence A. 1978. Problems in the reconstruction of Proto-Philippine construction markers. In Second International Conference on Austronesian Linguistics: Proceedings, Fascicle I, Western Austronesian, Stephen A. Wurm & Lois Carrington (eds), 33–66. Canberra: Pacific Linguistics. Reid, Lawrence A. 2002. Determiners, nouns, or what? Problems in the analysis of some commonly occurring forms in Philippine languages. Oceanic Linguistics 41: 295–309. Reid, Lawrence A. 2006. On reconstructing the morphosyntax of Proto-Northern Luzon. Philippine Journal of Linguistics 37: 1–63. Reid, Lawrence A. & Hsiu-chuan Liao. 2004. A brief syntactic typology of Philippine languages. Language and Linguistics 5: 433–490. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 281–337. Dordrecht: Kluwer. Ross, Malcolm. 2002. The history and transitivity of western Austronesian voice and voicemarking. In The History and Typology of Western Austronesian Voice Systems [Pacific Linguistics, No. 518], Fay Wouk & Malcolm Ross (eds), 17–62. Canberra: Australian National University. Rubino, Carl Ralph Galvez. 1997. A Reference Grammar of Ilocano. Ph.D. dissertation, University of California, Santa Barbara. Schachter, Paul. 1976. The subject in Philippine languages: Topic, Actor, Actor-Topic, or none of the above. In Subject and Topic, Charles N. Li (ed.), 491–518. New York NY: Academic Press. Schachter, Paul. 1977. Reference-related and role-related properties of subjects. In Syntax and Semantics, Vol. 8: Grammatical Relations, Peter Cole & Jerrold M. Sadock (eds), 279–306. New York NY: Academic Press. Schachter, Paul & Otanes, Fe T. 1972. Tagalog Reference Grammar. Berkeley CA: University of California Press. Thompson, Sandra A. & Mulac, Anthony. 1991. The discourse conditions for the use of the complementizer that in conversational English. Journal of Pragmatics 15: 237–251. Traugott, Elizabeth Closs. 1989. On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. Wrona, Janick. This volume. Non-derived stand-alone nominalization in Japanese. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha. This volume. Referential and non-referential uses of nominalizers in Malay. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha, Grunow-Hårsta, Karen & Wrona, Janick (eds.) This volume. Nominalization in Asian Languages: Diachronic and Typological Perspectives. Amsterdam: John Benjamins.
Referential and non-referential uses of nominalization constructions in Malay Foong Ha Yap
Hong Kong Polytechnic University This paper examines three versatile morphemes – yang, -nya and punya – that contribute to the formation of nominalization constructions in Malay. In particular, we examine how these morphemes give rise to both referential and non-referential uses of nominalization constructions. We also highlight that nominalizing elements in Malay can be derived from noun phrase markers such as third person genitive pronoun -nya (e.g. makan-nya [eat-3sg.gen] ‘his eating’) as well as general or semantically bleached nouns such as (em)punya (‘master, owner, possessor’) that can easily develop into light nouns equivalent to English indefinite pronoun one (e.g. tak rosak punya [neg spoil nmz] ‘the ones that are not spoiled’. We further show that various factors such as lexical source, morphosyntactic constraints, and in some cases language contact, help to shape the grammaticalization pathway(s) of each morpheme, particularly in terms of their range of grammatical and pragmatic functions.
1.â•… Introduction In the course of daily communication, we often need to refer not only to concrete entities such as animate beings, inanimate objects, and locations (i.e. first order ontological entities according to Lyons (1977: 442), or very simply the who, what, and where elements in news reports and the someone, something, and someplace elements in narratives); often we also need to refer to abstract entities such as events and propositions (i.e. second and third order ontological entities). This often calls for more complex nominal constructions such as gerundives, infinitives, and complement clauses. In some languages, these constructions (which are deverbalized or nominalized to varying degrees) can often be extended to express the speaker’s stance toward a given event, situation or proposition. Thus for example, while first order entities are often expressed as (lexicalized) noun phrases (e.g. a man, a ploughman, a fisherman), second and third order entities on the other hand are often expressed as nominalized clauses (e.g. flooding the fields, to mend the nets, (the fact) that they survived the storm), and expressions of speaker mood or stance (such as surprise or disbelief) are
 Foong Ha Yap
often realized via extended, non-referential uses of nominalized clauses, as in the Malay expression Rajin-nya budak ini (diligent-3sg.gen this child) ‘↜So diligent this child!’ or ‘↜Such diligence this child!’, where the nominalizing element -nya takes on a focus marking function to highlight its host constituent, in this case ‘the diligence of this child’. Numerous studies have found that in many languages the same morpheme that serves as a nominalizer frequently serves other functions as well, among them genitive and relative clause marking (Matisoff 1972; Delancey 1986; Herring 1991; Noonan 1997) and speaker mood or stance marking (e.g. Yap, Matthews & Horie 2004; Watters 2008; Yap & Matthews 2008). Detailed analyses of versatile nominalizers have been undertaken for many Tibeto-Burman languages, including Atong (van Breugel 2006), Burmese (Simpson 2008), Bodic and Tamangic languages (Noonan 2008, this volume; Grunow-Hårsta this volume), Kiranti (Bickel 1999), Mongsen Ao (Coupe 2006), Rawang (LaPolla 2008), Numhpuk Singhpo (Morey 2006, this volume), Nuosu Yi (Liu & Gu this volume), and Thulung Rai (Lahaussois 2003). Among the East Asian languages, similar analyses of versatile nominalizers have been undertaken for Chinese (e.g. Yap, Choi & Cheung 2010; Yap & Wang this volume), Japanese and Okinawan (Horie 2008, this volume; Shinzato this volume; Shibasaki to appear a & b), and Korean (Rhee 2008, this volume). The present paper examines nominalization phenomena from yet another language family, namely Austronesian. More specifically, we examine three highly versatile morphemes – namely, yang, -nya and punya – from the Malay variety spoken in the states of Perak and Selangor in peninsular Malaysia.1 These morphemes either signal (i.e. identify) or otherwise form nominalization constructions. For example, as a noun phrase marker (or referentiality marking device), third person genitive -nya can signal that the constituent (i.e. word, phrase, or clause) preceding it is a nominalization construction, as in (1). Deploying a different strategy, punya, derived from a general noun meaning ‘master, owner, possessor’, can function as a semantically light head noun (in this case, a nominalizer) when preceded by a modifying clause, as in (2).2 (1) rajin-nya diligent-3sg.gen ‘his diligence’
.╅ These three morphemes are fairly productive, and their usage is attested throughout the Malay-speaking region. Punya has received less attention in the literature, but its use is widespread in many other Malay varieties as well (see Gil 1999). .╅ Light nouns can function as nominalizers, like zhe in Classical Chinese (Aldridge 2008; Yap & Wang this volume); they can also function as classifiers.
(2)
Referential and non-referential uses of nominalization constructions in Malay 
(yang) nak basuh punya, letak dekat sini the.one want wash nmz put near here lit. ‘(the ones) to be washed ones, put (them) near here’3 ‘The ones to be washed, put them here.’
In addition to their nominalizing (hence referential) function, the three versatile morphemes yang, -nya and punya each serves a number of non-referential functions as well. The focus of this paper is to identify these various functions and their grammaticalization pathways, paying close attention to how they interact with each other, and in due course carve out distinct semantic niches for themselves. This paper is organized as follows: Section 2 focuses on the multiple functions and possible diachronic development of yang, based on a review of extant literature (including recent works such as van Minde 2008; Englebretson 2008 and Reid 2010) and a corpus analysis of a classical Malay text from the 14th century. Section 3 similarly discusses the range of functions and grammaticalization pathways of -nya, drawing on insights from Englebretson’s (2003) conversational analysis of colloquial Indonesian. Section 4 examines punya in similar fashion, based on descriptive fieldwork on colloquial Malay (e.g. Gil 1999) and data from classical texts made electronically available via the Malay Concordance Project housed at the Australian National University (courtesy of Ian Proudfoot); in this section, we will also focus on the impact of language contact on the rise of punya as a nominalizer and stance marker. Section 5 compares these three versatile morphemes in terms of their functional distribution, noting in particular the morphosyntactic characteristics that give rise to the semantic uniqueness of each morpheme. Section 6 concludes the paper.
2.â•… Nominalizer yang Previous scholars (e.g. Gerth van Wijk 1909; van Ophuijsen 1910; Mees 1969; Verhaar 1983; Kaswanti Purwo 1983; Simin 1988; Steinhauer 1992; van Minde 2008; Englebretson 2008) have identified a wide range of grammatical functions
.â•… As will be further elaborated in §4, native speakers of Malay often add nominalizer yang, producing a double nominalization construction of the type [yang VP punya], literally ‘the one(s) that VP one(s)’. While the use of yang is native to the Malay language, the use of punya is arguably induced by extensive contact with southern Chinese dialects, which explains why punya occurs with high frequency as a nominalizer in Bazaar (i.e. ‘market variety’) Malay.
 Foong Ha Yap
for yang, including relative clause marker, complementizer, theme/topic marker, and ‘defining article’ (particularly in contrastive contexts), as illustrated in (3) to (6) respectively.4 (3) Buang saja bunga yang sudah layu itu. throw just flower rel perf wither dem:distal ‘Just throw away those flowers that have withered.’ (4) Dia tahu (yang) orang ini tak boleh di-percaya-i. 3sg know comp person this neg can pass-trust-trans ‘S/he knows (that) this person cannot be trusted.’ (5) Yang dia, sedikit pun dia tak kesah. top 3sg little.bit also 3sg neg care ‘As for him, he didn’t even care the slightest bit.’ (6) Yang bapa ketawa; yang ibu menangis. ‘art’ father laugh ‘art’ mother cry ‘The father laughed; the mother cried.’5
Not much is known about the origin of yang. Mees (1969) has argued for a link between Classical Malay yang and pre-Classical Malay yam, the latter identified as a definiteness marker in seventh-century Old Malay inscriptions from the Palembang and Bangka region in Indonesia. Adelaar (1985) has suggested the possibility that yang is derived from ia + -ŋ (where ia is the third person pronoun while -ŋ is a ligature or ‘linker’ preceding a nominal expression). This analysis suggests a possible pronominal source for yang.6 More recently, Reid (2010) reconstructs the following development for yang:7
.â•… Except where specified, all the Malay examples used in this paper are of the Straits variety spoken in Perak and Selangor, West Malaysia. Terminology sometimes differs across scholars. For example, Englebretson (2008) uses the term ‘referring expression’ where we use ‘theme/ topic marker’ and ‘defining article’ (see (5) and (6) respectively). .â•… Possibly an extension of the topic marking use of yang, as shown below: (i)
Yang bapa, ketawa; yang ibu, menangis. top father laugh top mother cry ‘As for the father, (he) laughed; as for the mother, (she) cried.’
.â•… Other scholars have identified nan, nang, nyang, iang and jang as either dialectal cognates or orthographic variants of yang (see van Minde 2008 for a more detailed discussion). .â•… The following account is adapted from Lawrence A. Reid’s e-communication (October 12, 2010).
Referential and non-referential uses of nominalization constructions in Malay 
The nominalizer *yaŋ developed from three Proto-Malayo-Polynesian (PMP) morphemes, which now are reflected in Tagalog iyan ‘that’ (medial demonstrative):
PMP *ôi ‘case-marking preposition’ PMP *a ‘common noun nominal specifier’ PMP *=n ‘post-vocalic enclitic ligature’
In PMP the ligature *=n became *=ŋ by assimilation to a following *ka- counter in mensural numeral constructions (see Reid 2010, Table 10, p. 470), and was subsequently generalized as the (post-vocalic) ligature in other attributive constructions, and is today found frozen on many nominal specifiers (such as Tagalog ang) and demonstratives (such as Tagalog yong and Malay yang).
Reid’s (2010) account thus suggests that nominalizer yang has a demonstrative origin (i.e. a portmanteau formed with case marker *ôi, common noun nominal specifier *a, and post-vocalic ligature *=n). While some languages (e.g. the Sino-Tibetan, Japanese and Korean languages) often recruit general or semantically bleached ‘light’ nouns as nominalizers (see Yap & Wang this volume; see also DeLancey 1986; Horie 1998; Rhee 2008), other languages – including many from the Austronesian language family – often recruit noun phrase markers such as demonstratives and case markers to signal nominalization constructions (see especially Nagaya this volume; see also Sung this volume). Reid’s demonstrative *yaŋ > nominalizer yang account is thus consistent with robust crosslinguistic tendencies (see Malchukov 2006; Yap, Grunow-Hårsta & Wrona this volume). As seen in (7), yang is frequently used in equative constructions, often with contrastive focus and cleft-like effect, where the intended reading is ‘(S)he (not me/you/ her/someone else) is the one who is getting married tomorrow’.8 (7) Dia [yang nak nikah esok]. 3sg nmz fut marry tomorrow ‘(S)he is the one that’s getting married tomorrow.’
Equative constructions such as (7) above highlight another major function of yang, namely, nominalization.9 Note that in equative constructions, the presence of yang signals that the predicate is to be treated as nominal and hence referential. The nominal predicate introduced by yang bears the meaning ‘the one who VPs’, e.g. ‘the one who is getting married’.
.â•… Englebretson (2008: 13–14) also refers to this type of equative use of yang as a cleft construction. .â•… Mees (1969; see also van Minde 2008) identifies definiteness marking and nominalization as the two major functions for yang.
 Foong Ha Yap
As highlighted in cleft constructions such as (8), the yang nominalization construction could be an agent nominal, as in (8a), or a patient nominal, as in (8b). (8) a. [Yang nak nikah] bukan aku. nmz want marry neg 1sg ‘The one who wants to get married is not me.’ b. Sebaliknya, [yang di-paksa nikah ’tu] lah aku. on.the.contrary nmz pass-force marry dem foc 1sg ‘On the contrary, the one that is being forced to marry is me.’
These nominalization uses of yang were already attested in the earliest known Classical Malay texts, as seen in (9) from Hikayat Bayan Budiman (14th century). 10 (9) Adakah [yang sudah mati itu] boleh hidup pula? exst.q nmz already die/dead dem can live again ‘Is it (the case) that those that have died can live again?’
(Bayan 15: 2)
As noted in (3) earlier, yang is also used as a relative clause marker. An interesting question is what relationship the yang relative clause construction may have with the yang nominalization construction. It has been noted crosslinguistically that the relative clause construction often emerges via parataxis followed by syntaxis (i.e. subordination or embedding; see Givon 2009). In other words, two related clauses that are initially packaged under two intonation contours often become compressed or condensed under a single intonation contour. As illustrated in (10) below, where such compressions involve clausal integration of a yang nominalization (e.g. yang sudah layu itu ‘those that have wilted’) as an embedded construction following an object complement (e.g. bunga itu ‘those flowers’), a headed relative clause emerges (bunga (*itu) yang sudah layu itu ‘those flowers (*those) that have wilted’).11 This development involving intonational compression and clausal integration point to a
.â•… This hikayat (or story) is a collection of popular tales that originally came from the Sanskrit Sukasaptati (The Parrot’s Seventy Tales). The original Malay version was translated from a Persian adaptation by one Kadi Hassan in 773 AH (i.e. 1371 AD). The text I analyze comes from the Oxford University Press edition of Winstedt (1966), which is based on Manuscript L (1852) and Manuscript R (1849), now both lost. This text is made electronically accessible via the Malay Concordance Project at the Australian National University, with the spelling modernized. .â•… Note that Malay demonstratives are post-nominal (i.e. they follow the noun they specify) and we see retention of the demonstrative itu (‘that, the’) at the end of the relative clause construction, while English demonstratives are pre-nominal (i.e. they precede the noun they specify) and we thus see instead retention of the demonstrative those at the beginning of the relative clause construction.
Referential and non-referential uses of nominalization constructions in Malay 
close relationship between nominalization and relativization, with the latter arising as a secondary development from the former (see also Shibatani 2008; Shibatani & Awadh Khaled 2009). (10) Buang saja bunga itu. Yang sudah layu itu. → Buang saja bunga yang sudah layu itu.
(2 intonation units) (1 intonation unit)
‘Throw away those flowers. Those that have wilted. → Throw away those flowers that have wilted.’
(2 intonation units) (1 intonation unit)
There is textual evidence from Classical Malay that supports this clausal integration analysis involving a development from parataxis to syntaxis. As seen in (11) below, from narratives in Hikayat Bayan Budiman, (a) shows evidence of paratactic use of yang constructions, (b) shows evidence of a bridge context between parataxis and syntaxis, where the yang construction can be ambiguously interpreted as either an appositive nominalization construction or a non-restrictive relative clause, and (c) where there is clear evidence of relative clause usage of yang constructions. Note that paratactic readings arise from intonational pauses, often indicated by a comma, preceding the yang construction, as seen in (11a) and (11b). (11) a.
Hai saudagar, bunga apakah ini, yang saudagar pakai itu? intj merchant flower what.q this nmz merchant wear that ‘O merchant sir, what flower is this, that which you are wearing (pinned to your garment)?’ (Bayan 27: 4)
b. Hai raja segala burung di dalam alam ini, yang amat intj ruler all bird loc inside world this yang very
budiman dan memberi manfa`at segala pekerjaan! generous and give benefit every work
‘O ruler over all the feathered creatures in this world,
(i) one that is supremely generous and rewards every work!’
(ii) who is supremely generous and rewards every work!’ (Bayan 110: 31)
c.
Raja ini dari sebuah negeri yang asing; king this from one: cl country rel foreign
siapa akan tahunya? who will know=3sg.gen
‘This ruler is from a foreign country (lit. a country that is foreign); who will know him?’ (Bayan 30: 28)
The intimate link between nominalization and relativization can also be seen in constructions involving general nouns (e.g. orang ‘person’; barang ‘thing’) as ‘light’ head nouns followed by yang modifying clauses, with nominalizer yang reanalyzed as a relativizer, as seen in (12) below.
 Foong Ha Yap
(12) [barang yang sudah kita pertetapkan], ‘thing’ rel already 1pl.incl caus.fix.trans tiadalah dapat kita ubahkan lagi neg.exst=foc able 1pl.incl change.trans again ‘that (< ‘the things’) which we have already decreed, we cannot rescind’ (Bayan 231: 19)
It is interesting to note that the relativization process can be recursive, to some extent. As seen in (13) below, when apposed to a lexical head noun such as sahabat (‘friend’), the light noun barang ‘thing’ in the modifying expression that follows can be further grammaticalized – in combination with relativizer yang – to form a portmanteau relativizer barang yang ‘that’.12 (13) Maka anaknya saudagar itu pun … then child=3sg.gen merchant dem also pergi mencari sahabat [barang yang berkenan pada hatinya] go af.find friend ‘thing’ rel agreeable to heart=3sg.gen (i) ‘Then the merchant’s son went in search of friends, those that please him.’ (ii) ‘Then the merchant’s son went in search of friends that please him.’ (Bayan 221: 22)
In addition to the extension of yang constructions from its core nominalizing function to a secondary adnominal (or noun-modifying) function, we also see evidence of yang constructions being used in topic marking contexts, as in (14). (14) Janganlah tuan hamba syak hati lagi. do.not=prt master 1sg(humble form) suspicious.heart any.longer ‘Be not suspicious any longer, sir.’ Yang kami seorang diri ini top 1pl.incl one.person self dem tiadalah hamba ubahkan janji. neg.have-prt 1sg(humble form) change.trans promise ‘As for me, I will not go back on my promise.’
(Bayan 224: 16)
‘Headless’ yang constructions (i.e. yang nominalizations) are also very frequently found in cleft and other types of contrastive focus constructions, as in (15) below. In this excerpt from Hikayat Bayan Budiman, a Malay version of the Sanskrit Sukasaptati
.â•… Note here that barang (derived from ‘thing’) can also be used to refer to ‘person(s)’, clearly indicating that its use has extended to a very general domain, hence its categorization as a ‘light noun’.
Referential and non-referential uses of nominalization constructions in Malay 
(‘The Parrot’s Seventy Tales’), the parrot with its lucid observations of life recounts the story of a carpenter, a weaver, a goldsmith and an ascetic (or holy man) and their fight over a beautiful statue that had come to life. The carpenter had carved her, the weaver had dressed her in fine fabrics, the goldsmith had dressed her in fine jewelry, and the ascetic had prayed, if God be willing, that she be imbued with life. Captivated by her beauty once the statue came to life, each of the four is now fighting to claim her as his own. Note here their use of yang nominalizations in contrastive focus constructions – in the sense of “It is I (and not any of you three) that is her master, for I am the one that did W/X/Y/Z (to/for her).” But as we shall see below, there is again structural ambiguity, which then allows for a cleft focus interpretation as well. (15) a.
Maka kata serimala, then say carpenter
‘Akulah [yang empunya dia], 1sg=foc nmz own 3sg
karena aku [yang perbuat dia].’ because 1sg nmz caus.make 3sg
‘Then said the carpenter, “It is I who am her master (lit. the one that owns her), (i) for I am the one that carved her.”’ (ii) for it is I that carved her.”’
(Bayan 51: 31)
b. Maka jawab pandai bertenun, then reply one.skilled.at.weaving
‘Aku juga [yang empunya dia], 1sg too nmz own 3sg
karena aku [yang memberi pakaian].’ because 1sg nmz af.give clothing
‘Then replied the weaver, “I too am her master (lit. the one that owns her), (i) for I am the one that clothed her.”’ (ii) for it is I that clothed her.”’ (Bayan 52: 1–2)
c.
Maka kata pula pandai emas, then say in.addition goldsmith
‘Aku [yang empunya dia], 1sg nmz own 3sg
karena aku [yang memberi pakaian emas akan dia].’ because 1sg nmz af.give clothing gold to 3sg
‘Then the weaver added, “I am her master (lit. the one that owns her), (i) for I am the one that dressed her in gold.”’ (ii) for it is I that dressed her in gold.”’ (Bayan 52: 3–4)
 Foong Ha Yap
d. Maka kata zahid itu, then say holy.man dem
‘Bukan siapa pun [yang empunya dia], melainkan aku jua, neg anyone nmz own 3sg except 1sg only
karena aku [yang memohonkan nyawanya kepada Allah ta’ala].’ because 1sg nmz af.ask.trans life=3sg.gen to God the.exalted.one
‘Then the said the holy man, “None is her master (lit. the one that owns her), save me only, (i) for I am the one that asked God Almighty to give her life.”’ (ii) for it is I that asked God Almighty to give her life.”’ (Bayan 52: 5–7)
Worth noting is that the nominalizing function of yang can also extend its scope, in that we can obtain not only an argument (or participant) nominalization reading, as seen in interpretations (i) for (15a–d), but we can also obtain a cleft or contrastive focus reading, as in interpretations (ii) for the same set of examples. What interpretations (i) highlight is that focus effects can be achieved through stressed prosody; what interpretations (ii) further highlight is that nominalizer yang in cleft constructions can come to be associated with focus functions, though not quite as grammaticalized as the French (ne) … pas negator construction. That is, although nominalizer yang colludes with stressed prosody to reinforce the contrastive focus reading in cleft constructions, yang itself has not abandoned its role as nominalizer to become a focus marker. Rather, as in the case of complementizer yang, discussed earlier, there is both syntactic and semantic scope expansion, such that yang comes to be associated with signaling a complementation interpretation – that is, the focus could shift from the argument (or participant) nominalization reading in (i) and take on the complementation reading in (ii). In other words, we shift from ‘I am the one that did W/X/Y/Z’ to ‘It is I that did W/X/Y/Z’.13
.â•… It is worth noting that focus effects in cleft constructions can be signaled or reinforced by a wide range of strategies. As seen in (15a), the particle lah could serve as focus marker (Akulah yang empunya dia ‘It is I that owns her’. As seen in (15b), adverbial juga ‘too’ can ride on the wave of the focus effect in the preceding cleft construction (Aku juga yang empunya dia ‘I too am the one that owns her’). As seen in (15d), preposed sentential negator bukan also has a strong focus effect (Bukan siapa pun yang empunya dia, melainkan aku jua ‘It is not anyone that owns her (hence ‘None is her owner’), except me only.’ Crucially, however, as seen in (15c), the stressed prosody alone can also achieve the desired focusing effect. This is not to say that the role of the other constituents – e.g. focus particle lah, adverbs such as juga, preposed sentential negator bukan – is not important. On the contrary, it highlights that grammatical morphemes, particularly versatile ones, are good candidates for hosting stressed prosody, and they often go on to acquire pragmatic functions such as focus marking as well.
Referential and non-referential uses of nominalization constructions in Malay 
In sum, we see ample evidence of yang constructions serving both referential and non-referential functions. Perhaps derived from a demonstrative (which according to Reid (2010) could itself have been a combination of referentiality marking devices such as case marker *ôi and common noun nominal specifier *a, plus post-vocalic ligature *=n), yang was already a highly versatile nominalizer in Classical Malay. As highlighted in Figure 1, when apposed to a co-referential lexical noun phrase, the yang nominalization construction readily assumes a noun-modifying function, thus becoming a relative clause. When the yang construction is preposed to topic position, nominalizer yang assumes the function of theme/topic marker. Here yang gets to introduce an element of perspectivization, and with it an element of speaker subjectivity as well. Yang nominalization constructions are also very frequently found in equative constructions, often with cleft or other contrastive focus effects, and though yang remains a nominalizer still, it additionally acquires the pragmatic focusing function of its context. ‘Definiteness’ marker yang [yang NP] Topic marker yang [yang (NP) VP], NP VP [NP[yang (NP) VP]], NP VP Source uncertain (perhaps a demonstrative)
Nominalizer yang [yang [ __i VP]] [yang [(NP) [V __i]]] Yang in equative construction Relative clause marker yang NPi [yang [ __i VP]]
NP(lah) (no copula) [yang (NP) VP] NP(lah) (no copula) NP[yang (NP) VP]
NPi [yang [(NP) [V __i]]]
Figure 1.╇ Grammaticalization pathways of Malay nominalizer yang 14
3.â•… Nominalizer -nya In addition to yang, there is another nominalizing morpheme in Malay of ancient origin, namely, enclitic -nya. It occurs with high frequency and serves a wide range of
.╅ The thicker dotted arrows indicate that the relative clause constructions, though possible in contemporary colloquial Malay, were not attested in the topic marking and cleft constructions in Classical Malay texts such as Hikayat Bayan Budiman. The finer dotted arrows indicate that the definiteness reading associated with yang predates Classical Malay.
 Foong Ha Yap
functions. In his analysis of colloquial Indonesian conversations, Englebretson (2003) identifies the following major functions for -nya: third person genitive enclitic, ‘identifiability’ (i.e. definiteness) marker, nominalizer, and adverbial marker. These functions are here illustrated from (16) to (19) using colloquial Malay data. (16) Potong saja ekornya. cut just tail=3sg.gen ‘Just cut off its tail.’ (17) Kainnya terlalu halus. cloth=3sg.gen excessive fine ‘The fabric is too fine.’ (< ‘Its fabric is too fine.’) (18)
Budak ’ni makannya tak ikut masa. child this eat=3sg.gen neg follow time Lit. ‘As for this child, his eating is not according to schedule.’ Intended meaning: ‘This child is not eating regularly.’
(19) Biasanya dia awal. usual=3sg.gen 3sg early ‘Usually he’s early.’
Note the cline of abstraction in the above examples. In (16), -nya appears as a third person genitive enclitic (i.e. a possessive pronoun meaning ‘its’) following the possessee noun ekor ‘tail’, yielding a possessive construction ekor-nya ‘its tail’. In (17), -nya appears after a lexical noun kain ‘cloth’ in a context where the focus is now less on the possessive relationship and more on the identifiability of the possessee noun, thus yielding the interpretation ‘the cloth/fabric’ rather than ‘its cloth/fabric’.15 Englebretson (2003: 162) identifies this usage of enclitic -nya as a semantic extension from possessive to identifiability marker. In (18), -nya appears as an enclitic after the verb makan ‘eat’, and yields an event nominal interpretation makan-nya (equivalent to English gerundive ‘his eating’/ ‘the eating’). In this context, third person genitive -nya signals a nominalization construction. Such usage of -nya is productive in colloquial speech, which facilitates its reinterpretation as a nominalizer, given the right context. Unlike nominalizer yang, which focuses on signaling an argument referent (e.g. yang makan ‘the one that is eating’; yang di-makan ‘the thing that is being eaten’), nominalizer -nya is able to focus on an event referent (e.g. makan-nya ‘his eating’/*‘the eater’/*‘the thing eaten’). Nominalizer -nya thus provides us with a means to comment on events and situations as abstract
.â•… An anonymous reviewer has pointed out that there are parallels here with other languages, including Turkic languages (see for example Comrie 1988: 465; Siewierska, Rijkhoff & Bakker 1998: 811–12, Notes 14 and 33).
Referential and non-referential uses of nominalization constructions in Malay 
nominals. As Englebretson (2003: 168) puts it: “Because of its close association with nominal expressions, when it [i.e. -nya] is affixed to a lexical item from a different class, this item takes on the characteristics of a noun.” 16 Also noted in Englebretson (2003: 170–171), enclitic -nya often combines with various lexical items to form adverbials such as biasanya ‘normally, usually’ (< biasa ‘normal, usual’), as shown in (19) above. Other examples include akhirnya ‘finally’ (< akhir ‘end’), agaknya ‘presumably, perhaps’ (< agak ‘guess’), and silapnya ‘unfortunately’ (< silap ‘mistake’). Note that these adverbial -nya expressions often reflect speaker moods as well, particularly when found in clause-initial position, as seen in (20), (21) and (22), where enclitic -nya can readily serve as a convenient landing site for prosodic embellishments – such as vowel lengthening and emphatic rise-fall pitch contour – to reflect various shades of speaker’s emotive expressions, ranging from surprise or disbelief, to regret or resignation, or even annoyance or disgust. (20) Akhirnya::: kita kalah. end=3sg.gen 1pl lose ‘In the end, we lost.’ (21) Agaknya::: dia tak suka engkau. guess=3sg.gen 3sg neg like 2sg ‘Perhaps, s/he doesn’t like you.’ (22) Silapnya::: aku lupa beritahu mereka. mistake=3sg.gen 1sg forgot inform 3pl ‘Unfortunately/Regrettably, I forgot to mention it to them.’
Worth noting is that we can still retrieve a nominal interpretation in many of these adverbial uses of -nya – for example, ‘usually (< his/the usual practice), he comes early’ (19); ‘finally (< the final analysis), we lost’ (20); ‘perhaps (< my guess), (s)he doesn’t like you’ (21);17 ‘unfortunately/regrettably (< the mistake/error), I forgot to inform him’ (22). Here I suggest that absence of the copula, which is typical of Austronesian languages (Pustet 2003), may have facilitated a reanalysis of -nya nominal expressions in topic or clause-initial position into -nya adverbial ones. That is, we obtain reinterpretations
.â•… This could have come about through a process of coercion in grammar (see possible parallels in Huang and Ahrens (2003), in their discussion on noun phrase markers such as classifiers in Mandarin). .â•… Agaknya in (21) is interesting because we see here an extension of -nya from third person reference to first person reference (‘my guess’), and from thence to such readings as ‘I suppose’ and ‘maybe, perhaps’. Extensions from third person to second and first person pronominal usage was reported in Englebretson (2003), and is attested in other languages as well (e.g. Classical Chinese pronoun zhi).
 Foong Ha Yap
such as follows: ‘the usual practice (is), he comes early’ > ‘usually, he comes early’. Similar reinterpretations obtain for the other examples as well. As noted in Fischer (2007), sentential adverbs are often well-suited to express speaker mood functions. It is not surprising therefore that we find adverbials formed with enclitic -nya being heavily involved in the marking of speaker stance, arguably via the following grammaticalization pathway: third person genitive > (definiteness marking) > nominalizer > adverbial marker > stance marker. The use of -nya in the production of stance constructions is in fact very productive, and Englebretson’s analysis of colloquial Indonesian (2003: 172–185) reveals at least three major stance functions associated with -nya constructions. These include mental/emotional attitude, as in (23), speaker assessment of interactional relevance to ongoing discourse, as in (24), and marker of evidentiality, as in (25). Note that a nominalization interpretation can still be retrieved for bagus-nya in (23), where the strong nominal reading ‘the good thing (is), they came’ can give rise to an evaluative reading such as ‘it is good that they came’. Similarly, for maksud-nya in (24), we can still easily trace a nominalization link in ‘the meaning (is), we don’t have to worry at all’, with a delicate counter-balancing act between semantic bleaching and pragmatic strengthening giving rise to a discourse particle-like usage in the sense of ‘meaning (or that’s to say), we don’t have to worry at all’. A nominalization link is less direct but nevertheless still accessible for (25), with nampak-nya ‘apparently’ being derived from ‘the appearance (is)’ which in turn is derived from ‘what’s seen (is)’. (23) Bagusnya mereka datang. good=3sg.gen 3pl come ‘{The good thing is/What’s good is} they came.’ > ‘It is good that they came.’ (24) Maksudnya kita tak perlu risau lansung. meaning=3sg.gen 1pl.incl neg need worry completely ‘Meaning, we don’t have to worry at all.’ (25) Nampaknya enak juga. see=3sg.gen delicious also ‘It appears to be quite delicious, too.’
To briefly sum up our discussion thus far, we see an extended use of enclitic -nya from marking possession to marking nominalization and speaker stance. This development is reminiscent of the developments discussed within the Traugottian framework where constructions with propositional meaning take on textual and pragmatic functions (e.g. Traugott 1982, 1989, 1995, 2003, 2010). Data from Malay discussed here show that this development toward subjective and intersubjective uses can also take the path from nominal constructions to stance marking ones.
Referential and non-referential uses of nominalization constructions in Malay 
Based on his colloquial Indonesian data, Englebretson (2003) also identifies enclitic -nya as a third person pronominal object (or rather patient argument) following a transitive verb. This pronominal object (or patient argument) marking function for enclitic -nya is also attested in colloquial Malay, as seen in (26). (26) Jangan dibuangnya. do.not di-throw.away=3sg.gen ‘Don’t (you) throw it away.’ (Lit. ‘Don’t (you) be throwing it away.’)
An interesting question, however, is how third person genitive -nya, which is strongly associated with possessors as discussed earlier in (16) and also with agent arguments as seen in (27) below, comes to also be associated with patient arguments as in (26) above. (27) Dipukulnya saya. di-hit=3sg.gen 1sg ‘He hit me.’ (≠ ‘He was hit by me.’) (Lit. ‘He be hitting me.’)
An important clue to this extended use of -nya, where it could come to refer not only to third person agents but also third person patients, appears to be the negative imperative construction, where second person agents are involved, as seen in (28). In such contexts, the usual deictic reference of -nya to third person agent is neutralized. It is also worth noting that, in Malay, third person genitive -nya is also used to refer to second person referents as a marker of politeness (e.g. Di-sila-nya duduk ‘Please be seated’ di-welcome=3sg.gen(=>2sg.agt) sit).18 (28) Jangan di-pukul-nya saya. do.not di-hit=3sg.gen(=>2sg.agt) 1sg ‘Don’t (you) hit me.’ (Lit. ‘Don’t (you) be hitting me.’)
Note that in (27), enclitic -nya still refers to a third person agent in genitive case, which still yields a high transitivity event of ‘someone hitting me’ but with the focus not so much on the agent but on the event of hitting and on the affected patient, which is saya ‘me’. This ‘focusing away from the agent’ strategy is beautifully achieved through the use of third person genitive agent -nya. In negative imperative contexts such as (28), however, the implied causer is a covert second person agent such as
.â•… The word sila means ‘to sit cross-legged’. The expression Dipersilakan is often used as an invitation for the addressee to sit down (traditionally, cross-legged on a mat or rug for men). Literally, di-per-sila-kan means ‘to be seated’. It is used as a polite request, and extended its range of functions to a wide range of invitations, including Dipersilakan makan/minum/ masuk/etc. ‘Please (you) eat/drink/enter/etc.’ Sila is also used in more direct forms such as Di-sila-nya duduk/makan/minum/etc. ‘Please be seated/eat/drink/etc.’, where third person genitive -nya indirectly refers to the addressee (i.e. second person agent).
 Foong Ha Yap
engkau ‘you’, not a third person agent at all. This second person agent, however, is structurally supposed to ‘agree’ with third person genitive enclitic -nya. There appears to be a form-function mapping mismatch. A construction with a highly transparent form-function mapping would be utterances such as Jangan engkau pukul saya ‘Don’t you hit me!’ However, it is not uncommon crosslinguistically for third person pronouns to also extend to second and even first person pronoun referents, given the appropriate pragmatic context, for example, where the extended use of the third person pronoun is used as a politeness strategy (e.g. Shibasaki 2005). Social factors, then, could contribute to the retention of third person genitive -nya in negative imperative contexts such as Jangan dipukulnya saya (‘Don’t someone be hitting me’) in (28) above, where -nya clearly clearly relies on an undefined ‘third person’ referent, which for politeness reasons is used to indirectly refer to the potentially aggressive and offensive addressee (i.e. second person agent referent engkau ‘you’). Interestingly, elision of the patient NP creates ambiguity and can give rise to several possible interpretations, including (i) a second person agent, with an intended meaning similar to (28) above, in contexts where the first person speaker is telling the second person addressee to refrain from hitting the speaker; or (ii) a second person reflexive agent-patient, as in (29) below, in contexts where the speaker is telling the addressee not to hit himself/herself, i.e. where -nya is understood to be co-referential with diri-mu ‘yourself ’ (i.e. the addressee himself/herself). (29) Jangan di-pukul-nya (diri-mu). do.not di-hit-3sg.gen self-2sg.gen ‘Don’t (you) be hitting yourself.’
Elision of the patient NP could also give rise to a third interpretation, namely, (iii) a third person patient, as in (30), where the speaker is telling the second person addressee not to hit something or someone else. (30) Jangan di-pukul-nya (benda/orang itu). do.not di-hit-3sg.gen thing/person that ‘Don’t (you) hit that thing/person.’
(30) above is structurally and semantically similar to (26) above, reproduced as (31) below. In such constructions, with contexts where the patient argument is a third person referent (in this case, something that could potentially be (adversely) affected (e.g. being hit at or thrown away), third person genitive enclitic -nya could lose its weak association with the covert second person agent referent, and instead comes to associate with the third person patient referent instead. That is, -nya comes to be associated not with the person doing the hitting/throwing but rather with the thing in danger of being hit at or being thrown away. Negative imperative contexts then are potential sites for -nya to be extended from ‘third person agent referent’ to ‘second person
Referential and non-referential uses of nominalization constructions in Malay 
agent referent in prohibitive contexts’, and further on to ‘second person agent-patient referent in reflexive prohibitive contexts’. In some contexts, we also see the extension of -nya from ‘agent reference’ to ‘patient reference’.19 (31) Jangan di-buang-nya. do.not di-throw.away-3sg.gen ‘Don’t (you) throw it away.’ (Lit. ‘Don’t (you) be throwing it away.’)
To sum up our discussion in this section, Figure 2 below summarizes the versatility and grammaticalization pathways of enclitic -nya, highlighting in particular deictic and definiteness functions within the pronominal domain and extensions into speaker mood (i.e. stance marking) functions via the adverbial and intensifier pathways (see also Englebretson 2003: 157–186). The role of -nya as a nominalizing element is clearly tied to its noun phrase marking function, both as third person genitive enclitic and definiteness marker, and it is also via its nominalizing effect that we see the rise and prolific use of -nya constructions as stance adverbials. 3rd person patient referent -nya [Jangan di-V-nya] 2nd person reflexive patient referent -nya [Jangan di-V-nya] 2nd person agent referent -nya [Jangan di-V-nya NP] 3rd person agent referent -nya [Di-V-nya NP] 3rd person genitive -nya [N-nya]
Definiteness marker -nya [N-nya] ‘Action nominalizer’ -nya [V-nya] Adverbial marker -nya often with stance interpretation: attitudinal, evaluative, etc. [N/V/Adj-nya], VP Stance marker: Intensifier -nya (Begitu) [Adj-nya] (N)
Figure 2.╇ Grammaticalization pathways of Malay nominalizer -nya
.╅ This extension in the use of -nya, to originally refer to the agent but over time to increasingly refer to the patient, is consistent with diachronic and typological accounts which posit that there has been a shift across many Austronesian languages where the language is moving away from an older system oriented toward discourse transitivity and topicality (e.g. Old Malay) to one that is oriented toward thematicity (e.g. Modern Malay/ Indonesian) (see for example Cummings 1988; Wouk 1996; Huang 2002).
 Foong Ha Yap
Compared to the grammatical functions of yang, where we see a pivotal role for nominalizer yang in the rise of relativizer and topic marker yang (see Figure 1), the grammatical functions of -nya as highlighted in Figure 2 show strong evidence of pronominal uses as precursors of massive extensions into adverbial and stance functions. Indeed, in his corpus analysis of colloquial Indonesian, Englebretson (2003) notes that a significant one-third of -nya constructions convey speaker (inter)subjective interpretations, which include attitudinal, evaluative and evidential stances. Here I suggest that the head-final characteristic of enclitic -nya readily facilitates the hosting and encoding of speaker mood prosody, a position I continue to pursue in our discussion of head-final uses of punya in the following section (see also Xu 2004, for discussions of sentencefinal position being a natural site for in-situ focus realizations). Another interesting observation is that Malay can rely not only on nominalizer yang, which is typically used to form participant nominalization constructions; the language can also make use of third person possessive pronoun (i.e. genitive) -nya as a noun phrase marker that signals action nominalization constructions.
4.â•… Nominalizer punya: A case of language contact Malay has yet another nominalizer, namely punya. The use of punya is now largely restricted to colloquial Malay. Classical Malay texts show frequent use of empunya as a lexical noun meaning ‘master, owner’, as in (32), and eventually increasingly as a lexical verb mempunyai, often accompanied by affixes mem-…-i to mark voice and transitivity features, as in (33).20 (32) Dia empunya kebun ini. 3sg possessor orchard this ‘S/he’s the owner of this orchard.’ (33) Mereka mem-punya-i ilmu ajaib 3pl af-possess-appl knowledge magic ‘They possess magical powers.’
In the colloquial register, punya is extensively used as a possessive marker, as in (34), and as a possessive pronoun focusing on the possessee (rather than possessor), as in
.â•… Data from the Malay Concordance Project provide evidence of maampunyai (i.e. ma-ampunya-i) in some varieties in Malay. This is a variant of me-empunya-i, which easily reduces to mempunyai. The lexical noun empunya itself appears to have been formed from a noun empu ‘master’ and third person genitive -nya. Empunya can be used lexically either as a noun or as a verb. Thus Dia empunya kebun itu in (32) above can be interpreted not only as ‘He is the owner of the orchard’ but also as ‘He owns the orchard.’
Referential and non-referential uses of nominalization constructions in Malay 
(35), as well as an epistemic, attitudinal and interactional marker (i.e. speaker mood or stance marker), as in (36). (34) Jaga-jaga kau punya barang look.after 2sg gen thing(s) ‘Look after your stuff.’ (35) Jangan sentuh aku punya. do.not touch 1sg poss.pron ‘Don’t touch mine (< ‘my one’).’ (36) Ibu (pasti) akan marah punya/mya/mia. mother certainly fut be.angry stance ‘Mother will (surely) get upset (I can assure you).’
As noted in the literature (e.g. Gil 1999; Yap 2003; Yap, Matthews & Horie 2004), without sentence-final punya (or its phonologically reduced form mya or mia), utterances such as (36) would simply be a factual statement. The addition of sentence-final punya/ mya/mia, however, makes it a ‘trust me’ or ‘I’m telling you’ type assertion. That is, the intersubjective (or interactional) overtone becomes evident when sentence-final punya/mya/mia is added.21 Punya also produces an intensifying effect in pre-adjectival position, as shown in (37), where it expresses mirative expressions (or exclamations) that are laced with strong feelings, including feelings of surprise, incredulity, or even annoyance. A sense of counter-expectation is often discernible in these mirative expressions. (37)
Punya jauh rumah kau ’ni, Timah. stance far house 2sg this (name of person: female) Lit. ‘So far, this house of yours, Timah.’ Intended meaning: ‘This house of yours is incredibly far, Timah.’
Pre-adjectival uses of stance punya appear to have emerged as an extension of its genitive and associative linking functions, in particular via [begitu + punya + Adjective] constructions, as seen in (38a) and (38b) below. Crucially, the emphatic prosody accompanying the deictic adverb begitu ‘like that’ spreads across linking particle punya and extends over the descriptive adjective jauh ‘far’. When the deictic adverb begitu is elided, the intensifying prosody persists, and linking particle punya readily gets reinterpreted as an intensifier-type stance marker. This intensifier usage of punya is highly productive in colloquial speech.
.╅ Speaker mood or stance can always be conveyed by sentence-final prosody, but as is well-attested in the literature, sentence-final particles are ideally suited as markers of speaker mood and stance.
 Foong Ha Yap
(38) a.
Begitu punya jauh::: rumah kau ’ni, Timah. like.that lnk far house 2sg this (name of person: female) ‘So::: far, this house of yours Timah!’
b. Punya jauh::: rumah kau ’ni, Timah! stance far house 2sg this (name of person: female) ‘So::: far, this house of yours Timah!’
As noted in Yap, Matthews and Horie (2004), diachronic evidence reveals a gradual expansion in the grammatical functions of empunya as follows: lexical noun empunya > genitive/possessive pronominal (em)punya > stance marker punya/mya/ mia. As discussed above, the final stage involves the extended use of phonologically reduced punya as stance marker in sentence-final and pre-adjectival positions. As a sentence-final particle, punya is often further reduced to mya (or mia). Gil (1999) further reports the use of pun and nya in other Malay varieties as well. Similar developments involving the grammaticalization of lexical nouns to nominalizers and pragmatic markers (inclusive of sentence-final particles) are also attested in many other languages.22 Nominalizer punya is productive in colloquial Malay, as in (39), but not in the standard variety.23 The use of punya as a relative clause marker, as seen in (40), is marginal even in colloquial Malay, and is more typical of Bazaar (or ‘market variety’) Malay, pointing to the possible influence of language contact, in particular with dialects of Chinese origin. (39) (Yang) nak jahit punya letak ’kat sini. nmz want sew nmz put loc(< ‘near’) here ‘The ones you want to sew (or mend), put (them) here.’ (40) (Yang) mau jahit punya baju letak dekat sini. nmz want sew rel clothes put loc(< ‘near’) here ‘The clothes that you want to sew (or mend), put (them) here.’
