cover
next page >
Cover
title: author: publisher: isbn10 | asin: print isbn13: ebook isbn13: language: subject public...
43 downloads
916 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
cover
next page >
Cover
title: author: publisher: isbn10 | asin: print isbn13: ebook isbn13: language: subject publication date: lcc: ddc: subject:
Nominalizations Theoretical Linguistics (Routledge (Firm)) Koptjevskaja-Tamm, Maria. Taylor & Francis Routledge 0415060206 9780415060202 9780203214640 English 1993 P271.K67 1993eb 415
cover
next page >
< previous page
page_i
next page >
Page i Nominalizations Based on rich empirical material and clearly illustrated with examples from seventy languages, this book provides a thorough analysis of how nominalization types interact with other structural features. It focuses on action nominal constructions, and in particular, the comparison of their syntactic structure with that of finite clauses and of other noun phrases, a problem which has claimed much attention in current syntactic theories. The author’s research findings show that no language has syntactic means which are exclusively used in action nominal constructions, giving support to the view that constructions which appear later in language acquisition and language development are modelled on already existing constructions as far as possible. The book provides a detailed and consistent basis for constructing a typology of such derived constructions and for presenting cross-linguistic comparisons. The book also contains a detailed discussion of related phenomena, and the framework is applicable to the description of subordination, complementation and finiteness in various languages. By demonstrating the complexity in the relations between noun, verb and intermediate categories the results of the book also have important implications for the study of parts of speech. The book will be invaluable for linguists interested in typology, morphology and syntax. Maria Koptjevskaja-Tamm is Lecturer in Linguistics at the University of Stockholm. She is a member of the ESF Programme in Language Typology and a permanent contributor to the Swedish National Encyclopaedia.
< previous page
page_i
next page >
< previous page
page_ii
next page >
page_ii
next page >
Page ii This page intentionally left blank.
< previous page
< previous page
page_iii
next page >
Page iii Editorial Statement Theoretical Linguistics Chief Editor Professor John Hawkins, University of Southern California