Diachronic evidence in fact points to an important role for language contact in the rise of nominalizer and relative clause marker punya. As shown in Yap, Matthews and Horie (2004), corpus analysis of classical Malay texts from the 16th to 20th century
.â•… For typological discussions, see Yap & Matthews (2008), Yap & Grunow-Hårsta (2010), and Yap, Grunow-Hårsta & Wrona (this volume); see also Noonan (1997, 2008, this volume) for Bodic and Tamangic languages; Grunow-Hårsta (this volume) for Magar; Simpson (2008) for Burmese; Rhee (2008, this volume) for Korean; Horie (2008, this volume) for Japanese; Shinzato (this volume) and Shibasaki (forthcoming a,b) for Okinawan; Yap, Choi & Cheung (2010) and Yap & Wang (this volume) for Old and Middle Chinese. .â•… It should be noted that native speakers often prefer to add nominalizer yang in constructions such as (39), giving rise to ‘double nominalization’ phenomena. Interestingly, non-native speakers using the Bazaar variety tend not to use nominalizer yang.
Referential and non-referential uses of nominalization constructions in Malay 
reveals that genitive and possessive pronominal uses of punya rose sharply in the 19th century, coinciding with a massive influx of immigrant workers from southern China. The rise of these constructions appears to have facilitated the rise of nominalizer and relative clause uses of punya; these latter two constructions were not attested in Classical texts but were (and still are) evident in colloquial and Bazaar (or ‘market variety’) Malay. Here I suggest that the availability of nominalizer punya in turn facilitated the rise of sentence-final stance uses of punya; this stance marker usage is still in evidence in contemporary colloquial Malay.24 The examples in (41a–b) illustrate how nominalizer punya can easily be reanalyzed as a sentence-final particle (i.e. stance marker punya) via elision of the subject and verb in the matrix clause, leaving behind the complement clause as a stand-alone nominalization construction.25 (41) a.
Aku pasti [dia akan datang punya]. 1sg certain 3sg fut come nmz/sfp ‘I’m sure that he will come.’
b. Dia akan datang punya. 3sg fut come sfp ‘He will come (I assure you).’
Structural parallels between Malay punya and Cantonese ge3 constructions are highlighted in (42) below, the latter representing the southern Chinese dialects (e.g. Hakka, Hokkien and Teochew) with similar contact influence and widely spoken in the western coastal states of peninsular Malaysia. Among the parallel functions are the genitive, possessive pronominal, relative clause marker, nominalizer, and stance uses respectively. With particular reference to (42e), note that Malay nominalizers punya and Cantonese nominalizer ge3 both retain the assertive force of the matrix clause (‘I know’) even when this higher clause is elided. Such illicutionary force is encoded prosodically with the sentence-final nominalizers as host, which facilitates their reanalysis as stance markers. (42) Colloquial Malay a. Ini engkau punya kerja! a′. this 2sg gen doing ‘This is your doing!’
Cantonese ni1go3 (hai6) nei5 ge3 co3 this:cl cop 2sg gen mistake ‘This is your mistake.’
.â•… It is possible that a similar development has taken place at an even earlier date, particularly in places where Chinese traders have made contact with local Malays along the coast, not only on the Malayan peninsula but among the many islands of Indonesia as well, such as Ambon for example. Inclusion of Ambonese and other Malay texts in future analysis will help provide a fuller picture. .â•… The term ‘stand-alone’ nominalization construction was coined by the late David Watters (2008).
 Foong Ha Yap
b. Ini engkau punya. this 2sg poss.pron ‘This is yours.’ c.
b′. ni1go3 (hai6) nei5 ge3 this:cl cop 2sg poss.pron ‘This is yours.’
Bak balik aku beli punya benda. c′. waan4 faan1 ngo5 maai5 ge3 je5 give back 1sg buy rel stuff return back 1sg buy rel stuff ‘Give me back the stuff I bought.’ ‘Give me back the stuff I bought.’
d. Bak balik (yang) aku beli punya. d′. waan4 faan1 ngo5 maai5 ge3 give back nmz 1sg buy nmz return back 1sg buy nmz ‘Give me back the one(s) I bought.’ ‘Give me back the one(s) I bought.’ e. (Aku tahu) kau akan datang punya. e′. (ngo5 zi1) nei5 wui5 lai4 ge3 1sg know 2sg will come sfp 1sg know 2sg will come sfp ‘(I know) you will come (for sure).’ ‘(I know) you will come (for sure).’
Chronologically, given that pre-nominal relative clauses were already attested in Old and Middle Chinese (via zhi and zhe constructions), and given that the prenominal punya relative clause in Malay is a much more recent development, there is strong reason to believe that the southern Chinese dialects are the donor languages. Typologically, too, it has been shown that pre-nominal relative clauses are extremely rare among SVO languages (Greenberg 1966; Dryer 1992), with the Chinese dialects being rare exceptions. In fact, these Chinese dialects often rely on topicalization strategies to mitigate the frequently high processing costs of using pre-nominal relative clauses within an SVO configuration (Matthews & Yeung 2001). Malay, an SVO language, has actually long relied on post-nominal yang relative clause constructions (as noted earlier in §2). Thus, all things being equal, the emergence of punya as a prenominal relative clause marker would have been largely unmotivated, particularly in the case of lengthier and phonologically more bulky relative clauses, had it not been for a sociological development with sufficient force to overcome the high cognitive processing constraints. This sociological force, I suggest, came in the form of contact with Chinese dialects on a massive scale. Figure 3 summarizes the grammaticalization pathways for Malay punya. Not discussed earlier but included for the sake of comprehensiveness is the emergence of subordinator punya pasal, via the merger of adnominal (i.e. genitive and relativizer) punya and a lexical head noun pasal (which is a general noun with a wide range of meanings including ‘reason, matter, affair, business, problem, fault, etc.’), as shown in (43) below. 26 .â•… The term ‘adnominal’ refers to a linker between a modifying expression and its head noun, and is used extensively in linguistic descriptions related to languages with pre-nominal modification, which are typically verb-final languages such as Japanese, Korean, and Tibeto-Burman languages, with the Sinitic languages being rare exceptions from among the SVO type. Previous studies have identified general nouns (also referred to as formal nouns) meaning ‘person’, ‘thing’ or ‘place/location’ and demonstratives (e.g. Chinese zhi) as sources of
Referential and non-referential uses of nominalization constructions in Malay 
(43) a. b.
Semua ’ni dia punya pasal. all this 3sg gen fault ‘All this (is) his fault.’ Dia punya pasal, kita kena marah 3sg gen fault ipl:incl pass scold ‘Because of him (lit. ‘his fault’), we got scolded.’
c.
(Sebab) dulu takut gemuk punya pasal, because previously fear fat because.of ( adverbial adnominals and nominalizers (e.g. DeLancey 1986; Horie 2000; Rhee 2008; LaPolla 1994; Yap, Choi & Cheung 2010; Yap, Grunow-Hårsta & Wrona this volume).
 Foong Ha Yap
pathway, lexical noun (em)punya also developed a possessive pronominal usage, and this in turn gave rise to its nominalizing function (often as part of a (yang) … punya double nominalization construction). This second pronominal/nominalizer pathway further gave rise to the highly productive sentence final speaker mood punya/mya/mia marker. We thus see robust evidence of semantic extensions from referential to nonreferential uses in Malay, whereby a lexical noun develops more abstract grammatical and pragmatic functions over time. We also see lexical noun (em)punya being used as a lexical verb; this process of noun>verb conversion is not a grammaticalization phenomenon, for such conversions are spontaneous and generally is available to all languages, with many languages often further developing disambiguating strategies – and in the case of the Malay language, we see the development of the meN-…-i circumfix for lexical verb empunya, while the lexical noun use is often accompanied by noun phrase marker yang, which in time also came to accompany nominal clauses and thus came to be reinterpreted as a nominalizer, as noted in §2 earlier.
5.╅ Semantic niches of yang, -nya and punya Table 1 below highlights the functional distribution of each of the three versatile morphemes: yang, -nya and punya. Table 1.╇ Functional distribution of yang, -nya and punya in Malay Function
yang
Lexical noun Genitive Possessive Pronominal Definiteness marker Nominalizer Event Nominalizer/ Gerundive Relative Clause marker Preposition
√ high focus √
-nya
√ empunya √ (em)punya √ possessor-referring √ possessee-referring punya √ low focus √ punya (Bazaar ‘market’ variety) √
√ post-nominal
Subordinator Evidential/Attitudinal/ Epistemic/Stance marker
punya
√ mirativity marker √ stance adverbial
√ pre-nominal punya √ complex preposition punya pasal √ complex subordinator punya pasal √ sentence final particle punya/mya/mia
Referential and non-referential uses of nominalization constructions in Malay 
Overall, yang and -nya show very little functional overlap, each morpheme having found its own semantic niche over the course of time. In contexts where they do overlap, i.e. where they both function as identifiability (or definiteness) marker, yang generally yields a stronger focus interpretation and is often used for contrastive identification, as noted earlier in §2 and illustrated again in (44) below, while -nya is more subtle, and typically appears in contexts where the identity of the referent is already given or known, either in prior discourse or from shared knowledge, as in (45). (44) Aku nak yang biru, (bukan yang merah). 1sg want def blue neg gen red ‘I want the red one, (not the blue one).’ (45) Aku kena bayar sekarang, tapi duitnya tak cukup. 1sg must pay now but money=def neg sufficient Lit. ‘I have to pay now, but the money is insufficient.’ Intended meaning: ‘I have to pay now, but I don’t have enough money (with me).’
As highlighted in (46), in nominalization constructions, yang appears in clauseinitial position (e.g. [yang tumbang ke arah selatan] ‘the one(s) that fell toward the south’), as in (a), while -nya appears in enclitic position, as in (b), and often yields a different nominalization effect. For example, [tumbang-nya] ke arah selatan ‘↜the falling (inclusive of the way/direction it fell) was toward the south’. What we see then is that yang-nominalizations focus on participant arguments (the one(s) who VP), while nominalizations signalled by third person genitive enclitic -nya focus instead on event nominalizations (the VP-ing event itself). (46) a.
yang tumbang ke arah selatan nmz fall to direction south ‘the one(s) that fall/fell toward the south’ (referring for example to trees)
b. tumbang =nya ke arah selatan fall=3sg.gen to direction south ‘its/the falling (i.e. the way/direction it fell) was southward’
Given their morphosyntactic differences, it is not surprising that we find different types of constraints on the uses of yang and -nya. For example, clause-initial yang appears in post-nominal relative clauses (e.g. ikan [yang dijualnya] ‘the fish that he sells’) and in post-verbal factive complementations (e.g. aku tahu [(yang) dia kalah] ‘I know (that) he lost’), but enclitic -nya does not occur in these constructions. On the other hand, enclitic -nya can give rise to nominalization constructions that are readily reanalyzed as stance adverbials (i.e. it can help form attitudinal/evaluative/evidential/
 Foong Ha Yap
epistemic mood markers that occur sentence adverbially at the beginning of a clause) as in (47). These speaker mood or stance-marking functions, which are subjective and non-referential in nature, are not attested for nominalizer yang. On the whole, we see that yang and -nya tend to confine themselves to different functional domains. (47) a.
Agaknya, dia tak suka cadangan kau. guess=3sg.gen 3sg neg like plan-nmz 2sg ‘It seems (< one’s guess) s/he does not like your plan.’
There is slightly more overlap between the grammatical uses of punya with those of yang and -nya. Even so, we see evidence of division of labor and each morpheme has its own characteristic semantic signature. The functional difference between punya and yang largely hinges on register: yang is more formal, and punya is often used in speech produced by or directed toward the non-native interlocutor. Another difference is structural: nominalizer yang is head-initial while nominalizer punya is head-final, and it naturally follows that relativizer yang occurs post-nominally, while relativizer punya occurs pre-nominally. As noted earlier, this difference in word order between yang and punya can be attributed to differences in etymology, and in the case of punya constructions, to contact influence with southern Chinese dialects as well. There is greater semantic difference between the uses of punya and -nya. For example, in possessive pronominal constructions such as (48), third person genitive enclitic -nya refers to the possessor, while pronominal punya stands for the elided possessee.27 (48) a.
anaknya child=3sg.gen ‘his/her child’
b. dia punya 3sg poss.pron Lit. ‘his/her possession/thing/stuff ’/‘his/her one’ > ‘his/hers’
Functional overlap is also found in the stance-marking domain (which include attitudinal, evaluative, evidential, and epistemic intepretations), but even then there still are subtle differences. For example, as highlighted in (49), epistemic -nya in (49a) is more explicit in revealing the speaker’s state of mind, since the semantic value of its lexical host is known (e.g. agak-nya ‘probably, presumably’; from agak ‘guess’), while epistemic punya in (49b) leaves the reader to infer the speaker’s state of mind from the context, because its semantic referent is not explicitly expressed, though this in no way inhibits epistemic punya from expressing fine shades of speaker mood given its enviable access to sentence-final prosodic cues.
.â•… Othewise also referred to in the literature as ‘possessum’, i.e. the thing possessed.
(49) a.
Referential and non-referential uses of nominalization constructions in Malay 
Agaknya dia belum kahwin lagi guess=3sg.gen 3sg not.yet marry yet ‘Probably (s)he’s not married yet.’
b. Dia belum kahwin punya 3sg not.yet marry stance ‘(S)he’s not married yet, (I tell you/I bet you/ trust me/believe me/etc).’
There is good reason then for the Malay language to embrace stance markers formed with the help of different nominalizing elements, like -nya and punya illustrated in (49) above, since their subtle differences provide native speakers with a choice of explicit vs. implicit ways to express their subjective and intersubjective moods (e.g. Nampaknya dia tak akan datang ‘It looks like he’s not coming’ vs. Dia tak akan datang punya ‘He’s not coming, (trust me)’, often with socio-interactive implications such as ‘So let’s not wait any longer’). As we have seen in this section, competition among the versatile morphemes yang, -nya and punya has resulted in a subtle negotiation of semantic and functional turfs and their boundaries.
6.â•… Conclusion In this paper we have examined three nominalizers in Malay, namely yang, -nya and punya, in terms of their functions, structural differences, and to the extent possible, their diachronic development as well. The analysis presented here highlights some robust grammaticalization patterns that are consistent with typological observations reported elsewhere, including a robust nominal/pronominal > nominalizer > stance marker development. The notion of stance or speaker mood discussed in this paper is a broad one, and has included (contrastive) focus functions (particularly with respect to yang constructions), and attitudinal, evaluative, evidential, epistemic and other speaker mood/stance-marking functions (as seen in the case of enclitic -nya and sentence-final punya constructions). Crucially, this paper has shown how three different nominalizers in Malay – namely, yang, -nya and punya – give rise to a wide range of constructions that capture various shades of speaker mood. Their different stance effects can be traced back to differences in etymological origin and preferred structural configurations, and in the case of punya can be attributed to language contact as well. Their development from referential to expressive marker is consistent with a general tendency for linguistic items to often grammaticalize into pragmatic markers (e.g. Traugott 1982, 1989, 1995, 2003, 2010; Traugott & König 1991; inter alia), and thus points to strong cognitive motivations underlying a process of semantic extension to meet pragmatic needs.
 Foong Ha Yap
Acknowledgments I wish to gratefully acknowledge generous research funding from the Chinese University of Hong Kong (Faculty of Arts Direct Grant 2006–2007 and Departmental Research Grant (Linguistics) 2008–2009) and from the Hong Kong Polytechnic University (Faculty of Humanities Start-Up Research Fund 2010–2012). I also wish to thank an anonymous reviewer, and also Karen Grunow-Hårsta, Daniel Kaufman, František Kratochvíl, Hongyong Liu, Naonori Nagaya, Eric Potsdam, Lawrence Reid, Seongha Rhee, Masayoshi Shibatani, Andrew Simpson, Joanna Sio, and Janick Wrona for constructive comments and helpful discussions. Earlier versions of this paper were presented at the Conference on Beyond ‘Focus’ and Ergativity: Towards a More Comprehensive View of Austronesian Morphosyntax, organized by the Centre for General Linguistics, Typology and Universals Research (ZAS), Berlin (September 12–16, 2007) and the Workshop on Definiteness and Referentiality: Theory and Description, organized by the Australian Linguistic Society, University of Adelaide, Australia (September 26–28, 2007), and I wish to thank participants at these conferences for their helpful feedback. This paper has at various stages been inspired by the earlier works of David Gil, Robert Englebretson, and Don van Minde, whose research and earlier insights on punya, -nya and yang respectively paved the way for my present analysis.
Abbreviations 1 2 3 af agt appl art assoc caus cl comp def dem pl pn poss
First person Second person Third person Actor Focus Agent Applicative (also Transitivizer) Article/Definiteness marker Associative Causative Classifier Complementizer Definite Demonstrative Plural Pronoun/Pronominal Possessive
dist excl exst foc fut gen incl intj loc neg nmz pass perf q rel sfp
Dista Exclusive Existential verb ‘be/have’ Focus Future marker Genitive Inclusive Interjection Locative Negator Nominalizer Passive marker Perfective aspect Question marker Relativizer Sentence Final Particle
Referential and non-referential uses of nominalization constructions in Malay 
pron Pronoun prox Proximal prt Particle
sg Singular top Topic marker trans Transitivizer
References Adelaar, Karl Alexander. 1985. Proto-Malayic: The Reconstruction of Its Phonology and Parts of Its Lexicon and Morphology. Ph.D. dissertation, Leiden University. Bickel, Balthassar. 1999. Nominalization and focus constructions in some Kiranti languages. In Topics in Nepalese Linguistics, Yogendra P. Yadava & Warren G. Glover (eds), 271–296. Kathmandu: Royal Nepal Academy. van Breugel, Seino. 2006. Nominalisation in Atong. Paper presented at the Workshop on Nominalization and Its Discontents, Tibeto-Burman Special Interest Group, La Trobe University, Melbourne, November 2. Comrie, Bernard. 1988. General features of Uralic languages. In The Uralic Languages: Description, History and Foreign Influences [Handbuch der Orientalistik, Achte Abteilung, Handbook of Uralic Studies] Denis Sinor (ed.), 451–477. Leiden: E. Brill. Coupe, Alec R. 2006. Nominalization, relativization and genitivization in Mongsen Ao and Chang. Paper presented at the Workshop on Nominalization and Its Discontents, SinoTibetan Special Interest Group, La Trobe University, August 24. Cummings, Susanna. 1988. Syntactic Function and Constituent Order Change in Malay. Ph.D. dissertation, University of California, Los Angeles. DeLancey, Scott. 1986. Relativization as nominalization in Tibetan and Newari. Paper presented at the Nineteenth International Conference on Sino-Tibetan Languages and Linguistics, Ohio State University, Columbus, OH, September 11. Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68: 81–138. Englebretson, Robert. 2003. Searching for Structure: The Problem of Complementation in Colloquial Indonesian Conversation [Studies in Discourse and Grammar 13]. Amsterdam: John Benjamins. Englebretson, Robert. 2008. From subordinate clause to noun-phrase: Yang constructions in colloquial Indonesian. In Crosslinguistic Studies of Clause Combining, Ritva Laury (ed.), 1–33. Amsterdam: John Benjamins. Fischer, Olga. 2007. Morphosyntactic Change: Functional and Formal Perspectives [Oxford Surveys in Syntax and Morphology]. Oxford: OUP. Gil, David. 1999. The grammaticalization of punya in Malay/Indonesian dialects. Proceedings of the 9th Annual Meeting of the South-East Asian Linguistics Society, University of California, Berkeley, USA, May 22. Givon, T. 2009. The Genesis of Syntactic Complexity: Diachrony, Ontogeny, Neuro-Cognition, Evolution. Amsterdam: John Benjamins. Greenberg, Joseph H. 1966. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Language, Joseph H. Greenberg (ed.), 73–113. Cambridge MA: The MIT Press. Grunow-Hårsta, Karen. This volume. Innovation in nominalization in Magar, a Tibeto-Burman language of Nepal. In Yap, Grunow-Hårsta & Wrona (eds).
 Foong Ha Yap Herring, Susan C. 1991. Nominalization, relativization, and attribution in Lotha, Angami, and Burmese. Linguistics of the Tibeto-Burman Area 14(1): 55–72. Horie, Kaoru. 1998. On the polyfunctionality of the Japanese particle no: From the perspectives of ontology and grammaticalization. In Studies in Japanese Grammaticalization: Cognitive and Discourse Perspectives, Toshio Ohori (ed.), 169–192. Tokyo: Kurosio. Horie, Kaoru. 2000. Complementation in Japanese and Korean. In Complementation: Cognitive and Functional Perspectives, Kaoru Horie (ed.), 11–32. Amsterdam: John Benjamins. Horie, Kaoru. 2008. The grammaticalization of nominalizers in Japanese and Korean: A contrastive study. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 169–187. Amsterdam: John Benjamins. Horie, Kaoru. This volume. Versatility of nominalizations: Where Japanese and Korean contrast. In Yap, Grunow-Hårsta & Wrona (eds). Huang, Shuanfan. 2002. The pragmatics of focus in Tsou and Seediq. Language and Linguistics 3(4): 665–694. Huang, Chu-Ren & Ahrens, Kathleen. 2003. Individuals, kinds and events: Classifier, coercion of nouns. Language Sciences 25(4): 353–373. Kaswanti Purwo, Bambang. 1983. Kata yang sebagai pengetat. Majalah Pembinaan Bahasa Indonesia 4(3): 175–185. Kaufman, Daniel. 2009. Austronesian typology and the nominalist hypothesis. In Austronesian Historical Linguistics and Culture History: A Festschrift for Robert Blust, Alexander Adelaar & Andrew Pawley (eds). Canberra: Pacific Linguistics. Lahaussois, Aimée. 2003. Nominalizations and its various uses in Thulung Rai. Linguistics of the Tibeto-Burman Area 26(1): 33–57. LaPolla, Randy J. 1994. Parallel grammaticalizations in Tibeto-Burman: Evidence of Sapir’s drift. Linguistics of the Tibeto-Burman Area 17(1): 61–80. LaPolla, Randy J. 2008. Nominalization in Rawang. Linguistics of the Tibeto-Burman Area 31(2): 45–65. Lyons, John. 1977. Semantics, Vol. 2. Cambridge: CUP. Matisoff, James A. 1972. Lahu nominalization, relativization, and genitivization. In Syntax and Semantics, Vol. 1, John Kimball (ed.), 237–258. New York NY: Seminar Press. Matthews, Stephen & Yeung, Lousia Y.Y. 2001. Processing motivations for topicalization in Cantonese. In Cognitive-functional Linguistics in an East Asian Context, Kaoru Horie & Shigeru Sato (eds), 81–102. Tokyo: Kurosio. (Chinese version in New Ideas about Topic and Focus, Xu Liejiong & Liu Danqing (eds), 2003, 145–63. Shanghai Educational Publishing). Mees, C.A. 1969. Tatabahasa dan Tatakalimat. Kuala Lumpur: University of Malaya Press. van Minde, D. 2008. The pragmatic function of Malay yang. Journal of Pragmatics 40(11): 1982–2001. Morey, Stephen. 2006. Nominalisation and its discontents: Clausal nominalisation in Numhpuk Singpho. Paper presented at the Workshop on Nominalization and Its Discontents, Tibeto- Burman Special Interest Group, La Trobe University, Melbourne, August 3. Morey, Stephen. This volume. Nominalization in Numhpuk Singpho. In Yap, Grunow-Hårsta & Wrona (eds). Noonan, Michael, 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins.
Referential and non-referential uses of nominalization constructions in Malay 
Noonan, Michael. 2008. Nominalization in Bodic languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 219–237. Amsterdam: John Benjamins. Noonan, Michael. This volume. Aspects of the historical development of nominalizers in the Tamangic languages. In Yap, Grunow-Hårsta & Wrona (eds). van Ophuijsen, Charles Adriaan. 1910. Maleische Spraakkunst. Leiden: S.C. van Doesburgh. (cited in van Minde 2003). Pustet, Regina. 2003. Copulas: Universals in the Characterization of the Lexicon. Oxford: OUP. Rhee, Seongha. 2008. On the rise and fall of Korean nominalizers. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 239–264. Amsterdam: John Benjamins. Rhee, Seongha. This volume. Nominalization and stance marking in Korean. In Yap, GrunowHårsta & Wrona (eds). Sheikh Othman bin Sheikh Salim (Chief Editor). (1989). Kamus Dewan, new edn. Kuala Lumpur: Dewan Bahasa dan Pustaka. Shibasaki, Reijirou. 2005. Personal Pronouns and Argument Structures in Japanese: Discourse Frequency, Diachrony and Typology. Ph.D. dissertation, University of California, Santa Barbara. Shibasaki, Reijirou. Forthcoming a. From nominalizer to stance marker in the history of Okinawan. In Japanese/Korean Linguistics, Vol. 18, 101–113. Stanford CA: CSLI. Shibasaki, Reijirou. Forthcoming b. From relativization to nominalization and more: Evidence from the history of Okinawan. In Proceedings of the 36th Annual Meeting of the Berkeley Linguistics Society, 101–113. Shibatani, Masayoshi. 2009. Elements of complex structures, where recursion isn’t: The case of relativization. In Syntactic Complexity: Diachrony, Acquisition, Neuro-cognition, Evolution [Typological Studies in Language 85], T. Givon & Masayoshi Shibatani (eds), 163–198. Amsterdam: John Benjamins. Shibatani, Masayoshi & Khaled Awadh bin Makhashen. 2009. Nominalization in Soqotri, a South Arabian language of Yemen. In Endangered Languages: Contributions to Morphology and Morpho-Syntax, W. Leo Wetzels (ed), 9–31. Leiden: Brill. Shinzato, Rumiko. This volume. Nominalization in Okinawan: From a diachronic and comparative perspective. In Yap, Grunow-Hårsta & Wrona (eds). Siewierska, Anna, Rijkhoff, Jan & Bakker, Dik. 1998. Appendix – 12 word order variables in the languages of Europe. In Constituent Order in the Languages of Europe, Anna Siewierska (ed.), 783–812. Berlin: Mouton de Gruyter. Simin, Azhar M. 1988. Discourse-syntax of ‘yang’ in Malay (Bahasa Malaysia). Kuala Lumpur: Dewan Bahasa dan Pustaka. Simpson, Andrew. 2008. The grammaticalization of nominalizers in Burmese. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 265–288. Amsterdam: John Benjamins. Steinhauer, Hein. 1992. On the meaning of yang in Indonesian. In The Language Game: Papers in Memory of Donald C. Laycock, Tom Dutton, Malcolm Ross & Darrell Tryon (eds), 427–439. Canberra: Pacific Linguistics C-110.
 Foong Ha Yap Traugott, Elizabeth C. 1982. From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization. In Perspectives on Historical Linguistics [Current Issues in Linguistic Theory 24], Winfred P. Lehmann & Yakov Malkiel (eds), 245–271. Amsterdam: John Benjamins. Traugott, Elizabeth Closs. 1989. On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31–55. Traugott, Elizabeth Closs. 1995. Subjectification in grammaticalization. In Subjectivity and Subjectivisation, Dieter Stein & Susan Wright (eds), 37–54. Cambridge: CUP. Traugott, Elizabeth Closs. 2003. From subjectification to intersubjectification. In Motives for Language Change, Raymond Hickey (ed.), 124–139. Cambridge: CUP. Traugott, Elizabeth Closs. 2010. (Inter)subjectivity and (inter)subjectification: A reassessment. In Subjectification, Intersubjectification and Grammaticalization, Kristin Davidse, Lieven Vandelanotte & Hubert Cuyckens (eds), 29–71. Amsterdam: John Benjamins. Traugott, Elizabeth Closs & Konig, Ekkehard. 1991. The semantics-pragmatics of grammaticalization revisited. In Approaches to Grammaticalization, Vol. 1 [Typological Studies in Language 19], Elizabeth C. Traugott & Bernd Heine (eds), 189–218. Amsterdam: John Benjamins. Verhaar, John W.M. 1983. On the syntax of yang in Indonesian. In Papers from the Third International Conference on Austronesian Linguistics, Vol. 4: Thematic Variation, Amran Halim, Lois Carrington & Stephen A. Wurm (eds), 43–70. Canberra: Pacific Linguistics C-77. Watters, David E. 2008. Nominalization in the Kiranti and Central Himalayish languages of Nepal. Languages of the Tibeto-Burman Area 31(2): 1–44. Wouk, Fay. 1996. Transtitivity in Toba Batak and Tagalog. Studies in Language 10: 391–424. van Wijk, Gerth D. 1909. Spraakleer der Maleische Taal. Batavia: G. Kolff en Co. (1st edn 1889; cited in van Minde 2007). Xu, Liejong. 2004. Manifestation of informational focus. Lingua 114(3): 277–299. Yap, Foong Ha. 2003. On native and contact gramaticalization: A diachronic analysis of Malay empunya. Paper presented at the 13th Annual Meeting of the South-East Asian Linguistics Society, University of California, Los Angeles, May 2–4. Yap, Foong Ha, Choi, Pik-ling & Cheung, Kam-siu. 2010. Delexicalizing di: How a Chinese noun has evolved into an attitudinal nominalizer. In Formal Evidence in Grammaticalization Research [Typological Studies in Lanuage 94], An van linden, Jean-Christophe Verstraete & Kristin Davidse (eds), 63–91. Amsterdam: John Benjamins. Yap, Foong Ha, Grunow-Hårsta, Karen & Wrona, Janick. This volume. Nominalization strategies in Asian languages. In Yap, Grunow-Hårsta & Wrona (eds). Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José López-Couso & Elena Seoane (eds) in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins. Yap, Foong Ha, Matthews, Stephen & Horie, Karou. 2004. From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective. In Up and Down the Cline: The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins. Yap, Foong Ha & Wang, Jiao. This volume. From light noun to nominalizer and more: The grammaticalization of zhe and suo in Old and Middle Chinese. In Yap, Grunow-Hårsta & Wrona (eds).
Expressing exclamatives in Malagasy* Eric Potsdam
University of Florida This paper explores the syntax and semantics of the exclamative clause type in the Austronesian language Malagasy. Cross-linguistic considerations, distributional facts, and language-internal morphosyntax provide evidence that Malagasy exclamatives are syntactically nominal. All exclamatives are noun phrases, as in the English The things that he says! Unlike English, Malagasy has no clausal exclamatives. Based on the Malagasy data, the paper proposes that the exclamative force of nominals, in Malagasy and other languages, is located in the determiner layer of the structure.
1.â•… Introduction It is conventional in speech act theory to distinguish between an utterance’s sentential force and its illocutionary force (Austin 1962; Searle 1969; Chierchia and McConnellGinet 1990; Zanuttini & Portner 2003; Collins 2005; and others). Illocutionary force is the force associated with the speaker’s intention in producing a sentence, independent of the utterance’s form. It relates to the utterance’s communicative intent. The sentential force is the force conventionally associated with a sentence by virtue of its form. Sentential force “describes the way a syntactically-determined clause type (declarative, interrogative, imperative, etc.) is used in a certain communicative context” (Heusinger & Schwabe 2002). A sentence’s sentential force can reinforce its illocutionary force, or not. For example, Pass the salt! has both the sentential and illocutionary force of a command. The intention of the speaker is to command the addressee to do something, giving it the illocutionary force of a command. Its sentential force is that of a command as well because syntactically, it is an imperative clause type (Potsdam 1998), which is the canonical expression of a command in English. Could you pass the salt?, on the other hand, has only the illocutionary force of a command. In the relevant use, it (politely)
*I would like to thank my Malagasy consultants, Bodo and Voara Randrianasolo. I am also grateful to two anonymous reviewers, the volume editors, and the audience at the Beyond Focus and Ergativity in Austronesian Workshop at ZAS Berlin for insightful comments. This work was supported by NSF grant BCS 0131993.
 Eric Potsdam
commands. It is not an imperative however and does not have the syntactic form conventionally associated with commanding. Its sentential force is that of a question due to its form. This distinction between sentential and illocutionary force can be applied to exclamatory utterances. I define exclamations as expressive utterances that have the illocutionary force of exclaiming, regardless of their form. They “express a speaker’s emotive attitude towards a certain state of affairs that is not in accordance with his or her expectations” (D’Avis 2002: 26). Exclamations that additionally take a certain canonical syntactic form in a language have the sentential force of exclaiming and I call them exclamatives. In English, exclamatives have a range of syntactic expression (see Elliott 1974; Michaelis & Lambrecht 1996a; Huddleston & Pullum 2002; and others for discussion) but clausal exclamatives are typically marked with such, so, what a, how, or subject-auxiliary inversion: (1)
a. b. c. d. e. f. g. h. i. j.
He’s such a good cook! It’s so hot in here! What a day (I had)! How ugly that is! Can he eat a lot! Are you in for it! The things I do for that boy! You wouldn’t believe [the bickering that goes on]. Its amazing [the difference]. The cost of these clothes!
The examples illustrate not only diverse syntactic forms but also different types of exclamatives. Exclamatives need to be kept distinct from the larger, also not well defined, class of expressives, which may include interjections, declaratives, and other forms. I take as a defining property of exclamatives that they have propositional content that is assumed by the speaker to be true (Elliott 1971, 1974; Grimshaw 1979; Sadock & Zwicky 1985; Michaelis & Lambrecht 1996a, 1996b; Zanuttini & Portner 2000, 2003; Michaelis 2001; Villalba 2003; Portner & Zanuttini 2004; Collins 2005; Miró 2006; Rett 2006; and many others). That is, exclamatives presuppose that the proposition they express is a background belief, known by the speaker and addressee. This presupposition arises because exclamatives express an emotional reaction to or evaluation of a proposition, which is therefore likely to already be in the common ground. Declaratives, in contrast, are informative about the content of a proposition. Exclamatives are not prominent in typological work (with the exceptions of Elliott 1974; Sadock & Zwicky 1985; and Michaelis 2001). Nevertheless, I believe that there are tendencies in the grammatical expression of exclamatives. One pattern that is evident from the English data is that exclamatives are expressed with either clauses, (1a–f), or noun phrases, (1g–i). This, I believe, is cross-linguistically true
Expressing exclamatives in Malagasy 
and numerous languages allow both clausal and nominal exclamatives with seemingly identical interpretation: (2) English a. What strange things he says! b. The strange things (that) he says! (3) French a. (Qu’est-ce) que tu as de beaux biceps! ‘What fine-looking biceps you have!’ b. Les beaux biceps que tu as! ‘The fine-looking biceps you have!’
This paper explores certain aspects of the syntax and semantics of nominal exclamatives, a “clause” type that has received little attention in the literature. The exploration is based on the Austronesian language Malagasy, a language that shows only nominal exclamatives. The paper is organized as follows. I begin with an exploration of exclamatives in Malagasy. Section 2 discusses syntactico-semantic characteristics of exclamatives and shows that the construction under investigation in Malagasy has the sentential force, not just the illocutionary force, of an exclamative. Section 3 investigates the syntactic structure of these exclamatives and argues that, despite surface appearances, they are nominal exclamatives. Clausal exclamatives are unavailable in Malagasy. Section 4 turns to more specific questions regarding the interpretation of nominal exclamatives. Section 5 concludes.
2.â•… Malagasy exclamatives Malagasy is an Austronesian language spoken by approximately 14 million people on the island of Madagascar. Malagasy is a predicate-initial language, with the subject appearing in clause-final position. A variety of phrase types, VP, NP, AP, or PP, can be predicates and there is no copula:1 (4) a.
[mamaky boky]VP ny mpianatra read book the student ‘The student is reading a book.’
.╅ I use the following abbreviations in glossing: dem demonstrative 1/2/3 person acc accusative det determiner art article dir directional
dat dative deic deictic excl exclamation
 Eric Potsdam
b. [vorona ratsy feo]NP ny goaika bird bad voice the crow ‘The crow is a bird with an ugly voice’ c.
[faly amin’ ny zanany]AP Rasoa proud prep the child.3poss Rasoa ‘Rasoa is proud of her children.’
d. [any an-tsena]PP Rakoto prep acc-market Rakoto ‘Rakoto is at the market.’
Typical exclamatives in Malagasy are given in (5) and take the general form in (6). (5) a.
izany sotrohin’ i Paoly! dem drink.pass det Paul ‘What/How much Paul drinks!’
B
izany ngidi-n’ ireo voankazo ireo! dem bitterness-lnk dem fruit dem ‘How bitter that fruit is!’
c.
izany ahaingan’ ny f-amakia-ny boky! dem quickness the nmlz-read-3sg book ‘The quickness of her book-reading!’ ‘How quickly she reads!’
d. izany ao ambanin’ ny fandriana! dem loc under the bed ‘The things under the bed!’
(6) demonstrative (izany)
phrase (VP, NP, PP)
They consist of the demonstrative izany followed by a phrase. In (a), this phrase is a VP, in (b,c) it is an NP, and in (d) it is a PP. Malagasy has a large number of anaphoric demonstrative pronouns/determiners. They indicate the distance of the speaker from the object and whether or not the object is visible. Izany is the vaguest of these demonstratives, indicating that the object is invisible and an indefinite distance away (see Section 4 for further discussion).2
foc gen lnk loc neg nmlz nom
focus genitive linker locative negative nominalizer nominative
obj pass poss pref prep prog prox
object passive possessor prefix preposition progressive proximate
q rel sg/pl spec suff top
question relativizer number specifier suffix topic.
.â•… Rajaonarimanana (1995: 48) asserts that the object is “invisible, singulier, distance indéfinie—une zone plus ou moins étendue”.
Expressing exclamatives in Malagasy 
It is not immediately evident what the grammatical analysis of this construction is. It has both plausible nominal and clausal syntactic parses. In Section 3, I turn to this question and argue that the examples in (5) are noun phrases. Malagasy has no clausal exclamatives. Before doing this however, I show that the data under investigation are true exclamatives. That is, they have exclamative sentential force and not simply exclamatory illocutionary force. This will be accomplished by appealing to widely accepted semantic characteristics of exclamatives. These diagnostics are not particular to Malagasy and can be used quite generally to identify exclamative clauses. It is widely recognized that exclamatives have two properties: factivity and scalarity (Elliott 1971, 1974; Grimshaw 1979; Michaelis & Lambrecht 1996a, 1996b; Michaelis 2001; Zanuttini & Portner 2000, 2003; Villalba 2003; Portner & Zanuttini 2004; Collins 2005; Miró 2006; Rett 2006; and others). Factivity is the property that exclamatives presuppose their propositional content. Scalarity refers to the property of degree exclamatives that they express extreme degree evaluation.3 These two criteria will confirm that the Malagasy construction under investigation is a dedicated exclamative clause type. Exclamatives presuppose their propositional content. This is a consequence of the fact that exclamatives express the speaker’s emotional attitude towards a situation. The situation needs to be assumed to also be known by the hearer so that the speaker can issue an affective response to it. It is thus presupposed and not asserted. There are a number of diagnostics in the literature based on this property that can be used to test whether something is an exclamative. Grimshaw (1979) discusses the fact that exclamatives can often be embedded.4 They are restricted to being embedded under factive predicates like be amazing, which presuppose the truth of their complements, (7). Embedding of exclamatives under other kinds of predicates which do not require a presupposed complement is ill-formed, (8). (7) mahatalanjona/mahatsiravina/maharikoriko [izany sotrohiny] amazing/shocking/disgusting dem drink.pass.3sg ‘It’s amazing/shocking/disgusting how much he drinks.’ (8) *antenaina/heverina/inoana [izany sotrohiny] hope.pass/think.pass/believe.pass dem drink.pass.3sg ‘*It is hoped/thought/believed how much he drinks.’
.╅ See Michaelis and Lambrecht (1996) and Michaelis (2001) for further functionalpragmatic criteria. .╅ It is unclear whether this is true universally. Daniel Kaufman (personal communication) reports that it is not possible in Tagalog.
 Eric Potsdam
In addition, exclamatives are ill-formed as an answer to a question. This follows because the propositional content of an answer is asserted, not presupposed. The utterances in (9) are individually acceptable but they do not form a coherent discourse: (9) a.
nisotro toaka betsaka ve Rabe omaly? drink alcohol much Q Rabe yesterday ‘Did Rabe drink a lot yesterday?’
b. #izany nosotrohiny! dem past.drink.pass.3sg ‘How much he drank!’
Scalarity indicates that exclamatives quantify over a degree which is high relative to some contextually determined value. Exclamatives express emotion on the part of the speaker regarding this high degree. The Malagasy utterances under consideration show this property. First, they allow three types of degree interpretations found in degree exclamatives. These are labeled in Rett (2006) as amount, gradability, and kind interpretations. The amount interpretation quantifies over quantities, (10). The gradability interpretation quantifies over degrees of a gradable property such as height, goodness, badness, etc., (11). Lastly, the kind interpretation quantifies over kinds of individuals, (12) below. In each case, the most salient interpretation depends largely on context. Amount, gradability, or kind are not explicit in the linguistic signal.5 (10) a.
izany ronono sotrohiny! dem milk drink.pass.3sg ‘How much milk he’s drinking!’
b. izany herim-pon’ ny miaramila! dem courage the oldier ‘How much courage the soldiers have!’ (11) a.
izany ahavon’ ny tendrombohitra! dem height the mountain ‘How high the mountain is!’
b. izany boky tsara vakia-ko! dem book good read.pass-1sg ‘What a good book I’m reading!’ (12) a.
izany boky vakiany! dem book read.pass.3sg ‘The (kinds of) books he reads!’
.╅ Most of the examples are thus ambiguous. I do not exhaustively gloss them but, instead, simply give one possible interpretation supplied by my consultants.
Expressing exclamatives in Malagasy 
b. izany ao ambanin’ ny fandriana! dem loc under the bed ‘The (kinds of) things under the bed!’
One consequence of the scalarity property of exclamatives is that they are not acceptable with non-gradable predicates. Examples are infelicitious if the property is not a gradable one. This is the case for English and Malagasy: (13) a.
#izany fahavitana! dem nmlz.be.completed ‘#How completed it is!’ (ok: What a completion!’)
b. #izany fahavakian’ ilay fitaratra! dem nmlz.break that mirror ‘#How broken that mirror is!’ (ok: ‘How that mirror shattered!’)
Finally, scalarity entails that true exclamatives will not be embeddable under predicates which don’t imply an extreme degree. This also holds for the Malagasy construction: (14) a.
mahatsiravina [izany sotrohiny] incredible dem drink.pass.3sg ‘It’s incredible how much he drinks.’
b. #tsy mahatsiravina [izany sotrohiny] neg incredible dem drink.pass.3sg ‘#It’s not incredible how much he drinks.’
I conclude that the Malagasy data are true exclamatives. The construction has the sentential force of an exclamative and is not just a nominal/clause with exclamatory illocutionary force. In the next section, I turn to the syntax of these utterances.
3.â•… The nominal status of Malagasy exclamatives In many languages, the clausal or nominal status of an exclamative is straightforward. Morphological or syntactic cues make the syntax clear. For example, nominal exclamatives may have an unambiguous determiner, some kind of exclusively nominal morphology such as gender concord, or noun phrase word order. Clausal exclamatives cross-linguistically use wh-phrases or complementizer-like elements. In some languages, however, the underlying syntax is not immediately evident. This is the case for the Malagasy data, repeated below, with additional examples. (15) a.
izany sotrohin’ i Paoly! dem drink.pass det Paul ‘What/How much Paul drinks!’
 Eric Potsdam
b. izany ronono sotrohin’ i Paoly! dem milk drink.pass det Paul ‘What/How much milk Paul drinks!’ c.
izany ngidi-n’ ireo voankazo ireo! dem bitterness-lnk dem fruit dem ‘How bitter that fruit is!’
d. izany aha-mena-n’ ny masoandro! dem nmlz-red-lnk the sun ‘How red the sun is!’ e.
izany a-haingan’ ny f-amakia-ny boky! dem nmlz-quick the nmlz-read-3sg book ‘The quickness of her book-reading!’ ‘How quickly she reads!’
f.
izany ao ambani-n’ ny fandriana! dem loc under-lnk the bed ‘The things under the bed!’
(16) demonstrative (izany)
phrase (VP, NP, PP)
Such examples have two possible syntactic parses, as clauses or nominals. A clausal analysis of these data would work as follows. As schematized in (17), the demonstrative pronoun izany is the subject and the remainder phrase is the predicate. (17) [[izany]subject
[vp/np/pp]predicate]
Examples (15 a,c) would have the analysis in (18). (18) a.
[[izany]subject [sotrohin’ i Paoly]predicate] dem drink.pass det Paul Lit.: “That is drunk by him!’
b. [[izany]subject [ngidin’ ireo voankazo ireo]predicate ] dem bitterness dem fruit dem Lit.: “That is the bitterness of those fruit!’
As we saw above, Malagasy allows non-verbal predicates so there is nothing unusual about the lack of a copula. What is perhaps unusual is the subject-initial word order. Canonical clauses in Malagasy are predicate-initial as seen in (4) and (19a). Subjectinitial clauses are generally ill-formed, (19b). (19) a. miasa ny zazalahy work the boy ‘The boys are working.’ b. *ny zazalahy miasa the boy work
Expressing exclamatives in Malagasy 
Subject-initial word order is not completely impossible however. It is seen in a number of environments: small clause complements to perception verbs, (20) (Pearson 2005), after certain clausal conjunctions, (21), and in contrastive environments, (22) (Keenan 1976; Paul 2001; Pearson 2001). (20) nahita [ny mpianatra namaky boky] Rabe see the student read book Rabe ‘Rabe saw the student read(ing) a book.’ (21) a.
tsy miasa androany Rabe… neg work today Rabe ‘Rabe is not working today …’
b. satria [ny vadiny marary] because the spouse.3sg sick c.
satria [marary ny vadiny] because sick the spouse.3sg ‘because his spouse is sick.’
(22) a.
inona no ataon’ ny ankizy? what foc do.pass the children ‘What are the children doing?’
b. ny zazalahy miasa, ny zazavavy matory the boy work the girl sleep ‘The boys are working and the girls are sleeping.’
The exclamative subject-predicate word order in particular could be assimilated to the construction in (22), which is used when two subjects are contrastive (although not exclamative). The alternative, nominal analysis assigns a structure to the exclamatives as in (23). The initial demonstrative izany is a determiner that is followed by a noun phrase. (23) [[izany]D [vp/np/pp]NP]DP
The analysis assigned to (15a,c) is as in (24). In the clear cases, the determiner is followed by something that looks like a noun phrase headed by a noun, (15b–e) and (24b). This noun may be either a lexical noun such as ronono ‘milk’ and ngidy ‘bitterness’ in (15b,c) or a derived noun such as aha-mena ‘nmlz-red’ a-haingana ‘nmlzquick’ in (15d,e). In cases like (15a,f) or (24a), there is possibly a null noun head or direct complementation between the determiner and the VP/PP. I briefly return to this issue below. (24) a.
[[izany]D [sotrohiny]NP]DP dem drink.pass.3sg Lit.: “that amount/stuff drunk by him!”
 Eric Potsdam
b. [[izany]d [ngidin’ ireo voankazo ireo]NP]DP dem bitterness dem fruit dem Lit.: “that bitterness of the fruit!”
In what follows, I present evidence in favor of the nominal analysis for Malagasy. The evidence is of two types. Section 3.1 presents cross-linguistic considerations that favor a nominal approach. Section 3.2 presents syntactic evidence internal to Malagasy.
3.1â•… Cross-linguistic considerations Weak evidence against a clausal analysis for Malagasy exclamatives comes from the observation that cross-linguistically, clausal exclamatives often contain a wh-phrase or complementizer like element. This is illustrated in (25) for several languages. The wh-phrase/complementizer is bold-faced.6 (25) a. How happy she looks! b. What a jerk he is! c.
Italian
Che caro che è! What expensive that is ‘How expensive it is!
d. German
Wen der alles eingeladen hat! who he all invited has ‘Who all he has invited!’
e.
Chinese
Ta shi duo(me) hao de yi ge ren aa! he is how.much good de one cl person excl ‘What a nice person he is!’
f.
Japanese
John-wa nante kasiko-i-no-des-yoo! John-top wh intelligent-pres-fin-foc-mood ‘How intelligent John is!’
(D’Avis 2002:â•›5)
(Oda 2002:â•›99)
(Ono 2006:â•›6)
g. Russian
Kakoi xorošij čelovek Ivan! what good person Ivan ‘What a nice person Ivan is!’
(Oda 2002:â•›100)
.╅ Old Japanese, Classical Tibetan (Janick Wrona, personal communication), and Mandinka (Miro 2006) have clausal exclamatives that do not have wh-phrases or complementizers.
Expressing exclamatives in Malagasy 
h. Thai
Nit dii ?aray kànàat nán! Nit nice what size that ‘What a nice person Nit is!’
i.
Turkish
Ne hoş bir insane John! what nice a person John ‘What a nice person John is!’
j.
French
Que tu as de beaux that you have nice biceps ‘What nice biceps you have!’
(Oda 2002:â•›100)
(Oda 2002:â•›100) biceps!
k. Catalan
Quin pastís tan dolç que ha fet en Ferran! what cake so sweet that has done the Ferran ‘What a sweet cake Ferran made!’
(Miró 2006:â•›2)
In Malagasy, however, I was unable to elicit any exclamations that involved wh-phrases or complementizers. The Malagasy wh-phrases are listed in (26) and its complementizers are in (27). It is evident that the demonstrative izany is not related to any of these elements.7 (26)
a. b. c. d. e. f.
inona ahoana, manahoana ohatrinona iza oviana (t)aiza
(27) a. fa b. raha c. hoe
‘what’ ‘how’ ‘how much’ ‘who’ ‘when’ ‘where’
‘that’ ‘if ’ ‘whether’
.â•… The only possibility is that it is a combination of iza ‘who’ + ny ‘the’. This is unlikely as iza ‘who’ does not occur with a determiner and, if it did, the determiner would be initial: *ny iza. An anonymous reviewer suggests an explanation for the lack of clausal exclamatives employing a wh-phrase: On one view, Malagasy does not have embedded wh-interrogatives of the kind familiar from English and European languages. Embedded wh-questions are expressed either with a headless relative clause (Keenan 1976: 266) or a main clause wh-interrogative preceded by the complementizer-like element and quotative particle hoe (Potsdam 2007). If clausal exclamatives cross-linguistically are built on embedded whinterrogatives, then Malagasy has nothing from which to create a clausal exclamative.
 Eric Potsdam
At the same time, nominal exclamatives are well-attested cross-linguistically and do not use a wh-phrase:8 (28) a.
English
The things that he eats!
b. French
Les beaux biceps que tu as! the nice biceps that you have ‘What nice biceps you have!’
c.
Japanese
hikkirinashi-ni naru denwa! continuously-dat ring telephone ‘The telephones that don’t stop ringing!’
d. Catalan
Les bestieses que diu aquest còmic! the stupid.things that says this comedian ‘The stupid things this comedian says!’
e.
old japanese
punapito-wo miru-ga tomosi-sa boat.people-acc see-gen enviable-nmlz ‘How enviable it is to see the boat people’
f.
Turkish
gittikleri yerler! go.past.rp.pl.obj place.pl ‘The places they go!’
(Miró 2006:â•›3)
(Wrona, this volume:â•›428)
(Michaelis 2001:â•›1048)
g. Setswana
mo-dumo o ba o dirang! cl-noise rel they om make.prog ‘The noise they make!’
(Michaelis 2001:â•›1048)
Nominal exclamatives seem to be particularly common within Austronesian. MoyseFaurie (2011:â•›128) indicates that “Nominalised exclamative clauses may be considered the prototypical expression of exclamation in Polynesian languages, and in some Kanak languages”. The nominal status of such exclamatives is indicated in two ways in the data below, from Moyse-Faurie (2011). There is either a determiner (glossed art
.╅ Some examples contain a relativizer which is part of the relative clause inside the nominal.
Expressing exclamatives in Malagasy 
or spec below, following the original source), as in (29), or nominalizing morphology (glossed pref), as in (30).9 (29) a.
Maori
te ataahua o tērā kōtiro! art beauty poss deic girl ‘What a beautiful girl! (Lit.: the beauty of that girl)’ (Foster 1987:â•›118; cited in Moyse-Faurie 2011:â•›120)
b. Samoan le aka a le ali’i! spec laugh poss spec guy ‘Look, how the guy is laughing! (Lit.: the laughing of the guy)’ (Mosel & Hovdhaugen 1992:â•›521; cited in Moyse-Faurie 2011:â•›120) c.
Tuvaluan
te mmoko mai o te matagi! spec cold dir poss spec wind ‘The wind feels so [nice and] cool! (Lit.: the coolness of the wind)’ (Besnier 2000:â•›40; cited in Moyse-Faurie 2011:â•›119)
(30) a.
Iaai
caa û-nââk-iny miitr! neg pref-tough-suff meat ‘Isn’t this meat tough! (Lit.: not the toughness of the meat)’ (Ozanne-Rivierre 1984:â•›115; cited in Moyse-Faurie 2011:â•›128)
b. West Uvean
fale-leu fagamalie de fua nei pref-ripe well art fruit deic ‘How ripe this fruit is!’
We can see the use of both elements in the Malagasy example in (31). Izany, I am proposing, is a determiner, and the phrase following the determiner is a nominalized adjective with a possessor. (2) izany aha-mena-n’ ny masoandro! dem nmlz-red-lnk the sun ‘How red the sun is! (Lit.: That redness of the sun!)’
3.2â•… Syntactic evidence Syntactic evidence internal to Malagasy also supports a nominal analysis. The exclamatives under investigation have the distribution of nominals, not clauses. As we have already seen, exclamatives may occur embedded under factive predicates. This is also .â•… Examples (29a–c) and (30a) were also cited in a paper on “Nominalisation and exclamation” presented by Claire Moyse-Faurie at the Conference on Oceanic Linguistics (COOL-7) in 2007.
 Eric Potsdam
true of Malagasy, where exclamatives may occur in a variety of subordinate positions. Example (32) demonstrates that they may appear in subject position. (32) a. b.
mahagaga an-dRasoa [izany sotrohiny] surprise acc-Rasoa dem drink.pass.3sg ‘How much he drinks surprises Rasoa.’ heno-ko [izany ahaingan’ ny famakiany boky] hear.pass-1sg dem quickness det read.nmlz.3sg book ‘I heard about how fast she reads.’
c.
mahatalanjona [izany ditran’ ny alika] amazing dem stubbornness det dog ‘How stubborn the dog is is amazing.’
Clauses may also appear as subjects in Malagasy; however, they usually have an introductory complementizer, which is impossible here: (33) *mahagaga an-dRasoa [fa/hoe izany sotrohiny] surprise acc-Rasoa that/whether dem drink.pass.3sg ‘How much he drinks surprises Rasoa.’
More telling, exclamatives can occur as objects of the general preposition amy, (34). Observe the preposition in the English translation. Clauses, as far as I am aware, do not occur in this position. (34) a.
tsiravina ami-n’ [izany sotrohiny] aho shocked prep-lnk dem drink.pass.3sg I ‘I am shocked at how much he drinks.’
b. zendana aho ami-n’ [izany ahavon’ ny tendrombohitra] astonished I prep-lnk dem height the mountain ‘I am astonished at how high the mountain is.’
A further piece of distributional evidence in favor of the nominal analysis is that exclamatives can be fronted. Malagasy has two fronting constructions in which the subject of the clause is preposed and followed by either the particle dia to mark topic or no to mark focus. Both constructions are possible with exclamatives: (35) a.
nampalahelo ahy [izany fahantran’ ny olona] sadden me dem poverty the people ‘How poor the people are saddened me.’
b. [izany fahantran’ ny olona] dia nampalahelo ahy dem poverty the people topic sadden me ‘How poor the people are, it saddened me.’ c.
[izany fahantran’ ny olona] no nampalahelo ahy dem poverty the people focus sadden me ‘It’s how poor the people are that saddened me.’
Complement clauses, in contrast, do not front with these constructions:
Expressing exclamatives in Malagasy 
(36) a. nampalahelo ahy [fa lasa Rabe] sadden me that gone Rabe ‘That Rabe is gone saddened me.’ b. *[fa lasa Rabe] dia/no nampalahelo ahy that gone Rabe topic/focus sadden me ‘That Rabe is gone, it saddened me. /It’s that Rabe is gone that saddened me.’
I conclude that the external distribution of exclamatives supports a nominal analysis over a clausal one. Exclamatives also do not have the internal structure of a clause. For example, contrary to the clausal analysis, izany in exclamatives does not behave like a subject. It cannot front with the topic and focus particles introduced above, (37). These constitute well-known subjecthood tests in Malagasy (Keenan 1976). This is expected if izany is only a determiner within a noun phrase. (37) a. b.
izany (ronono) nosotrohiny! dem milk past.drink.pass.3sg ‘How much (milk) he drank!’ *izany (ronono) dia/no nosotrohiny! dem milk topic/focus past.drink.pass.3sg
Lastly, the construction as a whole does not behave like a clause in that sentence-level particles like tena ‘really, indeed’ and toa ‘seem’ cannot appear in exclamatives: (38) a.
tena nisotro toaka betsaka izy really past.drink alcohol much he ‘He really drank lots of alcohol.’
b. *izany tena nosotrohiny! dem really past.drink.pass.3sg (‘How much he really drank!’) (39) a.
toa nisotro toaka betsaka izy seem past.drink alcohol much he ‘He seems to have drunk lots of alcohol.’
b. *izany toa nosotrohiny! dem seem past.drink.pass.3sg (‘How much he seems to have drunk!’)
I conclude that Malagasy exclamatives do not have the internal syntax of a clause.
3.3â•… The internal structure of nominal exclamatives Accepting that Malagasy exclamatives are nominal, what is their internal structure? In the transparent cases, there is a determiner followed by an NP with an overt noun head. This NP may consist of just a noun, (40), a noun with a possessor, (41), or a noun
 Eric Potsdam
with a relative clause, (42).10 Note that the optional relativizer izay is generally possible in the latter examples. (40) a.
izany trano! dem house ‘What a house!’
b. izany korontana! dem mess ‘What a mess!’ c.
izany Rasoa! dem Rasoa ‘How [pragmatically appropriate adjective] Rasoa is!’
(41) a.
izany ngidin’ ireo voankazo ireo! dem bitterness dem fruit dem ‘How bitter that fruit is!’
b. izany ahaingan’ ny f-amakia-ny boky! dem quickness the nmlz-read-3sg book ‘How quickly she reads!’ c.
izany aha-tsara-n-tarehi-n-dRasoa! dem nmlz-good-lnk-face-lnk-Rasoa ‘How pretty Rasoa is!’
d. izany ditran’ ny alika! dem stubbornness the dog ‘The stubbornness of the dog!’ (42) a.
izany sakafo [rel. clause nohaniny ]! dem food past.eat.pass.3sg ‘The food he ate!’
b. izany zavatra [rel. clause sotrohiny ]! dem thing drink.pass.3sg ‘The things that he’s drinking!’ c.
izany boky tsara [rel. clause (izay) vakiako ]! dem book good rel read.pass.1sg ‘What a good book I’m reading!’
The NP may not have another determiner, since izany functions as the determiner for the whole phrase:
.╅ I assume that these examples are genuine exclamatives based on their interpretation and native speaker intuitions.
Expressing exclamatives in Malagasy 
(43) a. *izany ny f-amakian’ ny mpianatra boky dem the nmlz-read the student book ‘How quickly the students read!’ b. *izany ilay/ny trano! dem that/the house ‘What a house!’
To this point, we have a noun phrase with the unexceptional structure in (44). (44)
DP D izany
NP NP
(RC)
Many of the exclamatives do not have a head noun however, (45). In these cases, there are at least two options. There could be a null head noun modified by a relative clause (RC), (46a). Alternatively, the determiner izany could directly take the relative-clause like element as its complement, (46b), with no mediating head noun. (45) a.
izany sotrohin’ i Paoly! dem drink.pass det Paul ‘What/How much Paul drinks!’
b. izany ao ambanin’ ny fandriana! dem loc under the bed ‘The things under the bed!’ (46)
a.
DP D izany
b. NP
NP
(RC)
DP D izany
RC/PP
N Ø
Both options are a priori plausible and I will not decide between them here. The issue is not specific to exclamatives. Malagasy shows wide use of “headless” nominalizations throughout the grammar. There is currently very little work on nominal structure in Malagasy (see Ntelitheos 2006). For simplicity, I will assume the null head structure in (46a) in what follows.
 Eric Potsdam
4.â•… Localizing the exclamative force A primary question that arises in the analysis of nominal exclamatives is where the exclamatory sentential force comes from, if anywhere. The literature suggests three lines of thought: Michaelis & Lambrecht 1996a, 1996b offer a Constructional approach. Nominal exclamatives are representative of at least two constructions: a Metonymic NP Construction, which “licenses a reading of a definite NP in which a definite NP ‘stands for’ a proposition that invokes some degree of a scalar property” (Michaelis & Lambrecht 1996b: 237), and the Abstract Exclamative Construction, which requires that exclamatives display a small set of semantico-pragmatic properties (Michaelis & Lambrecht 1996b: 239). Takahashi (1997) critically evaluates their analysis and I will not discuss it here. An alternative suggested, but rejected, in Michaelis and Lambrecht (1996a) is an ellipsis analysis in which nominal exclamatives are derived from full clauses. Thus, they propose that (47) could be derived from either (48a) or (48b). A problem that they point out is that there are multiple sources for a given nominal exclamative; however, such an exclamative does not seem to be genuinely ambiguous, only vague. The ellipsis operation that would be necessary to get from (48) to (47) seems unmotivated. (47) The nerve of that man! (48) a. I can’t believe the nerve of that man! b. It’s incredible the nerve of that man!
Instead, they suggest that the affective judgment in (47) could be “pragmatically inferred” (Michaelis & Lambrecht 1996a: 388) without deletion. I will investigate a third alternative, pursued in Portner and Zanuttini (2004) (henceforth P&Z). They propose that a clause-like meaning is achieved through a compositional semantics. The exclamative interpretation is due to an exclamative morpheme E present in the syntactic structure. Portner and Zanuttini locate E in the relative clause that is included in the nominal: (49) The things [E that he says]!
They argue that English nominals are only exclamative when they contain a relative clause. Their judgments are that only (50a) can be an exclamative; that is, it is exclamatory by virtue of its structure. (50b), in contrast, is an exclamation only. Its force is determined pragmatically, via the context. (50) a. The strange people who come from Italy! b. The/Those strange Italians!
To illustrate further, they claim that The height of the building! can be an exclamation expressing surprise or astonishment over the height of the building; however, it is an
Expressing exclamatives in Malagasy 
elliptical sentence fragment and, hence, not an exclamative (P&Z 2004: 5). Examples like (50a), by contrast, are not felt to be elliptical. They support their claims with agreement and pronominalization facts in embedded contexts. When a true exclamative is embedded it behaves morphosyntactically like a proposition. In English, exclamatives embedded in post-predicate position are pronominalized with the singular subject it because of their propositional meaning. This is true not only for clausal exclamatives, (51a), but also nominal ones, (51b). In support of the claim that nominals without a relative clause are not exclamatives, P&Z (2004: 6) claim that they do not license it as a subject, (51c). They conclude from these data that the relative clause is necessary to turn a nominal into an exclamative. It is the exclamative morpheme E in the relative clause that gives the nominal a propositional interpretation.11 (51) a. It’s amazing [what strange things he says]. b. It’s amazing [the strange things he says]. c. *It’s amazing, the Italians.
I would like to propose an alternative analysis for the location of the E morpheme based on the Malagasy data. Unlike English, a relative clause does not seem to be required for a nominal exclamative. The examples in (52) appear to be simple noun phrases.12 (52) a.
izany herim-pon’ ny miaramila! dem courage the soldier ‘The courage of the soldiers!’
b. izany korontana! dem mess ‘What a mess!’ c.
izany Rasoa! dem Rasoa ‘How [pragmatically appropriate adjective] Rasoa is!’
d. izany ao amin ny paosi-tranony! dem loc prep the closet.3sg ‘The things in his closet!’
This suggests that the exclamative morpheme cannot necessarily reside in the relative clause and it must be possible to locate it within the nominal itself. I propose that the
.╅ I, and others, do not share this intuition that a relative clause is necessary for a true exclamative interpretation. For the purposes of what follows, however, I accept the generalization. .╅ As before, I assume that the examples are genuine exclamatives based on their interpretation and native speaker intuitions. (d) might contain a reduced relative clause; however, this is much less likely for the other noun phrases.
 Eric Potsdam
exclamative interpretation lives in the determiner. The main argument for this view is that the determiner in nominal exclamatives is restricted cross-linguistically. This is clearly the case for Malagasy, which I illustrate below, but it is also the case for other, unrelated languages.13 Malagasy has a large number of demonstratives, given in Table 1. These elements can be used as determiners introducing a noun phrase, (53a), or they can be used alone, as pronominals, (53b). Table 1.╇ Malagasy demonstratives (Rahajarizafy 1960: 24) Visible
Near Far Very far Indefinite
sg
pl
ito, ity itsy iry io, iny
ireto iretsy irery ireo, ireny -re- infix
–
(53) a.
Invisible or vague
izato, izaty izatsy izary izao, izany -za- infix
faly izany zaza happy dem child ‘That child is happy.’
b. ratsy izany bad dem ‘That’s bad.’
The demonstratives differ along three dimensions: number (singular vs. plural), distance of the object from the speaker, and whether or not the object is visible. As can be seen in the last row of the table, the two final columns are derived from the visible singular forms by infixation of some CV element. Malagasy also has a determiner ny, which roughly translates as ‘the’. Rajemisa-Raolison (1969: 54) indicates that only the bold-faced demonstratives can be used in exclamatives. The other demonstratives are not possible. Although there is some pattern, the choice seems somewhat arbitrary. For example, of the two near singular demonstratives, ity is a possible exclamative determiner but ito is not. When the infix -za- ‘invisible’ is inserted however, i〈za〉ty is not
.â•… P&Z (2004: 5–6, 9) in fact suggests that English nominal exclamatives can only contain the, citing the contrast in (i) with an exclamative interpretation.
(i)
The/??Those strange people who come from Italy!
If this is a real contrast, it would support the claim that determiners are restricted in nominal exclamatives.
Expressing exclamatives in Malagasy 
possible as an exclamative determiner but i〈za〉to is. (54) confirms that the definite determiner ny ‘the’ is also not possible with nominal exclamatives, nor is the absence of a determiner. (54) a. *ny ahavon’ ny tendrombohitra the height the mountain b. *ahavon ny tendrombohitra height the mountain ‘The height of the mountain!’ not acceptable as an exclamative
In addition, Malagasy has a framing demonstrative construction in which any of the above demonstratives occurs at the beginning and the end of the noun phrase: (55) a.
ireo olona ireo dem person dem ‘those people’
b. izao holazaiko izao dem fut.say.pass.1sg dem ‘what I will say’ c.
izany vaovao mampalahelo izany dem news sadden dem ‘the sad news’
d. izany tenin’ ny Ntaolo izany dem word the ancestor dem ‘the words of the ancestors’
This framing use of the demonstratives is also impossible with an exclamative meaning: (56) *izany herim-pon’ ny miaramila izany dem courage the soldiers dem ‘The courage of the soldiers!’ not acceptable as an exclamative
I conclude that the choice of determiner, as well as its particular syntax, determines the availability of an exclamative meaning. This suggests that the exclamative morpheme is associated with the determiner. This determiner restriction in nominal exclamatives occurs in other languages. In Spanish, there is an invariant determiner-like element la de which occurs with nominal exclamatives of amount, (57). La de, which may be a contraction of la cantidad de ‘the quantity of ’, does not inflect for gender or number and is restricted to exclamatives.14 One might expect to find exclamative-specific determiners in other languages as well.
.╅ I thank Maria Biezma for discussions about the Spanish data.
 Eric Potsdam
(57) a.
la de cosas que come Juan! det thing.fem.pl that ate Juan ‘How much Juan ate!’
b. la de libros que leyó Juan! det book.masc.pl that read Juan ‘How many books John read!’
In Russian, there are both proximal and distal demonstratives that inflect for case, number, and gender. Nominal exclamatives require the proximate nominative forms, (58). Nominals with distal demonstratives or no demonstrative cannot be used as exclamatives, (59).15 (58) a.
(ox už) èti deti! oh yet dem.prox.nom.pl children.nom.pl ‘The children!’
b. (ox už) èta kurica! dem.prox.f chicken ‘The chicken!’ c. (ox už) èto suščestvo! dem.prox.n creature ‘The creature!’ (59) a. *deti! children.nom not acceptable as an exclamative, only as a vocative b.
*te deti! dem.dist.nom.pl children.nom.pl ‘those children’ not acceptable as a exclamative, only as an answer to a question
As in Malagasy, the demonstrative pronouns do not have deictic force in the nominal exclamatives. In Russian, demonstratives and possessives cannot co-occur: (60) a.
ja videl *ètix/?tex ejo detej I saw dem.prox.acc.pl/dem.dist.acc.pl her children.acc ‘I saw her children.’
b. *èti ejo deti ploxo sebj veli dem.prox.nom.pl her children badly self conducted (‘Her children were badly behaved.’)
.╅ I thank Maria Polinsky for discussions about the Russian data.
Expressing exclamatives in Malagasy 
This restriction is lifted when the demonstrative occurs in a nominal exclamative: (61) èti ejo deti! dem.prox.nom.pl her children.nom.pl ‘These children of hers!’
This very small sample of languages suggests that there is indeed something special about the determiner in nominal exclamatives that contributes the exclamative force. If this hypothesis is correct, it sets up a reasonable parallel with clausal exclamatives. The exclamative force of a clausal exclamative is often taken to reside in the complementizer domain, which is the clausal parallel to the determiner domain in nominals. At the same time, it is not unexpected that there may be languages that do not have nominal exclamatives. This can be attributed to them not having the lexical resources in their determiner system. Hindi seems to be one such language.16
5.â•… Conclusion This paper has used exclamative constructions in Malagasy to point to the importance of nominal exclamatives to the study of clause types. Although largely ignored in the literature, nominal exclamatives are the dominant, and sometimes only, realization of exclamatives in some languages. Such nominals are true exclamatives given the semantic and syntactic tests in the literature. It is clearly not an accident that nominals and nominalizations are so widely used to encode exclamatives. One question for investigation is why clauses and nominals are so consistently co-opted for this function. In starting to address this question, I briefly explored the issue of where in a nominal exclamative the sentential force was located. There are numerous possible answers to this question and I suggested a compositional approach in which the exclamative force was isolated in the determiner. Additional work is needed to flesh out this proposal and evaluate it against a wider range of languages and data.
References Austin, J.L. 1962. How to Do Things with Words. Cambridge MA: Harvard University Press. Besnier, Niko. 2000. Tuvaluan: A Polynesian Language of the Central Pacific. London: Routledge. Chierchia, Gennaro & McConnell-Ginet, Sally. 1990. Meaning and Grammar: An Introduction to Semantics. Cambridge MA: The MIT Press.
.╅ I thank Rajesh Bhatt for discussions about the Hindi data.
 Eric Potsdam Collins, Peter. 2005. Exclamative clauses in English. Word 56: 1–17. D’Avis, Franz-Josef. 2002. On the interpretation of wh-clauses in exclamative environments. Theoretical Linguistics 28: 5–31. Elliott, Dale E. 1971. The Grammar of Emotive and Exclamatory Sentences in English. Ph.D. dissertation, Ohio State University. Elliott, Dale E. 1974. Toward a grammar of exclamations. Foundations of Language 11: 231–246. Foster, John. 1987. He whakamārama. A Full Self-Help Course in Māori. Auckland: Reed. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10: 279–236. von Heusinger, Klaus & Schwabe, Kerstin. 2002. Preface to the special edition on information structure and sentence type. Theoretical Linguistics 28: 1–4. Huddleston, Rodney & Pullum, Geoffrey K. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Keenan, Edward L. 1976. Remarkable subjects in Malagasy. In Subject and Topic, Charles N. Li (ed.), 247–301. New York NY: Academic Press. Michaelis, Laura A. 2001. Exclamative constructions. In Language Typology and Language Universals, Martin Haspelmath, Ekkehard Koenig, Wulf Oesterricher & Wolfgang Raible (eds), 1038–1050. Berlin: Mouton de Gruyter. Michaelis, Laura A. & Lambrecht, Knud. 1996a. The exclamative sentence type in English. In Structure, Discourse and Language, Adele Goldberg (ed.), 375–389. Stanford CA: CSLI. Michaelis, Laura & Lambrecht, Knud. 1996b. Toward a construction-based theory of language function: The case of nominal extraposition. Language 72: 215–247. Miró, Elena Castroviejo. 2006. Wh-exclamatives in Catalan. Ph.D. dissertation, Universitat de Barcelona. Mosel, Ulrike & Hovdhaugen, Even. 1992. Samoan Reference Grammar. Oslo: Scandinavian University Press. Moyse-Faurie, Claire. 2011. Nominalization and exclamation in Oceanic languages. In Topics in Oceanic Morphosyntax, Claire Moyse-Faurie & Joachim Sabel (eds), 111–136. Berlin: Mouton de Gruyter. Ntelitheos, Dimitrios. 2006. The Morphosyntax of Nominalizations: A Case Study. Ph.D. dissertation, UCLA. Oda, Toshiko. 2002. Exclamatives and negative islands. In Chicago Linguistic Society 38–2: The Panels, Mary Andronis, Erin Debenport, Anne Pycha & Keiko Yoshimura (eds), 97–110. Chicago IL: Chicago Linguistic Society. Ono, Hajime. 2006. An Investigation of Exclamatives in English and Japanese: Syntax and Sentence Processing. Ph.D. dissertation, University of Maryland, College Park. Ozanne-Rivierre, Françoise. 1984. Dictionnaire Iaai-Français (Ouvéa, Nouvelle-Calédonie). Paris: Peeters-Selaf. Paul, Ileana. 2001. Ve as a second-position clitic. Oceanic Linguistics 40: 135–142. Pearson, Matthew. 2001. The Clause Structure of Malagasy: A Minimalist Approach. Ph.D. dissertation, UCLA. Pearson, Matthew. 2005. The Malagasy subject/topic as an A΄ element. Natural Language and Linguistic Theory 23: 381–457. Portner, Paul & Zanuttini, Raffaella. 2004. The semantics of nominal exclamatives. In Ellipsis and Nonsentential Speech, R. Elugardo & Robert J. Stainton (eds), 57–67. Boston MA: Kluwer. Potsdam, Eric. 1998. Syntactic Issues in the English Imperative. New York NY: Garland. Potsdam, Eric. 2007. Malagasy sluicing and its consequences for the identity requirement on ellipsis. Natural Language and Linguistic Theory 25: 577–613.
Expressing exclamatives in Malagasy 
Rahajarizafy, R.P. Antoine. 1969. Essai sur la grammaire malgache. Antananarivo: Imprimerie Catholique. Rajaonarimanana, Narivelo. 1995. Grammaire moderne de la langue malgache. Paris: L’Asiathèque. Rajemisa-Raolison, Régis. 1969. Grammaire malgache. Fianarantsoa: Librairie Ambozontany. Rett, Jessica. 2006. Exclamatives are degree constructions. Handout, Rutgers Linguistics Conference 1. Sadock, Jerrold & Zwicky, Arnold. 1985. Speech act distinctions in syntax. In Language Typology and Syntactic Description, Vol. 1, Timothy Shopen (ed.), 155–196. Cambridge: CUP. Searle, John. 1969. Speech Acts: An Essay in the Philosophy of Language. Cambridge: CUP. Takahashi, Mineko. 1997. A critical review of a construction grammatical analysis of the concealed exclamation. Sophia Linguistica 41: 149–169. Villalba, Xavier. 2003. An exceptional exclamative sentence type in Romance. Lingua 113: 713–745. Wrona, Janick. This volume. A case of non-derived stand-alone nominalization. In Yap, Grunow-Hårsta & Wrona (eds). Zanuttini, Raffaella & Portner, Paul. 2000. The characterization of exclamative clauses in Paduan. Language 76: 123–132. Zanuttini, Raffaella & Portner, Paul. 2003. Exclamative clauses: At the syntax-semantics interface. Language 79: 39–81.
Nominalizations in Toqabaqita and closely related languages* Frantisek Lichtenberk University of Auckland
The paper has two major focuses. One is the nominalizing suffixes found in Toqabaqita, an Oceanic language, and its near relatives, in particular those nominalizers that are of some antiquity and one recent replacement. The other one is certain aspects of nominalizations in Toqabaqita. One feature of Toqabaqita is double nominalizations, where the nominalizing suffix occurs twice. Double nominalizations are also found in a few other Oceanic languages. Toqabaqita nominalizations can be used with a stance function. And one also finds “cognate nominalization” in Toqabaqita, where a noun phrase is headed by the nominalization of the verb that heads the predicate. There is also “cognate relativization”, where the noun modified by a relative clause is the nominalization of the verb in the relative clause.
1.â•… Introduction The Toqabaqita language is a member of the Longgu/Malaita/Makira group of languages, also known as the Cristobal-Malaitan group. The name Longgu/ Malaita/Makira (henceforth LMM) has been adopted from Lynch et al. (2002). There are between 20 and 30 LMM languages, spoken in the south-eastern part of the Solomon Islands. They form a low-level subgroup within Oceanic, which in turn is a subgroup within Austronesian. In this study most of the data will come from a single LMM language, Toqabaqita, simply because this is the language for which I have the most detailed data.
*I am grateful to participants at the Workshop on Nominalizers and Copulas in East Asian and Neighboring Languages for comments on the version of this paper delivered there and to the three referees for their comments on a subsequent version. I am also grateful to Br. Athanasius Faifu for the data on South Lau, obtained from him on various occasions in 1989. The Toqabaqita data come from my field notes. The sources of the other data are as acknowledged. In some cases the morpheme breaks and morpheme glosses have been added by me.
 Frantisek Lichtenberk
This paper has two main goals. One is to survey some nominalization strategies in languages that belong in the LMM group. This investigation is hampered by paucity of detailed information on nominalization in these languages; nevertheless, some patterns can be discerned. As a rule, the LMM languages exhibit more than one type of nominalization strategy; however, here the focus will be on one of the types found in Toqabaqita and on the corresponding forms in the other languages. The historical and comparative study of nominalizations in the LMM languages is dealt with in Section 2. The other main goal is investigation of nominalizations in Toqabaqita, with special focus on some of their characteristics, specifically double nominalizations, the use of nominalizations with a (speaker) stance function, and cognate nominalizations and their use in cognate relativization. These matters are treated in Section 3. The concluding Section 4 provides a summary.
2.â•… Nominalizations in the LMM languages 2.1â•… Background Before investigating nominalization patterns in the LMM languages, it is useful to consider some historical and comparative background. Lynch et al. (2002: 70) reconstruct three productive nominalizing affixes for Proto Oceanic. These are listed in (1): (1) a. *i- instrument nominalizer b. *-in- patient and general nominalizer c. *-an location and general nominalizer
Lynch et al. do not reconstruct any “action” nominalizer (for action nominalizations see further below), although it is possible that the label “general nominalizer” subsumes action nominalizers. All three productive nominalizers reconstructed by Lynch et al. for Proto Oceanic had been inherited from Proto Malayo-Polynesian, Malayo-Polynesian being a firstorder subgroup of Austronesian. With respect to Proto Oceanic *-an (1c), Lynch et al. (2002: 70) point out “phonological complications”: “we find forms that reflect not only *-an but also *-aŋ, *-ana, *-aŋa and *-ŋa”. It is the forms in LMM languages that point to *-ŋa that will be particularly relevant to the discussion that follows. There is no comprehensive study of nominalization in Oceanic; nevertheless, there are certain patterns that can be observed. Oceanic languages typically have more than one nominalization strategy, whereby a noun is derived from a verb. One of these is action nominalization, and there are one or more other nominalization patterns. “Action nominalization” is used here as a cover term for nominalizations that express, following Comrie and Thompson (1985: 350), “the fact, the act, the quality, or occurrence” of the state of affairs designated by the source verb. For example, Toqabaqita makes
Nominalizations in Toqabaqita and closely related languages 
a grammatical distinction between action nominalizations and deverbal nouns. Deverbal nouns have the syntactic properties of basic, non-derived nouns. On the other hand, action nominalizations exhibit properties both of nouns and of verbs (Section 3.2). Formation of action nominalizations is fully productive: any verb can be turned into an action nominalization. Action nominalizations are semantically regular: they express the facts, the acts, the qualities, or occurrences of the states of affairs designated by their source verbs (see the quote from Comrie & Thompson 1985 above). And they readily permit, in certain regular ways, the expression of one or both of the core arguments of the source verb, the subject and/or the direct object (Section 3.2). On the other hand, the formation of deverbal nouns is lexically restricted, their meanings are not always transparent, and with many of them neither of the core arguments of the source verb is expressed. The characteristic semantic properties of action nominalizations and deverbal nouns and the expression and non-expression of the core arguments with the two types are illustrated in (2) and (3). The construction in (2) contains an action nominalization. The noun phrase suka ‘sugar’, which is grammatically the possessor of the nominalization, corresponds to the direct object of the verb uusi ‘buy’. (The personal suffix on the nominalization indexes the possessor.) (2) uusi-la-na suka buy-nmlz-3sg.pers sugar ‘(the act of) buying sugar’, ‘buying of sugar’
On the other hand, the form in (3), also based on uusi ‘buy’, is a deverbal noun. It designates not an act of buying (something), but the place where buying takes place: (3) uusi-a buy-dvn ‘market’
Action nominalizations are formed by means of a suffix morpheme that has three variant forms. In (2) above the suffix has the form -la, and in (4) below -laa: (4) ngata-laa speak-nmlz ‘speaking’
Action nominalizations in Toqabaqita are discussed in detail in Section 3. Section 3.2 contains a brief mention of the third form of the nominalizer, -le. Deverbal nouns are formed by means of the suffix -a. They can designate a variety of concepts: i.
the type of event, state, condition or characteristic designated by the source verb:
(5) a.
firu-a b. maqu-a fight-dvn be.afraid-dvn ‘(a) fight, war’ ‘fear’
 Frantisek Lichtenberk
ii. the outcome or result of the event designated by the source verb: c.
manata-a d. qaa-qafu-a think-dvn rdp-wrap-dvn ‘thought, idea’ ‘human corpse wrapped in a mat for burial’
iii. the place or, less commonly, the time where or when the state of affairs denoted by the source verb obtains: e.
uusi-a f. arakwa-a buy-dvn (moon)be.full-dvn ‘market’ ‘time of full moon’
iv. means, instruments, tools:
g. laungi-a h. decorate.os-dvn ‘personal decorations, ornaments’
nono-m-a rub.sb.with.healing.leaves-tr-dvn ‘leaves used in healing’ (the patient’s body is rubbed with such leaves)
v. the performer of the event designated by the source verb: i.
thuu-fa-a cover-tr-dvn ‘protective, guardian spirit’, ‘Saviour (Christ)’
There is a minor pattern where a deverbal noun may designate a type of action or the performers of the action. In the latter case the deverbal noun always signifies a group of people acting collectively. When the same deverbal noun refers to the action, the normal significance is that of a collectivity of people involved, although this is not necessarily the case. j.
luqu-a
shift.residence-dvn (i) ‘movement, relocation of people to a different area’ (ii) ‘group of people moving, relocating to a different area’
k. lae-a
go-dvn (i) ‘going; occasion of going somewhere’ (ii) ‘group of people going somewhere’
Besides action nominalizations and derived deverbal nouns as just discussed, there are also cases in Toqabaqita where there is no (overt) morphological difference between a verb and the corresponding noun; for example, butha ‘chew betel quid’, ‘juice produced when chewing a betel quid’; raraqu ‘make string figures, play cat’s cradle’, ‘string figure, cat’s cradle’. In the absence of historical evidence one can’t tell whether such cases involve formation of the noun from the verb or vice versa.
Nominalizations in Toqabaqita and closely related languages 
Even though Oceanic languages typically contain more than one nominalizing strategy, even recent grammatical descriptions of Oceanic languages do not always comment in detail on the differences between the types of nominalizations. (For a cross-linguistic discussion of different types of nominalization in several Oceanic languages see Moyse-Faurie [2007].)
2.2â•… Nominalizers in the LMM languages The purpose of this section is to consider nominalizing morphemes, suffixes, in the LMM languages. However, only those nominalizers are considered in some detail that are directly relevant to the subsequent discussion of the Toqabaqita action nominalizer -la/-laa, illustrated in (2) and (4), respectively, in Section 2.1. One of these nominalizers has the orthographic forms -nga and/or -ngaa/-ngā in some languages and -na in others. These will be referred to as “nasal-initial nominalizers”. The nasal-initial nominalizers Austronesian
Oceanic
Southeast Solomonic
Bugotu/Gela/Guadalcanal
Longgu/Malaita/Makira
Malaita/Makira
Longgu
Arosi
Ulawa
Toqabaqita Lau South Lau Kwara‘ae Kwaio Marau Oroha ‘Are‘are
Figure 1.╇ The position of the LMM languages in the Oceanic family (after Lynch et al. 2002), and those LMM languages from which data are cited here (the language names are in italics; the internal composition of the Malaita/Makira subgroup has been simplified)
 Frantisek Lichtenberk
are (direct) descendents of Proto Oceanic *-ŋa (Section 2.1). The other nominalizer has the forms -laa/-lā, -la and/or -le/-lä in some languages, and -raa and possibly also -ra in others. These will be referred to as “liquid-initial nominalizers”. The historical relation between the nasal-initial and the liquid-initial nominalizers will be briefly considered in Section 2.2.11. The languages from which data are presented below and their position in the LMM subgroup are given in Figure 1. The languages are listed, approximately, in the order of their genetic distance from Toqabaqita, both in terms of the subgroups within Longgu/Malaita/Makira and within the subgroup of which Toqabaqita is a member, Longgu being the most distant, and Lau the closest.
2.2.1â•… Longgu In his brief sketch of the grammar of the language, Ivens (1934: 607) mentions the existence of two “gerundival endings”, -a and -la. He says he has only one instance of the form -la in the texts available to him: (6) na soko-la-na na maramana art? finish-nmlz-3sg.poss art? earth ‘the end of the earth’
(Ivens 1934: 608)
(Ivens lists na as an article/demonstrative. Hill [1992: 95] [see further below] gives a as the article used with common nouns, while Ivens considers a to be a personal article. There are various inconsistencies in Ivens’ brief description of the language, and some of his analysis and examples need to be treated with caution.) In (6) above, the nominalizing suffix is followed by the third person singular possessive suffix -na. As will be seen later, in some LMM languages there are two nominalizing variants -la and -laa/-lā, with the former used before possessive suffixes and the latter in word-final position. The other “gerundival ending”, -a, is illustrated in (7), where it is followed by the third person singular possessive suffix -na: (7) na nai vaolu-a-na art? ? new-nmlz-3sg.poss ‘the renewal of it’
(Ivens 1934: 608)
The nominalizer -a is cognate with the Toqabaqita suffix -a, used to form deverbal nouns; see the examples in (5) in Section 2.1. Cognate forms are also found in other LMM languages. Ivens also mentions a “noun ending” of the form -nga but does not say anything about its function; see (8). For the meaning ‘danger’ of a nominalization based on the verb ‘die’ see also Arosi maeraa in Section 2.2.2. (8) mae-mae-nga rdp-die- nmlz? ‘danger’
(Ivens 1934: 607)
Nominalizations in Toqabaqita and closely related languages 
In a more recent comprehensive description of Longgu, Hill (1992) mentions neither -la nor -nga as nominalizing suffixes. According to her, nominalization of verbs is achieved by means of one of two noun phrase clitics, -i singular or -gi plural; or by means of the suffix -a, followed by a possessive suffix; or by means of reduplication. Example (9) contains the suffix -a; cf. (7) above from Ivens (1934): (9) vuni-a-na begin-nmlz-3sg.poss ‘its beginning’
(Hill 1992: 112)
2.2.2â•… Arosi Fox (1978, but compiled in the early part of the 20th century) gives -nga “suffix forming nouns, added to verbs, adjectives and nouns” (p. 319); and -ha “gerund. suffix” (p. 148). The nominalization in (10) contains the suffix -nga and the one in (11), based on the same verb, contains -ha: (10) tahi-nga live/flourish-nmlz ‘life’
(Fox 1978: 416)
(11) tahi-ha live/flourish-nmlz gerund (no translation given)
(Fox 1978: 416)
Fox gives one form that may contain a liquid-initial nominalizing suffix: maeraa ‘danger, sickness’ (p. 263); cf. mae ‘be numb, unable to move, ill’, ‘be unconscious, faint’, ‘die’. The final raa would be in regular phonetic correspondence with the nominalizing suffix -laa found in some of the other LMM languages; see (4) from Toqabaqita in Section 2.1 and examples from other languages further below. As shown in (8) in Section 2.2.1, in Longgu, according to Ivens (1934), the (reduplicated) verb ‘die’ can be made into a nominalization whose meaning is ‘danger’. And although Fox does not mention a nominalizing suffix -a, he lists maea (pp. 262–263), whose senses include, among others, ‘a small particle of food, smaller than an areca nut, taken and breathed on, with the words of a spell repeated and then added secretly to the food in a bowl, the man who eats shortly falls sick’, ‘the swelling caused by maea’, ‘a spell or charm, usually with words and lime’, and ‘a snake whose bite causes illness or death, a charmed snake, a poisonous snake (in Arosi belief the same thing)’. Capell (1971: 38) mentions -ha as a “noun-forming suffix”, which, according to him, is productive. Capell does not mention a -nga suffix, but he does say this (p. 38): “Arosi has fewer derived nouns than the Malaita languages, in that suffixation is less common in Arosi, and the -ana, -anga suffixes found in Malaita as formants of verbal nouns are absent from Arosi.” It is not clear whether -anga in the quote should be -nga. In the languages of Malaita one typically finds -nga rather than -anga (but see -‘anga in [28b] in Section 2.2.8 from Kwara‘ae.)
 Frantisek Lichtenberk
2.2.3â•… Ulawa For Ulawa, Ivens (1929a) lists -nga, and -lä and -laa (and also -hä “termination added to verbs to make nouns”, -ha “gerundive, added to transitive verbs” [p. 259] [but see mae-ha ‘sickness’ further below], and -haa “noun ending added to verbs” [p. 60] [see (68) in Section 3.3]), but it is not clear what the differences are between their functions, if any. Ivens characterizes -nga as a “noun ending; added to verbs, adjectives, to other nouns and to composite verbs” (p. 227). For example: (12) mae-nga die/be.ill-nmlz ‘sickness’, ‘death’
(Ivens 1929a: 180)
There are several -lä entries in the dictionary, two of which relate to nominalizations. Ivens characterizes one of them as a “noun ending, added to verbs” (p. 154), and as an example he gives the form mae-la-na in (13), which like mae-nga in (12) above, is based on mae ‘die’, ‘be ill’: (13) mae-la-na die-nmlz-3sg.poss ‘his death’
(Ivens 1929a: 180)
(Besides mae-nga ‘sickness’, ‘death’ and mae-la- ‘death’, Ivens also lists [p. 180] mae-ha ‘sickness’.) The other -lä is characterized by Ivens as a “gerundive ending, to which the suffixed [possessive] pronouns are always added” (p. 154), which suggests that its function is to form action nominalizations. One example that Ivens gives is the form in (14): (14) säu-ni-lä(-ku) kill/beat-tr-nmlz-1sg.poss gerund (no translation given)
(Ivens 1929a: 306)
About the suffix -laa Ivens says that it is a “noun ending added to verbs” (p. 154). As an example he gives the form in (15): (15) hou-laa proclaim/be.acclaimed-nmlz ‘a feast’
(Ivens 1929a: 112)
Besides hou-laa Ivens also gives houlana, which is most likely hou-la-na: (16) hou-la-na proclaim/be.acclaimed-nmlz?-3sg.poss? ‘[his?] fame, rule’
Ivens uses, although not consistently, the orthographic symbol ä for e in words whose cognates in other languages have /a/ in that position, when the next syllable contains a
Nominalizations in Toqabaqita and closely related languages 
high vowel. -la, -lä also Â�-le, and -laa evidently represent one and the same morpheme, in spite of the variety of orthographic forms. Furthermore, it is likely that the form -laa is used if no other suffix follows, while -la and -lä/-le are used if a possessive suffix does follow. As will be seen further below, parallel alternations between forms with a double/long vowel and a single/short vowel are found in other LMM languages.
2.2.4â•… Oroha Ivens (1927) mentions several suffixes used to form verbal nouns, -na, and -ra and -raa among them. -na is the expected cognate of -nga and -ngaa in other LMM languages, and -ra and -raa the expected cognates of -la and -laa. (17) mae-na die/be.ill-nmlz ‘death’
(Ivens 1927: 591)
(18) mae-raa die/be.ill-nmlz ‘sickness’
(Ivens 1927: 607)
(19) hana-raa eat-nmlz ‘food, especially vegetable food’
(Ivens 1927: 591)
As an example of the suffix -ra Ivens gives the expression in (20): (20) pa’u-ra i suna (fire)smoke-nmlz of fire ‘fire-smoke’
(Ivens 1927: 591)
Ivens considers i to be a genitive preposition. However, elsewhere in the sketch of the grammar of Oroha he gives pa’uraisuna ‘smoke of fire’, which can serve as a base to derive the adjective pa’uraisuna-’a ‘smoky’, -’a being one of two “adjectival terminations” (p. 600). It is conceivable that pa’uraisuna ‘smoke of fire’ is a compound/lexicalized phrase, where the nominalizer is not word final, and so it is the short form -ra, rather than the long form -raa, that is used, as is the case with nominalizers in other LMM languages.
2.2.5â•… Marau For this language, Ivens (1929b) gives -na and -raa as two out of five nominalizing suffixes. In Marau too -na is the expected cognate of -nga/-ngaa in other LMM languages, and -raa the expected cognate of -laa/-la. (21) hahi-na cook.in.oven-nmlz ‘cooking’
(Ivens 1929b: 348)
 Frantisek Lichtenberk
(22) hou-raa be.famous-nmlz ‘public feast’
(Ivens 1929b: 348)
2.2.6â•… ‘Are‘are In his dictionary of the language, Geerts (1970) mentions two nominalizing suffixes: -na “noun ending”, “verbal noun ending; suffixed to verbs and nouns” (p. 69) and -ha “a noun ending; verbal noun ending; adjective ending to form nouns” (p. 18). The form -na is the expected cognate of -nga/-ngaa in other LMM languages. As an example of a nominalization with -na Geerts gives the form in (23): (23) haihoroi-na fight.hand.to.hand-nmlz ‘hand to hand fighting’
(Geerts 1970: 25)
(Hai in haihoroi is most likely the reciprocal prefix hai-, used in conjunction with the suffix -i, as is the case in other LMM languages; cf. horo ‘beat, strike, kill, win in games’.) There is no mention of a nominalizer -ra or -raa in the dictionary, which would be the expected cognates of -la and -laa in other LMM languages.
2.2.7â•… Kwaio For Kwaio, Keesing (1985) gives -nga/-ngaa as the most common nominalizer, and -laa as a less common nominalizer. Besides -laa he also discusses -la. (Keesing also gives -a as a nominalizer, less common than the other two.) The forms -nga/-ngaa serve to derive “verbal nouns or abstract nouns” (p. 77). The distribution of the two forms is conditioned phonologically: -nga is used with disyllabic bases, and -ngaa with trisyllabic bases; see (24a, b) and (24c, d), respectively (24) a. eno-nga b. leka-nga sleep-nmlz go-nmlz ‘sleeping’ ‘trip’ (Keesing 1985: 77) c.
to‘oru-ngaa d. live-nmlz ‘living’ (Keesing 1985: 77)
nanama-ngaa (ancestor)convey.efficacy-nmlz ‘ancestrally conferred efficacy’ (Keesing 1985: 78)
The form -la functions to form “gerunds” (p. 80), and from the examples given it appears that -la and -laa too are in complementary distribution: -laa is used when the nominalizer is word-final, while -la is used when another suffix follows, as in (25a, b) and (25c, d), respectively:
Nominalizations in Toqabaqita and closely related languages 
(25) a. kwasi-laa b. ogu-laa decorate-nmlz gather.together-nmlz ‘ornaments’ ‘gathering’ (Keesing 1985: 79) c. kwa‘i-la-na wela d. kwa‘i-la-i hit-nmlz-3sg.poss child hit-nmlz-impers ‘hitting of the child’ ‘hitting, killing’ (Keesing 1985: 80)
Besides kwasi-laa ‘ornaments’ (25a) Keesing also gives kwasi-nga ‘ornaments’ (p. 79). The “impersonal pronoun ending” -i (25d) is used in the absence of an expression of a possessor which corresponds to the direct object of the source verb.
2.2.8â•… Kwara‘ae For Kwara‘ae Deck (1934) lists several nominalizing suffixes, including -nga, -la and -‘a. There is evidence of a distinction between action nominalizations and deverbal nouns. The suffix -la occurs before possessive suffixes and appears to form action nominalizations: (26) mae-la-ku die-nmlz-1sg.poss ‘my act of dying’
(Deck 1934: 12)
On the other hand, -‘a appears to form deverbal nouns. If there is an expression of a possessor, it is by means of an independent phrase rather than by means of a possessive suffix. In (27) the possessor is expressed by means of an independent personal pronoun: (27) mae-‘a nau die-nmlz 1sg ‘my death (in the abstract)’
(Deck 1934: 12)
It is not clear what the function of the nominalizer -nga is. Besides -nga, there is also -‘anga, which is used with stems that end in a or a‘i. (The acute accent mark in [28a] signals “a shifting of the accent” [p. 13] when -nga is added to a verb.) (28) a. tata‘é-nga b. suasuala-‘anga rise.up-nmlz be.touchy-nmlz ‘arising up’ ‘touchiness’ (Deck 1934: 14)
2.2.9â•… Lau For Lau, Fox (1974, but compiled in the early part of the 20th century) lists -nga and -ngā “suffix[es] making verbal nouns” (p. 145), and -la and -lā also “suffix[es] making verbal
 Frantisek Lichtenberk
nouns”, and also “suffix making gerundives” for -la (p. 110). It is not clear how the forms -nga and -ngā are used. Fox gives examples of both with the same verb and gives somewhat different glosses to the two nominalizations: (29) fale-nga give-nmlz ‘food given by bridegroom’s relations to the bride’s relations’
(Fox 1974: 55)
(30) fale-ngā give-nmlz ‘food at a wedding’
(Fox 1974: 55)
As far as the forms -la and -lā are concerned, from the examples given in the dictionary it appears that -lā is used word-finally, while -la is used when another suffix follows: (31) a.
bae-lā b. speak/talk/say/tell-nmlz ‘speech, saying, word’ (Fox 1974: 25)
ade-la-na do-nmlz-3sg.poss ‘his doing of it’ (Fox 1974: 110)
Besides fale-nga ‘food given by bridegroom’s relations to the bride’s relations’ (29) and fale-ngā ‘food at a wedding’ (30), Fox also gives the nominalization fale-lā ‘a gift’ (p. 55). Fox (1974: 1) also lists -a “suff. forming vbl. ns”: (32)
mae-a die/be.very.ill-nmlz ‘death’, ‘sickness’ (Fox 1974: 125)
Besides mae-a ‘death’, ‘sickness’, Fox also gives mae-la-na ‘danger’, ‘dying, death’ (-na 3sg possessive suffix), mae-la ‘death feast’ and mae-lā ‘accident’ (p. 125). Anna Margetts (pers. comm., September 2006) mentions -la and -laa (and also -a) as nominalizers in Lau, but she has no evidence of -nga or -ngaa/-ngā. The South Lau (Walade) dialect has the nominalizing forms -laa and -la, which function to form action nominalizations. (The South Lau data come from Athanasius Faifu, pers. comm., 1989.) The form -laa is used word-finally, while -la is used if another suffix follows: (33) a.
firu-laa b. firu-la-mu fight-nmlz fight-nmlz-2sg.poss ‘fight, fighting’ ‘your fighting’
South Lau also has the suffix -a, which serves to form deverbal nouns: (34) fanga-a eat-dvn ‘feast’
Nominalizations in Toqabaqita and closely related languages 
There is no evidence of nominalizers -ngaa or -nga in South Lau.
2.2.10â•… Toqabaqita Toqabaqita has the forms -laa, -la and -le, which function to form action nominalizations, and -a, which functions to form deverbal nouns. The liquid-initial suffixes are discussed in detail in Section 3.2. For examples of deverbal nouns formed by means of the suffix -a see (5a–k) in Section 2.1. There is no suffix of the form -ngaa or -nga with a nominalizing or any other function in Toqabaqita. 2.2.11â•… Interim summary The main points of the discussion above can be summarized in the following way: In the LMM group of languages one finds both nasal-initial and liquid-initial nominalizers. In some of the languages one or both of the nominalizers have two forms, one with a short/single a and one with a long/double ā/aa.1 In at least some of those languages the forms with the short/single vowels are used before possessive suffixes, while the forms with the long/double vowels are used word-finally. There is some evidence that these nominalizers are used to form action nominalizations. Proto-LMM had the nominalizing suffix *-nga, inherited from the ancestor language, probably with *-ngā/-ngaa as a variant. Later, still in Proto-LMM, a new nominalizer appeared,*-la, probably with *-lā/-laa as a variant. If both *-lā/-laa and *-la existed, then it is likely that the former was used in word-final position, and the latter when a possessive suffix followed. From the data available it is not possible to tell whether the nasal-initial and the liquid-initial nominalizers were in free variation with each other, whether there were phonological conditions on their use, whether their use was conditioned lexically, or whether they had different functions. In some of the daughter languages, the liquid-initial nominalizer remained only marginal (Arosi) or disappeared later on (‘Are‘are, Longgu). In others, both the nasal-initial forms and the liquid-initial forms are, or appear to be, relatively common (Ulawa, Oroha, Marau, Kwaio, Kwara‘ae). And in still others, the liquid-initial forms have completely replaced the nasal-initial forms (Lau, South Lau, Toqabaqita). In the remainder of this paper, the focus will be on Toqabaqita, where only the liquid-initial nominalizer forms exist, besides the suffix -a, which serves to form deverbal nouns. The latter suffix will not be considered further. Where relevant, data from other Oceanic languages will be briefly considered as well.
.╅ It is not clear from the descriptions of some of the languages whether double orthographic symbols for vowels, such as aa, represent two vowels or single long vowels. For Toqabaqita, double vowel symbols represent two vowels.
 Frantisek Lichtenberk
3.â•… Nominalizations in Toqabaqita 3.1â•… Grammatical background Before discussing the properties of Toqabaqita nominalizations, it is necessary to outline the relevant aspects of the grammar of the language. (More information can be found in Lichtenberk 2008.) Toqabaqita is a nominative-accusative language, almost exclusively head-marking. The basic constituent orders are SV and AVO. In verb phrases verbs are accompanied by various particles, some of which are preverbal and others postverbal. Among the preverbal particles are subject markers, which simultaneously mark the person, number and clusivity (inclusive, exclusive) of the subject, and tense (future, nonfuture), aspect (imperfective, signalled by the same set of subject markers that express the future tense), negation, sequentiality or dehortation. (The dehortative set is archaic.) There is only one set of the negative subject markers; the tense/aspect/sequentiality contrasts are suspended in negation. The sentence in (35) contains the first person dual exclusive nonfuture subject marker mere, and the one in (36) the first person singular negative subject marker kwasi: (35) Kamareqa mere fanga sui naqa. 1du(excl) 1du(excl).nfut eat compl prf ‘We have finished eating.’ (36) Nau kwasi ngata bii ta wane. 1sg 1sg.neg speak com some person ‘I did not speak with anybody.’
Example (37) contains the ventive directional mai, one of the postverbal particles. (Second person singular imperatives usually do not have a subject marker.) (37) Lae mai! go vent ‘Come here!’
There are two classes of transitive verbs in the language, referred to here as Class 1 and Class 2, respectively. Both types of verbs take suffixes to index their direct objects, although Class 1 transitive verbs have object-marking suffixes only for the third person (singular, dual and plural). With Class 1 verbs, objects other than third person can be expressed only by means of the independent personal pronouns. In the third person, either an object suffix or an independent pronoun, but not both simultaneously, can be used. (38) a.
rongo-a hear-3sg.obj ‘hear him/her/it’
Nominalizations in Toqabaqita and closely related languages 
b. rongo-a teqe wela hear-3.obj2 one child ‘hear a child’ c.
rongo qoe hear 2sg ‘hear you’
Class 2 transitive verbs have object suffixes for all grammatical persons and numbers. Objects cannot be expressed by means of independent personal pronouns. (For the gloss 3.obj in [39b] see Footnote 2.) (39) a.
thaitoqoma-na know-3sg.obj ‘know him/her/it’
b. thaitoqoma-na wane baa know-3.obj man that ‘know that man’ c.
thaitoqoma-mu know-2sg.obj ‘know you’
The same set of object suffixes used with Class 2 transitive verbs also serves to mark the possessor in one type of possessive construction; see further below. With Class 1 transitive verbs the (third person) object suffixes are not used under certain conditions: with incorporated objects and when a verb occurs as the first/nonfinal component of a serial verb construction. The latter type of case is illustrated in (40). The verb rongo ‘hear’ is followed by the verb leqa ‘be good’. The stimulus (that which was heard) is expressed not as a direct object but as an oblique object, the complement of the “general” preposition qani. (Qani is one of several prepositions in the language that derive historically from transitive verbs and, with one type of exception, index their objects in the same way that Class 1 transitive verbs index their direct objects. The preposition will be relevant in Section 3.2.) (40) rongo leqa qani-a hear be.good genp-3sg.obj ‘hear him/her/it well’
.╅ In the presence of a lexical direct object inside the verb phrase, the third person object suffix -a is used regardless of the grammatical number of the object (singular, dual or plural); hence the absence of specification of number in such cases. The same is true of the third person object suffix -na used with Class 2 transitive verbs; see examples (39a, b). In the absence of a lexical direct object, the two suffixes have, or may have, a specifically singular value. Cf. Footnote 3.
 Frantisek Lichtenberk
Cf. (38a), where rongo ‘hear’ carries the object suffix -a. On the other hand, Class 2 transitive verbs always retain the direct-object suffix. This is illustrated in (41), where thaitoqoma ‘know’ is followed by the verb leqa ‘be good’. The noun phrase wane baa ‘that man’ is the direct object of the verb-serializing compound: (41) thaitoqoma-na leqa wane baa know-3.obj be.good man that ‘know that man well’
Cf. (39b). There are two types of attributive possessive constructions in the language, suffixing and bare. In the suffixing type, the possessum noun carries a personal suffix that indexes the possessor. In addition, with a third person possessor, there may be a lexical noun phrase expressing the possessor. The personal suffixes are the same as those indexing the direct objects of Class 2 transitive verbs. Compare (42a, b, c) below and (39a, b, c) above: (42) a.
thaina-na mother-3sg.pers ‘his/her mother’
b. thaina-na wela mother-3.pers3 child ‘the child’s mother’ c.
thaina-mu mother-2sg.pers ‘your mother’
The suffixing construction expresses inalienable possession. However, as (43a) below shows, not all cases of inalienable possession select this construction. The personal suffixes correspond to, and many of them are cognate with, the possessive suffixes in the other LMM language (see Section 2.2). However, the Toqabaqita suffixes have other functions besides indexing possessors. They are used to index the direct objects of Class 2 transitive verbs, and they are also used with some postverbal particles in verb phrases, where they index the subject. For these reasons the label “personal suffixes” is preferred over “possessive suffixes”.
.╅ In the presence of a possessor phrase inside the possessive noun phrase, the personal suffix -na is used regardless of the grammatical number of the possessor (singular, dual or plural); hence the absence of specification of number. In the absence of a possessor phrase, the suffix has, or may have, a singular value. Cf. Footnote 2.
Nominalizations in Toqabaqita and closely related languages 
In the bare possessive construction, the possessum does not carry a personal suffix to index the possessor: (43) a.
maka nia father 3sg ‘his/her father’
b. maka wela naqi father child this ‘this child’s father’ c.
maka qoe father 2sg ‘your father’
The bare possessive construction is used in cases of alienable possession, but with some lexical items and also under certain other conditions when the possessum– possessor relation is of the inalienable type (Lichtenberk 2005). The language also has associative noun phrases, where the head noun is modified by a non-referential noun phrase. The modifier phrase expresses notions such as the type of entity designated by the head noun, as in (44), or the type of contents of a container designated by the head noun. The head noun carries the associative suffix -qi/-qe. The form -qi is used if the final vowel of the noun is high, as in (44), and -qe is used if the final vowel of the noun is non-high (see [46] and the immediately following text in Section 3.2). (44) thaalu-qi thaqaro egg-assoc bird ‘bird egg’
With this background we can now move to considering nominalizations in the language.
3.2â•… The basic structure of nominalizations In Toqabaqita, noun phrases headed by action nominalizations can be used in the following syntactic functions: subject (example [48]), direct object (51), object of a preposition (45), head of the predicate of a verbless sentence (45), possessor in a possessive noun phrase (55), and with the ligature ni (47). Nominalizations, possibly with complements, adjuncts and/or particles, can also constitute clauses (Section 3.4), and nominalizations can function as heads/possessums in possessive noun phrases (55). In (45) the nominalization heads the predicate of a verbless clause. There is another nominalization in the sentence, raqa-futa-laa ‘possum hunting’, which functions as a purpose modifier of the head nominalization.
 Frantisek Lichtenberk
(45) Raa keeroqa ni bo=naqa4 lae-laa work 3du lig intens=intens go-nmlz uri-a raqa-futa-laa. purp-3.obj climb-possum-nmlz ‘Their work was to go possum-hunting.’
(Raqa-futa-laa is a nominalization of the transitive verb raqa ‘climb’ with an incorporated object. The resulting verb is intransitive.) Action nominalizations, henceforth simply “nominalizations”, are formed in Toqabaqita by means of the variants -laa, -la or -le of the nominalizer. The -le variant is used before the associative suffix -qe (cf. [44] in Section 3.1 and the paragraph that precedes it). The -le form is, in fact, a modification of the -la variant: the vowel of the nominalizer assimilates to the vowel of the associative suffix. (46) manta-le-qe doo think-nmlz-assoc thing ‘thinking’, ‘a/the thought’ (lit.: ‘thing thinking’)
(Replacement of a by e before the associative suffix is also found in nouns that are not nominalizations; for example baeke-qe kafara ‘bagful of copra’; baeka ‘bag’, kafara ‘copra’.) The -le variant of the nominalizer is not relevant to the discussion that follows and will not be discussed any further. The variant -laa is used in word-final position: (47) kafo ni siisiu-laa fresh.water lig bathe-nmlz ‘water for bathing’
The variant -la is used when a nominalization functions as the head (possessum) in a suffixing possessive construction, where the possessum noun carries a personal suffix indexing the possessor (Section 3.1); see, for example, (48) below. However, as illustrated further below, nominalizations can also head the other, bare type of possessive construction, in which case the nominalizer has the form -laa. With respect to the expression of the core arguments of the base verbs, the Toqabaqita system can be characterized, in Koptjevskaja-Tamm’s (1993) typology, as involving a mixture of the Ergative-Possessive and the Possessive-Accusative types. In the former type, the S and the O (P) arguments are expressed as possessors, while
.╅ The equal sign signifies phonological fusion of two or even three words, in which case one or both/all of the words appear(s) in a reduced form. Bo= is a reduced form of the intensifier boqo.
Nominalizations in Toqabaqita and closely related languages 
in the latter type S and A are. That is, with intransitive verbs the S argument is always expressed as possessor, while with transitive verbs it may be the O argument (the more common pattern) or the A argument. Example (48) contains a nominalization of an intransitive verb. The S argument is expressed as possessor in a suffixing possessive construction: (48) Ngata-la-da qe leqa. speak-nmlz-3pl.pers 3sg.nfut be.good ‘Their speaking was good.’ (For example, the way they spoke was good, or what they said was good.)
With nominalizations based on transitive verbs there are two basic patterns. In one, the nominalization occurs in the suffixing possessive construction, and the possessor is the O argument. There is no expression of the A argument. This is the more common pattern. (49) = (2) uusi-la-na suka buy-nmlz-3sg.pers sugar ‘(the act of) buying sugar’, ‘buying of sugar’
In the other pattern, both the A and the O arguments are expressed. The nominalization occurs in the bare possessive construction, without a personal suffix indexing the possessor, and so it is the -laa variant of the nominalizer that is used here. Here it is the A argument that is expressed as possessor, while the O argument is expressed in a prepositional-phrase complement of the nominalization. The complement is headed by the general preposition qana/qani. The form qana is used with lexical objects; the qani form is used elsewhere. (For an example of qani with an object suffix see [40] in Section 3.1.) This preposition is used with a variety of functions. It can mark locations, goals, purpose, among others, and it introduces the complements of many intransitive verbs. In (50) the prepositional phrase headed by qana corresponds to the O argument of the source verb; the A argument is expressed as the possessor of the nominalization: (50) fasi-laa qoe qana baqu qena ki plant-nmlz 2sg genp banana that pl ‘your planting (of) those bananas’
One of the referees has pointed out that this pattern, where the A is encoded as possessor and the O as prepositional-phrase complement, is not common cross-linguistically: in her sample of nominalization in 168 languages Koptjevskaja-Tamm (2005: 255) mentions it as a minor pattern with “a very restricted cross-linguistic distribution”. As discussed in Section 3.1, there are two basic types of transitive verbs in Toqabaqita, Class 1 and Class 2. The two examples of nominalizations of transitive verbs given in (49) and (50) above both involve Class 1 transitive verbs. The two classes of verbs behave differently when nominalized. Nominalizations of Class 1
 Frantisek Lichtenberk
verbs can occur in the possessum position of a suffixing possessive construction, as in (49). They carry a personal suffix that indexes the possessor, the O argument. In addition, that participant can be expressed in the possessor position by means of a lexical noun phrase. In (51) the woman is indexed on the nominalization ngali-la- ‘taking’ by means of the third person singular personal suffix -na and the lexical possessor phrase kini qeri ‘the/that woman’: (51) Ma toqa qe=ki keka mangosi-a ngali-la-na and people that=pl 3pl.seq not.be.able-3.obj take-nmlz-3.pers kini qeri …. woman that ‘But (lit.: and) the people could not take the woman (with them in their canoes, because…).’
Nominalizations of Class 2 transitive verbs, unlike those of Class 1, contain, in addition to the nominalizing suffix, the relevant object suffix that would be present on the base, non-nominalized verb. As pointed out in Section 3.1, Class 2 verbs always retain the object suffix. The object suffix precedes the nominalizer. Example (52) contains the non-nominalized Class 2 transitive verb soe-toqo ‘ask sb. (about sb., st.)’ with the second person singular object suffix -mu. The direct object encodes the addressee. (52) …kwa soe-toqo-mu, “Taa na qo manata-toqo-na?”.5 1sg.seq ask-test-2sg.obj what? foc 2sg.nfut think-test-3sg.obj ‘… (and) I asked you, “What are you thinking about?”.’
In (53) the verb soe-toqo ‘ask sb. (about sb., st.)’ has been nominalized but still carries the second person singular object suffix: (53) Wane naqi seqe-n=e noni uri-a man this body-3sg.pers=3sg.nfut be.shy purp-3.obj soe-toqo-mu-laa uri-a wela qoe ki. ask-test-2sg.obj-nmlz purp-3.obj child 2sg pl ‘This man is (too) shy to ask you about your children.’ (Lit.: ‘This man, his body is shy for asking you for your children.’)
Nominalizations of Class 2 transitive verbs can also carry a personal suffix, in addition to an object suffix and the nominalizer. The personal suffix can only be third person singular. It follows the nominalizer, and consequently the nominalizer has the form -la.
.â•… The test suffix -toqo derives transitive verbs. In most, but not all, cases the derived verb has the meaning ‘test X by V-ing it’, where X is the referent of the direct object, for example kuqu-toqo-na sofusofu ‘taste soup by drinking it’ (kuqu ‘drink’, -na third person object suffix, sofusofu ‘soup’).
Nominalizations in Toqabaqita and closely related languages 
In (54) the nominalization of soe-toqo ‘ask sb. (about sb., st.)’ carries the third person plural object suffix, the nominalizer, and the third person singular personal suffix -na: (54) B: Qoo, nau lio-ku qe dora oh 1sg mind-1sg.pers 3sg.nfut forget
qana soe-toqo-da-la-na. genp ask-test-3pl.obj-nmlz-3sg.pers
[A: ‘When will those people arrive?’] B: ‘Oh, I forgot to ask them.’ (Lit.: ‘Oh, I, my mind forgot its asking them.’)
Example (55) below too contains a combination of an object suffix, here third person singular (but see Footnote 2 in connection with example [39b] in Section 3.1 for the absence of marking of number), the nominalizer, and the third person singular personal suffix (see Footnote 3). (As mentioned in Section 3.1, with Class 2 transitive verbs the object suffixes have the same forms as the corresponding personal suffixes.) In addition to the nominalization of the Class 2 transitive verb thaitoqoma ‘know’, the sentence also contains nominalizations of the intransitive verbs keekeda ‘write’ and teqemaqi ‘read’ as possessors of the nominalization of ‘know’: (55) Thaitoqoma-na-la-na keekeda-laa ma teqemaqi-laa know-3.obj-nmlz-3sg.pers write-nmlz and read-nmlz si doo taingaqi neri. prtt thing be.important vivid ‘Knowing how to write and read is important.’ (Lit.: ‘Its knowing writing and reading (is) an important thing.’)
In nominalizations of Class 2 transitive verbs of this type, the third person singular personal suffix is semantically empty: it does not index a participant. Even though the literal translation of (54) further above is ‘Oh, I, my mind forgot its asking them’, the personal suffix does not index lio-ku ‘my mind’. And in (55) too, the personal suffix on the nominalization has no referent: literally ‘Its knowing writing and reading (is) an important thing.’ This type of construction is used only if the object suffix on the nominalization is third person, not when the object is other than third person; cf. (53) with the second person singular object suffix. Toqabaqita nominalizations exhibit properties both of nouns and of verbs. As nouns, they head noun phrases, as amply illustrated above. They can take possessor modifiers (see [55] above) and other modifiers that occur with nouns, such as demonstratives ([77] in Section 3.5), the interrogative quantifier fita ‘how many?, how much?’, classifiers (see [88] in Section 3.5 for both) and relative clauses (various examples in Section 3.5). Like verbs they can be accompanied by some of the particles that
 Frantisek Lichtenberk
occur with verbs in verb phrases, in the same function. In (56) the nominalization is accompanied by the ventive directional mai: (56) Nia qe tofe fasi-a lae-laa mai. 3sg 3sg.nfut cancel abl-3.obj go-nmlz vent ‘He changed his mind about coming (here).’ (Lit.: ‘He cancelled his coming here.’)
Cf. example (37) in Section 3.1, where the ventive directional occurs with the verb lae ‘go’ in a verb phrase. However, while nominalizations can occur with certain particles that occur with verbs, they do not express the future-nonfuture tense distinction and cannot be negated. As discussed in Section 3.1, the tense distinction and negation are expressed by means of preverbal subject markers, but these do not occur with nominalizations. There are three aspects of Toqabaqita nominalizations that the following discussion concentrates on: (i) double nominalizations; (ii) a stance function of nominalizations; and (iii) cognate nominalizations, in particular their use in cognate relativization. It is important to keep in mind that the latter two functions are functions of nominalizations, not of the nominalizing morpheme. The nominalizer has only one function, that of deriving nominalizations.
3.3â•… Double nominalizations By “double nominalizations” are meant here nominalizations that contain two instances of the nominalizer. However, the second instance of the nominalizer does not create a new, functionally different nominalization. It is added to a form that is itself a nominalization, and the result too is a nominalization. The use of double nominalizations is discussed further below. Double nominalizations can be formed in Toqabaqita only from transitive verbs and only if the nominalization carries a personal suffix indexing the the O argument. One instance of the nominalizer occurs before the personal suffix and so has the form -la, and the other occurs after the personal suffix, in word-final position, and so has the form -laa: (57) thau-ngi-la-da-laa kill-tr-nmlz-3pl.pers-nmlz ‘killing (of) them’
Furthermore, double nominalizations are possible only in the absence of a possessor noun phrase. If there is a possessor noun phrase present, only simple nominalizations are grammatical: (58) a.
gani-la-na-laa ask.sb.(for.st./to do st.)-nmlz-3sg.pers-nmlz ‘asking him/her (for st., to do st.)’
Nominalizations in Toqabaqita and closely related languages 
b. *gani-la-na-laa wane ask.sb.(for.st./to do st.)-nmlz-3.pers-nmlz man (‘asking the man [for st., to do st.]’) c.
gani-la-na wane ask.sb.(for.st./to do st.)-nmlz-3.pers man ‘asking the man (for st., to do st.)’
Double nominalizations normally occur only in oblique-object positions. However, they are never obligatory. Simple nominalizations are always possible instead. There is no difference between simple and double nominalizations in terms of objective content, but in oblique-object positions double nominalizations are often preferred and are judged “stronger”, “more forceful”, “better” than simple nominalizations, but this is so only with double nominalizations based on Class 1 transitive verbs. For double nominalizations based on Class 2 transitive verbs see further below. The next example shows free variation between a simple and a double nominalization of a Class 1 transitive verb in person B’s response: (59) A: Qo gani-a naqa wane baa ura seleni? 2sg.nfut ask.sb.(for.st.)-3.obj prf man that all money B: Qe aqi boqo. Seqe-ku qe noni 3sg.nfut not.be.so asrt body-1sg.pers 3sg.nfut be.ashamed
qana {gani-la-na / gani-la-na-laa}. genp ask.sb.(for.st.)-nmlz-3sg.pers / ask.sb.(for.st.)-nmlz-3sg.pers-nmlz
A: ‘Have you asked the man for money?’ B: ‘No, I haven’t. I am ashamed to ask him.’ (Lit.: ‘…. My body is ashamed about asking him.’)
As discussed in Section 3.1, nominalizations of Class 1 transitive verbs do not carry object suffixes, while those of Class 2 verbs do; see (49) and (51), and (53)–(55), respectively. The examples of double nominalizations given above all involve Class 1 transitive verbs. It is possible to form double nominalizations of Class 2 transitive verbs, where the verb base carries four suffixes: verb.base-object.suffix-nominalizer3sg.personal.suffix-nominalizer. The third person singular personal suffix between the two instances of the nominalizer is semantically empty (Section 3.2). (60) B: Qoo, nau lio-ku qe dora oh 1sg mind-1sg.pers 3sg.nfut forget
qana soe-toqo-da-la-na-laa. genp ask-test-3pl.obj-nmlz-3sg.pers-nmlz
[A: ‘When will those people arrive?’] B: ‘Oh, I forgot to ask them.’ (Lit.: ‘Oh, I, my mind forgot its asking them.’)
 Frantisek Lichtenberk
However, double nominalizations of Class 2 transitive verbs are not normally used, possibly because of their complexity, simple nominalizations being preferred. Thus, the simple nominalization in (54) in Section 3.2 is preferred over the synonymous double nominalization in (60). Double nominalizations are grammatical in positions other than oblique-object, but there it is simple nominalizations that are normally used. In (61) the nominalizations occur in the direct-object position, and the construction with the simple nominalization is strongly preferred: (61) Ili-toqo-na {qani-la-na / qani-la-na-laa}. do-test-3.obj eat-nmlz-3sg.pers / eat-nmlz-3sg.pers-nmlz ‘Try eating it.’
Two more examples of double nominalizations of Class 1 transitive verbs in obliqueobject positions are given below: (62) Keka kwaqe nau qana faqa-qolo-si-la-ku-laa, …. 3pl.seq beat 1sg genp caus-be.straight-tr-nmlz-1sg.pers-nmlz ‘They [the speaker’s parents] would give me whippings to straighten me, ….’ (That is, when the speaker was little, his parents would give him whippings to make him behave.) (63) Qe aqi qosi maqu qana kwaqi-la-mu-laa, …. 3sg.nfut negv 2sg.neg be.afraid genp cut-nmlz-2sg.pers-nmlz ‘Don’t be afraid of being operated on, (they will put you to sleep [with an anaesthetic] first and only then will they operate on you).’ (Lit.: ‘Don’t be afraid of your cutting, ….’)
Toqabaqita is not the only LMM language in which double nominalizations are possible, although as a rule little is known about their use in those languages. One such language is South Lau, a very close relative of Toqabaqita. Unlike in Toqabaqita, in South Lau, double nominalizations are possible even in the presence of a possessor noun phrase (cf. the ungrammatical example [58b] above from Toqabaqita). Of the two examples of double nominalizations in South Lau available, one occurs in an oblique-object position (64) and one in the direct-object position (65). (On the other hand, in Toqabaqita, double nominalizations do not normally occur in the directobject position.) (64) lea mee ôadomi nau fua-na ŋali-la-na-laa ia go vent help 1sg purp-3sg.poss take-nmlz-3sg.poss-nmlz fish nee fua-na lao-na ola this all-3sg.poss inside-3sg.poss canoe ‘Come (and) help me to take (all of) these fish into the canoe.’
Nominalizations in Toqabaqita and closely related languages 
(65) nau asifaini-a ôigini-la-na-laa fou nee ma 1sg try-3sg.obj lift-nmlz-3sg.poss-nmlz stone this and ka ôafetai seq be.difficult ‘I tried to lift this stone, but (lit.: and) it was (too) difficult.’
Double nominalizations are also possible in Ulawa. Ulawa has the nominalizers -nga, -lä/-le and -laa, and -hä, -ha and -haa (Section 2.2). In the entries for -hä and -nga Ivens (1929a: 59 and 227, respectively) says that “[i]n some words [the suffixes are] added to a noun which already has a noun ending”. Unlike in Toqabaqita, in Ulawa double nominalizations the two suffixes can occur next to each other. It may be the same suffix occurring twice, or two different suffixes may be combined. In (66) the suffix -nga occurs twice: (66) mala oho-nga-nga ? attempt/strive.with-nmlz-nmlz ‘temptation, trial’
(Ivens 1929a: 236)
There are several mala entries in the dictionary, but it is not clear which one, if any, is present in (66). Besides the double nominalization oho-nga-nga Ivens also gives the simple nominalization oho-nga: (67) mala oho-nga ? attempt/strive.with-nmlz ‘tempt’
(Ivens 1929a: 236)
On the other hand, in (68) -nga is immediately followed by the nominalizer -haa: (68) älida-nga-haa travel.by.sea-nmlz-nmlz ‘making journeys overseas’
(Ivens 1929a: 19)
Besides the double nominalization, Ivens also gives a simple nominalization with only the suffix -nga: (69) älida-nga travel.by.sea-nmlz a. ‘sea journey’ b. ‘canoe load of voyagers’ (Ivens 1929a: 19)
It is not clear what differences there are between simple and double nominalizations in Ulawa, but the glosses for (68) and (69) suggest a difference between an action nominalization in the case of the double nominalization and a deverbal noun in the case of the simple nominalization.
 Frantisek Lichtenberk
Double nominalizations are also possible in Tuvaluan, a Polynesian language, only fairly distantly related to the LMM languages. Besnier (2000: 512–513) characterizes double nominalizations in Tuvaluan thus: Occasionally, two nominalizing suffixes are concatenated. Any of the regular allomorphic forms may appear as the first suffix in such sequences, but the second suffix must be the form -ga, i.e. without gemination of the last segment of the element to which it is attached[.] … Double suffixed constructions are slightly more emphatic than regular suffixed nominalized constructions, and are characteristic of informal conversation and other casual interactional contexts; otherwise, they do not differ from their regularly suffixed counterparts.
The nominalizing suffix -ga is phonologically /ŋa/ and is cognate with the nasal-initial nominalizers found in some of the LMM languages. As an example of double nominalization Besnier gives the form in (70), without a translation: (70) i te sigaa-ga-ga ki lalo at the fall-nmlz-nmlz to down
Compare the (presumably synonymous) simple nominalization in (71): (71) i te sigaa-ga ki lalo at the fall-nmlz to down ‘when [he] fell down’
There are parallels between double nominalizations in Tuvaluan and in Toqabaqita: in both languages they are emphatic but otherwise synonymous with their simple counterparts.
3.4â•… The use of nominalizations with a stance function In Toqabaqita, nominalizations can be used with a “(speaker) stance” function, that is to signal the speaker’s attitude toward the state of affairs expressed in the clause (Yap & Matthews 2008, also Noonan 1997).6 Toqabaqita nominalizations can be used to express disapproval of, annoyance with, the relevant state of affairs. A nominalization and its complements or adjuncts, if any, and particles, if any, constitute a clause. Although there is no verb phrase, such constructions are complete, not elliptical,
.╅ As pointed out to me by Robin Hooper (pers. comm., February 2007), it is possible to use English noun phrases, even without nominalizations, with a stance function; for example: Children, the noise! (the speaker wants the children to stop being noisy).
Nominalizations in Toqabaqita and closely related languages 
clauses. Two examples of nominalizations with the stance function are given in (72) and (73): (72) Liu-laa ni boqo neri. walk.aimlessly-nmlz lig asrt npast.here ‘Walking/running all over the place (all the time).’
The sentence in (72) could be said, for example, to a child who has been moving, running around to make him/her stop; or it could be said to a person who has had an accident while walking from place to place without a goal in mind. (73) Qaa! Ngata-laa sula doo qoro qeri. hey! talk-nmlz prol thing be.many that ‘Hey! All this talking about the things./Hey! Too much talking about the things./Hey! That’s enough talking about the things.’(Lit.: ‘Hey! Talking about those many things.’)
In (74) below, fau-la-ngaqi-la-na has the form of a nominalization, but there is no verb fau-la-ngani in the present-day language, although there is fau-ngani ‘cut with (e.g. a knife) in an idle manner, not really cutting, just playing around with the implement’. (The transitive suffix regularly has the form -ngaqi before the nominalizer and in some other contexts, and -ngani elsewhere.) The first -la in fau-la-ngaqi-la-na is a “stem extender”, not the nominalizer. Stem extenders, of which -la is one, are found in a number of transitive verbs. In a few cases, there are pairs of synonymous verbs, one with the stem extender and one without; for example: futa-la-ngani/futa-la-ngaqi and futa-ngani/futa-ngaqi ‘beget’, ‘bear (a child’) (futa ‘be born’, -ngani/-ngaqi transitive suffix). It is likely that at one point there was a verb fau-la-ngani/fau-la-ngaqi, possibly in free variation with fau-ngani/fau-ngaqi, but it has survived only in the nominalized form. (74) Fau-la-ngaqi-la-na naifa nau ada ka foo. cut.idly.with-ext-tr-nmlz-3.pers knife 1sg tim 3sg.seq be.blunt ‘Cutting around with my knife; it might get blunt./Stop cutting around with my knife; it might get blunt.’
It appears that nominalizations can be used with a stance function also in South Lau: (75) alifii-laa! make.noise/be.noisy-nmlz ‘Be quiet!’ (Lit.: ‘Making noise!/Being noisy!’) (76) alifii-laa ada wela ka ada make.noise-nmlz tim child seq wake.up ‘(Be) quiet, or the child will wake up./(Be) quiet; the child might wake up./(Be) quiet, so that the child may not wake up.’
 Frantisek Lichtenberk
Judging by the lack of mention of the use of nominalizations with a stance function in other LMM languages (and in other Oceanic languages), it is either non-existent there, or it is infrequent and so has escaped attention.
3.5â•… Cognate nominalization and cognate relativization By “cognate nominalization” is meant here a situation where a nominalization occurs in a clause whose predicate is headed by the same verb as the one that is the base of the nominalization. In the two examples below, the nominalizations and the corresponding verbs are in bold. (77) … kwa fula kwa riki-a na muki too 1sg.seq arrive 1sg.seq see-3.obj comp 2pl.ipfv stay ba-muluqa qan=teqe si too-laa qeri bana, nau kwai lim-2pl.pers genp=one prtt stay-nmlz that lim 1sg 1sg.fut ngali kamuluqa i laa lokap. take 2pl loc in jail ‘[If, after the 14 days of warning to you to abandon the Marching Rule movement] I arrive and see that you haven’t budged [i.e. that the addressees have not abandoned the movement], I will put you in jail.’ (Lit.: ‘… I arrive and see that you are staying in just that one staying, I will take you in jail.’)
In (78) thaame-qe ‘father-assoc’ and the nominalization qai-laa ‘shouting’ have fused into one phonological word, and the final vowel of the associative suffix has been deleted: (78) Manga n=e raa-raqa ka fula=i time rel=3sg.nfut rdp-climb 3sg.seq arrive=loc fuufuur-a qai baa, nia ka qai qana teqe crown.of.tree-3.pers tree that 3sg 3sg.seq shout ins one thaame-q=qai-laa n=e baqita. father-assoc-shout-nmlz rel=3sg.nfut be.big ‘When, after climbing and climbing, he reached the crown of the tree, he gave out a very loud shout.’ (Lit.: ‘When he climbed and climbed and arrived at the crown of the tree, he shouted with a father of a shout that was big.’)
However, more common than cognate nominalization of the type exemplified above is a construction that can be referred to as “cognate relativization”. In cognate relativization, the head noun modified by a relative clause is a nominalization based on the same verb that heads the predicate of the relative clause. In the examples of cognate relativization given below, the head nominalizations and the corresponding verbs in the relative clauses are in bold.
Nominalizations in Toqabaqita and closely related languages 
In (79) the head nominalization is qai-laa ‘shouting’, and the relative clause contains the verb qai ‘shout’. The sentence in (79) is from the same text as the sentence with the cognate nominalization qai-laa ‘shouting’ in (78) above. In the text the sentence in (79) directly follows the one in (78), and the two instances of qai-laa have the same referent. (79) Qai-laa n=e qai qani-a e shout-nmlz rel=3sg.nfut shout ins-3sg.obj 3sg.nfut quna qeri, “….”. manner this ‘[When, after climbing and climbing, he reached the crown of the tree, he gave out a very loud shout.] This is what he shouted, “….”./He shouted this, “….”.’ (Lit.: ‘The shouting that he shouted with was thus, “….”.’)
In cognate relativization the head nominalization does not supply any information beyond that supplied by the verb in the relative clause. The nominalization merely serves as a structural head for the relative clause. The position relativized is either oblique object or direct object. When the position relativized is oblique object, the relative clause contains a preposition that carries a suffix that indexes the head nominalization. The preposition is either the general one, qaniÂ�, used with an instrumentmarking function, (as in [79]), or another preposition specifically selected for by the verb (and the head nominalization). Since in cognate relativization the general preposition has an instrument marking function (for supporting evidence see Lichtenberk 2006), it is glossed ins(trument) rather than genp (general preposition). In (79) above and in (80) below, the relative clauses contain the general/instrument-marking preposition. In both sentences the verbs in the relative clauses are intransitive. In (80) there are two instances of the general/instrument-marking preposition. The first one, in the form qani-a, indexes the head nominalization and has the specifically instrumentmarking function. The second one, in the form qana, has the general-prepositional function. The preposition has the “reduced” form qana because it has an adjacent lexical (as opposed to a pronominal) complement. (80) … kai manta fasi suli-a raa-laa na kai raa 3sg.fut think prec prol-3.obj work-nmlz rel 3sg.fut work qani-a qana biqu qeri qi laa manta-na tootoqonaqo, …. ins-3sg.obj genp house that loc in mind-3sg.pers earlier.time ‘[When a man wants to build a house,] he will first of all think about the work he will do on the house in his mind [its size, etc.], …. (Lit.: ‘… he will first think of the working that he will work with on the house in his mind first, ….’) (That is, the man will first make a mental plan of the house.)
In (81) the relative clause contains the prolative preposition suli with the third person singular object suffix -a that indexes the head nominalization. In this example, both
 Frantisek Lichtenberk
the verb in the relative clause and the head nominalization are reduplicated to express the fact that the events of running/jogging are temporally extended. (This is one of the functions of reduplication of verbs in Toqabaqita.) (81) Fii-fita-laa qeri na kera kere fii-fita suli-a, rdp-run-nmlz that rel 3pl 3pl.nfut rdp-run prol-3.obj kera kera qoo-qono ki …. 3pl 3pl.nfut rdp-sit pl ‘… the running [here: jogging] that they do, they/those who are always sitting (in offices all day long and get a little fat).’ (Lit.: ‘… the runnings that they run along, they who sit ….’)
The verb fita ‘run’ selects the prolative preposition for its oblique objects that express the place one runs on or along. In (82) the prolative preposition occurs in its reduced form sula with a lexical object. (Compare the reduced form qana of the general preposition in (80) further above, also with a lexical object.) (82) … kere fita ba-da sula tala qe=ki, …. 3pl.nfut run lim-3pl.pers prol road that=pl ‘… they just run [here: jog] on/along those roads, ….’
When the verb in a relative clause is transitive, there are two patterns. In one there is no lexical expression of the patient, theme, or stimulus, which would be the direct object of the verb otherwise. That participant is expressed only by means of an object suffix on the verb. Here too the relative clause contains the general preposition qani in its instrument-marking function. The head nominalization is indexed on the preposition by means of the third person singular object suffix -a. This pattern is illustrated in (83): that which had been seen is indexed by the object suffix on the verb riki ‘see’ in the relative clause. (83) Nau ku uqunu suli-a riki-laa na ni nau 1sg 1sg.nfut narrate prol-3.obj see-nmlz rel profore 1sg ku riki-a qani-a neri. 1sg.nfut see-3sg.obj ins-3sg.obj npast.here ‘I am talking about what I saw.’ (Lit.: ‘I am talking about the seeing that I saw it with.’)
In the other pattern there is a lexical expression of the patient, theme or stimulus, and a special construction of ‘instrument inversion” is used. In instrument inversion the participant that would otherwise be expressed as the direct object of a transitive verb is expressed by means of an oblique object, and the participant that would normally be expressed by means of an oblique object is expressed as the direct object. Instrument inversion takes place, under certain conditions, in relative clauses, whether those
Nominalizations in Toqabaqita and closely related languages 
involve cognate relativization or not, and in some other environments. (For more detail on instrument inversion see Lichtenberk 2006.) For example, in the basic pattern, without instrument inversion, the verb furi ‘cut’ takes as its direct object a patient noun phrase: (84) Furi-a berete qena. cut-3.obj bread that ‘Cut the bread.’
The sentence fragment in (85) below contains cognate relativization with the verb furi ‘cut’ in the relative clause and the nominalization furi-laa ‘cutting’ as the head. Instrument inversion has taken place in the relative clause: the verb furi carries the suffix -a, which marks direct objects, but the patient ‘this ongi bamboo’ is realized as an oblique object. It is the head nominalization that has the direct-object relation to the relative clause. (85) … ma ka toqola-ngani-a si furi-laa na keki and 3sg.seq explain-tr-3.obj prtt cut-nmlz rel 3pl.fut furi-a qana fo ongi naqi …. cut-3sg.obj genp clf bamboo.sp this ‘… and he explained [to them] how they would cut the ongi bamboo ….’ (Lit.: ‘… and he explained the cutting they would cut concerning this ongi bamboo ….’)
Similarly, instrument inversion has taken place in (87) further below. As (86) shows, the verb qaru-toqo ‘use’ takes in its basic pattern a theme noun phrase as its direct object. (The verb contains the test suffix -toqo but is semantically opaque; see Footnote 5.) (86) Qaru-toqo-na gete. fall-test-3.obj gate ‘Use the gate (to go in).’
In (87), with cognate relativization, the object suffix on the verb qaru-toqo ‘use’ indexes the head nominalization qaru-toqo-na-laa ‘using’, while the theme is expressed as an oblique object. Qaru-toqo is a Class 2 transitive verb, and so the object suffix -na is retained in the nominalization (Section 3.2). (87) … qaru-toqo-na-laa na miki qaru-toqo-na qana fall-test-3.obj-nmlz rel 1pl(excl).ipfv fall-test-3.obj genp qota, na ni nau ku thathami-a kwai uqunu areca.nut foc profore 1sg 1sg.nfut want-3.obj 1sg.fut narrate suli-a si manga qeri. prol-3sg.obj prtt time this ‘… the uses that we have for areca nuts (lit.: the using that we use concerning areca nuts), that’s what I want to talk about at this time.’
 Frantisek Lichtenberk
A construction akin to cognate nominalization is found in focusing. In such “cognate focusing”, the noun phrase in focus is headed by a nominalization of the same verb that occurs in the clause that expresses the presupposition. There are other parallels between relativization and focusing in Toqabaqita, such as the use of the same element na/n= to introduce relative clauses and the clause(s) of presupposition in the focus construction; see (79) and (87) above for the former and (88) below for the latter. And instrument inversion takes place both in relative clauses and in the clauses of presupposition in the focus construction. In (88) cognate focusing occurs in an open interrogative. The verb in the presupposition clause is intransitive, and the nominalization in the focus position is indexed by means of the object suffix on the general/ instrument-marking preposition qani, just as may be the case in cognate relativization. (88) Fita fa lae-laa n=o lae qani-a uri Honiara? how.many? clf go-nmlz foc=2sg.nfut go ins-3sg.obj all Honiara ‘How many times have you been to Honiara?’ (Lit.: ‘How many goings is it that you have gone with them to Honiara?’)
4.â•… Summary The main points of the present study can be summarized as follows. As a rule, the LMM languages (as well as other Oceanic languages) have multiple nominalizing strategies. Unfortunately, not many descriptions of the languages provide information that goes much beyond the formal aspects of the derivational processes, such as the type of nominalizer involved. Often very little or nothing is said about the productivity of the processes and about the functions of the different types of nominalizations. The immediate ancestor of the LMM languages inherited the Proto Oceanic nominalizer *-ŋa, probably with *-ŋā/-ŋaa as a variant, but this suffix came into competition with a new nominalizer, *-la, probably with *-lā/-laa as a variant. In some languages, such as Kwaio, both nominalizers are still present. In others, namely Lau and Toqabaqita, the liquid-initial nominalizer has completely supplanted the original nasal-initial one. In Toqabaqita and at least some other LMM languages, and in at least one other Oceanic language, double nominalizations are possible. In Toqabaqita and in Tuvaluan, double nominalizations are synonymous with the corresponding simple nominalizations in terms of objective content, but are judged to be emphatic, “stronger”. Possibly, it is their greater phonological bulk that makes double nominalizations emphatic. It may be that in one language discussed here, Ulawa, double nominalizations have meanings that are different from the corresponding simple nominalizations, but owing to insufficient information this is merely a hypothesis at this time. Whether the existence of double nominalizations in the languages discussed here is due to a single development
Nominalizations in Toqabaqita and closely related languages 
in a proto-language or to multiple developments is a question that cannot be answered without information on the presence or absence of double nominalizations elsewhere in Oceanic. In Toqabaqita and its close relative South Lau, nominalizations can be used with a stance function. In Toqabaqita, on which more detailed information is available, this use of nominalizations appears infrequent. Grammars of Oceanic languages, even relatively comprehensive ones, do not mention any stance functions. It is not clear whether this is due to its absence in those languages or to its rarity. Finally, in Toqabaqita there is cognate nominalization, and, in particular, cognate relativization (and cognate focusing). In the latter, the function of a nominalization is merely to serve as the syntactic head for the relative clause (or to encode the referent in focus with respect to the clause of presupposition). The nominalization does not provide information beyond that provided by the verb in the relative clause (or the clause of presupposition). The origins of the nominalizers are not known. Even though the liquid-initial nominalizer is a relatively recent innovation in Proto LMM, there is no obvious lexical source. It is conceivable that its rise is due to an idiosyncratic phonological process of replacement of the initial /ŋ/ with /l/, for unknown reasons. At least in the languages considered here, the nominalizers in question are not characterized by polysemy or heterosemy. (For the latter concept see Lichtenberk 1991.) The elements always form nominalizations, with the proviso that in double nominalizations the form to which the nominalizer is suffixed is a nominalization, but so is the derived form. Clearly, there is much more to be said on the subject of nominalizations in Oceanic, but more detailed descriptive information is required.
Abbreviations 1 2 3 abl all art asrt assoc caus clf com
first person second person third person ablative allative article assertive associative causative classifier comitative
comp compl du dvn excl ext foc fut genp impers in
complementizer completive dual deverbal noun exclusive (stem) extender focus future general preposition impersonal inessive
 Frantisek Lichtenberk
ins intens ipfv lig lim loc neg negv nfut nmlz npast.here obj os pers pl poss prec
instrument intensifier imperfective ligature limiter locative negative negative verb nonfuture nominalizer nonpast here object oneself personal plural possessive precedentive
prf profore prol prtt purp rdp rel sb seq sg st test tim tr vent vivid
perfect pronominal foregrounder prolative partitive purpose reduplication relative somebody sequential singular something test timitive (‘lest’) transitive ventive vivid
References Besnier, Niko. 2000. Tuvaluan: A Polynesian Language of the Central Pacific. London: Routledge. Capell, A. 1971. Arosi Grammar. Canberra: Pacific Linguistics. Comrie, Bernard & Thompson, Sandra A. 1985. Lexical nominalization. In Language Typology and Syntactic Description, Vol. III, Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 349–398. Cambridge: CUP. Deck, Norman C. 1934. Grammar of the Language Spoken by the Kwara‘ae People of Mala, British Solomon Islands [Reprint No. 5]. Wellington: Polynesian Society. Fox, Charles E. 1974. Lau Dictionary. Canberra: Pacific Linguistics. Fox, Charles E. 1978. Arosi Dictionary (revised edition with English-Arosi index prepared by Mary Craft). Canberra: Pacific Linguistics. Geerts, P. 1970. ‘Āre‘āre Dictionary. Canberra: Pacific Linguistics. Hill, Deborah. 1992. Longgu Grammar. Ph.D. dissertation, Australian National University. Ivens, Walter G. 1927. A study of the Oroha language, Mala, Solomon Islands. Bulletin of the School of Oriental Studies 4: 587–610. Ivens, Walter G. 1929a. A Dictionary of the Language of Sa‘a (Mala) and Ulawa, South-east Solomon Islands. London & Melbourne: OUP & Melbourne University Press. Ivens, Walter G. 1929b. A study of the language of Marau Sound, Guadalcanal, Solomon Islands. Bulletin of the School of Oriental Studies 5: 345–358. Ivens, Walter G. 1934. A grammar of the language of Longgu, Guadalcanal, British Solomon Islands. Bulletin of the School of Oriental Studies 7: 601–621. Keesing, Roger M. 1985. Kwaio Grammar. Canberra: Pacific Linguistics. Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge.
Nominalizations in Toqabaqita and closely related languages 
Koptjevskaja-Tamm, Maria. 2005. Action nominal constructions. In The World Atlas of Language Structures, Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds), in collaboration with Hans-Jörg Bibiko, Hagen Jung & Claudia Schmidt, 254–257. Oxford: OUP. Lichtenberk, Frantisek. 1991. Semantic change and heterosemy in grammaticalization. Language 67: 475–509. Lichtenberk, Frantisek. 2005. Inalienability and possessum individuation. In Linguistic Diversity and Language Theories [Studies in Language Companion Series 72], Zygmunt Frajzyngier, Adam Hodges & David S. Rood (eds), 339–362. Amsterdam: John Benjamins. Lichtenberk, Frantisek. 2006. Instrument inversion in Toqabaqita. Linguistics 44–4: 749–780. Lichtenberk, Frantisek. 2008. A Grammar of Toqabaqita. Berlin: Mouton de Gruyter. Lynch, John, Ross, Malcolm & Crowley, Terry. 2002. The Oceanic languages. Richmond: Curzon. Moyse-Faurie, Claire. 2007. Les formes nominalisées du verbe dans quelques langues océaniennes. Faits de Langue 30: 97–116. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sandra A. Thompson (eds), 373–394. Amsterdam: John Benjamins. Yap, Foong Ha & Matthews, Stephen. 2008. The development of nominalizers in East Asian and Tibeto-Burman languages. In Rethinking Grammaticalization: New Perspectives [Typological Studies in Language 76], María José & Elena Seoane (eds), in collaboration with Teresa Fanego, 309–341. Amsterdam: John Benjamins.
Exclamatives and temporal nominalizations in Austronesian Daniel Kaufman Cornell University
Two functions commonly fulfilled by nominalization are exclamative formation and temporal subordination. This article examines the morphology and syntax of these types of nominalizations across a wide range of Austronesian languages. I argue that the commonality between these two functions is found in the presuppositionality (both real and accommodated) involved in their interpretation. Special attention is paid to the mixed nominal-verbal properties of these constructions while offering evidence that nominalization can be a process which applies at several levels within the syntax.
1.â•… Introduction As is by now well known, nominalization strategies are employed for a strikingly wide variety of discourse functions across languages (Noonan 1997, inter alia). In this paper we examine just two functions of nominalization, exclamatives and temporal adjuncts, as they are attested commonly throughout Austronesian languages. We take a primarily diachronic-descriptive approach to the data and posit a new historical account for the genesis of a particular type of exclamative morpheme Proto-MalayoPolynesian *ka- and temporal adjunct marker *paR-.1,2 We also touch upon the idea of
.╅ The term Western Malayo-Polynesian (WMP) does not refer to a genetic subgroup but is rather composed of all the Malayo-Polynesian languages which cannot be subgrouped into the better defined Central-Eastern Malayo-Polynesian (CEMP) (Blust 1993). But because CEMP languages are generally understood to have undergone morphosyntactic restructuring, probably due in no small part to contact with non-Austronesian languages in East Indonesia, many of the patterns discussed here can likely be reconstructed for Proto-Malayo-Polynesian without recourse to external witnesses from CEMP. .╅ All Tagalog and Indonesian/Malay data comes from my own notes. Because the majority of the discussion is illustrated by Tagalog, this language is not specifically indicated on the right of each example.
 Daniel Kaufman
“high nominalization” as a possible explanation for some of the common properties between these two constructions. A brief note on lexical category is in order before setting out on our main topic. The question of identifying nouns and verbs in morphologically conservative Austronesian languages naturally looms large in any discussion of nominalization. Philippine verbs have been argued to possess many nominal properties (Himmelmann 1999; Naylor 1979 et seq; Kaufman 2007, 2008, in progress) which in turn has been argued to be the result of an older reanalysis of nominal categories into verbal ones (Starosta, Pawley & Reid 1982). What are typically referred to as patient voice and actor voice sentences in the Austronesianist literature are exemplified by Tagalog in (1). Sentence (1a), a canonical transitive construction with a definite agent and patient, employs the patient voice which corresponds with genitive case on the agent and nominative case on the patient. Sentence (1b), employs the actor voice, used for intransitives and transitives with indefinite patients. Here, nominative case is assigned to the actor and genitive case is assigned to the patient. As noted by earlier grammarians, these sentences are translatable as either verbal predications or copular sentences with nominal predicates. The latter interpretation is motivated by the noun-like distribution of these predicates and their assignment of genitive (possessor) case to non-subject arguments. (1) a.
súnog-∆ ni Maláyà ang bandílà burn-pv p.gen M. nom flag ‘Malaya burnt the flag’ ~ ‘The flag was Malaya’s burnt thing.’
b. nag-súnog si Maláyà nang bandílà burn-av p.nom M. gen flag ‘Malaya burnt a flag’ ~ ‘Malaya was the burner of a flag.’
For clarity, we forgo Tagalog orthographic conventions and spell out the genitive case marker as nang and plural marker as manga (orthographically, ng and mga) Abbreviations: abs absolutive accusative acc adjectival adj alrd already asp unidentified aspect actor voice av cmpl completive deic deictic det non-case marking determiner dist distributive erg ergative excl exclamative existential ext
gen genitive grnd gerundive incm incompletive intns intensive lnk linker loc locative mdrt moderative neg negation nmlz nominalizer nom nominative nvl non-volitional obl oblique p personal name case marker
pl plural prt unidentified particle pst past pv patient voice qm question marker rel relative marker rl realis soc sociative temp temporal adjunct marker top topic marker tr transitivity related affix
Exclamatives and temporal nominalizations in Austronesian 
Morphologically conservative Austronesian languages show up to four voices (patient, actor, locative, conveyance) which can be reconstructed to Proto-Austronesian (Wolff 1973; Ross 2002). These voices are used for a much wider variety of functions than is indicated by their names but all have in common that they assign nominative case to the argument bearing the proto-role with which they are associated. With this short background, we can compare nominalizations proper, the subject of the present investigation. While the canonical predicates above typically assign nominative case to a thematic argument in the clause, nominalizations do not assign nominative case but rather can only assign genitive and oblique case to arguments. This can be seen by comparing the canonical property predicate in (2a) with the corresponding exclamative in (2b). In the former the sole argument appears as a subject with nominative case and in the latter it takes genitive case and is syntactically restricted in the same manner as other genitive/possessor dependents in Tagalog. Note that the property predicate cannot assign genitive case to its subject and the exclamative cannot assign nominative case, as shown in (3). (2) a.
ma-ganda si Maria. adj-beauty p.nom M. ‘Maria is beautful.’
b. kay ganda ni Maria! excl beauty p.gen M. ‘How beautiful Maria is!’ (3) a. *ma-ganda ni Maria. adj-beauty p.gen M. ‘Maria is beautful.’ b. *kay ganda si Maria! excl beauty p.nom M. ‘How beautiful Maria is!’
Looking now at temporal adjuncts, in (4a) we find a canonical intransitive sentence with an actor voice predicate and in (4b) we find its corresponding temporal adjunct. Again, we find that the sole argument is expressed as a genitive/possessor rather than a subject. (4) a.
alis si Maria leave p.nom M. ‘Maria left.’
b. pag-alis ni Maria… ger-leave p.gen M. ‘When Maria leaves/left…’
 Daniel Kaufman
The underlying commonality between the constructions under discussion here is that they lack assertive force and do not constitute predications on their own. In all cases, a predication is packaged as a presupposition. Seen in this way, both temporal adjuncts and exclamatives seem to share a relationship to anaphoric expressions. A typical usage of temporal adjuncts in Tagalog is shown in (5). Note that, just as the referent Pedro in the first clause is referred to anaphorically by the genitive pronoun niya in the following clause, the predicate dumating in the first clause, marked by voice and aspect, is referred to nominally in the following clause (as gerundive pag-dating). Pronominal anaphora can thus be seen as a model for temporal anaphora via nominalization. (5) Datingi si Pedroj. Pag-datingi niyaj … arrive p.nom p. ger-arrive 3s.gen ‘Pedro arrived. When he arrived…’ (More literally ‘Upon his arrival…’)
This is expected in the case of the temporal adjuncts as they are truly part of the background at the point of utterance, as in (5) but perhaps not so obvious in the case of exclamatives. However, Michaelis & Lambrecht (1996) and Michaelis (2001) argue that the presuppositional nature of exclamatives can be the key commonality tying together the various ways in which they are expressed cross-linguistically. That exclamatives contain presuppositions was already noted by Sadock & Zwicky (1985: 162) (cited by Michaelis 2001: 1040), emphasis mine: Exclamations are intended to be expressive whereas declaratives are intended to be informative. Both represent a proposition as being true, but in an exclamation, the speaker emphasizes his strong emotional reaction to what he takes to be a fact, whereas in a declarative, the speaker emphasizes his intellectual appraisal that the proposition is true.
The actual exclamatory force of nominalized exclamatives is probably best considered as arising from the pragmatic bridging which the hearer must undertake for the exclamative to be interpretable. The speaker is moving one step ahead of the hearer in assuming the property attributed to the subject and making a claim as to its degree.3 Following the analogy of (5) above, we can also conceive of property exclamatives as forcing accommodation on the part of the hearer, as indicated by the coindexing in (6), where the discourse antecedent is overt. Again, an indefinite predicational antecedent is referred back to with a nominal. In a typical exclamative, the discourse antecedent
.╅ Portner & Zanuttini (2004) attribute this to a null exclamative morpheme but this cannot capture the intuition that the exclamative quality results from forcing the hearer to accommodate a referent without previous introduction (see also Potsdam this volume for an additional criticism).
Exclamatives and temporal nominalizations in Austronesian 
is not overt and requires accommodation, underscoring the mirative nature of the speech act. (6) Ma-bilisi si Pedroj. Kay bilisi niyaj talaga! adj-speed p.nom p. excl speed 3s.gen truly ‘Pedro is fast. He’s so fast!’ (More literally, ‘His speed!’)
In this way, both temporal adjuncts and exclamatives can be analyzed as anaphoric expressions, the latter of which attains its illocutionary force, at least in part, by triggering pragmatic accommodation. The rest of this paper is organized as follows. A development from existentials to adjectivals and exclamatives with the PMP *ka- formative is proposed in Section 2.1; some synchronic properties of Tagalog exclamatives and the split nominal/adjectival features of exclamatives are explored in Section 2.2; the generalization of the *ka- morpheme to new environments is taken up in 2.3; an interim summary is presented in 2.4. In Section 3.0 we proceed to temporal adjuncts and discuss the development of PMP *paR- from a gerund marker of sorts to a temporal subordinator. The morphology and syntax of gerunds is sketched out in 3.1 and split nominal-verbal features of gerunds and temporal adjuncts are discussed in 3.2. Section 4.0 concludes and offers some speculations on the syntax-semantics correspondences in the constructions discussed.
2.â•… Tagalog exclamatives from an Austronesian perspective 2.1â•… The exclamative-existential connection Most Philippine languages have a number of morphological and syntactic strategies for forming exclamatives from property denoting roots. Tagalog, for instance, has the three morphological exclamative formations in (7a)–(c). One defining feature of the exclamatives in (7a)–(c) is their assignment of genitive case to the intransitive subject in contrast to their adjectival predicate counterparts which assign nominative case, as seen in (8). (7) a.
kay ganda niya! b. pagka-(ganda~)ganda niya! excl beauty 3s.gen excl-intns~beauty 3s.gen ‘how beautiful she is’ ‘how beautiful she is’
c.
nápaka-ganda niya! excl-beauty 3s.gen ‘how beautiful she is’
(8) ma-ganda siya adj-beauty 3s.nom ‘She’s beautiful’
 Daniel Kaufman
All three exclamatives formations above contain what is in all likelihood a frozen prefix ka-, as highlighted. As discussed in detail by Blust (2003), Proto-Austronesian (Pan) *ka- was a multifunctional morpheme which had an essential role in the paradigm of stative verbs and is often found to be in a paradigmatic relation with ma- in statives and adjectives in modern languages.4 In Proto-Malayo-Polynesian (PMP) – a primary branch of PAn and the progenitor of all Austronesian languages outside of Taiwan – this affix appears to have taken on several more functions, such as ‘formative for abstract nouns’, ‘manner in which an action is carried out’ (Blust 2003: 473) in addition to a (probably unrelated) ‘co-’ function (e.g. Tagalog ka-silid co-room ‘roommate’). It seems that *ka- additionally (or underlyingly) functioned as an existential marker in at least PMP, if not PAn.5 It survives in this putatively original function in examples such as (9a), from Tagalog, and also possibly underlies the common Austronesian locative circumfix ka- -an, also exemplified here with Tagalog in (9b).6 (9) a.
mag-ka-pérà b. ka-batu-han av-ext-money ext-stone-loc ‘to have money’ ‘stony place’
.â•… See Evans & Ross (2001) for the multifunctionality of Proto-Oceanic *ma- (used primarily for forming stative and experiential verbs and adjectives) where the prefix shares many of its characteristics with its cognate forms in Western Malayo-Polynesian languages albeit with much less productivity. .â•… If the existential function can be reconstructed for PAn *ka-, then this may have underlied its role in the stative paradigm. There is a wealth of evidence that predicates of possession and obtaining – functions for which the reflexes of *ka- are regularly employed in modern languages – are easily reanalyzed as passives. One need only compare the widespread get > passive marker in East Asian languages and English (see also Zeitoun & Teng’s (2009) arguments for a similar reinterpretation of *ki- acquisitive having been reinterpreted as a passive marker in Rukai, Paiwan and Puyuma). The similar function of stative predicates in expressing non-volitional and agentless actions in Austronesian makes this connection imminently plausible but this line of inquiry must be left to further investigation. .â•… Because it is not entirely productive or transparent, the ka element in magka- and ka- -an is not typically recognized as an independent morpheme and is usually left unglossed. Naylor (2005: 429, fn.26) treats Tagalog ka- as an inchoative although this seems to be an inappropriate gloss for most of its functions. Herein, glosses will follow the existential *kahypothesis unless clearly unwarranted in a particular language. Example (7b) further contains pag-, a polysemic affix which is difficult to ascribe a precise function but which will be discussed in further detail in Section 3. Example (7a) contains nápa- which is most probably the perfective stative na- combined with the causative pa-, although this is no longer part of a productive paradigm and must thus be considered frozen morphology.
Exclamatives and temporal nominalizations in Austronesian 
As shown in (10)–(16), reflexes of PAn *ka- can further be seen in this function in Bolaang Mongondow ko-; Pendau ‘o; Timugon Murut maka; Babuza/; Bugis әŋka; Wolio ko-,7 to only name a few. Clearly, much work is to be done to reconstruct this function for *ka- although this preliminary genealogical distribution appears promising. (10) Bolaang-Mongondow a.
ko-iput b. ko-mata ext-tail ext-eye ‘having a tail’ ‘having eyes’
(Usup et. al. 1981: 20)
(11) Pendau a’u ndau ‘o-piso 1s.nom neg ext-machete ‘I don’t have a machete.’
(Quick 2003: 139)
(12) Murut ati-ati pulu′-rali, maka-ulun noyo bagu whichever headland-det.pl av.ext-person alrd prt ‘As for whichever were the headlands (i.e. every single headland), there were people (there)’ (Prentice 1971: 252) (13) Murut aa, ginio panauntu nano, tuan, maka-putor hest that young_foliage 3s.gen sir av.ext-sago_palm_grub ‘Er, as for that also, its young foliage, sir, (it) has sago palm grubs.’ (Prentice 1971: 265) (14) Bugis Nakko әngka tau-pa-sala if ext person-tr-sin ‘If there is a guilty person.’
(Sirk 1996: 170)
.â•… Unlike in Bolaang-Mongondow and Pendau, the vowel in Wolio ko- shows an irregular *a>o vowel correspondence (cf. Wolio adjectival ma-). It is not clear if this is due to borrowing, sporadic change, or inheritance from PMP *ka-R- with the addition of the PMP * infix of unclear function (cf. Bikolano ka-harong ext-house ‘owner of the house’). Mead (2003) discusses the sporadic *a>o change in Sulawesi.
 Daniel Kaufman
(15) Wolio a.
ko-bulu b. ko-bake c. ext-hair ext-fruit ‘to be hairy’ ‘to bear fruit’
d. ko-baju e. ext-jacket ‘to wear a jacket’
ko-oni f. ext-voice ‘to speak’
ko-tawa ext-leaves ‘to have leaves’ ko-nami-gara ext-taste-salt ‘to have a salty taste’ (Anceaux 1988: 14)
Ross (1995) reconstructs stative/adjectival PAn prefix *ma- as a reduction of *ka- which we can now view here as the combination of the existential prefix and the actor voice infix.8 In many languages, the prefixes ma- and ka- are still in a paradigmatic relationship in which the former is found in finite contexts (loosely construed) and the latter in non-finite contexts (e.g. imperatives, complements of negation, “narrative tense”). While in Tagalog, this ka- emerges in morphologically complex exclamatives, as seen earlier in (7a)–(c), other Philippine languages employ the prefix ka- unadorned in the same context, as shown in (17) for Cebuano, (18) for Mansaka and (19) for Wolio.9 The ‘finite’ adjectival counterparts are shown in the corresponding (b) examples (see also Zorc 1975: 142 for other Bisayan examples). (16) Cebuano a.
ka-tambok niya! b. ma-tambok siya ext-fat 3s.gen adj-fat 3s.nom ‘How fat he is!’ ‘He’s fat.’
.â•… The gloss adj will be maintained in languages where there is little synchronic evidence for this morphological decomposition. .â•… There is, however, one unexpected difference between kay and may in Tagalog which groups kay together with the adjectival ma- instead of the existential may. This involves the possibility of number agreement with ma- and kay but not with may, as shown in (i)–(iii). CV reduplication is possible in (i) and (ii) but not in (iii), the putative actor voice existential. Plural reduplication is again possible when property words are used in argument position, as in (iv). I have no explanation for this fact at present. (i)
Ma-ga~ganda sila (ii) adj-pl~beauty 3p.nom ‘They are beautiful.’
Kay ga~ganda nila! ext pl~beauty 3p.gen ‘How beautiful they are!’
(iii)
May (*ga~)ganda sila av.ext pl~beauty 3p.nom ‘They have beauty.’ (i.e. ‘They are somewhat beautiful’)
(iv)
Na-bighani ako sa ga~ganda nila pv.nvl.rl-enchant 1s.nom obl pl~beauty 3p.gen ‘I was enchanted by their beauty.’
Exclamatives and temporal nominalizations in Austronesian 
(17) Mansaka a.
ka-pora da agaw naan ext-red alrd now 3s.gen ‘Oh, how red it is!’ (Svelmoe & Svelmoe 1974: 63ff, cited in Blust 2003:fn.10)
b. ma-pora yang atup nang baray… adj-red nom roof gen house ‘The roof of the house is red’
(Svelmoe & Svelmoe 1990: 291)
(18) Wolio a.
ka-luntu-na o mia sii ext-lazy-3.gen det person this ‘How lazy this person is!’
(Anceaux 1988: 53)
b. ma-luntu o mia sii adj-lazy det person this ‘This person is lazy.’
Understanding both adjectives and exclamatives to be built off of existentials, the Tagalog exclamative marker kay seen in (7) above can be analyzed as the basis of the existential marker may.10 The morphemes kay and may can be decomposed to several component proto-morphemes at some earlier stage of the language with kay containing the existential element ka- followed by what may be reconstructed as a (possibly indefinite) determiner *i (cf. Reid 1978), as shown in (20a).11 The existential may can be reconstructed to the same combination of morphemes plus the PAn actor voice morpheme , as shown in (20b). (19) a.
*ka i > kay ext det ext
b. *ka i > may ext det av.ext
.â•… Although these are both represented orthographically as separate words in Tagalog, they both behave as proclitics or prefixes in being inseparable from their complements. For instance, they may not be followed by second position clitics, as full words generally can. .â•… It is not clear to what historical stage this determiner can be reconstructed with a nominative function. It is found in several Cordilleran and Bisayan languages (Wolff 1963; Sityar 2000 for Cebuano; Rubino 2005 for Utudnon; and Zorc 1975) and is employed on the complements of existential and negative existential predicates, e.g. Cebunao duna=y and walà=y, respectively. It probably also gave rise to the i- initial indefinite nominative case markers in Waray-Waray. A locative marker *i has been reconstructed by Blust (1995) for PAn and identified with Tagalog y in may by Naylor (2005). However, we expect that a locative marker would precede the possessor and not the possessum in an existential, and it is thus more likely that Tagalog y should be identified as a frozen indefinite nominative marker rather than a locative.
 Daniel Kaufman
The presence of the actor voice in the may existential is reflected syntactically in the fact that the possessor is expressed in the nominative case, as can be seen in (21), similar to regular intransitive subjects. This pattern is different from the common syntactic pattern found with reflexes of PAn *wada, an independent existential predicate, in which the possessor is expressed as a genitive modifier of the possessum, as exemplified by Ilokano in (22) and Kimaragang Dusun in (23). Note that in the latter language, the affixal existential ki- (possibly from *ka- i), which also crucially lacks the bilabial nasal element typically indicative of actor voice morphology, follows the same pattern. (20) may aso siya av.ext dog 3s.nom ‘He has a dog.’ (21) Ilokano Adda aso=na ext dog=3s.gen ‘He has a dog.’
(Rubino 1997: 118)
(22) Kimaragang a.
waro tanak nuh oy? ext child 2s.gen qm ‘Do you have any children?’
b. ki-tanak nuh oy? ext-child 2s.gen qm ‘Do you have any children?’
(Kroeger 2005: 411)
From a synchronic perspective, it is thus not unreasonable to still consider exclamative kay as a voiceless existential and may as an actor voice existential, as the case marking of coocurring arguments corresponds to that of unmarked and actor voice predicates, respectively. In particular, the actor voice existential assigns nominative case to the most agentive argument (Foley & Van Valin’s (1984) “Proto-Agent”) across the board, which can correlate to the possessor in existential constructions. Unmarked predicates on the other hand, typically assign genitive case to this argument. Compare for instance the different case marking patterns with the unmarked bivalent predicate in (24) and its actor voice counterpart in (25). (23) dala ni Ligaya ang pitákà carry p.gen L. nom wallet ‘Ligaya carries the wallet.’ (24) nag-dá~dala si Ligaya nang pitákà av.rl-incm~carry p.nom L. gen wallet ‘Ligaya carries a wallet.’
Exclamatives and temporal nominalizations in Austronesian 
The bare predicate in (26) can also be argued to have a fairly direct analog in another type of exclamative construction in Philippine languages. This type employs a bare property denoting root embedded under a nominative determiner. As above, genitive case is assigned to the sole argument. Similar examples in Botolan Sambal and Ivatan are shown in (27) and (28).12 (25) ang ganda mo! nom beauty 2s.gen ‘How beautiful you are!’ (26) Botolan Sambal a.
hay ganda nin babayi! nom beauty gen woman ‘How beautiful the woman is!’
b. hay bitil ko! nom hunger 1s.gen ‘How hungry I am!’
(Antworth 1979: 50)
(27) Ivatan ay so pia na no tao excl acc good 3s.gen gen person ‘How good the man is!’
(Reid 1972: 58)
Relatedly, Amurrio (1970: 36) describes a Pangasinan construction employing agáy la or aláy as expressing ‘wonder, compassion or indignation’. This marker may take property denoting complements with the ka- -an circumfix, as in (29a), or in their bare form, as in (29b); in both forms the (notional) subject is expressed in the genitive case.13 (28) Pangasinan a.
aláy ka-abig-an to! excl nmlz-good-nmlz 3s.gen ‘How good he is!’ (lit. ‘what goodness of him’)
b. agáy_la=y sulit mo excl=det cruel 2s.gen ‘How cruel you are!’ (lit. ‘What cruelty of you!’)
.â•… Ivatan unexpectedly uses accusative case for certain types of fronted arguments, as with exclamatives. .â•… In keeping with Amurrio’s descriptive practice for Pangasinan, I gloss the ka- -an circumfix as a nominalizer here rather than existential plus locative as done elsewhere.
 Daniel Kaufman
In this section, we have examined the possibility of historical and synchronic connection between existentials and property denoting words in Austronesian. Both have been argued to be formed with the existential ka- prefix. While property predicates were formed with the addition of the actor voice * infix, exclamatives employed the non-finite form, i.e. the bare ka- form. This form can be considered as a nominalization of sorts as it expresses the subject in the genitive case. A more direct manifestation of nominal syntax in exclamatives is the bare predicate construction in which a bare property denoting root is embedded under a nominative determiner with the subject again being expressed in the genitive case. In the following section we take a closer look at the synchronic morphosyntax of exclamatives in Tagalog, concentrating on the mixed nature of exclamatives as semi-adjectival and semi-nominal constituents.
2.2â•… Synchronic properties of Tagalog exclamatives Interestingly, despite the difference in their case assignment properties when functioning as predicates, the unmarked existential and actor voice existential have the same basic syntax when used as modifiers. Compare these two formations in predicate function in (30) and modifier function in (31).14 (29) a.
kay ganda nang áwit! b. may ganda ang awit ext beauty gen song av.ext beauty nom song ‘How beautiful the song is!’ ‘The song has beauty.’
(30) a.
ang áwit na kay ganda b. ang áwit na may ganda nom song lnk ext beauty nom song lnk av.ext beauty ‘The song which is so beautiful.’ ‘The song which has beauty.’
The external argument can also be topicalized, as in (32). When topicalized, this argument appears in the nominative and not the genitive case, following the Austronesian general ban on extraction of genitive arguments. (31) ang áwit ay kay ganda! nom song top ext beauty ‘The song is so beautiful!’
The exclamative is also notable in that it is one of the few constructions in Tagalog which imposes selectional requirements on root type. In contrast to the forms in (7), the forms in (33a)–(c) are unacceptable because the exclamative is combined with an action-denoting root, takbo ‘run’, instead of a property-denoting one. These formations
.╅ This must thus be added to the list of exceptional syntactic environments which license relativization of genitives in Tagalog, as discussed by Cena (1979).
Exclamatives and temporal nominalizations in Austronesian 
do not have the “derivational strength” of ma-, as shown in (33), which can convert almost any root into a property denoting word. (32) a. *kay lákad! b. *pagka-lákad~lákad! ext walk excl-intns~walk c. *nápaka-lákad! d. ma-lákad excl-walk adj-walk ‘has a walk’15
The Tagalog exclamative can, however, be used to modify a verbal predicate in the same way adjectivals may take verbal complements in Tagalog. Compare the sentence in (34a) where the agent of the following verbal predicate is assigned genitive case by the exclamative with its non-exclamative counterpart in (34b) in which the subject is assigned nominative case by the adjective. (33) a.
kay bilis niya=ng takbo! ext speed 3s.gen=lnk run ‘How fast he runs!’
b. ma-bilis siya=ng takbo adj-speed 3s.nom=lnk run ‘He runs fast.’
Despite the existential analysis of exclamatives in Tagalog argued for here, these constructions should not be mistaken for existential predications. Rather, it is clear that the existential exclamatives are robustly non-predicational. The pivotal feature found in all the above examples, including those containing embedded verbal predicates as in (34), is the inability of expressing an assertion. This can be diagnosed by the (unâ•‚)grammaticality of adding a question marker to the two constructions in (34). All predicational statements can be converted into yes-no questions by the addition of the question marker ba, but this is disallowed with exclamatives, as seen in (35).16
.â•… Although not particularly common, this adjective can be used to describe a place which requires much walking, a person who has many things to do (lákad ‘walk, errand’). Additionally, it has a more specialized meaning, as described by Newsbreak magazine (12/20/07), “‘Malakad’ is a commonly used term in legal circles, referring to a judge’s or justice’s penchant to use connections to get promoted.” .â•… The ‘bare’ exclamative in Tagalog differs here in marginally allowing the question marker, as in (i). However in this case, it is perhaps only truly felicitous as an echo question, i.e. in a context in which a speaker has already uttered an exclamative of the same general form. (i)
ang ganda ko ba? nom beauty 1s.gen qm ‘Am I so beautiful?’
 Daniel Kaufman
(34) a. *kay bilis ba niya=ng takbo? ext speed qm 3s.gen=lnk run b. ma-bilis ba siya=ng takbo? av.ext-speed qm 3s.nom=lnk run ‘Does he runs fast?’
Finally, another interesting synchronic aspect of Tagalog exclamatives emerges in their plural marking which suggests that despite seemingly nominal type syntax, these constructions cannot be completely assimilated to nouns. Several types of plural marking exist in Tagalog: argument plurality and nominal predicate plurality are indicated by the proclitic manga and adjectival number agreement is indicated by CV-reduplication. CV-reduplication on an entity denoting lexeme is ungrammatical as shown in (36). Conversely, the use of manga with property denoting lexemes, as in (37a), is awkward and requires special context. As shown in (38), exclamatives pattern with property lexemes in taking CV-reduplication to indicate plurality rather than the proclitic manga. This is unexpected given that they display two hallmarks of nominal syntax; they are embedded under a case marking determiner and assign genitive case to their subject. (35) a.
manga gúrò sila pl teacher 3p.nom ‘They’re teachers.’
b. *gu~gurò sila pl~teacher 3p.nom (36) a. ??manga ma-ganda sila pl av.ext-beauty 3p.nom b. ma-ga~ganda sila (cf. *ma-ga~ganda siya) av.ext-pl~beauty 3p.nom av.ext-pl~beauty 3s.nom ‘They are beautiful.’ (37) a. *?ang manga ganda ninyo! nom pl beauty 2p.gen b. ang ga~ganda ninyo! (cf. *ang ga~ganda mo!) nom pl~beauty 2p.gen nom pl~beauty 2s.gen ‘How beautiful you are!’
However, this mixed behavior is expected if we understand nominalization as a process which can apply on several different levels in the morphosyntax (cf. Malchukov 2004; Alexiadou 2001; Ntelitheos 2006; Yanagida & Whitman 2008). If the order of phrases within the nominal domain is DetP > NumP > NP, in accordance with the standard surface order of these elements, nominalization on the level of DetP excludes the possiblity of nominal-type number marking, as the material beneath the locus of
Exclamatives and temporal nominalizations in Austronesian 
nominalization retains its “pre-nominalized” category. Because bare root exclamatives only obtain syntactic nominality from the case marking determiner, nominal type plural marking, which is internal to the case marker, is illicit, as seen above in (38). On the other hand, there are some morphological facts regarding reduplication in these constructions which do not obtain an easy structure-based explanation. As already hinted by the above, Tagalog has an exceedingly rich arsenal of reduplicant morphemes employed with a wide variety of meanings. Some of these different morphemes possess different phonological forms (e.g. 1σ, 2σ) and can also be distinguished by their different interpretations in different morphosyntactic contexts. 2σ-reduplication, when applied to ma- adjectivals, yields a moderate degree interpretation of the adjective, as seen in (39), but the same reduplication in the bare exclamative construction yields the opposite, an intensive interpretation, as seen in (39) (Wolff; Centeno & Rau 1991:§6.742). These facts are somewhat puzzling considering that this reduplication does not occur with entity denoting roots at all and thus cannot be understood as resulting from the split adjectival-nominal nature of property exclamatives in Tagalog. Similarly, full-word reduplication is found with ma- adjectivals with an intensive interpretation but is ungrammatical with bare exclamatives, as seen in (40). (38) a.
ma-ligá~ligáya ang babáe adj-mdrt~joy nom woman ‘The woman is somewhat joyful.’
b. ang ligá~ligáya nang babáe! nom intns~joy gen woman ‘How incredibly joyful the woman is!’ (39) a.
ma-ligáya=ng ma-ligáya ang babáe adj-joy=lnk adj-joy nom woman ‘The woman is very joyful.’
b. *ang ligáya=ng ligáya nang babáe nom joy=lnk joy gen woman ‘The woman is somewhat joyful.’
The intensive reduplication found in (39b) is furthermore not found when the property denoting root is used as an argument, as in (41). It is thus an exclusive feature of property of exclamatives. (40) Hindi niya na-intindi-han ang (??liga~)ligaya ko neg 3s.gen pv.nvl.rl-understand-lv nom â•… intns~joy 1s.gen ‘He didn’t understand my joy’
We have seen in this section that while bare exclamatives share properties of both canonical ma- adjectivals and canonical nominals, they also show unique features which cannot be ascribed to either component, as summarized in (42). A proposal
 Daniel Kaufman
for accounting for the split nature of exclamatives was briefly sketched out. This involved treating them as adjectivals which have undergone “high nominalization”. A fuller analysis which also takes into account their unique properties will have to await further work. (41) Summary of exclamative properties in Tagalog: nominal
adjectival
unique
genitive subjects nom case marking
1σ redp pl. agreement
2σredp intensive
2.3â•… Th e generalization of *ka- to a multifunctional exclamative in Austronesian A synchronic link between *ka- derived existentials and what are analyzed here as existential + actor voice *ma- formations need not exist in order for derivations with the former to have exclamative force. Many languages retain no synchronic traces of the existential function of *ka- and have additionally lost the adjectival *ma- prefix but maintain a *ka- reflex in a nominalizing and exclamative function. Muna is one such language as can be seen in the alternation below. A regular predication is shown in (43a) with its exclamative counterpart (43b). (Note that, unlike Philippine languages, Muna shows person agreement on verbal and adjectival predicates, as seen with the no- prefix in (43b) which agrees with the grammatical subject lalo-ku ‘my heart’.) (42) Muna a.
no-pana lalo-ku 3s.rl-hot heart-1s.gen ‘I am angry.’ (Lit. ‘My heart is angry’)
b. ka-pana-no lalo-ku! nmlz-hot-3s.gen heart-1s.gen ‘How angry I am!’ (Lit. ‘The anger of my heart!’) (van den Berg 1989: 173)
In (43a), a canonical predication, the property predicate pana ‘hot’ carries subject agreement while the (idiomatic) subject lalo ku ‘my heart’ follows. In (43b), the predicate head is marked with the nominalizer ka- while the subject agreement is replaced by genitive agreement. As van den Berg (1989) shows, Muna constructions headed by predicates nominalized with ka- are ambiguous between nominalizations and exclamatives, as he illustrates with the alternative interpretations of (44). (43) Muna ka-bhari-no anahi-hi-mu (!) nom-many-3s.gen child-pl-2s.gen (i) ‘the number of your children’ (ii) ‘how many children you have!’
(van den Berg 1989: 173)
Exclamatives and temporal nominalizations in Austronesian 
Other languages have generalized the exclamative function of *ka- with property denoting roots to new contexts. Ivatan (Batanic, N. Philippines) allows attachment of ka- to event denoting roots. The result is interpreted as an event-oriented exclamative, as in (45a), which can be compared to its predicational counterpart in (45b). (44) Ivatan a.
machi-nanaw sa du kayskuyan av.soc-study 3p.nom obl school ‘They study in school.’
b. su ka-pachi-nanaw da! acc excl-soc-study 3p.gen ‘How hard they study!’
(Hidalgo & Hidalgo 1971: 148)
(Hidalgo & Hidalgo 1971: 148)
Another, unrelated exclamative morpheme in Ivatan, ja-, alternates with adjectival ma-, as shown in (46a) and (46b). Interestingly, this prefix can also be attached to event denoting and entity denoting roots with interpretations varying accordingly. When attached to event denoting roots with aspect marking it yields the meaning ‘how often pred’, as in (46). When attached to entity denoting roots it yields the meaning ‘how many root’, as in (46). Although we do not know the provenance of this affix, its complementary distribution with adjectival ma- is similar to reflexes of *ka- and its usage appears to have spread throughout the lexicon such that it can now apply to almost any lexical category. (45) Ivatan a.
ma-vid sya b. adj-beauty 3s.nom ‘She is beautiful!’
ja-vid na! excl-beauty 3s.gen ‘How beautiful she is!’
c.
ja-ta~tañis-en d. excl-asp~cry-pv ‘how often x cries’
ja-savung excl-blossom ‘how many blossoms’ (Hidalgo & Hidalgo 1971: 70, 95)
We have seen in this section how *ka- has expanded its role in other Austronesian languages to form nominalized exclamatives from other types of roots, notably, event denoting and entity denoting roots. The commonality maintained with these *ka- derivations throughout Austronesian is their use of genitive case for the subject (when present) and their lack of assertive force. There also exist constructions which lack nominalizing morphology and which show semi-verbal properties but which still assign genitive case to the subject and possess an exclamative function. In the next section we take a look at some of these “hidden” nominalizations and their properties.
2.4â•… Split nominal features in exclamatives Some Austronesian languages appear to have generalized other elements of the nominal syntax associated with ka- existentials to non-ka- forms. Often this involves
 Daniel Kaufman
treating ma- adjectivals as nominalizations as well. This change can be viewed as the reinterpretation of a “low” morphological nominalization to a “high” syntactic one. An example can be seen in Bugis, which employs canonical adjectival predicates with genitive pronouns instead of nominative ones for exclamative/exclamatory purposes. This is shown in (47a) with the ma- adjective mapanre ‘dexterous’ and in (48) with the unaffixed (but reduplicated) adjective sommeng ‘arrogant’. Compare this to the canonical adjectival prediction in (47b), in which the intransitive subject is assigned nominative/absolutive case. (46) Bugis a.
ma-panre=na=ritu jemma! adj-dexterous=3s.gen=deic person ‘What a dexterous man that one is!’
(Sirk 1996: 149)
b. ma-panre=i adj-dexterous=3s.nom ‘S/he is dexterous.’ (47) Bugis somme~ssommem-mu, le to-sunra-e! intns~arrogant-2s.gen rt person-sunra-det ‘How arrogant you are, oh Sunra people!’
(Sirk 1996: 149)
The exclamative construction in Tukang Besi, as seen in (49) and (50), also employs regular adjectival predicates with genitive subjects.17 (48) Tukang Besi Ke ‘eka-su i aba! and fear-1s.gen obl earlier ‘I was really frightened earlier!’ (49) Ke to’oge nu ana-‘u! and big gen child-3.gen ‘Hasn’t your son grown up!’
(Donohue 1999: 480)
As discussed earlier for Tagalog, Tukang Besi exclamatives also display part-adjective and part-nominal syntax. The underlying property denoting nature of the adjective is still visible in its ability to license degree adverbs like saori ‘very’, as shown in (51). Following the proposal sketched out above, the exclamative may be analyzable as a bona fide adjective at the point of modification. Nominalization takes place
.â•… Donohue (1999: 156) further states that genitive subjects are also occasionally heard for canonical adjective predicates but not for any other type of predication, although this usage was ‘univesally proscribed’ by speakers when queried about it.
Exclamatives and temporal nominalizations in Austronesian 
in the syntax above modification thus allowing possessor agreement on the entire constituent saori-nini. (50) I wudo monini-‘a di Walanda ke saori-nini-no! obl season cold-nmlz obl Holland and very-cold-3.gen ‘In winter in Holland it’s very cold!’ (Donohue 1999: 459)
Another example can be seen in the Malay constructions in (52) employing betapa, an exclamative marker for property denoting words. Here, the adjective is typically marked with =nya, which is both the third person genitive and a more general marker for a possessum (see Englebretson 2003; Yap 2007). The subject is thus typically introduced as a possessor as shown in (52a). Introducing the subject as a canonical nominative argument, as in (53b), is awkward, although not ungrammatical. (51) a.
betapa cantik=nya Ayu! excl beautiful=3s.gen Ayu ‘How beautiful Ayu is!’
b. ?betapa cantik Ayu! excl beautiful Ayu ‘How beautiful Ayu is!’
In all of these cases, adjectival predicates maintain their basic predicate form but adopt aspects of nominal syntax in taking genitive subjects instead of expected nominative ones. These languages thus follow the same general pattern of exclamative nominalization despite having lost the inherited morphological ma- ~ ka- alternation to express it.
2.5â•… Summary: Austronesian exclamatives The reconstruction of *ka- as an existential marker in PAn must still be regarded as tentative and more supporting data from Formosan languages is necessary to secure it.18 Assuming, however, that this reconstruction can be substantiated, the functional development of the affix in Figure 1 appears natural. The adjectival and stative functions consistently appear in those adjectives which can be construed as possessing the root, as opposed to coming about as the result of the root (cf. Himmelmann’s 1999 Type 1 and Type 2 adjectives in Tagalog) and are thus easily derived from an existential interpretation. From the adjectival function, the property exclamative is derived as the non-predicational variant of the adjective proper, which additionally contained actor voice morphology. Finally, from the property exclamative function, we find languages
.╅ At this point, projecting an existential meaning for *ka- at the level of PAn has been motivated by reconstruction of *ka- as a PMP existential affix and internal reconstruction from the other functions of *ka- attested in Formosan languages, especially stative/adjectival.
 Daniel Kaufman
such as Ivatan which have gone a step further and expanded the function of *ka- to marking exclamatives of any lexical category. EXISTENTIAL EXISTENTIAL
ADJ FORMATIVE
PROPERTY EXCLM
GENERAL EXCLM (Ivatan)
STATIVE
ADJECTIVAL *ma- (with AV)
Figure 1.╇ Proposed development of PAn *ka-
Having observed one type of nominalization, we now move on to our second topic: gerundive constructions and their use in temporal adjuncts.
3.â•… Gerundives In the second part of this paper we investigate a particular use of gerunds (eventnominalizations) as when-clauses in Philippine languages along with certain functional cognates in other Austronesian languages.
3.1â•… Morphology and syntax Gerundive is understood here as a general term subsuming both gerunds proper, i.e. event nominals, in addition to temporal adjuncts (Schachter & Otanes 1972: 160, 446) which, as will be seen shortly, differ slightly from event nominals in Tagalog. Gerundives are distinguished from voice marked forms in not assigning nominative case to any argument. Rather, all (non-directional) arguments are expressed in the genitive case (i.e. the ‘nominal’ pattern in Koptjevskaja-Tamm’s (1993) typology). This general case frame is exemplified by three languages in (53)–(55) (with Ilokano in (55) marking the object with oblique case). (52) Tagalog ang pag-da~dala nila nang asáwa ko sa Manílà nom ger-tr~bring 3p.gen gen spouse 1s.gen obl Manila ‘Their bringing of my wife to Manila’ (53) Sarangani Manobo peg-ibing te esawa ko doton te Davao ger-bring gen wife 1s.gen there gen Davao ‘The bringing of my wife to Davao’
(DuBois 1976: 94)
Exclamatives and temporal nominalizations in Austronesian 
(54) Ilokano …ti panag-subalit=na kadagiti surat=na ╇╛╛╛↜det ger-answer=3s.gen obl.pl letter=3s.gen ‘…his answering of her letters’
(Rubino 1997: 105)
In Tagalog, some gerunds (event nominals) are formed from stems by CV reduplication of the root. Derivational verbal morphology (e.g. paki- sociative, pa- causative, pang- distributive), but not voice morphology is included in gerund formations. In Table 1, we can see actor voice forms with three derivational prefixes compared with their corresponding gerunds. Table 1.╇ Actor voice forms with three derivational prefixes and corresponding gerunds
Actor voice infinitive Gerund
pagtr
pakisoc
pangdist
mag-húli ‘to catch’ pag-hu~húli ‘catching’
maki-húli ‘to catch with others’ paki~ki-húli ‘catching with others’
mang-húli ‘to catch many of ’ pang-hu~húli ‘catching many of ’
The initial nasal in the actor voice forms can be thought of as derived from the addition of the actor voice affix to a p- initial stem, e.g. + pag- à mag-. In all of the examples in Table 1, the gerunds are thus derived simply by the addition of the prefix plus reduplication of the first CV. Two major exceptions to this pattern are found in gerunds of verbs which do not take further verbal morphology in the actor voice, that is, plain actor voice verbs, and ma- verbs (glossed here as patient voice non-volitional). In these two cases, shown in Table 2, we find the addition of pag- even though it is absent in the corresponding actor voice form. The CV reduplication seen in Table 1 is also notably absent here. Table 2.╇ Addition of pag- to actor voice verbs, and ma- verbs
Actor voice infinitive Gerund
av
mapv.nvl
húli ‘to catch’ pag-húli ‘catching’
ma-húli ‘to get caught’ pag-ka-húli ‘getting caught’
 Daniel Kaufman
Note that, due to the differences in gerund formation between the two types of verbs it is impossible to isolate a single morphological exponent for the category of gerund in Tagalog; gerunds are formed sometimes by reduplication and sometimes by the addition of pag-.19 We return to this point below. In addition to forming event nominals, it seems that gerunds were also historically employed for introducing temporal adjuncts with the function of ‘when’ clauses. Evidence for this can be seen in (56) and (57) from Tagalog, and in (58) and (59) from Sarangani Manobo and Botolan Sambal, respectively. (55) Tagalog pag-pások nang pulis sa bangko… ger-enter gen police obl bank ‘When the police entered the bank …’ (56) Tagalog pag-bili ko nang isdà… ger-buy 1s.gen gen fish ‘When I bought a fish …’ (57) Sarangani Manobo Peg-dineg te amay din kenyan ger-hear gen father 3s.gen hat.obl ‘When his father heard that.’ (DuBois 1976: 94) (58) Botolan Sambal pama-ka-lengè nin arì ha halità-∆ nin gowardya… ger-nvl-hear gen king obl speak-pv gen guard ‘The king, upon hearing what the guard said.’ (Antworth 1979: 105)
Although it may have been the case historically that one morphological paradigm was employed both for event nominals and temporal adjuncts, these two functions have taken on separate lives in certain languages. As shown above in Table 1 and Table 2, CV reduplication is required for the gerund forms of all Tagalog verbs except those with plain actor voice and stative ma-. Temporal adjuncts, on the other hand, never employ reduplication regardless of the verb form. This can be seen by comparing the when-clause in (60a) with the gerund in (60b). The root bukas ‘open’ takes the pagprefix in the actor voice and therefore requires reduplication in the gerund but does not take this reduplication as a temporal adjunct. The presence of reduplication thus
.╅ We adopt a glossing convention here for Tagalog by which pag- is glossed as the gerund morpheme and cv- reduplication is marked as a transitivity related morpheme but nothing crucial hinges upon this choice of interpretations.
Exclamatives and temporal nominalizations in Austronesian 
serves to distinguish temporal adjuncts from event nominal gerunds with verbs which take pag- in the actor voice. (As above, the pag- prefix continues to be glossed with the more general label gerundive for both of these functions.) (59) a.
pag-bukas ko nang isa=ng pintò… ger-open 1s.gen gen one=lnk door ‘When I opened one door…’
b. ang pag-bu~bukas ko nang isa=ng pintò nom ger-tr~open 1s.gen gen one=lnk door ‘my opening of one door’
The lack of correspondence between the gerund proper and the temporal adjunct form appears to be a consequence of the degrammaticalization of pag- from a gerund marker in an earlier stage of Tagalog to an independent temporal subordinator. Recall from Table 2 that pag- turns up in two common Tagalog paradigms where it is absent in the finite verb form. If true gerunds were historically used for introducing temporal adjuncts, then it is clear how the pag- prefix – with its wider than expected distribution in gerunds, appearing in the and ma- paradigms – could have been reinterpreted as a temporal subordinator itself. A more obvious consequence of the degrammaticalization of pag- is its separability from the verb in its temporal adjunct function, but not its gerund function.20 This can be seen with the addition of adverbial material such as negation as in (61a). Note, furthermore, that the verb is not a gerund at all but is rather inflected as a regular actor voice predicate and assigns nominative case to its subject. Splitting the prefix from its host in a real gerund is ungrammatical, as seen in the comparison between (61b) and (61c). (60) a.
pag hindì ka pások… temp neg 2s.nom enter ‘When you don’t enter…’
b. *pag hindì mo pások ger neg 2s.gen enter c.
ang hindì mo pag-pasok nom neg 2s.gen ger-enter ‘your not entering’
Interestingly, however, pag in temporal adjuncts cannot always be treated as an independent subordinator as there also exist constructions in which it clearly functions
.╅ Note that in its degrammaticalized manifestation, pag is glossed here as a temporal subordinator.
 Daniel Kaufman
as a bound prefix, just as in true gerunds. This is seen in “immediate when-clauses”, as shown in (62), where the gerundive is reduplicated with the corresponding ‘just as x happened’ reading. It is clear that pag- forms a single word with the root in such constructions as only words can be subject to reduplication with the linker. (61) pag-dating na pag-dating ni Mario ger-arrive lnk ger-arrive p.gen M. ‘As soon as Mario arrived’
In connection to the previous discussion of split properties in exclamatives, it is also worth noting that the reduplication in (62) is characteristic of property predicates but not verbal ones. The corresponding construction with a non-gerundive verb form is ungrammatical, as shown in (63). (62) *dating na dating si Mario arrive lnk arrive p.nom M.
Note also that if this reduplication is analyzed on par with the intensive reduplication of the same form applying to property predicates, we can derive its meaning by having the intensive function applying to the temporal semantics itself, i.e. ‘when’ + intensive = ‘just as’. We saw in this section that although temporal adjuncts probably developed from gerunds in Philippine languages, some divergences between the two constructions exist in present day Tagalog. In particular, the pag- prefix has been degrammaticalized into a temporal subordinator with the consequence that an associated transitive distinction expressed by reduplication in the gerund is lost in the temporal adjunct and external elements can now intervene between temporal subordinator pag and its complement. In the next section we turn our attention to several Austronesian languages outside the Philippines in order to observe nominal properties in temporal adjuncts which lack morphological signs of nominalization.
3.2â•… Split nominal features in temporal adjuncts Just as we saw for exclamatives, temporal adjuncts also display split nominal features in a wide range of Austronesian languages. In particular, we commonly find forms retaining certain verbal properties but expressing their subjects in the genitive case. In Indonesian/Malay, the temporal adjunct is formed with the prefix se- and the 3rd sg. genitive pronoun/determiner -nya, as in (64a), although the external argument can also be expressed directly on the adjunct as a genitive pronoun, as shown in (64b) (but this usage is somewhat antiquated). The cognate construction in Wolio is shown in (65).
Exclamatives and temporal nominalizations in Austronesian 
(63) Indonesian/Malay a.
se-tiba=nya aku b. se-tiba=ku temp-arrive=3s.gen 1s temp-arrive=1s.gen ‘When I arrive.’ ‘When I arrive.’
(64) Wolio sa-tuwu-na o bulu-na… temp-grow-3.gen det feather-3.gen ‘when the feathers (of his wings) had grown…’
(Anceaux 1988: 55)
The Indonesian/Malay se- -nya temporal adjunct appears to be restricted to verbs as opposed to property predicates, but this is by no means a particularly widespread restriction cross-linguistically.21 As shown by Mead (2006), the cognate construction in Mori Bawah also allows adjectives, as seen in (66). (65) Mori Bawah Sa mokula-no wua m-petiba andio… when hot-3s.gen fruit LG-winged.bean this ‘When the winged bean seeds were hot …’
(Mead 2006: 13)
All the above examples show the nominal property of expressing their subjects as possessors but these forms cannot be treated as regular nominalizations. For one, they are sensitive to a transitivity distinction which is not present in other nominalizations. While gerunds can be formed from both transitive and intransitive verbs in Indonesian/Malay, transitive verbs are impossible in temporal adjuncts. Compare the ungrammatical temporal adjunct of the verb dengar ‘to hear’ in (67a) with its perfectly acceptable gerund in (67b). (66) Indonesian/Malay a. *Se-dengar-nya… b. pen-dengar-an-nya temp-hear-3s.gen ger-hear-ger-3s.gen ‘his/her hearing’
.â•… As Foong Ha Yap (p.c.) reminds me, se- takes on a different meaning with adjectival roots in Indonesian/Malay, namely, ‘as x as’, e.g. se-panas-panas-nya ‘as hot as’, as seen in (i). Se- also receives an ‘as’ interpretation with a small number of verbs such as tahu ‘know’ in the construction shown in (ii). (i)
se-panas-panas-nya hari itu… (ii) as-hot~redup-3s.gen day that ‘As hot as it was that day…’
se-tahu aku… as-know 1sg ‘As far as I know…’
 Daniel Kaufman
Transitive verbs in the temporal adjunct construction in Mori Bawah also appear to require a different form. Although Mead is not explicit about this, he offers an example of a transitive temporal adjunct in (68), in which the genitive pronoun is attached directly to the temporal marker and the verb appears in the actor voice (glossed by Mead as participle). (This is strikingly similar to the degrammaticalization of Tagalog pag- observed above, where the verb also appears in its voice marked form.)22 (67) Mori Bawah sa-no ronge-o mia andio motae… when-3s.gen <part>hear-3s.abs person this that ‘When this person heard that…’
(Mead 2006: 14)
Split properties of temporal adjuncts are nowhere more visible than in the South Sulawesi languages. Just as we saw earlier with the Bugis exclamatives in (48) and (47), temporal adjuncts in South Sulawesi languages often express the subject in the genitive without showing any nominalizing morphology on the verb. Transitive verbs in all South Sulawesi languages (except those of the Seko subgroup) are distinguished by ergative prefixes, historically derived from genitive enclitics, and absolutive second position clitics, historically derived from nominative pronouns. Under the proposal sketched out in Section 2.2 above, temporal adjuncts resemble high nominalizations in maintaining the ergative prefix from the verbal form as is. If nominalization is conceived of as applying at a particular point within the phrase structure, the verb can pick up its nominal properties above the point at which ergative case is assigned but below the point at which absolutive case is assigned. In this way, nominalization effects the type of case assigned to transitive patients but not to transitive agents (see Ntelitheos 2006; Yanagida & Whitman 2008 and references therein for details). This is seen in the comparison of the Bugis temporal adjunct with its corresponding predication in (69a) and (b), respectively. An example from the related Mamasa language is shown in (70), where the logical subject of the first clause (highlighted) is again expressed as a genitive.
.╅ Another morphological form for expressing temporal adjuncts in Mori Bawah employs a reflex of PAn *ka-, as seen earlier with Tagalog. Unlike Tagalog, a genitive pronominal in this construction can also express patients. In (i), the genitive marked pronoun refers to the patient while in the second verb it refers to the intransitive subject. (i)
Mori Bawah Ko-tidu-ku, ko-tebangku-ku. temp-punch-1s.gen temp-fall.over-1s.gen ‘Immediately I was punched, I fell over.’
(Mead 2006: 20)
Exclamatives and temporal nominalizations in Austronesian 
(68) Bugis a.
mu-tikkәn-na 2s.erg-seize-3s.gen ‘when/because you seized him’
b. mu-tikkәn=i 2s.erg-seize=3s.abs ‘you seized him’ (Sirk 1996: 94) (69) Mamasa na-lambi′-na ade’ na-anda′-i bakku′-na 3.erg-find-3.gen rprt 3.erg-pick.up-loc backpack-3s.gen ‘When she reached him, she picked up his knapsack’ (Matti 1994: 80)
The intuition that nominalization is taking place here at an intermediate point in the clause structure is strongly supported by data from the geographically proximate Rampi language (Pamona-Kaili, Central Sulawesi). Unlike the South Sulawesi languages, Rampi indicates temporal adjuncts morphologically in addition to the changes in case assignment. Rampi temporal adjuncts employ a reflex of PAn *ka- (ko-, glossed here as nmlz in accordance with Friberg 1990) which is crucially positioned outside of ergative agreement as shown in (71). This is the expected position if nominalization takes place above ergative case assignment. (70) Rampi Ko=no=oki’=na=lu datu wulehu’ to mahi=mo… nmlz=3s.erg=look=3s.gen=pst king mouse rel die=cmp ‘The king seeing the already dead mouse…’ (Friberg 1990: 61)
While having a derivational affix external to agreement marking may be surprising from a typological perspective on affix ordering (cf. Bybee 1985), it is expected given the possibility of high nominalization, which yields split verbal-nominal properties. This suggests a basic structure of the form: abs/gen > (nmlz) > erg > root, in which the choice of genitive or absolutive marking is governed by the presence or absence of the nominalizer. To sum up this section, we have seen that temporal adjuncts, more specifically, when-clauses, are formed by nominalization in a variety of Austronesian languages. In Philippine languages, this appears to have been done on the basis of the gerund. In Tagalog, a prefix common in gerund formation has been reanalyzed as a temporal subordinator via a process of de-grammaticalization. While this has led to finite verb forms being licensed in subordinate when-clauses, we also witnessed evidence from intensive reduplication that true gerunds still play a role in this area. Outside of the Philippines we observed several examples of split nominal properties in when-clauses.
 Daniel Kaufman
In the case of the South Sulawesi languages and Rampi, it was suggested that these constructions could be fruitfully analyzed as high nominalizations, on par with the exclamatives discussed earlier.
4.â•… Conclusion A priori, we would be at odds to find semantic commonalities between such disparate phenomena as exclamatives and when-clauses. Yet, we have seen considerable evidence here from Austronesian languages that both phenomena are typically expressed via nominalization. This could either be a family particular fluke or there could be a cognitive basis for treating these phenomena in a structurally similar fashion. It has been argued here that there does exist such a basis and that this basis finds an analogy in anaphoric relations. Returning to examples (5) and (6), repeated here as (72) and (73), both when-clauses and exclamatives refer back to a predicate (whether overt or not) from which they derive their reference. (71) Datingi si Pedroj. Pag-datingi niyaj … arrive p.nom p. ger-arrive 3s.gen ‘Pedro arrived. When he arrived…’ (More lit. ‘Upon his arrival…’) (72) Ma-bilisi si Pedroj. Kay bilisi niyaj talaga! adj-speed p.nom p. ext speed 3s.gen truly ‘Pedro is fast. He’s so fast!’ (More lit. ‘His speed!’)
As anaphoric elements, both when-clauses and exclamatives are most appropriately expressed as nominals. Nominals, although by no means requiring presuppositionality, have been argued to inherently possess referential properties either by virtue of their lexical category (Baker 2003) or due to their proto-typical functions in discourse (Croft 1991). They are thus uniquely suited for anaphoric functions. Following this line of thought, we can now explain why Philippine gerunds are interpreted specifically as when-clauses rather than hypotheticals (i.e. if-clauses) or as indicating simultaneous action (i.e. as-clauses).23 The answer is that when-clauses are semantically distinct from these other types of adjuncts in being presuppositional, i.e. requiring a previously introduced or pragmatically bridged referent, as originally argued by Heinämäki (1978). Hypotheticals on the other hand are quite the opposite, typically
.╅ Foong Ha Yap (p.c.) brings to my attention the fact that nominalization for marking conditional clauses has been reported for Old Chinese (Yap & Wang this volume). Further typological research should be able to uncover just how widespread a pattern this is.
Exclamatives and temporal nominalizations in Austronesian 
disallowing presuppositional complements. We can see this quite clearly in the incompability of Tagalog kay exclamatives in conditional sentences. In (74a), we see a property predicate felicitously embedded under the conditional and in (74b) we see the ungrammatical result of embedding the kay exclamative under the same operator.24,25 (73) a.
Kung ma-ganda si Maria… if adj-beauty p.nom M. ‘If Maria is beautiful…’
b. *Kung kay ganda ni Maria… if ext beauty p.gen M.(For, ‘If Maria is so beautiful…’)
The presuppositionality inherent in pag- gerunds has also been retained in the degrammaticalized temporal subordinator, pag. This is clear when we compare the minimal pair in (75). In the first member of the pair, (75a), we find a typical conditional reading; we are being told that Maria is the designated substitute without any implication that Maria has ever replaced the speaker. In contrast, in (75b) with degrammaticalized pag, we additionally find a strong implication that Maria has already served as the speaker’s replacement on at least one occasion, a difference which is captured well by the use of ‘if ’ versus ‘when’ in the translations. (74) a.
Kung walà ako ríto, si Maria ang kapalit ko if neg.ext 1s.nom here p.nom M. nom replacement 1s.gen ‘If I’m not here, Maria is my replacement.’
b. Pag walà ako ríto, si Maria ang kapalit ko temp neg.ext 1s.nom here p.nom M. nom replacement 1s.gen ‘When I’m not here, Maria is my replacement.’
.â•… According to the glossing here of kay as a “bare existential”, this may appear strange, as existentials are understood to introduce a variable into the discourse and are thus inherently non-presuppositional. It appears however that this characterization only applies to existential predicates. Crucially, Philippine languages appear to make a distinction between predicational and non-predicational existentials: the former are produced with the existential marker plus voice while the latter simply involve the existential marker, hence the term “bare existential”. .â•… Note that a simple syntactic account which claims that the exclamative marker and the conditional operator occupy the same position and are thus in complementary distribution – a story which could account for the ungrammaticality of English (i) – cannot be correct for Tagalog, as kay exclamatives have no connection to the left-periphery and are rather positioned similarly to non-exclamative adjectives, as shown in (ii). (i) *‘If how beautiful Maria is…’ (ii)
umaga=ng kay ganda morning=lnk ext beauty ‘a morning so beautiful’
 Daniel Kaufman
Presuppositionality can thus be seen as the common semantic denominator in both constructions. Nominal type temporal adjuncts are also presupposed in that they always refer back to a predication, as seen in (72). Nominal type exclamatives derive their illocutionary force in part by compelling the hearer to accommodate a presupposition, i.e. corresponding to the adjectival predication in (73) preceding the exclamative, which in this atypical example, happens to be overt. As a final note, it should be mentioned that nominalization appears to be a necessary but not sufficient condition for the type of presuppositionality inherent in the when-clauses and exclamatives examined here. Event nominals can, of course, also have non-presuppositional uses, as in English (76) and Tagalog (77).26 (75) A killing could happen at any time here. (76) Kung gusto nila nang pag-na~nákaw, na~nakáw-an din sila. if like 3p.gen gen tr-ger~steal incm~steal-lv also 3p.nom ‘If they like stealing/thievery, they will also be stolen from.’
Although there does appear a tendency in discourse for nominalizations to be employed in presuppositional and definite contexts, it is perhaps only in their grammaticalization with more specific functions that this tendency becomes categorical.
References Adelaar, Alexander and Himmelmann, Nikolaus P. (eds). 2005. The Austronesian Languages of Asia and Madagascar. London: Routledge. Alexiadou, Artemis. 2001. Functional Structure in Nominals: Nominalization and Ergativity. Amsterdam: John Benjamins. de Amurrio, Fidel. 1970. Pangasinan Grammar. Bugallon, Pangasinan: Fidel de Amurrio. Anceaux, Johannes.C. 1988. The Wolio Language. Dordrecht: Foris. Antworth, E.L. 1979. A Grammatical Sketch of Botolan Sambal. SIL: Manila. Blust, Robert. 1993. Central and Central-Eastern Malayo-Polynesian. Oceanic Linguistics 32(2): 241–293. Blust, Robert. 1995. Austronesian comparative dictionary. Computer files. Honolulu: University of Hawai’i. Blust, Robert. 2003. Three notes on early Austronesian morphology. Oceanic Linguistics 42(2): 438–478. Cena, R. 1979. Tagalog counter-examples to the Accessibility Hierarchy. Studies in Philippine Linguistics 3(1): 119–124. Donohue, Mark. 1999. A Grammar of Tukang Besi. Berlin: Mouton de Gruyter.
.╅ Thanks to an anonymous reviewer for emphasizing the importance of this point.
Exclamatives and temporal nominalizations in Austronesian 
DuBois, Carl D. 1976. Sarangani Manobo. Manila: SIL. Englebretson, Robert. 2003. Searching for Structure: The Problem of Complementation in Colloquial Indonesian conversation. (Studies in Discourse and Grammar 13). Amsterdam: John Benjamins. Evans, Bethwyn and Ross, Malcolm. 2001. The history of Proto-Oceanic *ma-. Oceanic Linguistics 40(2): 269–290. Friberg, Barbara. 1990. Sulawesi Language Texts. Language Data Asia-Pacific Series No. 15. Dallas: SIL. Foley, William A. and Van Valin Jr., Robert D. 1984. Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Heinämäki, Orvokki. 1978. Semantics of English Temporal Connectives. University of Indiana: IULC. Hidalgo, Cesar A. and Hidalgo Araceli C. 1971. A Tagmemic Grammar of Ivatan. Manila: Linguistic Society of the Philippines. Himmelmann, Nikolaus P. 2008. Lexical categories and voice in Tagalog. In Voice and Grammatical Functions in Austronesian Languages, Simon Musgrave & Peter K. Austin (eds), 247–293. Stanford: CSLI Publications. Himmelmann, Nikolaus P. 2004. Tagalog. In Morphology: A Handbook on Inflection and Word Formation. Gert Booij, Christian Lehmann and Joachim Mugdan (eds), Vol. 2: 1473–1490. Berlin: de Gruyter. Kaufman, Daniel. (in progress). Aspect, individuation and lexical category in Tagalog. ms. Kaufman, Daniel. 2007. The nominalist hypothesis in Austronesian. Presentation at the Workshop on Beyond Focus and Ergativity: Towards a More Comprehensive View of Austronesian Morphosyntax, Zentrum für allegemeine Sprachwissenschaft, Humboldt-Universität Berlin, September 13–15. Kaufman, Daniel. 2008. Austronesian voice as thematic nominalization. Paper presented at the 15th Meeting of the Austronesian Formal Linguistics Association (AFLA-15), University of Sydney, June 30-July 2. Koptjevskaja-Tamm, Maria. 1993. Nominalizations. London: Routledge. Kroeger, Paul. 2005. Kimaragang. In The Austronesian Languages of Asia and Madagascar, Alexander K. Alexander & Nikolaus P. Himmelmann (eds), 397–415. London: Routledge. Malchukov, Andrej. 2004. Nominalization, Verbalization: Constraining a Typology of Transcategorial Operations (Lincom Studies in Language Typology 8). München: Lincom Europa. Matti, David F. 1994. Mamasa pronoun sets. NUSA 36: 65–88. Mead, David. 2003. Evidence for a Celebic supergroup In Issues in Austronesian Historical Phonology, John Lynch (ed), 115–141. Pacific Linguistics 550. Canberra: Pacific Linguistics. Mead, David. 2006. When to use a genitive pronoun in Mori Bawah. Paper presented at the Tenth International Conference on Austronesian Linguistics, Linguistic Society of the Philippines and SIL International, Palawan, Philippines, January 17–20. http://www.sil.org/asia/philippines/ical/papers.html Michaelis, Laura A. and Lambrecht, Knud. 1996. The exclamative sentence type in English. In Conceptual Structure, Discourse and Language, Adele Goldberg (ed), 375–389. Stanford: CSLI. Michaelis, Laura A. 2001. Exclamative constructions. In Language Typology and Language Universals, Martin Haspelmath, Ekkehard Koenig, Wulf Oesterricher & Wolfgang Raible (eds), 1038–1050. Berlin: Mouton de Gruyter.
 Daniel Kaufman Naylor, Paz B. 1979. Linking, relation-marking, and Tagalog syntax. In Austronesian Studies: Papers from the Second Eastern Conference on Austronesian Languages, Paz Buenaventura Naylor (ed), 33–50. Michigan Papers on South and Southeast Asia No. 15. Ann Arbor, Michigan: Center for South and Southeast Asian Sudies, University of Michigan. Naylor, Paz B. 2005. On the stative predicate: Tagalog “existentials” revisited. In Current Issues in Philippine Linguistics and Anthropology, Hsiu-chuan Liao & Carl R. Galvez Rubino (eds), 419–435. Manila: Linguistic Society of the Philippines and SIL. Noonan, Michael. 1997. Versatile nominalizations. In Essays on Language Function and Language Type. Dedicated to T. Givón, Joan Bybee, John Haiman & Sarah Thompson (eds), 373–394. Amsterdam: John Benjamins. Ntelitheos, Dimitrios. 2006. The Morphosyntax of Nominalizations: A Case Study. Ph.D. dissertation, University of California, Los Angeles. Ogawa, Naoyoshi. 2003. English-Favorlang vocabulary. In English-Favorlang Vocabulary by Ogawa, Paul Li (ed). Tokyo: Research Institute for Languages and Cultures of Asia and Africa, Asia-African Lexicon Series 43. Pittmann, Richard. 1966. Tagalog -um- and mag-: An Interim Report. Papers in Philippine Linguistics 1: 9–20 Canberra: Pacific Linguistics, Series A8. Prentice, D.J. 1971. The Murut Languages of Sabah. Canberra: Pacific Linguistics. Quick, Phil A. 2003. A Grammar of the Pendau Language. Ph.D. dissertation. Australian National University. Reid, Lawrence A. 1966. An Ivatan Syntax. Honolulu: University of Hawai’i Press. Reid, Lawrence A. 1978. Problems in the reconstruction of Proto-Philippine construction markers. In Second International Conference on Austronesian Linguistics: Proceedings, Fascicle I, Western Austronesian, S.A. Wurm & Lois Carrington (eds), 33–66. Canberra: Pacific Linguistics. Ross, Malcom D. 1995. Reconstructing Proto-Austronesian verbal morphology: Evidence from Taiwan. In Austronesian Studies Relating to Taiwan, Paul Jen-kuei Li, Cheng-hwa Tsang, Ying-kuei Huang, Dah-an Ho & Chiu-yu Tseng (eds), 727–791. Symposium Series of the Institute of History and Philology, Academia Sinica, Number 3. Taipei: Academia Sinica. Ross, Malcolm. 2002. The history and transitivity of Western Austronesian voice and voicemarking. In The History and Typology of Western Austronesian Voice Systems, Fay Wouk & Malcolm Ross (eds), 17–62. Canberra: Pacific Linguistics. Rubino, Carl. 1997. A Reference Grammar of Ilokano. Ph.D. dissertation, University of California, Santa Barbara. Rubino, Carl. 2005. Utudnon, an undescribed language of Leyte. In Current Issues in Philippine Linguistics and Anthropology, Hsiu-chuan Liao & Carl R. Galvez Rubino (eds), 306–337. Manila: Linguistic Society of the Philippines and SIL. Sadock, Jerrold M. and Zwicky, Arnold M. 1985. Speech act distinctions in syntax. In Language Typology and Syntactic Description, Voume 1, Timothy Shopen (ed), 155–196. Cambridge: Cambridge University Press. Schachter, Paul and Otanes, Fe T. 1972. Tagalog Reference Grammar. Berkeley: University of California Press. Sirk, Ü.1996. The Buginese Language. Moscow: Nauka. Sityar, Emily. 2000. The topic and y indefinite in Cebuano. In Formal Issues in Austronesian Linguistics, Ileana Paul, Vivianne Phillips & Lisa deMena Travis (eds), 145–166. Studies in Natural Language and Linguistic Theory. Dordrecht: Springer.
Exclamatives and temporal nominalizations in Austronesian 
Starosta, Stanley, Pawley, Andrew and Reid, Lawrence A. 1982. The evolution of focus in Austronesian. In Papers from the Third International Conference on Austronesian Linguistics, Volume 2: Tracking the Travelers, Amran Halim, Lois Carrington & S.A. Wurm (eds), 145–170. Canberra: Pacific Linguistics. (Pacific Linguistics C-75). Svelmoe, Gordon and Svelmoe, Thelma. 1974. Notes on Mansaka grammar. Language Data, Asian-Pacific Series No. 6. Huntington Beach, California: SIL. Svelmoe, Gordon and Svelmoe, Thelma. 1990. Mansaka Dictionary. Dallas: SIL. Usup, H.T., Rompas, H., Kuhon, J., Ny, S.V. Moningkey-Rumambi, M.M., Toding Datu, A.B.G. Rattu. 1981. Morfologi dan Sintaksis Bahasa Bolaang Mongondow. Jakarta: Pusat Pembinaan dan Pengembangan Bahasa Departmen Pendidikan dan Kebudayaan. van den Berg, René. 1989. A Grammar of the Muna Language. Dordrecht: Foris. Wolff, John. 1962. A Description of Cebuano Visayan: Morphology. Ithaca: Division of Modern Languages, Cornell Univ. Wolff, John. 1973. Verbal inflection in Proto-Austronesian. In Parangal kay Cecilio Lopez Andrew Gonzalez (ed), 71–91. Quezon City: Linguistic Society of the Philippines. Wolff, J, Centeno, M.T.C. and Rau, D.V. 1991. Pilipino through Self-Instruction. 4 vols. Ithaca: Cornell Southeast Asia Program. Yanagida, Yuko and Whitman, John. 2008. Alignment and word order in Old (8th century) Japanese. ms. University of Tsukuba and Cornell University. Yap, Foong Ha. 2007. What punya can do, that yang and -nya cannot: An analysis of three nominalizers in Malay. Paper presented at the Workshop on Beyond ‘Focus’ and Ergativity: Towards a More Comprehensive View of Austronesian Morphosyntax. Zentrum für Allegemeine Sprachwissenschaft, September 13–15. Zeitoun, Elizabeth and Teng, Stacy F. 2009. From ki-N ‘get N’ in Formosan languages to ki-V ‘get V-ed’ (passive) in Rukai, Paiwan and Puyuma. In Discovering History Through Language: Papers in Honour of Malcom Ross, Bethwyn Evans (ed.), 479–500. Canberra: Pacific Linguistics. Zorc, R. David. 1975. The Bisayan Dialects of the Philippines: Subgrouping and Reconstruction. Ph.D. dissertation. Cornell University.
 Daniel Kaufman
Appendix 1.╇ Austronesian languages
Language Key: Bolaang Mongondow 1, Botolan Sambal 2, Bugis 3, Cebuano 4, Favorlang 5, Ilokano 6, Ivatan 7, Kimaragang 8, Mamasa 9, Mansaka 10, Mori Bawah 11, Muna 12, Murut 13, Pangasinan 14, Pendau 15, Rampi 16, Sarangani Manobo 17, Tukang Besi 18, Wolio 19 National Languages: Tagalog (Philippines), Malay (Malaysia, Brunei), Indonesian (Indonesia)
part vi
Papuan languages
Discourse-structuring functions of Abui demonstratives* František Kratochvíl
Nanyang Technological University In a number of East and South-East Asian languages, certain grammatical elements such as pronouns, generic nouns, or demonstratives (e.g. one, thing, this) have acquired additional pragmatic functions. Well-documented examples of this grammaticalization process are the Mandarin de, the Malay punya/nya/mia and the Japanese no (cf. Yap, Matthews et al. 2004); the grammaticalized element occurs in the sentence-final position encoding speaker’s certainty about the proposition. A similar development has taken place in Abui (a Papuan language of Eastern Indonesia); markers describing speaker’s attitude towards a proposition (evidentiality and assertion) are recruited from two sources: (i) demonstratives and (ii) the utterance verb ba ‘say’.1
1.â•… Introduction This paper gives an overview of uses of Abui demonstratives and of the marker ba. It attempts to clarify the grammaticalization into pragmatic markers, exemplified in (1). (1) a.
di me do 3a come prx ‘he is coming/came/will come (as can be seen)’
*Acknowledgements: The data (naturally occurring and elicited) presented in this paper has been collected by the author during fieldwork in Alor between 2003–2008. Benidiktus Delpada and Waksi Maufani assisted with transcribing, translating and analysing the data. The research on Abui was supported by Leiden University through a grant from the Dutch Science Council (NWO), and by La Trobe University. We thankfully acknowledge the valuable input we have received in various stages from the editors of this volume, three anonymous reviewers, Joanna Sio, Robert Borsley, Ger Reesink, Marian Klamer and the members of RCLT (La Trobe University). .â•… ‘Papuan’ is a convenience term used to refer to non-Austronesian languages of New Guinea and neighboring islands. It does not necessarily imply genetic affiliation among various ‘Papuan’ languages.
 František Kratochvíl
b. di me nu 3a come spc ‘he came some time ago/will come once’ c.
di me ba 3a come say ‘(is is said/they say that) he is coming/came/will come’
All three constructions in (1) have the same intonation contour with the pitch rising on the penultimate syllable and dropping on the final syllable. Depending on the context, the sentence-final markers (underlined) encode tense, evidentiality, or assertion. As the glossing indicates, the sentence-final markers were recruited from demonstratives and from the speech report verb ba ‘say’, respectively. The proximal demonstrative do (prx) in (1a) may indicate the present or very recent occurrence of the event, first hand evidence, or speaker’s assertion of the proposition. The distal demonstrative nu (spc) in (1b) marks either distal temporal location of the event, or remote but reliable evidence. The marker ba (say) in (1c) indicates reported information that the speaker does not want to vouch for. In Malay, Mandarin, and Japanese, a single form was grammaticalized in a variety of functions that each deserves an autonomous treatment. The same approach is problematic for Abui data. Not a single item, but the entire demonstrative paradigm (six forms) has acquired additional functions such as markers of tense, mood, and evidentiality or as subordinators and clause linkers. It has been pointed out that grammaticalization of open-class etyma may result in heterosemy.2 The present data shows that closed-class etyma, such as demonstratives may grammaticalize into grammatical and pragmatic markers. Although the grammaticalization is motivated by a perceived similarity between two meanings, its outcome is not predictable: markers originating from other sources can be included in the new paradigm. The underlying metaphors of the tense and evidential paradigms (space → time, space → source.distance) do not motivate the grammaticalization of the marker ba which originates in the verb ‘say’. The diverse origin of the markers in the grammatical categories under discussion suggests that each category possesses a unique set of values to be marked; the set of values in each category is not necessarily restricted by any of the underlying metaphors (such as space → x).
.╅ The term heterosemy refers to cases in which two or more historically related (derived from the common source) meanings/functions belong in different morphosyntactic categories (cf. Lichtenberk 1991: 476).
Discourse-structuring functions of Abui demonstratives 
In addition to the sentence-final uses, the demonstratives and ba are also used as subordinating markers, marking adverbial and noun-modifying clauses. The subordinating function of the demonstratives is parallel to the function of nominalizations in other languages, except that the entire paradigm is used. A short note is in place about the glossing. As demonstratives belong to a closed class, that does not correspond to the English demonstratives, we use the glosses prx, md and dst for the proximal, medial and distal forms respectively. The gloss of addressee-based forms includes an additional ad. The same glosses are used also for the grammaticalized forms. The form ba is glossed as ‘say’ in its original verbal meaning. The grammaticalized forms of ba are glossed as say (for evidential), top (for topic), qt (quotative), sim (simultaneous linker), etc. The paper is structured as follows: Section 2 provides a brief sociolinguistic and typological profile of Abui. Taking into account the different origins of the sentence-final markers presented in (1), demonstratives and ba are discussed separately. Section 3 discusses the various functions of Abui demonstratives. In the noun phrase, demonstratives indicate spatial (Section 3.1.1) or discourse location (Section 3.1.2) of the referent. Demonstratives also mark the noun-modifying clauses (Section 3.1.3). In the clausal, demonstratives indicate spatial or temporal location of an event (Section 3.2). Finally, in the sentence, demonstratives serve as subordinators, marking adverbial clauses or as evidentiality and assertion markers in the sentence-final position (Section 3.3).3 Section 4 discusses the functions of the marker ba. Firstly, ba marks topical noun phrases (Section 4.1.1), or links noun-modifying clauses (parallel to relative clauses in other languages, Section 4.1.2). In the clause, ba marks speech reports/quotations (Section 4.2.1), and links complement clauses (Section 4.2.2). In complex sentences ba links clauses describing simultaneous events (Section 4.3.1), purpose clauses (Section 4.3.2), and encodes reported evidentiality in the sentence-final position (Section 4.3.3).
2.â•… Abui and its speakers Abui is a non-Austronesian (Papuan) language spoken by about 16.000 people in the central valley and surrounding mountains of the Alor Island in Eastern Indonesia. Abui together with another two dozen Papuan languages in the Timor-Alor-Pantar
.╅ The term clause is used in this paper to refer to a syntactic unit that is headed by a verb phrase that is inflected for aspect and tense. The sentence on the other hand is a syntactic unit that may consist of multiple clauses. The evidential inflections are associated with the sentence, appearing on the final clause.
 František Kratochvíl
area are claimed to be members of the Trans-New Guinea family.4 A linguistic description of the northern Abui dialects is available (Stokhof 1984; Kratochvíl 2007). Alorese, Alor Malay, and Indonesian are the only Austronesian languages spoken in the area.
2.1â•… Linguistic situation Most Abui speakers are nowadays bilingual. Monolingual speakers are found only among the oldest generation. Alor Malay (the local lingua franca) is the target of a language shift; the youngest generation is learning Malay as their first language. The language shift has a larger impact in the coastal areas where many children are raised in Alor Malay and have only passive knowledge of Abui at first. In the traditional mountainous settlements and in isolated areas the language shift has lesser impact.
2.2â•… Typological profile Abui has a simple phonemic inventory with 16 native and 3 loan consonants. There are 5 short vowels each of them having a long counterpart. In a number of cases lexical tone is found. Abui is a head marking language with a semantic alignment.5 Controlling and volitional participants are realized as the a argument in both transitive and intransitive construction and are expressed with nps and free pronouns. The affected participants are realized as the u argument, expressed by nps and pronominal prefixes on the verb. There are three types of pronominal prefixes distinguishing the following types of u arguments: patients (pat), recipients or goals (rec), and benefactives or locations (loc). Verbal morphology is complex and further includes aspect inflection. Verb compounding and serialization are heavily relied on. The nominal morphology is restricted to marking of the possessor. Inflections for number, case or gender do not appear. Abui syntax is characterized by strict constituent order. The nominal modifiers (mod) follow the head noun (n) with the exception of deictic demonstratives (demS) and possessors (poss), as illustrated in (2):
.â•… Pawley (2001) and Ross (2005) group Timor-Alor-Pantar languages with the Trans-New Guinea family. The affiliation and internal subgrouping of the TAP languages are currently under investigation (cf. Holton, Klamer & Kratochvíl 2009). .â•… In semantic alignment, the semantic features of the participants drive the argument realization and are directly reflected in the morphological forms chosen. This type of alignment is sometimes referred to as fluid-S system.
Discourse-structuring functions of Abui demonstratives 
(2) np: dems/nmcs (poss-) n n/adj/v/quant ba + nmc dema
The deictic demonstrative indicates the spatial location of the referent and together with the possessor marking precede the head (n). Adjectives (adj), stative verbs (v) and quantifiers (quant) follow the head. The final constituent of an NP is usually an anaphoric demonstrative (demA) that indicates the ‘discourse location’ of the referent. Noun-modifying clauses (nmc) normally occur following the head linked with ba.6 However, a nmc elaborating on the location of the referent (nmcS) occurs in the same position as the deictic demonstrative, preceding the head noun. The clause template in (3) shows that the arguments precede the predicate complex. The topical arguments can be left-dislocated. The predicate complex however always remains in clause final position.
(3) clause: adv np proA adv/demS npU vp asp neg demt
Note that the deictic demonstrative (dems) indicating the spatial location of the event always precedes the predicate. The demonstrative (demt) indicating the temporal location of an event is the final clause constituent. In a complex sentence, the main clause (mc) may contain marking of tense, aspect and mood. In subordinate clauses (sc), the marking of tense, aspect and mood is reduced. The position of a sc with respect to the mc is determined by its semantic type. Adverbial scs precede the mc. scs expressing non-factive complements or purpose follow the mc. Abui discourse relies on clause chains in which only the final clause is fully inflected. In narratives, strategies such as tail-head linkage are relied on. More details can be found in Kratochvíl (2007).
3.â•… Functions of Abui demonstratives This section discusses functions of Abui demonstratives in distinct syntactic positions in the NP, clause and (complex) sentence. In each of these positions, Abui demonstratives encode a distinct grammatical category. Section 3.1 focuses on the noun phrases, in which demonstratives indicate spatial and discourse location of a referent. Section 3.2 deals with the clause, where demonstratives indicate spatial or temporal location of an event. Section 3.3 discusses the sentence, where demonstratives encode assertion and evidentiality. The assumption is made here that the spatial use of
.╅ The term noun-modifying clause is borrowed from Matsumoto (1997). The term refers to relative clauses and clauses where no constituent relationship can be found between the head noun and the modifying clause.
 František Kratochvíl
demonstratives in the nominal domain is primary, while the other uses are extensions based on the metaphor space → time.
3.1â•… Functions of demonstratives in the noun phrase In the noun phrase, demonstratives indicate spatial or discourse location of a referent. Either referential property is associated with a distinct syntactic position: (i) demonstratives that precede the head noun indicate the spatial location of the referent; they are deictic (see Section 3.1.1), (ii) demonstratives that follow the head noun indicate the location of the referent in the discourse; they are anaphoric (Section 3.1.2).
3.1.1â•… Spatial location of the referent As mentioned in 3.1, Abui demonstratives indicating spatial location (deictic demonstratives) precede the head noun and are accompanied by a pointing gesture of some kind. The pointing can be done by index finger, lips, eye movement, or chin. Deictic demonstratives are organized in a three-way (proximal-medial-distal) system presented in Table 1. Both medial and distal positions allow an additional vertical parameter. For the proximal and medial forms, the frame of reference is ‘relative’: it alternates between the speaker and the addressee (cf. Levinson 2003: 24–61). Speaker’s viewpoint is the default choice; addressee’s viewpoint is pragmatically marked. Note that the viewpoint alternation does not apply to the distal location. For vertically oriented locations it is not possible to alternate viewpoint. Speaker’s viewpoint is always applied. Note also that the vertical parameter dictates the tone. Table 1.╇ Abui deictic demonstratives horizontal location
viewpoint (v) speaker
proximal medial distal
addressee
do (prx) to (prx.ad) o, lo (md) yo (md.ad) oro (dst)
vertical location low * o (md.l) wo (dst.l)
high * ó (md.h) wó (dst.h)
In (4), non-vertical deictic demonstratives are illustrated: (4) a.
do fala b. o fala prx house md house ‘this house (near me)’ ‘that house there (further from me)’
c.
to fala d. yo fala prx.ad house md.ad house ‘this house (near you)’ ‘that house there (further from you)’
e.
Discourse-structuring functions of Abui demonstratives 
oro fala dst house ‘that house over there (far from us)’
The vertically oriented deictic demonstratives are illustrated in (5). (5) a.
o moku b. ó moku md.l kid md.h kid ‘that child below’ ‘that child above’
c.
wo moku d. wó moku dst.l kid dst.h kid ‘that child below over there’ ‘that child above over there’
Deictic demonstratives may be used pronominally as in (6), where the medial demonstrative o (md) referring to the sea water, located further from the main character. The proximal do (prx) indicates the discourse location of the referent (for more details see the next section): (6) ah,# ╛╛╛o do,# tama do baai ba ya, hare na buuk-o! oh [md prx] sea prx also say water so 1s consume-pnct ‘oh, the thing there, the sea is reportedly also water, so let me drink!’
In the following fragment, the medial demonstrative o (md) refers to the giant snake, that attempting to climb into the tree house from under: (7) di faling do sei, # hen baai, o la marang 3a axe prx come.down.cnt then even md be.md come.up.icp ‘they threw down the axe, and even then that (snake) still kept climbing’
Note that in (7), it is not clear whether the pronominally used medial demonstrative o (md) comes from its deictic or anaphoric use. The evidence comes from the distal demonstratives: only the spatial forms have pronominal uses, as illustrated in (8). The distal oro refers to an old man who is asked to explain a story to a group of younger people. (8) e-ratala we di he-ni-tahang-d-i-a mai, 2s.al-g:c 3p 3a 3ii.loc-1pe.pat-ask-hold-pfv-dur when ‘ah, ╛╛╛oro nu ri ha-tahang-d-i-a!’ oh [dst spc] 2p 3ii.pat-ask-hold-pfv-dur ‘when your granchildren asked us about it, (we said:) ‘ask him over there’
3.1.2â•… Discourse location of the referent In the final position of the np, Abui demonstratives indicate the discourse location of the referent: they are anaphoric. The full paradigm is presented in Table 2. The proximal and medial forms occur also in the deictic demonstrative paradigm, but the distal forms cannot be used deictically.
 František Kratochvíl
Table 2.╇ Abui anaphoric demonstratives viewpoint (v)
discourse location
[+proximal] [+medial] [+distal]
speaker
addressee
do (prx) o (md) nu (spc)
to (prx.ad) yo (md.ad) hu (spc.ad)
As the final constituent of the NP the anaphoric demonstratives are associated with a change in intonation contour and develop to right edge markers. Anaphoric demonstratives have a discourse structuring role. They either identify the actual participants or refer back to participants that already occurred in discourse (cf. Mithun 1987: 187–191). Demonstratives have an orienting role, telling the addressee which participants are in focus (proximate), pointing out the centre of interest. Speakerbased anaphoric demonstratives are the default choice when one refers to actual or just mentioned participants. For instance, asking about the quality of water source one can say: (9) ya do buuk kaang? water prx consume good ‘is the water drinkable?’
Although the utterance in (9) does not include any pointing, the referent of the np ya do ‘the water’ is obvious and identifiable for all interlocutors. In narratives, do (prx) is used to introduce prominent participants or to refer to very recently mentioned participants. Another example is given in (10). In the initial part of the narrative, the giant snake and its victims are referred to with the proximal demonstrative do (prx). (10) ╛╛╛Mon Mot mon do, # di yaa oro it-i Lelawi,# [name snake prx] 3a go dst lie.on-pfv place hen sei,# ╛╛╛amakaang do tafuda nee-i there come.down.cnt [person prx] all eat-pfv ‘the snake Mon Mot, he went over there to Lelawi he came down there, the people, he ate all of them’
As the narrative moves on, the snake is continuously being referred to with the proximal demonstrative do (prx) as the prominent participant: (11) ni-ya,# ma ╛╛╛Mon Mot mon do 1pe.al-mother be.prx [name snake prx] he-melang te=he-r-i? 3ii.al-village where=3ii.loc-reach-pfv ‘mother, (as for) the snake Mon Mot, how (do we) reach his village?’
Discourse-structuring functions of Abui demonstratives 
The medial forms are used to refer to objects in relatively recent discourse. In (12), the medial o (md) indicates that the boats have been mentioned before. (12) ╛╛╛kapal loku o marang tut taha he-ta-top-i [boat pl md] come.up shore on.cnt 3ii.loc-dstr.pat-drop.cpl-pfv ‘those boats disembarked on the shore’
The distal demonstrative nu (spc) points to referents that occurred very early in the discourse or introduces new referents into the discourse. In (13) the demonstrative nu (spc) refers to fish that occurred in the discourse about 30 seconds earlier. The fish is referentially distal also because it is not a prominent participant that is immediately identifiable for the interlocutors. (13) di ╛╛╛afu nu mi ba ho-pa do=ng we-i 3a [fish spc] take sim 3ii.rec-touch.cnt prx=look leave-pfv ‘she took that fish and shifted it towards him’
In (14), interlocutor A introduces a certain type of car in a distant location. Interlocutor B confirms that he can actually see it: (14) ╛╛╛oro oto nu # a he-wahai kaang? [dst car spc] 2s 3ii.loc-look.at good ‘that car over there, can you see it?’
The speaker-based demonstratives presented in Table 2 have addressee-based counterparts. The speaker chooses the addressee’s perspective to refer to a referent mentioned in the addressee’s previous discourse or to draw the addressee’s attention to new or unexpected information. Consider the following scenario. A betel nut was offered for chewing in the previous discourse. Assuming that the guests will also ask for betel vine (usually chewed together with betel nut), in (15), the host anticipates their question saying: (15) meeting to,# ri wala mi, ui h-iéng… betel.vine prx.ad 2p just take back 3ii.pat-look ‘the betel vine (that you might want), just take it, at the back…’
In Abui, there is also a specialized topic marker ba (discussed in Section 4.1.1) which is used to mark topical arguments. The function of the addressee-based demonstratives differs from the topic marker in the sense that the former are always used in interaction with what the speaker assumes is part of addressee’s knowledge or expectation. The latter is used to structure speaker’s own discourse. The addressee-based forms have undergone subjectification to serve increasingly abstract, pragmatic and interpersonal functions (cf. Traugott 1995: 32). In (16), A is urging B to remember to buy something. B cannot remember. Therefore A explicitly states what B ought to remember, using the medial addressee-based
 František Kratochvíl
demonstrative yo (md.ad). Using the addressee-based form, the speaker implies addressee’s previous knowledge of what ought to be bought. (16) A: ╛╛╛nala nuku bel ba mi ne-r=te! [something one] buy purp take 1s.loc-reach=incp.c ‘buy one thing for me!’ B: nala eh? what inter ‘what?’ A: ╛╛╛mur yo [citrus md.ad] ‘oranges (that you heard about)’
In the following narrative fragment, the giant snake is being challenged about his killing of the entire village that occurred about two decades earlier. The challenger of the giant serpent has not witnessed the killing. (17) wah,# ╛╛╛ne-serang ne-yai yo baai# a tafuda nee-i! inter [1s.al-people 1s.al-folk md.ad] even 2s all eat-pfv ‘what! even my people and my tribe (as you ought to remember) you have eaten them all!’
The distal demonstrative hu (spc.ad) indicates a specific referent that is somehow related to addressee’s previous discourse or experience. It may introduce new referents in statements (18), or in questions (19). In (18), B uses hu (spc.ad) to contradict A’s expectation that it is meat that is being prepared. (18) A: e-mayol di mahiting maal? 2s.al-woman 3a meat cook ‘is your wife preparing meat?’ B: naha,# di ╛╛╛afu hu maal not 3a [fish spc.ad] cook ‘no, she is preparing that fish’
In (19), the demonstrative hu (spc.ad) combines with the question word nala ‘what’ to refer to the addressee’s choice of transportation means. (19) a ╛╛╛nala hu tai mit ba miyei? 2s [what spc.ad] on sit sim come.cpl ‘by what (transportation means) did you come?’, lit.: ‘what did you sit on to come?’
3.1.3â•… Noun-modifying constructions (NMC) As mentioned in 2.2, Abui nouns can be modified by a range of modifiers (clauses, nps, adverbs, etc.) linked with ba. The nominal head and the nmc form a single
Discourse-structuring functions of Abui demonstratives 
prosodic unit, unlike the topic construction, where ba is separated by a pause from the subsequent comment. The phonological unity indicates that both the modified noun and nmc are conjoined in a single referential constituent (see also 4.1.2). The nmc may be followed by a demonstrative that marks the end of the nmc, as the distal nu (spc) in (20). (20) ╛╛╛kawen ba ╛╛╛topa nu# ong bul-r-i=te re-i! [machete lnk {blunt spc}] make sharp-reach-pfv=incp.c reach.icp-pfv ‘that machete that is blunt, sharpen it finally!’
While in (20) the nmc consists of a single constituent, in (21) the nmc contains a verb and its argument. As in (20), the end of the nmc is marked by the anaphoric demonstrative. The reference of the demonstrative is ambiguous. It may encode the time but also the discourse location of the entire constituent. (21) ╛╛╛fe ba ╛╛╛he-toku ha-kor-i yo# [pig lnk {3ii.al-leg 3ii.pat-bind.cpl-pfv md.ad}] do-tik-i ba taki-i 3i.rec-stretch-pfv sim escape-pfv ‘the pig that had its legs bound (according to you), stretched itself from the ropes and escaped’
There is no conclusive evidence to determine to which part of the complex np the demonstrative belongs. Although the nmc may be preceded by a quantifier, the demonstrative can never occur between the modified noun and the nmc. Further, if the nominal head is obvious from the context, it can be omitted. The nmc such as hetoku hakori yo ‘the one with bound legs’ is sufficient to refer to the pig. We consider such constructions to be instances analogous to free relatives in other languages. Stand-alone nominalizations are formed with the individualized plural quantifier loku, as exemplified in (56) in Section 4.1.1. Alternatively, verbal roots can be nominalized by possessive prefixes, such as ne- (1s.al) in ne-kariang ‘my work, my working’.
3.1.4â•… Interim summary Each of the anaphoric demonstratives may occur as final constituent in the np, following the nmc. The demonstrative is associated with the right edge of the np; it never occurs between the head noun and the nmc but marks the end of the [noun-ba-nmc] constituent. The noun-modifying constituent does not necessary stand in argument relationship with the modified head noun; constituents such as adverbials or other nouns modifying the head noun are also linked with ba. It follows then, that the [noun-ba-modifier] constituents do not always correspond to relative clauses in other languages (cf. Section 4.1.2).
 František Kratochvíl
The functions of Abui demonstratives in the np are summarized in (22). The demonstrative preceding the head indicates the spatial location of the referent: it is deictic (dems). The demonstrative following all other adnominal modifiers indicates the ‘discourse location’ of the referent: it is anaphoric (dema). In the [noun-ba-nmc] constructions (ii), the final demonstratives are ambiguous, often indicating either the ‘discourse location’ of the entire referential constituent or the ‘temporal location’ of the nmc: (22) np: i. dems (poss-) n n/adj/v/quant/ (23)
dema
ii. dems (poss-) n n/adj/v/quant ba + nmc dema/t
In the discourse, speaker-based demonstratives are the default choice. Addresseebased forms have undergone subjectification and are used when the speaker wants to introduce a new referent into the discourse or to direct addressee’s attention to an earlier-mentioned but unexpected referent.
3.2â•… Functions of Abui demonstratives in clause The constituent order in the Abui clause is strict. It is schematically represented in (3) and repeated here as (24). The constituents available and accessible in the preceding context are omitted. (24) clause: adv np proA adv/dems npU vp asp neg demt
In the clause, demonstratives indicate spatial or temporal location of the event. As in the noun phrase, each referential property is associated with a distinct syntactic position. Demonstratives that precede the vp indicate the spatial location of the event (dems): they are spatial (Section 3.2.1). Demonstratives that follow the vp indicate the location of the event in the time (demt): they have temporal reference (Section 3.2.2).
3.2.1â•… Spatial location of an event In (25), the demonstrative do (prx) indicates that the addressee’s location is ‘proximal’ to the deictic centre (located in the speaker): (25) a do mi-a maiye,# ama e-l feng kaang 2s prx in-dur if person 2s.loc-give injure good ‘if you stay here, people can harm you’
In (26), the demonstrative o (md.l) indicates that the comb fell down on the ground through a gap in the elevated platform: (26) ai,# ket o ha-yei to, na sei oh comb md.l 3ii.pat-fall prx.ad 1s come.down.cnt mi-i=se ye! take-pfv=incp.i inter ‘ah, the comb just fell down (as you see), I will go down and take it!’
Discourse-structuring functions of Abui demonstratives 
In (27), the demonstrative oro (dst) indicates that the event takes place in a location distant from the speaker. (27) Fani oro kamai ha-d-a mui-l-a name dst cat 3ii.pat-hold-dur game-give-dur ‘Fani is playing with a cat over there’
Note that all deictic demonstratives, given in the Table 1, may be used to indicate the spatial location of an event. More details can be found in Kratochvíl (2007: 269–270).
3.2.2â•… Temporal location of an event Abui verbs do not carry any tense inflection, but inflect exclusively for person and aspect. The temporal location of an event can be indicated by adverbs such as el ‘before’, ko (fut), or by nominals such as afeida ‘yesterday’ or akun ‘tomorrow’. Anaphoric demonstratives offer another way of indicating the temporal location of an event (in fact tense). Based on the metaphor space → time, demonstratives in the clause-final position come to indicate the temporal location of an event with respect to the coding time.7 This position is available to all demonstratives given in Table 2. The speaker-based demonstratives are the default choice to indicate temporal location of an event: tense. The proximal forms indicate that an event has just occurred, is still ongoing or is about to occur very soon. A simple example was given in (1a) and is repeated here as (28). Depending on its context, the following statement can have each of the three meanings indicated in the translation. (28) di me do 3a come prx ‘he is coming/just came/will come (shortly)’
In Abui, in addition to non-completive stems, many verbs also have completive stems (cpl) indicating that the reported event has an end point. The completive stem followed by a demonstrative usually refers to an event that occurred in the past. The addresseebased forms are used to appeal to the addressee’s knowledge of the event. This is illustrated in the difference between (29a) and (29b).
.â•… The term coding time is borrowed from Levinson (1983): ‘Time deixis concerns the encoding of temporal points and spans relative to the time at which an utterance was spoken (or a written message inscribed). This time (…) we shall call coding time or CT (…). Thus, just as place deixis encodes spatial locations on co-ordinates anchored to the place of utterance, so time deixis encodes times on co-ordinates anchored to the time of utterance. Time deixis is commonly grammaticalized in deictic adverbs of time (like English now and then, yesterday and this year), but above all in tense.’
 František Kratochvíl
(29) a. di de-melang da-wai yaar do 3a 3i.al-village 3i.pat-turn go.cpl prx ‘he just went back to his village’ b. di de-melang da-wai yaar to 3a 3i.al-village 3i.pat-turn go.cpl prx.ad ‘you know that he just went back to his village’
The medial forms in (30) indicate a longer distance between the coding time and the temporal location of an event. Again, addressee-based forms are used to appeal to the addressee’s knowledge of the event. The completive stem of the verb yaar ‘go’ combines with perfective suffix -i (pfv) to indicate that the event of going back is finished: (30) a.
di de-melang da-wai yaar-i o 3a 3i.al-village 3i.pat-turn go.cpl-pfv md ‘he went back to his village some time ago’
b. di de-melang da-wai yaar-i yo 3a 3i.al-village 3i.pat-turn go.cpl-pfv md.ad ‘you should know that he went back to his village some time ago’
The distal forms are used when the speaker is uncertain about the exact temporal location of an event that occurred in the past.8 The temporal use of demonstratives overlaps here with their evidential and mood use discussed in Section 3.3.2. This is illustrated in (31b) where the addressee-based form hu (spc.ad) marks irrealis. In natural speech, the ambiguity is resolved by the context; the elicited examples illustrating the paradigm are ambiguous: (31) a. di de-melang da-wai yaar nu 3a 3i.al-village 3i.pat-turn go.cpl spc ‘he went back to his village (a certain time, long time ago)’ b. di de-melang da-wai yaar hu 3a 3i.al-village 3i.pat-turn go.cpl spc.ad ‘it would be better if he had gone back to his village’
As (31) shows, Abui demonstratives do not only mark past tense. They indicate the temporal ‘distance’ from the coding time. In (32), the medial demonstrative o (md) refers to a nearby future event. Another example, illustrating the tense marking use of o (md) is given in (56). (32) na o-pa=ng marang o 1s 2s.rec-touch.cnt=look come.up md ‘I will come near to you’
.â•… Similar extension of demonstratives to indicate temporal location of an event is reported for Iaai, an Oceanic language of New Caledonia (Ozanne-Rivierre 2004: 134–135).
Discourse-structuring functions of Abui demonstratives 
Another example is given in (33) where nu (spc) points to an unspecified future time: (33) ko na sei la mi-a nu fut 1s come.down.cnt be.md take-dur spc ‘I will soon come down there and take it’
3.3â•… Functions of Abui demonstratives in sentence Depending on their position, Abui demonstratives have a number of functions in sentence. Demonstratives can function as subordination markers, marking certain types of subordinate clauses (see Section 3.3.1). In the sentence-final position, demonstratives indicate source of information (i.e. evidentiality, see Section 3.3.2), or have other pragmatic functions, such as assertion markers (see Section 3.3.3). The ambiguity is resolved by larger context.
3.3.1â•… Subordination markers As mentioned above, a subset of Abui demonstratives (see Table 2) marks adverbial clauses; i.e. subordinate clauses (sc) expressing relative time of the event described in the main clause (mc). Anaphoric demonstratives are associated with the right edge; they occur in the clause-final position of scs and are followed by a pause. The adverbial scs appear in the same position as left-dislocated nps or adverbials.9 In (34), the proximal do indicates that the event of shooting is the moment when the addressee has to watch: (34) ╛╛╛na ha-tak do, # a he-roa {1s 3ii.pat-shoot prx} 2s 3ii.loc-watch.cnt ‘when I shoot (with the bow), you watch it’
In (35), the proximal demonstrative do indicates that the pregnant woman kept fetching water until shortly before she gave birth to her children: (35) ╛╛╛di ya do he-taki-a bang mi {3a water prx 3ii.loc-loosen-dur carry.on.shoulder take sei buuk-buuk do, di moku come.down.cnt red[consume] prx} 3a kid do ha-yar-i prx 3ii.pat-give.birth.cpl-pfv ‘she was continuously bringing water (and) drinking it, when she gave birth to her children’
.â•… The marking of some types of subordinate clauses as NPs is reported also for other Papuan languages (Foley 1986: 202–203; Reesink 1994: 100–102; Vries 2006). Demonstratives have a discourse structuring function in clause chains also in Malay and Pileni (Næss 2004: 89–91).
 František Kratochvíl
The addressee-based yo (md.ad) in (36) indicates that the man freed himself from the ropes some time before he put up the ladder and climbed in the house. The addresseebased form contradicts addressee’s expectation that the man is still tied up. Another example of the subordinating use of yo (md.ad) is given in (54a). (36) ╛╛╛do-tik-i kaan-r-i yo,# {3i.rec-stretch-pfv good.cpl-reach-pfv md.ad} hen di awering do ha-b-i ya then 3a ladder prx 3ii.pat-join-pfv seq mara fala=ng mara go.up.cnt house=look go.up.cnt ‘after he actually untied himself, he put up the ladder and climbed in the house’
The distal form nu (spc) marks a sc that very loosely describes the relative time of the main event. In (37), nu is translated as ‘whenever’: (37) ╛╛╛no-mi=ng maran nu,# na kul Kalang-Fat yaar=te {1s.rec-in=look come.up.cpl spc} 1s must place go.cpl=incp.c ‘whenever I feel like that, I must go to Kalabahi (lit. whenever it comes up in me, I first must go to Kalabahi’
The other distal form hu (spc.ad) is very similar to nu (spc). It marks a sc that very loosely describes the temporal frame of the event reported in the mc and appeals to addressee’s knowledge, as illustrated in (38). (38) ╛╛╛motai ayoku mit ba de-ananra-ananra hu,# {spouses two sit sim 3i.loc-red[narrate.cnt] spc.ad} mayol de-teh ha-tama-d-i ba buuk woman 3i.al-tea 3ii.pat-make-hold-pfv purp consume ‘while the husband and wife sat down chatting, the wife prepared their tea to drink’
Abui anaphoric demonstratives are primarily used to track nominal referents in the discourse. Examples (35)–(38) demonstrate that adverbial clauses (scs expressing relative time) can also be marked with demonstratives. The adverbial clauses are formally treated as other referential constituents such as nps. This is summarized in (39): (39) {adverbial.clause dem}# main.clauseâ•…â•… relative temporal location
3.3.2â•… Evidentiality In Section 3.2.2, we saw that Abui demonstratives can indicate temporal location of an event when they occur in the clause-final position. The previous section discussed how the same set of demonstratives marks subordinate clauses that indicate the relative
Discourse-structuring functions of Abui demonstratives 
time of the event described in the main clause. Demonstratives listed in Table 2 may occur in the sentence-final position to encode evidentiality (source of information). In (40), the proximal demonstrative do (prx) indicates speaker’s immediate experience. (40) na nala nee=ti beek-a do 1s something eat=phsl.c bad-dur prx ‘I couldn’t eat up (swallow) anything’
In (41), A is not aware of the poor health of B’s mother. B considers her poor health obvious as he uses to (prx.ad) to indicate that A has sufficient evidence to verify the proposition. (41) A: mangmat,# ma e-ya yo? foster.child be.prx 2s.al-mother md.ad ‘child, what about your mother?’ B: ni-ya ha-rik to! 1pe.al-mother 3ii.pat-hurt prx.ad ‘my mother is sick (as you could see)’
The medial addressee-based demonstrative yo (md.ad) is used when a proposition is based on earlier evidence available to the addressee; in (42), the speaker stresses that the addressee knew about the funeral: (42) pi yaar-i ni-ya do nabuk yo 1pi go.cpl-pfv 1pe.al-mother prx bury md.ad ‘we went to bury our mother (as you could have seen)’
Another example of the evidential use of yo (md.ad) is given in (52). The distal demonstrative nu indicates that an event, such as the past handed down by the ancestors, was not witnessed by the addressee, but should be considered as truth. The reportative ba (discussed in Section 4) would not be in place in (43), because the speaker believes in the truth of his statement: (43) nuku oro Mali do=ng we-i nu one dst place prx=look leave-pfv spc ‘one (ancestor) went to Mali over there’
The mapping of evidentiality is based on the metaphor space → source.distance. The immediately witnessed events are marked with proximal demonstratives. The medial forms indicate events witnessed or experienced in the past. Distal forms indicate events that are remembered, handed down and believed in, coming from a reliable source. The reportative evidential ba marks second-hand sources.
3.3.3â•… Assertion This section deals with the pragmatic function of demonstratives as assertion markers. A short note is needed here about the terms used in this section. Each discourse is
 František Kratochvíl
placed in a context which is (at each particular point) ‘the body of information that is presumed, at that point, to be common to the participants in the discourse’ (cf. Stalnaker 1998: 3).10 An assertion can be seen as ‘a proposal to alter the context by adding the information that is the content of the assertion to the body of information that defines the context’ (cf. Stalnaker 1998: 6).11 In Abui, sentence-final demonstratives, in addition to assertion, may indicate tense and evidentiality, as discussed in Sections 3.2.2 and 3.3.2. For a correct interpretation of sentence-final demonstratives context is needed. In (44), both interlocutors use do (prx). A boy is instructed as to how to use a bow. Tired of learning how to shoot, the boy (A) complains to his grandfather (B) about pain in his hand. In his turn, grandfather (B) expresses surprise about boy’s (A) discomfort. (44) A: do,# do na-rik baai do! prx prx 1s.pat-hurt even prx ‘here, this even hurts me!’ B: te=wi-r=te a-rik-e?# where=like.md.cpl-reach=incp.c 2s.pat-hurt-ipfv ‘how come it is hurting you?’
no-mi he-uk-d-a do! 1s.rec-in 3ii.loc-withdraw-hold-dur prx ‘that really surprises me!’
The distal nu (spc) indicates weak assertion. In (45) A asks B to pluck one banana for him. B answers reluctantly using nu (spc). This can be interpreted that B dropped some bananas earlier on, or might drop some later. From the context we know that B did not drop any bananas so far and is not planning to drop any. (45) A: ne-feela,# ma nuku=ng sik ba ne-l 1s.al-friend be.prx one=look sever purp 1s.loc-give
baai ne-l=re! also 1s.loc-give=reach.inc
‘friend, pluck also one for me and give it to me’
.â•… ‘We can represent the information that defines the context in which a speech act takes place with a set of possible situations or possible worlds – the situations that are compatible with the information’ (Stalnaker 1998: 5). .â•… Assertion can alter the context also by ‘eliminating from the context set – the set of possible worlds available for speakers to distinguish between – those possible worlds in which the proposition expressed in the assertion is false’ (Stalnaker 1998: 6).
Discourse-structuring functions of Abui demonstratives 
B: ma iti,# na ong ha-yei nu! be.prx that 1s make 3ii.pat-fall spc ‘those (bananas) here, I dropped them (some time ago)/I will drop them!’
As stated earlier, addressee-based demonstratives are pragmatically marked. The speaker uses addressee-based demonstratives when he wants to contradict the actuality of the event taken for granted by the addressee or to correct previous context. In (46), the use of to (prx.ad) implies that the addressee should understand why the speaker and his companion have to leave. (46) hai, # ni kul yaar to! inter 1pe must go.cpl prx.ad ‘oh, we really must go (as you can see/understand)’
The distal addressee-based demonstrative hu is used in situations like (47). A boy is instructed as to how to use a bow. The speaker indicates the wished result of the activity that is different from boy’s achievements: (47) d-i-a he-kira he-i=se hu! hold-pfv-dur 3ii.loc-hard 3ii.loc-put=incp.i spc.ad ‘hold it (a bow) so that it would become tight’
There are a few cases, showing that evidentiality and assertion are encoded independently in distinct syntactic positions. In the surface structure it appears as if the final demonstrative is doubled, as in (48). In fact each of the demonstratives encodes a different grammatical category. The addressee-based to encodes the evidentiality, while do indicates the assertion. (48) na-táng do,# di namur to do! 1s.inal-hand prx 3a wound.cpl prx.ade prxas ‘my hand really got hurt just a while ago (as you know)!’
The demonstrative in the slot closer to the vp encodes the source of information. The sentence-final demonstrative marks assertion, proposing to the addressee to alter the context. In (49), the first proximal do encodes the source as first-hand evidence, the second do marks assertion: (49) ma iti,# it de-i do do be.prx that lie.on 3i.loc-put prxe prxas ‘yes, it is really in its place now’
Demonstratives may be added to imperatives to indicate urgency, as in (50). The demonstrative indicates whether the imperative is to be interpreted as a request, advice or suggestion. The addressee-based demonstratives appeal to addressee’s understanding of the imperative.
 František Kratochvíl
The proximal demonstrative do stresses the urgency of the imperative in (50a). The medial demonstrative o (md) weakens the appeal; it is considered a polite form.12 (50) a. ko e-neng ru-fal ri-melang yaa do! fut 2s.al-man 2p.rec-separate 2p.al-village go prx ‘you will go with your husband to your village indeed!’ b. marang mit-i o! come.up.icp sit-pfv md ‘come (up) sit down’ c.
pi Afena Hapong pa te-bikeng-r-a o! 1pi place go.down.cnt dstr.loc-louse-reach-dur md ‘we (could) go to Afena Hapong and search each other’s lice’
The medial yo is used in imperative such as (51) to stress that the addressee ought to understand: (51) a-ran ba taa yo! 2s.pat-quiet sim lie md.ad ‘you should calm down and sleep!
3.4╅ Summary As summarized in Table 3, the function of Abui demonstratives varies depending on their syntactic position: Table 3.╇ Functions of Abui demonstratives in different syntactic domains domain
structure
function
np
dems (poss-) n n/adj/v/quant dema
clause
a
sentence
i. {adverbial.clause demrt} {main.clause} ii. {final.clause}deme demas
dems dema dems demt demrt deme demas
dems u
pred demt
space anaphora space temporal location relative temporal location evidentiality assertion
In the np, demonstratives indicate spatial and discourse location of the referent. In the clause, they indicate spatial or temporal location of an event. In the sentence, demonstrative may function as subordinators of adverbial clauses; in sentence-final position, they mark evidentiality and assertion. When a single demonstrative is found
.â•… The form o is reported as imperative in some Papuan languages (cf. Foley 1986: 164–165). Abui form o is not restricted to imperatives but has also other functions.
Discourse-structuring functions of Abui demonstratives 
in the sentence-final position, it may encode tense, evidentiality or assertion. This ambiguity can be only resolved when the relevant context is considered. In some cases, two demonstratives occur in sentence-final position; the first indicates evidentiality, the second assertion. Addressee-based demonstratives are used when the speaker wants to interact with the addressee’s knowledge.
4.â•… Functions of the marker ba From the data presented in Section 3, a conclusion could be drawn that in Abui grammatical categories of anaphora, tense, and evidentiality are structured in the same way as space. However, such a conclusion would be precipitous, as will be demonstrated in this section. Although spatial co-ordinates can be applied to other grammatical categories, the structure of a more abstract grammatical category and the set of distinct values to be marked remain independent of the structure of space. In fact, additional markers can be recruited from sources other than the demonstrative paradigm if a particular grammatical category requires so. To demonstrate this point, we will consider markers grammaticalized from the verb ba ‘say’ that encode values of the same grammatical categories as the demonstrative paradigms, discussed in the previous section. The primary function of the speech verb ba ‘say’ is to introduce a quote. In (52) taken from a narrative, ba introduces a reply: (52) “wonge?”# he-ni-l ba:# hallo? 3ii.loc-like.prx.cpl-give say “Is anybody there?” so he says:’ “ai la ne-l ong ha-yei yo,# oh be.md 1s.loc-give make 3ii.pat-fall md.ad “oh, when it seemed that I fell, awering do sobak-d-i yo!” ladder prx slip-hold-pfv md.ad the ladder slipped away! (as you have heard)”
The position of the speech verb ba ‘say’ with respect of the quotation can vary. In (53), the verb follows the quotation, linked with the indexical verb henil ‘do like this’: (53) ‘ah ne-d-e pa ha-k-o!’ oh 1s.loc-hold-ipfv go.down.cnt 3ii.pat-throw-pnct he-ni-l (*di) ba 3ii.loc-like.prx.cpl-give ╇╛3a say ‘I will go down feed it!’ (he) says like this’
 František Kratochvíl
The verb ba ‘say’ combines with a single core argument which refers to the speaker. The speaker is not treated as a typical a argument. This can be seen in (53), where ba cannot combine with the free pronoun di (3a). Sometimes, as in (54), the speaker is expressed with the undergoer prefix.13 (54) ‘ah, nu-kuta ko beeka-d-i!’ (*di) he-ba oh 1pe.al-g:p fut bad-hold-pfv ╇╛3a 3ii.loc-say ‘oh, grandfather, it will be impossible’ he said’
The speech report introduced by the verb ba ‘say’ is not its core argument; it is never marked on the verb with a pronominal prefix. Instead, it is treated as an adjunct, often overtly linked with the indexical form henil ‘do like, do so’, as in (52) or (53). The form henil is also used anaphorically to refer to the speech report. The verb ba ‘say’ typically follows onomatopoeic words that represent various sounds (ideophones), as in (55). In these contexts, the verb ba may be inflected for aspect. Note that bai is translated as ‘it went’ to capture the impersonal subject and perfective aspect. (55) a.
na nala ong ha-yei yo, ‘pak’ ba-i 1s something make 3ii.pat-fall md.ad sound.of.hitting say-pfv ‘when I dropped something, it went (saying) ‘pak’’
b. ‘sobak’ ba-i ba saai sound.of.slipping say-pfv sim come.down.cpl ‘it went ‘swak’ as he slipped down’
The restricted valence of ba ‘say’ and its ability to adjoin quotations are crucial for understanding its grammaticalization, discussed in the following sections.
4.1â•… Functions of ba related to the noun phrase The verb ba ‘say’ is the source of the topic marker (Section 4.1.1) and of the nounmodifying construction linker (Section 4.1.2). The nmc-linking function is possibly an extension of the topic marking function.
4.1.1â•… Topic marker When ba follows an np, the phrase is marked as being topical. The entire constituent is distinguished by prosodic features such as a rising contour on the topic marker
.â•… The single argument of the verb ba ‘say’ in Abui cannot be expressed with a free pronoun because free pronouns are exclusively used for controlling and volitional participants. The relation between the speaker and utterances seems to be that of attribution, as the prefix he- (3ii.loc) suggests. Similar valence restrictions on the verb ‘say’ are found crosslinguistically (cf. Munro 1982: 316–317).
Discourse-structuring functions of Abui demonstratives 
followed by a pause (indicated with #). In syntax, topical NPs are left-dislocated, as the NP sameng ‘cement’ in (56).14 Note also the use of the clause-final o (md), indicating that the event occurred some time ago. (56) sameng ba,# Pak Kas di mi ba Takpala maran-i o cement top Mr name 3a take sim place come.up.cpl-pfv md ‘as for the cement, Mr Kas brought it up to Takpala (some time ago)’
In (57), the topic marker ba delimits the topical argument which is in this case a standalone nominalization. The topic is followed by a resumptive pronoun he- prefixed to the equational copula ne ‘like this’ (cf. Dryer 2007: 233). The copula links the topic with the copula complement denoting the Abui warrior clan Kapitang. (57) te-l feng kaang loku ba,# he-n-e Kapitang dstr.loc-give injure good plNMLZ top 3ii.loc-like.prx-ipfv name ‘those who know how to kill, those are Kapitang’
Note that the individuated plural quantifier loku (pl), parallel to colloquial English ‘guys’ can derive words denoting actors such as ‘fishermen’ or ‘bricklayers’ illustrated in (58). (58) a. afu tahai loku b. fish search pl ‘fishermen’
fala ong kaang loku house make good pl ‘bricklayers’
As mentioned in Section 3.1.3, the quantifier loku (pl), together with possessive prefixes, are the only devoted nominalizers. The demonstratives, as illustrated in Section 3, are multi-functional markers.
4.1.2â•… Noun-modifying construction The second function of ba is that of a noun-modifying construction linker. This use probably originates in the topic construction. The marker ba indicates that the head noun is a local topic that will be elaborated on. However, there are a number of important distinctions. The noun and the nmc form a single prosodic constituent; there is no pause after ba. The nmc is interpreted referentially (not predicatively, as in the topic construction) and forms thus part of a larger referential constituent, often translated with the relative clause construction in English. The Abui construction is different from relative clause, because also non-clausal constituents can be linked in the same way, such as nouns, adverb, or numerals. The modified head noun is in no argument
.╅ Left dislocation and paratactic NPs and adverbial clauses preceding clause chains are so common in Papuan languages that they are claimed as an areal feature (cf. Vries 2006: 814).
 František Kratochvíl
relationship with such noun-modifying constituents (cf. Sections 3.1.3–3.1.4). In (59), ba links the noun fala ‘house’ with a nmc. (59) fala ba ╛╛╛tayoka ong fak-i nu ama da-wai house lnk {earthquake make break-pfv spc} person 3i.pat-turn ong kaan-r-a make good.cpl-reach-dur ‘that house that was destroyed by the earthquake, people are again repairing it’
As illustrated in (60), the linked modifier can be a numeral, such as heyeting-ayoku ‘seventh’, or another modifier noun, such as mayol ‘woman, female’. (60) e-ratala ba ╛╛╛he-yeting-ayoku ba ╛╛╛mayol,# 2s.al-g:c lnk [3ii.loc-seven] lnk [woman] he-n-u fa h-ieng mok 3ii.loc-like.prx-prf be.md.ad 3ii.inal-eye sleepy ‘your seventh female grandchild, that one is actually already sleepy’
4.2â•… Functions of ba in clause In the clause, ba has been grammaticalized into a quotative marker (Section 4.2.1) and complement linker (Section 4.2.2).
4.2.1â•… Quotative marker As illustrated in (61), the marker ba may occur as the last constituent of a quotation. The speech-report function is taken over by the verb fanga ‘speak’, which is fully inflected. The marker ba indicates the end of the quotation; it has been grammaticalized as a quotative marker. The marked quotation contains the features of direct speech, such as the adverbials ma and ka.15 (61) di he-fanga=ti# ╛╛╛‘ma ka do mi-a ba afu 3a 3ii.loc-tell.cnt=phsl.c {be.prx irr prx in-dur sim fish mi-a beeka’ ba take-dur bad} qt ‘he just said: ‘here you would not be able to get any fish’
.â•… A grammaticalization path of this type is well-attested in other languages such as Nepali, Newari, Sherpa, Methei, Jirel, Magar (Saxena 1988: 376, 378); Kambera, Buru, Tukang Besi (Klamer 2000), Nama, Twi, Vai (Heine & Kuteva 2002: 267–8).
Discourse-structuring functions of Abui demonstratives 
As illustrated in (62), the use of the quotative ba extends to indirect speech: (62) ne-nahaa di he-fanga=ti, ╛╛╛ha-rik ba 1s.al-y:si 3a 3ii.loc-tell.cnt=phsl.c {3ii.pat-hurt} qt ‘my younger sibling just told (about it) he is ill’
4.2.2â•… Complement linker While in (62) the indirect speech complement is not overtly linked with the matrix clause, in (63), ba links a non-factive complement.16 (63) di he-fangi ba ╛╛╛di ko me kaang ba 3a 3ii.loc-tell.cpl lnk {3a fut come good} qt ‘he told (about it) that he could come’
Ba is restricted to linking complements of verbs of speech.17 It does not occur with non-speech verbs. Moreover, verbs that combine with factive complements, such as ‘hate’, ‘notice’, ‘hear’, or ‘see’, require the complement to precede the matrix clause and do not allow the linker ba. In Abui syntax, a systematic distinction is made between factive and non-factive complements (cf. Kiparsky & Kiparsky 1970).
4.3â•… Functions of ba in sentence In the sentence, ba links (i) clauses expressing simultaneous events (Section 4.3.1), (ii) purpose clauses (Section 4.3.2), or (iii) encodes reported evidence in sentence-final position (Section 4.3.3).
4.3.1â•… Simultaneous linker As illustrated in (64), ba has been grammaticalized into a clause linker of clauses describing events that occurred simultaneously. The chained clauses linked with the simultaneous ba are prosodically integrated to a greater degree than clauses linked with the sequential ya.
.â•… The only reason for the quotative marker analysis of the second ba is analogy with examples (60) and (61). Janick Wrona suggested another plausible analysis: the second ba might be a part of a framing complementizer construction. .â•… The verb ‘say’ was grammaticalized into a complementizer in Methei, Ladakhi (Saxena 1988: 382–3), Thai, Ewe, Krio, Uto-Aztecan, Yoruba, Akan, Tamil, Telugu, Kera (Ransom 1988: 371), Tukang Besi, Buru (Klamer 2000: 70), Egyptian, Koranko, Vai, Baka, Bemba, Chamling, Lezgian (Heine & Kuteva 2002: 261–5).
 František Kratochvíl
(64) a-ran ba taa yo! 2s.pat-quiet sim lie md.ad ‘you should sleep calmly!
Another example is given in (65), where the clauses containing the verbs ‘mumble’ and ‘go’ are linked with ba: (65) no-dang-no-dang-d-i ba yaa he-ya si-a red[1s.rec-mumble]-hold-pfv sim go 3ii.al-water scoop.cpl-dur ‘mumbling I went to fetch their water’
Although simultaneous linkers are found in many Papuan languages (cf. Foley 1986: 180–185), little is known about their origin.18 The first clause is grammatically dependent, sharing the outer aspectual inflection (imperfective, durative, inchoative and inceptive, but not perfective and completive) and usually also the actor argument with the final clause of the chain. Without the shared inflections the first clause is not grammatical.
4.3.2â•… Purpose clause linker As illustrated in (66), ba may link purpose clauses. Grammaticalization of the verb ‘say’ into a purpose linker is relatively common:19 (66) a. na Fanmalei ho-l ba ╛╛╛di taa 1s name 3ii.rec-give purp {3a lie} ‘I am cradling Fan Malei to make him sleep’ b. na fui saai ba ╛╛╛afu la tahai 1s plain come.down.cpl purp {fish be.md search} ‘I came down to the coast to search there for fish’
4.3.3â•… Reportative evidential marker In narratives, the reportative marker ba indicates a second hand (hear-say) source of the information.20 This is illustrated in (67), taken from the beginning of a narrative:
.â•… Grammaticalization of the verb ‘say’ into a simultaneous marker is attested for Nigerian languages such as Izon and Igbo (cf. Güldemann 2008: 331–332). .â•… Heine and Kuteva (2002: 91, 265–267). .â•… Pragmatic extension of the verb ‘say’ to encode evidentiality is reported for TibetoBurman languages such as Jinghpaw, Sherpa, Jirel, Magar, Methei (Saxena 1988: 377, 381), and for Lezgian, Southern Min (Heine & Kuteva 2002: 265).
Discourse-structuring functions of Abui demonstratives 
(67) Kulwalakul to,# he-ya he-maama do tafuda name prx.ad 3ii.al-mother 3ii.al-father prx all de-mok ba 3i.loc-bring.together say ‘this Kulwalakul (that you mentioned), (it is said that) his parents and all (relatives) would get together (for a ceremony)’
When the speaker is not sure whether an event really occurred as described, the reportative marker can be used to create ‘distance’ from the message as in (68): (68) a. di de-kaai ha-liel marei ba takata do=ng 3a 3i.al-dog 3ii.pat-lift go.up.icp sim dry.land prx=look
marei ba go.up.icp say
‘reportedly he lifted up his dog and climbed on the shore’
b. moku nuku oro raharak-d-i-a do yang kid one dst shiver-hold-pfv-dur prx perhaps
ha-rik ba 3ii.pat-hurt say
‘the child that is shivering over there, reportedly it might be ill’
The reportative marker ba is obligatory in reports of cognitive acts such as dreams. The performing participant staged inside the cognitive act or process, characterized by the lack of control, is recorded only as the possessor of ‘spirit, soul’, similarly to English ‘my memory fails me’. (69) na piei-l-a=ti # na-noting Kalang Fat yaa ba 1s dream-give-dur=phsl.c 1s.inal-spirit place go say ‘I dreamt that I went to Kalabahi’, lit.: ‘I dreamt (that) my soul (reportedly) went to K.’
As illustrated in (70), the marker ba can be preceded by the proximal demonstrative do that indicates the very recent temporal location of the event (marked with the subscript t), as in (70a). This example illustrates that the position to encode tense is closer to the verb phrase than the position for encoding source of information. In (70b), the demonstrative do indicates the assertion of the reported information (marked with the subscript a). The meaning of do is rendered with the English ‘indeed’. In (70c), a layered structure is presented, where ba again encodes the reported source (marked with the subscript e), the first do encodes tense and the sentence-final encodes assertion.
 František Kratochvíl
(70) a. ne-pikaai do it-i namur do ba 1s.al-head prx lie.on-pfv wound.cpl prxt saye ‘my head reportedly just got wounded’ b. ne-pikaai do it-i namur ba do 1s.al-head prx lie.on-pfv wound.cpl saye prxa ‘reportedly my head is wounded indeed’ c.
ne-pikaai do it-i namur do ba do 1s.al-head prx lie.on-pfv wound.cpl prxt saye prxa ‘my head reportedy just got wounded indeed’
Note that examples such as (70) are found mainly in conversations. The example given, however, is elicited. A systematic collection of conversation data will provide more information about the possible combination of the sentence final markers in Abui.
4.4â•… Summary The utterance verb ba ‘say’ is the source of grammatical markers such as topic marker, quotative marker, and linker of noun-modifying, complement, simultaneous and purpose clauses. The common denominator for the usages of ba, summed up in Table 4, is its syntactic position of a constituent introducing or following a quotation. Note that the various functions of ba in the clause and sentence are complementary to the functions of demonstratives in the same syntactic position. Table 4.╇ Functions of Abui marker ba in different syntactic domain domain
structure
function
np
dems (poss-) n baLNK+ nmc dema np baTOP
baLNK modifier linker baTOP topic marker
clause
a. {(in)direct speech} baQT} b. {mc} baLNK {non-factive complement}
baQT quotative marker baLNK complement linker
sentence
a. {dependent.clause} baSIM {final.clause} b. {main.clause} baPURP {purpose.clause} c. {sentence} bae
baSIM simultaneous linker baPURP purpose linker bae evidential marker
5.â•… Conclusion Grammatical categories of spatial deixis, tense, and anaphora have long been treated as different types of deixis in the spatial, temporal, and textual domain (cf. Fillmore 1975; Lyons 1977; Levinson 1983: 62–63). In many languages, spatial co-ordinates can be applied to more abstract domains such as time based on metaphors such as space → time. Metaphors indicate similarity between the conceptual domains and
Discourse-structuring functions of Abui demonstratives 
thus facilitate the mapping (cf. Gentner, Imai et al. 2002: 557). The points on the spatial co-ordinates applied to other domains (anaphora, tense, evidentiality, assertion) preserve a partly their value: a possible case of heterosemy. There is some evidence that the space→time metaphors are processed as mappings from space to time and not as two different coherent systems (cf. Boroditsky 2000: 26). However, the more abstract domains are not structured in the same way as space. The gaps that cannot be mapped to the spatial co-ordinates may be filled with markers from other sources, such as the marker ba originating in the speech verb ba ‘say’, as summarized in Table 5. Table 5.╇ Grammatical categories expressed by Abui demonstratives and marker ba space
discourse
time
source
proximal medial distal n.a.
recent mention earlier mention early mention/new topic ba
recent past or near future earlier or later long ago/unspecified simultaneous ba
witnessed (first hand) witnessed (past) remembered reported ba (say)
Demonstratives can be extended to map other domains such as evidentiality, or assertion (e.g. English so). Abui data show the extension may involve the entire paradigm. It is possible that similar systematic extension can be found in many other languages but has been overlooked in earlier studies. Abui data presents a challenging case of grammaticalization of demonstratives into anaphoric pronouns, tense, evidential and assertion markers. A short note is in order here considering the grammatical status of the demonstratives. While in other languages discussed in this volume a single morpheme, identified as nominalizer, can cover similar functions as Abui demonstratives, it seems problematic to speak of Abui demonstratives as nominalizers. The reason is that a complete paradigm is transferred to map more abstract grammatical domains. Nominalizations and Abui constituents marked with demonstratives (adverbial, noun-modifying, purpose clauses, assertions, clauses marked for source) are largely semantically parallel and fulfil similar functions.21 Unlike the typical predicative clauses, clauses marked with demonstratives have referential function (establishing temporal location, identifying a property, source of information, or adding information in context). These functions are also found in nominalizations, but the Abui demonstratives make the referential clauses more specific by applying the spatial grid to categories such as tense, evidentiality or assertion.
.â•… As one of the reviewers summarized, the anaphoric demonstratives have in fact taken on roles of clausal ‘specifiers’. In this new role the anaphoric demonstratives record the tense.
 František Kratochvíl
Abbreviations # 3a 3i 3ii al cnt cpl dst dstr dur e g:c i icp incp.c inal inter lnk loc md md.ad
pause 3rd person A 3rd ps. controlling U 3rd ps. non-controlling U alienable continuative completive distal distributive durative exclusive grandchild inclusive inceptive inceptive completive inalienable interjection linker location-type U medial addressee-based medial
neg nmc pat pfv phsl.c p pl pnct prx prx.ad purp qt rec red say s sim spc spc.ad top y:si
negator noun-modifying construction patient-type U perfective phasal completive plural plural quantifier punctual proximal addressee-based proximal purpose clause linker quotative recipient-type U reduplication hearsay evidential singular simultaneous specific addressee-based specific topic marker Younger sibling
References Boroditsky, Lera. 2000. Metaphoric structuring: Understanding time through spatial metaphors. Cognition 75: 1–27. Dryer, Matthew S. 2007. Clause types. In Language Typology and Syntactic Description. Volume I: Clause Structure, Timothy Shopen (ed.), 224–175. Cambridge: CUP. Fillmore, Charles J. 1975. Santa Cruz Lectures on Deixis, 1971. Bloomington IN: Indiana University Linguistics Club. Foley, William A. 1986. The Papuan Languages of New Guinea. Cambridge: CUP. Gentner, Dedre, Imai, Mutsumi & Boroditsky, Lera. 2002. As time goes by: Evidence for two systems in processing space → time metaphors. Language and Cognitive Processes 17: 537–565. Güldemann, Tom. 2008. Quotative Indexes in African Languages: A Synchronic and Diachronic Survey. Berlin: Mouton de Gruyter. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
Discourse-structuring functions of Abui demonstratives 
Holton, Gary, Klamer, Marian & Kratochvíl, František. 2009. The languages of Alor and Pantar (Eastern Indonesia): A (re)assessment. Paper presented at the Eleventh International Conference on Austronesian Linguistics (ICAL-11), Aussois, France, 22–26 June. Kiparsky, Paul & Kiparsky, Carol. 1970. Fact. In Progress in Linguistics: A Collection of Papers, Manfred Bierwisch & Karl Erich Heidolph (eds), 143–173. The Hague: Mouton. Klamer, Marian. 2000. How ‘report’ verbs become quote markers and complementisers. Lingua 110: 69–98. Kratochvíl, František. 2007. A Grammar of Abui: A Papuan language of Alor. Utrecht: LOT. Levinson, Stephen C. 1983. Pragmatics. Cambridge: CUP. Levinson, Stephen C. 2003. Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge: CUP. Lichtenberk, Frantisek. 1991. Semantic change and heterosemy in grammaticalization. Language 67(3): 475–509. Lyons, John. 1977. Semantics. Cambridge: CUP. Matsumoto, Yoshiko. 1997. Noun-modifying Constructions in Japanese: A Frame-semantic Approach [Studies in Language Companion Series]. Amsterdam: John Benjamins. Mithun, Marianne. 1987. The grammatical nature and discourse power of demonstratives. BLS 13: 184–194. Munro, Pamela. 1982. On the transitivity of ‘say’ verbs. In Studies in Transitivity, Paul Hopper & Sandra A. Thompson (eds), 143–160. New York NY: Academic Press. Næss, Åshild. 2004. Spatial deixis in Pileni. In Deixis and Demonstratives in Oceanic Languages, Gunter Senft (ed.), 81–97. Canberra: Pacific Linguistics. Ozanne-Rivierre, Françoise. 2004. Spatial deixis in Iaai. In Deixis and Demonstratives in Oceanic Languages. Günter Senft (ed), 129–39. Canberra: Pacific Linguistics. Pawley, Andrew. 2001. The Proto Trans New Guinea obstruents: arguments from top-down reconstruction. In The Boy from Bundaberg: Studies in Melanesian Linguistics in Honour of Tom Dutton, Andrew Pawley, Malcolm Ross & Darrell Trevor Tryon (eds), 261–300. Canberra: Pacific Linguistics Research School of Pacific and Asian Studies Australian National University. Ransom, Evelyn N. 1988. The grammaticalization of complementizers. BLS 14: 364–374. Reesink, Ger. 1994. Domain-creating constructions in Papuan languages. In Topics in Descriptive Papuan Linguistics, Ger P. Reesink (ed), 98–121. Leiden: Vakgroep Talen en Culturen van Zuidoost-Azië en Oceanië Rijksuniversiteit te Leiden. Ross, Malcolm. 2005. Pronouns as a preliminary diagnostics for grouping Papuan languages. In Papuan Pasts: Investigations into the Cultural, Linguistic and Biological History of the Papuan-speaking Peoples, Andrew Pawley, Robert Attenborough, Robin Hide & Jack Golson (eds), 15–66. Canberra: Pacific Linguistics. Saxena, Anju. 1988. On syntactic convergence: The case of the verb ‘say’ in Tibeto-Burman. BLS 14: 375–388. Stalnaker, Robert. 1998. On the representation of context. Journal of Logic, Language, and Information 7: 3–19. Stokhof, Willem A.L. 1984. Annotations to a text in the Abui language (Alor). Bijdragen tot de Taal-, Land- en Volkenkunde 140(1): 106–162. Traugott, Elizabeth. 1995. Subjectification in grammaticalization. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein & Susan Wright (eds), 31–54. Cambridge: CUP.
 František Kratochvíl de Vries, Lourens. 2006. Areal pragmatics of New Guinea: Thematization, distribution and recapitulative linkage in Papuan narratives. Journal of Pragmatics 38: 811–828. Yap, Foong Ha, Matthews, Stephen & Horie, Kaoru. 2004. From pronominalizer to pragmatic marker: Implication for unidirectionality from a crosslinguistic perspective. In Up and down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 137–168. Amsterdam: John Benjamins.
Index
A actor focus╇ 24, 500, 590, 595, 654 adnominal╇ 1, 19, 31, 75, 77, 98, 102, 396, 402, 411, 418, 448, 451, 457, 634, 648–649 adnominalization╇ 78, 103 genitivization╇ 49, 61, 73, 78, 83, 90, 98, 256, 289, 293, 323, 446 relativization╇ 26–27, 31, 34, 49, 61, 73, 78, 82, 85, 98, 101–102, 147, 152–153, 163, 166, 189–190, 216, 226, 241–242, 246, 251, 256, 270, 289–290, 293–294, 297, 308, 313, 323, 446, 499–503, 509–510, 514, 519–520, 523, 525–526, 531, 533, 540, 553, 555, 569, 633–634, 685–686, 706, 712–713, 715–717, 732 adverbial marker╇ 638, 640 adverbialization╇ 78 agentive noun╇ 180, 516–517, 519 analogy╇ 32, 66, 73, 206, 307–308, 724, 748, 781 anaphora╇ 367, 724, 776–777, 784–785 anaphoric demonstrative╇ 662, 761, 767 anaphoric pronoun╇ 785 anti-agentive particle╇ 290 article system╇ 591–592, 616, 618 defining article╇ 630 personal name article╇ 614 aspect completive╇ 38, 208, 284, 582, 717, 769–770, 782, 786 habitual╇ 34, 36, 209, 223, 232–233, 235, 240–241, 251,
264, 284, 508, 517, 519, 542, 554, 570–571 imminent aspect╇ 34, 229, 240, 251 imperfective╇ 35, 172, 180, 197, 210, 222–223, 229, 232, 247–248, 284, 698, 718, 782 inceptive╇ 210, 562, 782, 786 perfective╇ 4, 33, 37, 123, 148, 191, 197, 203, 206, 210–211, 274, 281, 285, 331, 345, 353, 373, 375, 420, 500, 511, 524, 531, 540, 548–549, 563–564, 570–571, 573–574, 576, 580–585, 654, 726, 770, 778, 782, 786 progressive╇ 208, 210–211, 247–248, 420, 452, 531, 548–549, 554, 662 reciprocal╇ 210, 524, 546, 549, 694 resultative╇ 37, 123, 210, 371, 375, 582–584 associative╇ 71, 90, 110, 113–114, 123, 326, 645, 654, 701–702, 712, 717 attributive╇ 112, 114, 118, 123, 148, 156–157, 163, 182–185, 199–202, 205, 210, 219, 225, 228, 240–241, 244, 251, 258, 381, 436, 474, 476, 486–490, 503, 505–506, 509, 631, 700 adjectival attributive╇ 114 nominal attributive╇ 118 non-relative attributive╇ 199–202, 205 attributive form (versus conjunctive form) C case-marking concord╇ 25 classifier╇ 14, 21, 80–81, 98, 103–104, 109–110, 112–115,
122–123, 129–130, 133, 145, 148–149, 153, 190, 261, 284, 317, 319–320, 325–326, 329–330, 500, 654, 717 clausal subordination╇ 44, 47, 61–62, 75, 86, 101–102, 281, 294 clause chaining structure (see converbal construction) cleft construction╇ 231, 332, 334, 445, 447, 450, 468, 604, 631, 636 clitic╇ 10–11, 22–23, 26, 41, 62, 79, 85–86, 95–96, 98, 100–101, 173, 175, 188–189, 296, 329, 350, 381, 528, 545–546 enclitic╇ 19, 33, 40, 258–259, 269, 376, 504, 631, 637–644, 651–653 proclitic╇ 306, 350–352, 356, 613, 734 pronominal clitic╇ 22–23, 26, 62, 79, 85–86, 95–96, 98, 100–101, 545–546 complement copula complement╇ 182, 274–275, 284, 306, 596, 779 factive complement╇ 781, 784 non-factive complement╇ 781, 784 complement clause╇ 48, 171–174, 178, 220, 260, 267, 282, 314, 329, 334, 340, 428, 451, 473, 476, 480–481, 483, 492, 516, 527, 536, 594, 607, 611–612, 614, 617, 620, 622, 647 nominal complement clause╇ 178 verbal complement clause╇ 163, 170
 Index complementation╇ 74, 163, 170–173, 199, 216, 241, 256, 290, 313, 323, 334–336, 338, 340, 528, 540, 636, 667 complementizer conditional╇ 44, 48, 62, 74–75, 79, 86, 97–98, 126–127, 134, 138–139, 175, 197, 207, 209–210, 251, 284, 313–314, 334–339, 469, 473, 536, 598, 748–749 factual╇ 248, 314, 334–336, 338, 645 purposive╇ 35, 73–74, 79, 86, 88–89, 97, 99, 103, 282, 301, 303, 480, 487, 511, 570, 576–577, 579 complementizer marking (versus subject marking)╇ 462 conclusive form (also final form; versus nonconclusive form)╇ 427, 429, 488–490 conjunctive form (versus attributive form)╇ 487, 489 converbal construction (or clause chaining structure)╇ 46, 245 copula╇ 14, 35–36, 39, 43–44, 46, 81, 92–96, 104, 119, 121, 123, 136–137, 154, 157–158, 170, 172, 182, 191, 203, 210, 222–223, 229–233, 235, 241, 243, 247, 251, 258–259, 274–275, 280, 284, 305–306, 309, 314, 331–334, 339, 343–356, 369, 371, 383, 388, 396–397, 413–414, 416, 420, 423–424, 428–431, 434–441, 446, 449–450, 467, 469, 491, 516, 526, 535, 542, 596, 604, 639, 661, 666, 779 copula ellipsis╇ 43, 439, 542 copula-obligatory╇ 43 copula-less╇ 43, 526, 542 copula-support╇ 535 copular subject╇ 594, 596–604, 606–607, 611–615, 618–622 copular complement╇ 594, 596–601, 604–607, 611, 613, 615, 620–622
zero copula╇ 437, 440, 443 copular clauses╇ 596–598 identificational clause╇ 597, 603, 613, 617, 620–621 predicational clause╇ 597–604, 613, 617–619, 621 pseudocleft clause╇ 597, 604–607, 613–614, 617, 620, 622 specificational clause╇ 597–600, 602–604, 613–614, 617, 620–622 D decategorialization╇ 561, 584–585 defective element defective indexical╇ 143–144 defective noun╇ 11, 395–398 deixis╇ 18, 49, 769, 784 distal╇ 49, 191, 239, 251, 258, 284, 299, 440, 532–533, 591–592, 612–613, 619, 621, 630, 680, 758–759, 762–767, 770, 772–775, 785–786 medial╇ 168, 170, 174, 181, 190, 452–453, 484, 492, 592, 612, 619, 631, 759, 762–765, 770, 773, 776, 785–786 proximal╇ 47, 49, 251, 285, 299, 592, 655, 680, 758–759, 762–764, 768–769, 771, 773, 775–776, 783, 785–786 defective noun╇ 11, 395–398 definite article╇ 314, 319–321, 323–325, 327–328, 612 definiteness marker (see also identifiability marker)╇ 2, 14, 16–19, 22, 49, 289, 293, 307–309, 321, 322–324, 334, 339, 608, 630, 638, 643, 650–651, 654 demonstrative (see also deixis)╇ 1–2, 14, 15, 17–19, 27, 31, 32, 38, 40, 47, 89, 95, 113, 127, 130, 143, 153, 169, 218 anaphoric demonstrative╇ 662, 761, 767
distal demonstrative╇ 191, 239, 251, 258, 440, 532, 591, 613, 621, 758, 765–766, 773 medial demonstrative╇ 631, 763, 770, 776 proximal demonstrative╇ 47, 251, 758, 764, 771, 773, 776, 783 de-referentialization╇ 618, 623 determiner╇ 14, 18, 211, 280, 319, 323, 328–329, 334, 380, 385, 527, 537, 659, 665, 667, 669–671, 673–675, 678–679, 681, 729, 731–732, 734–735, 744 deverbal noun╇ 294, 315–317, 687–688, 709, 717 deverbalization (versus substantivization; see also nominalization strategies)╇ 196, 487 diachronic development (see also historical development)╇ 49, 109, 110, 387, 431, 434–435, 629, 653 E elaboration╇ 215–218, 228, 241–242, 244–246 elimination╇ 215–217, 241–242, 250 ellipsis╇ 43, 151, 158, 265, 268, 402, 410, 412, 436, 439, 542, 676 exclamative (see also mirative)╇ 6, 19, 41, 43, 250, 305, 424, 426, 428, 430–434,537, 590, 594, 609, 614, 676, 679, 680, 725 nominal exclamative╇ 676–677, 681 clausal exclamative╇ 669, 681 exhaustive identification╇ 606 extension metonymic extension╇ 577–578, 584 semantic extension╇ 31, 47, 90, 216, 318, 533, 576, 578, 584, 638, 653 ezafe╇ 17–18, 34, 363–373, 375–383, 385–388 adnominal ezafe╇ 371
Index  anaphoric ezafe╇ 367–369, 371 demonstrative ezafe╇ 368–369, 373 tense ezafe╇ 370–373, 375–376, 385, 387–388
F finite construction╇ 343–344, 346, 351–352, 354 focus construction╇ 6, 265, 276–278, 333–334, 464, 716 focus system, Austronesian (see also voice system) actor focus╇ 24, 500, 590, 595, 654 circumstantial focus╇ 574, 595 instrumental focus (see instrumental voice) locative focus╇ 500, 510, 512, 590, 595 non-actor focus╇ 24 patient focus╇ 25, 36–38, 95, 500, 583, 590, 595 focus particle╇ 19, 81, 191, 281, 430, 636 focusing strategy╇ 23, 25–26 agreement-based focusing strategy╇ 26 contrastive focus╇ 93, 305, 631, 634–637 marked argument structure╇ 2, 22 formal noun (see defective noun and light noun) free relative╇ 767 G genitive agent╇ 147, 148, 165, 641 genitive subject╇ 19, 22, 23, 42, 83, 88, 94, 95, 255, 261, 265, 266, 271, 272 genitivization (see adnominalization) gerundive╇ 34, 83, 147–155, 157–158, 321, 516, 518–519, 537, 543, 551, 564, 638, 650, 692, 724, 740, 743–744 grammaticalization recursive (or cyclical)╇ 12
H head-internal relative (see headless relative clause) historical development (see also diachronic development)╇ 163, 167, 195–196, 202–205, 253, 364, 438, 528 I identifiability marker (see definiteness marker) infinitival╇ 220–221, 525, 607, 612 illocutionary force╇ 40–41, 44, 373, 400–401, 403–404, 659–661, 663, 665, 725, 750 inalienable possession╇ 20, 218, 227, 239, 240, 251, 750 instrumental noun╇ 563 insubordination╇ 483–485 J juxtaposition╇ 28, 31, 90, 482 K kakari musubi (see also focus construction)╇ 445, 448, 457, 461, 472, 517 keishiki meishi (see also formal noun)╇ 454 L language contact╇ 210, 217, 335, 439, 627, 629, 644, 646, 653 left-dislocation╇ 600–602, 604, 606 ligature (or linker)╇ 32, 562, 570, 630–631, 637, 701 light noun╇ 11, 22, 28–31, 44, 48, 61–67, 69, 72–79, 81–85, 87–89, 91–92, 95, 97–98, 100–103, 308, 318–319, 323–324, 339–340, 510, 634 locative noun╇ 22, 65, 79–82, 85–86, 90–92, 97, 165, 510 M main clause infinitive╇ 483 mental staging function╇ 27, 44, 78
mirative (see also exclamative)╇ 19, 26, 34, 39, 49, 199, 202, 225, 234, 348, 420, 434, 645, 725 mirativity (see also stance) amazement╇ 38 counter-expectation╇ 38, 284, 645 surprise╇ 8, 16, 34, 38–39, 76–77, 234–237, 426, 483, 525, 527, 537, 542, 544, 546–547, 627, 639, 645, 671–672, 676, 774 modality╇ 168, 208, 220, 276, 345, 401, 404, 484–485 mood╇ 1–2, 5–6, 8, 18, 26, 33–34, 36, 38–41, 43, 49, 62, 73, 76–78, 88, 98, 100, 120, 125, 209, 216, 218, 229–230, 249, 291, 293, 314, 331–333, 484, 489, 551, 595, 627–628, 640, 643–645, 650, 652–653, 668, 758, 761, 770 admonitive╇ 281, 284 affirmative╇ 209, 249, 372–373 assertive╇ 77, 120, 126–127, 135–140, 143–144, 276, 280, 331–333, 465, 647, 717, 724, 737 counterexpectation (see also mirativity)╇ 465 declarative╇ 112, 168, 191, 276–277, 280–281, 284, 305, 309, 331–332, 352–353, 355, 372–373, 400–401, 406, 408–409, 412, 420, 474, 659, 724 dismissive╇ 464–465 dubitative╇ 77 intentional╇ 209–210 irrealis (see also non-realis; versus realis)╇ 6, 33, 35, 169, 170, 207, 209, 218, 230, 266, 276, 280, 345, 355, 388, 511, 574, 770 imperative╇ 155, 197, 246, 251, 284, 309, 347, 400, 404, 450, 467, 469, 474, 486, 524, 530, 641–642, 659–660, 775–776
 Index interrogative╇ 29, 72, 76–77, 112, 118, 138, 197, 251, 277, 298, 332–333, 373, 383, 406, 415, 601, 659, 669, 705, 716 non-realis (see also irrealis; versus realis)╇ 34 realis╇ 34, 36, 263–264, 266, 268–270, 274, 276, 279, 285, 293–294, 302, 305–306, 308–309, 331–332, 355, 449, 568, 590, 595 suppositional╇ 210–211, 509 N nominal derived nominal╇ 3, 13, 150, 187–188, 431, 517, 542 gerundive nominal╇ 149–153, 155, 157–158, 321 predicate nominal╇ 39, 156, 424, 430–431, 601 unattached nominal╇ 594, 611, 613, 617 nominal complement clause╇ 178 nominalist hypothesis╇ 34 nominality╇ 523, 525, 531, 538–539, 541, 735 nominalization action nominalization╇ 3, 165–166, 296–297, 542, 644, 686–687, 709 agent nominalization╇ 22–23, 26, 66, 68, 74–75, 101–102, 186, 322, 508, 511, 516, 519 argument nominalization (see participant nominalization) clausal nominalization╇ 5–6, 15, 25, 44, 64, 101, 154, 163–166, 170–172, 174–176, 181, 190, 229, 244, 247, 255, 265, 272, 274, 276, 280, 283, 289–290, 293–294, 299–301, 306–309, 343, 426, 512, 523, 525–526, 531, 551, 554, 564–565 cognate nominalization (compare with cognate relativization)
derivational nominalization╇ 163–166, 181, 185, 188–190, 216, 289–290, 293–296, 298, 313, 315, 554 direct nominalization╇ 487–488 double nominalization╇ 12, 85–86, 100, 533, 629, 646, 650, 707–710 embedded nominalization╇ 7–8, 49, 170, 178–180, 241 event nominalization╇ 3, 19, 100, 117, 266–267, 270, 274, 282–283, 542 exclamative nominalization╇ 739 free-standing nominalization (see stand-alone nominalization) gerundive nominalization╇ 147–148, 151, 153, 155, 158 grammatical nominalization╇ 565 ‘headless’ nominalization╇ 675 ‘hidden’ nominalization╇ 737 ‘high nominalization’╇ 722, 736, 738, 746, 747, 748 inflectional nominalization╇ 437 instrument nominalization╇ 186 lexical nominalization╇ 5–7, 20, 64, 164, 265, 426–427, 500, 503, 511–513, 518, 526, 551, 561, 563–565, 571–574, 577, 581, 584–585 locative nominalization╇ 11, 22, 97, 186–187, 316 main-clause nominalization (see also stand-alone nominalization)╇ 423, 424, 434, 441 non-embedded nominalization (see also stand-alone nominalization)╇ 7, 43, 170, 180, 231, 240, 246, 253
participant nominalization (see also argument nominalization)╇ 3, 4, 7, 46, 166, 185, 187–189, 275, 296, 636 patient nominalization╇ 25–26, 62, 83–85, 90, 97–99, 101, 186–187, 322 stand-alone nominalization╇ 40–42, 49, 305, 331–333, 423, 431, 434, 437–438, 441, 647, 779 syntactic nominalization╇ 313, 561, 563–565, 571–574, 579, 581, 585 temporal nominalization╇ 54, 304 zero nominalization╇ 2, 13, 14, 22, 257, 262, 263, 298, 525, 532, 542, 546 nominalization strategies╇ 9 deverbalization╇ 169, 487 substantivization╇ 2, 14, 89, 528 nominalized main clause (see stand-alone nominalization) nominalizer agent nominalizer╇ 4, 19, 25, 30–31, 44, 62, 66–67, 72–73, 75, 77–78, 82, 100, 103, 315 affixal nominalizer╇ 32 circumfixal nominalizer╇ 10 erstwhile nominalizer╇ 16, 223, 233, 343 infixal nominalizer╇ 10 morphological nominalizer╇ 26, 33, 280, 315 patient nominalizer╇ 4, 62, 78–79, 81, 83, 97, 315 prefixal nominalizer╇ 10, 15 primary nominalizer╇ 264, 266 post-predicate nominalizer╇ 9, 33 secondary nominalizer╇ 11, 68, 264–266 sentential nominalizer╇ 313, 319, 335–336, 451
suffixal nominalizer╇ 8, 39 versatile nominalizer╇ 2, 4, 81, 243, 317, 337, 637 nominalizer accretion╇ 12 nominalizing derivative (see also derivational nominalization)╇ 117 non-conclusive form (also non-final form; versus conclusive form) infinitive╇ 74, 169, 208–209, 483, 525, 544, 741 gerund╇ 321–322, 474, 511, 516, 523, 525, 544, 590, 691–692, 725, 741–745, 747 participle╇ 191, 371, 374, 378, 388, 581, 746 non-referential╇ 1–2, 25–27, 33, 35, 38, 44–46, 49–50, 73–74, 92, 96–97, 589–591, 593–594, 597, 600–601, 603–604, 606–607, 609–614, 616–623, 627–629, 637, 650, 652, 701 noun phrase marker (see referentiality marking device) P passive construction╇ 99, 527, 529, 531, 533, 547 passive voice╇ 61, 96, 524, 530–531, 539, 546–547, 549, 555, 581, 583 passivization╇ 92–93, 97 pivot (see also topic)╇ 24, 113, 122, 379, 644, 733 possessee pronominal╇ 62, 72–73, 79, 89–92, 97–98, 103 pragmatic marker╇ 27, 646, 653, 757 presupposition╇ 127, 139–140, 274, 278, 660, 716–717, 724, 750 pronominalization╇ 600, 603, 606, 677 prosodic pause╇ 128 proximity concord╇ 532 pseudo-cleft╇ 23, 428
Index  Q quotative (see also reportative)╇ 171, 317, 340, 469, 669, 759, 798 R reduplication╇ 12, 111, 251, 337, 524, 531, 542, 548–549, 551–552, 554, 562–563, 565, 574–575, 578, 590, 610, 691, 714, 718, 728, 734–735, 741–742, 744, 747, 786 CV reduplication╇ 728, 741–742 lexical reduplication╇ 549 verbal reduplication╇ 531 referential╇ 261, 591, 622 referentiality╇ 22, 49, 102, 151, 589, 591–594, 597, 601, 603, 606, 612, 617–619, 628, 637, 654 definite╇ 24, 63, 90, 113, 122, 130, 140, 156, 210–211, 245, 251, 267, 289, 297, 299, 303, 306, 309, 314, 319–325, 327–328, 330, 333–334, 365, 379, 384, 388, 500, 593–594, 596, 600, 603, 612, 619, 622, 654, 676, 679, 722, 750 generic╇ 70, 75, 84, 122, 177, 243–244, 271, 294, 296, 307–309, 385, 528, 536, 593, 757 indefinite╇ 63, 72, 91, 112, 245, 251, 319–320, 322, 325, 330, 366, 378–379, 388, 589, 593, 596, 627, 662, 678, 722, 724, 729 referentiality marking device╇ 151, 628 case marker╇ 15–16, 24–26, 32–33, 35, 46–47, 62, 75, 83, 93–96, 98–99, 103, 280, 376, 380–381, 383, 439, 449, 501, 505–507, 511–516, 525, 527–528, 531–533, 538, 571, 580, 631, 637, 722, 735 classifier╇ 14, 21, 80–81, 98, 103–104, 109–110, 112–115, 122–123, 129–130, 133, 145, 148–149, 153, 190, 261, 284,
317, 319–320, 325–326, 329–330, 500, 654, 717 definiteness marker╇ 16–17, 19, 289, 293–294, 296–297, 299–300, 303–305, 307–309, 314, 317, 319, 321–322, 324, 328–329, 334, 339, 608, 630, 643, 650, 654 demonstrative╇ 16–19, 24–26, 31–34, 41, 47, 63, 66–67, 98, 102, 104, 112–115, 123, 129–130, 133, 142, 144, 153, 169, 191, 239, 245, 251, 258, 261, 285, 295, 299–300, 320–321, 326, 363–365, 367–369, 371–376, 383–388, 418, 440, 525, 528, 532–533, 538, 570–571, 591, 593, 607, 612–614, 618–619, 621, 631–632, 637, 654, 662, 666–667, 669, 679–681, 690, 758, 761, 763–771, 773, 775–777, 783 plural marker╇ 20, 590, 722 possessive pronoun╇ 10, 14, 20, 240, 638, 644 quantifier╇ 63, 123, 129, 177, 251, 380, 705, 767, 779, 786 relabeling (see also structural pruning)╇ 66 relative clause externally-headed relative clause╇ 28, 271 headless relative clause╇ 41, 122, 268, 301–302, 334, 445, 447–448, 452, 455, 457, 468, 503, 506–507, 510–511, 513, 516, 518–519, 533, 584, 604, 669 internally-headed relative clause (see also headinternal relative)╇ 101, 176, 261, 269 light-headed relative clause╇ 28 post-head relative clause (or postnominal relative clause)╇ 177 pre-head relative clause (or prenominal relative clause)╇ 302 primary relative clause╇ 28
 Index secondary relative clause╇ 28–30, 69 relative pronoun╇ 340, 383–386, 418 relativization (see also adnominalization) cognate relativization╇ 717 relativizer╇ 19, 21, 25, 28–33, 62–64, 67–72, 75, 81, 84–85, 98, 101–104, 115, 123, 153, 191, 298, 302, 323, 328–329, 333, 338–340, 363–364, 385–386, 466, 504, 532, 633–634, 644, 648–649, 652, 654, 662, 670, 674 rentaikei (or adnominal form)╇ 13, 15, 33, 445–449, 452–457, 459–460, 466–468 reportative (see also quotative)╇ 251, 340, 405, 773, 782, 783, 788 S semi-adjectival╇ 732 semi-nominal╇ 732 semi-verbal╇ 737 sentence final copula╇ 331 sentence final particle (see also sentential ending)╇ 8, 39–41, 43, 44, 65, 76–77, 81, 89, 104, 126, 135, 145, 148, 168, 180, 218, 306, 332, 333, 350–354, 441, 452, 462–464, 469, 645–647, 650, 654 sentence final prosody╇ 76–78 sentential adverb╇ 27, 590, 608, 612, 640
sentential ending (see also sentence final particle)╇ 397–402, 404, 406, 408, 411–412, 415, 417–420 shushikei (or finite form)╇ 13, 14, 33, 460 stance╇ 18, 627 epistemic mood╇ 652 evidentiality╇ 40, 168, 204, 345–346, 405, 640, 757–759, 761, 771–777, 782, 785 mirativity (see also exclamative)╇ 8, 20, 38, 202, 217, 228, 234, 240, 241, 249, 250, 402, 436, 650 speaker attitude╇ 8, 27, 39, 234, 249–250, 542 speaker perspective╇ 542–543 structural pruning (see relabeling) subject marking (versus complementizer marking) subordination (see clausal subordination; versus insubordination) subordinator╇ 44–46, 48, 62, 74–75, 86, 197, 304, 337, 473, 508–509, 536, 648, 650, 725, 743–744, 747, 749 substantivization (versus deverbalization; see nominalization strategies) syncretism╇ 35, 39, 49, 204, 217, 224–226, 241–242, 246, 251, 256, 283, 289–290, 294, 297, 308, 323, 446, 465
T temporal reference╇ 49, 155, 768 tense ezafe╇ 370–373, 375–376, 385, 387–388 topic (see also pivot)╇ 5, 31, 33, 74, 123, 233, 234 topic marker╇ 104, 175, 258, 272, 281, 294, 296, 299–300, 303–306, 317–318, 336–339, 528, 630, 637, 644, 655, 765, 778–779, 784, 786 V verb concatenation╇ 198 verb periphrasis╇ 436 verbality╇ 539, 541 voice system, Austronesian (see also focus system, Austronesian) actor voice╇ 34, 539, 562–563, 566, 568–569, 574, 722–723, 728–730, 732, 736, 739, 741–743, 746 circumstantial voice (see also circumstantial focus) conveyance voice╇ 34 instrumental voice╇ 562, 567, 577–578 locative voice╇ 510, 539, 562, 574–575, 579 non-actor voice╇ 563, 574 patient voice╇ 34, 539, 562, 566, 574–575, 579–581, 583, 722, 741 undergoer voice (see also patient voice or patient focus)
Language index
A Abui╇ 18, 22, 27, 34, 38, 40, 47, 50, 533, 757–766, 768–772, 774, 776–779, 781, 784–785 ‘Are‘are╇ 694, 697 Arosi╇ 690–691, 697 Athpare╇ 246, 248, 280 B Bahdini Kurdish╇ 34 Belhare╇ 243, 249–250, 305 Bantawa╇ 222–223, 244, 246–247, 249–250 Bolaang-Mongondow╇ 727 Botolan Sambal╇ 42, 731, 742, 754 Budai Rukai╇ 8, 15–16, 24–26, 32–33, 39, 45, 64, 523–524, 529 Bugis╇ 727, 738, 746–747, 754 Buru╇ 41, 780–781 C Cantonese╇ 3–4, 8–9, 33–34, 39–40, 48, 76, 114, 122–123, 125–126, 129, 131–132, 134, 139, 142, 144, 147–148, 150–158, 250, 439, 647, 649 Catalan╇ 669–670 Cebuano╇ 34, 42, 728–729, 754 Chantyal╇ 4–5, 8, 38, 43, 77, 195–203, 205–207, 209, 216–217, 225, 243, 245, 248–249, 256, 276, 307, 364, 426, 433–436, 445, 466–468 Chaozhou╇ 4, 21–22, 76, 81, 109–114, 116–117, 119–122, 250, 439 Chinese╇ 2, 4, 9, 11, 13, 18–19, 21–23, 25–31, 33, 42–45, 48, 50, 61–70, 72–73, 75–81, 85–103, 109, 112, 114, 117–120, 125, 135, 145, 147–151, 153–154, 156, 158,
313, 317, 320, 326, 328, 335–338, 364, 423, 439–440, 457, 482–483, 500, 510, 569, 578, 628–629, 639, 646–649, 652, 654, 668, 748 Old Chinese╇ 13, 18–19, 22–23, 27, 29–31, 42, 44–45, 61–67, 69–70, 72–73, 75–76, 78–79, 81, 85–87, 90–95, 100–102, 149, 510, 748 Middle Chinese╇ 19, 29–31, 61, 63, 68–70, 72–73, 77, 79, 81, 89, 91–92, 95, 97–98, 102, 646, 648 Modern Chinese╇ 29–31, 48, 62, 65–66, 73, 75, 77–78, 85, 89, 95–96, 98–99, 102–103, 149 D Dumi╇ 222, 245, 248 Dzala╇ 17, 22 G Galo╇ 5–6, 11–12, 166, 255, 257–271, 273, 275, 277–284, 305 Ghale╇ 195, 204 Gurung╇ 199–200, 202, 204–205, 207–209, 238 I Iaai╇ 671, 770 Ilokano╇ 38, 589, 591, 619, 621–623, 730, 740–741, 754 Indonesian╇ 15, 19, 22, 27, 40, 42, 50, 569, 589, 608, 629, 638, 640–641, 643–644, 721, 744–745, 754, 760 Iranian╇ 1–2, 17–18, 22, 34, 38, 50, 361, 363–364, 370, 376, 381–383, 385–387 Ivatan╇ 731, 737, 740, 754
J Japanese╇ 1–2, 4–5, 7, 11–14, 16, 27–29, 32–33, 39, 43–44, 46–47, 50, 65, 75, 86, 90, 101, 110, 120, 122, 125, 175, 250, 316, 337, 391, 402, 423–434, 436–438, 440–441, 445–452, 456–460, 462–464, 466, 468–469, 473–476, 478–492, 508, 532, 628, 631, 646, 648, 668, 670, 757–758 Old Japanese╇ 14, 427–429, 431–434, 437–438, 445–446, 451–452, 466, 469, 476, 668, 670 Middle Japanese╇ 14, 429–431, 433–434, 437, 446, 449, 451, 469 Modern Japanese╇ 13–14, 29, 46, 424–425, 433–434, 436, 445, 447–449, 451, 469, 476, 488 K Kimaragang Dusun╇ 730 Kavalan╇ 26, 45, 499–503, 508–514, 516–520, 529, 536, 540, 544 Kham╇ 200, 205, 216–217, 223, 225, 243, 245–247, 249, 305, 307, 356, 433–434 Takale Kham╇ 214, 247, 249, 433–434 Korean╇ 1, 4–7, 9–12, 27–29, 32–33, 39, 43–44, 46, 50, 65, 75, 77, 86, 90, 110, 120, 122, 175, 250, 320, 391, 393–395, 397, 402–403, 407, 411, 417, 419, 426, 433–434, 438, 440, 473–481, 483–492, 532, 537, 543, 628, 631, 646, 648 Old Korean╇ 29
 Language index Middle Korean╇ 394–395 Modern Korean╇ 28–29, 393–395, 397, 411, 417 Kuki-Chin╇ 247, 343–344, 346, 349–353, 356–357 Kwara‘ae╇ 691, 695, 697 L Lahu╇ 90, 164, 224, 229, 231, 246, 276, 290, 323, 344, 354, 423–424, 426, 433–434, 445, 466–468 Lhasa Tibetan╇ 4, 28–29, 31, 65, 243, 344, 439–440 Limbu╇ 7, 222, 243, 248, 353 Lolo-Burmese╇ 35, 314, 352, 354–356 Longgu╇ 685, 690–691, 697 M Magar╇ 4–8, 10–11, 15, 20–22, 34–35, 39, 41, 43, 155, 200, 205, 215–223, 225–226, 228–229, 231–234, 236, 238–247, 249–251, 646, 780, 782 Malagasy╇ 27, 41, 50, 659, 661–666, 668–669, 671–673, 675, 677–681 Malay╇ 12–13, 15, 19, 22, 27, 40, 42, 50, 77, 125, 337, 457, 608, 616, 627–634, 637–638, 640–641, 643–644, 646–650, 653, 721, 739, 744–745, 754, 757–758, 760, 771 Mamasa╇ 746–747, 754 Manange╇ 35, 45, 167, 169–170, 172–175, 179–185, 203–204, 209, 250, 290 Mandarin╇ 4, 9, 33–34, 48, 76, 112–114, 117, 120–123, 125, 129–132, 134, 147–158, 188, 250, 318, 320, 326, 328, 335–338, 364, 373–374, 439, 482–483, 500, 569, 578, 639, 757–758 Mansaka╇ 728–729, 754 Maori╇ 671 Marau╇ 693, 697 Mongsen Ao╇ 4, 7–8, 10–11, 35, 46, 164–165, 167–168, 170–171, 174–177, 179, 183, 185–187, 189, 244, 250, 270, 280, 290, 628
Mori Bawah╇ 745–746, 754 Muna╇ 10, 736, 754 Murut╇ 727, 754 N Nar-Phu╇ 197–198, 203–204, 206–209 Nepali╇ 195, 198, 206, 209, 211, 218, 236, 241–242, 248–249, 251, 346, 348, 780 Newari╇ 9, 27, 46, 780 Dolakha Newar╇ 246 Numhpuk Singpho╇ 4, 16–17, 22, 90, 289–294, 300–301, 306–309 Nuosu Yi╇ 4, 17, 22, 64, 307, 313–318, 320–326, 328–329, 331, 333–340, 628 O Okinawan╇ 43, 50, 250, 445–446, 451–465, 468–469, 628, 646 Old Okinawan╇ 445, 451, 454, 458, 462, 469 Modern Okinawan╇ 445, 451–452, 455, 458, 461–462, 469 Oroha╇ 693, 697 P Pangasinan╇ 619, 731, 754 Pazeh╇ 47, 565, 569, 575 Pendau╇ 727, 754 Persian (Farsi)╇ 364, 383 R Rampi╇ 747–748, 754 Rawang╇ 20, 22, 43, 225, 227, 306, 440, 628 S Samoan╇ 16, 671 Saisiyat╇ 10, 12, 15, 22, 24, 32, 34, 36–37, 64, 518, 529, 561–571, 574–576, 580, 582, 584–585 Sarangani Manobo╇ 740, 742, 754 Seke╇ 197, 203 Setswana╇ 670 Sunwar╇ 35, 225, 233, 247, 249, 343, 346–349, 355–357
T Tagalog╇ 24–25, 27, 38, 42–43, 77, 138, 501, 537, 589–597, 599–600, 604, 608–610, 612, 615–619, 621–623, 631, 663, 721–726, 728–729, 732–736, 738–744, 746–747, 749–750, 754 Tamang╇ 197–198, 203, 206–207, 209 Thai╇ 669, 781 Thakali╇ 197, 203, 209 Tibetan╇ 1–2, 4, 27–29, 31, 33, 35, 38, 43, 65, 102, 110, 164, 167–170, 172, 176–177, 179, 181, 184, 188–190, 195–196, 200, 204–205, 209, 222, 239, 242–243, 246–248, 250, 255–258, 290, 306, 314, 319, 343–345, 348, 352–357, 433–434, 439–441, 532, 631, 668 Classical Tibetan╇ 31, 35, 204, 352–356, 433–434, 439, 668 Dongwang Tibetan╇ 4, 167–168, 170, 172, 176–177, 179, 181, 184, 188–190, 250, 290 Lhasa Tibetan╇ 4, 28–29, 31, 65, 243, 344, 439–440 Toqabaqita╇ 3, 5, 12, 50, 685–691, 697–698, 700–703, 705–706, 708–710, 714, 716–717 Tshangla╇ 195, 204, 247, 356 Tukang Besi╇ 738, 754, 780–781 Turkish╇ 138, 364, 669–670 Tuvaluan╇ 12–13, 16, 671, 710, 716 U Ulawa╇ 12–13, 692, 697, 709, 716 W Wambule╇ 222, 243, 245, 248–250 West Uvean╇ 671 Wolio╇ 42–43, 727–729, 744–745, 754 Z Zhuokeji rGyalrong╇ 4, 10, 167–168, 172, 174, 176–177, 184, 186–189, 250, 290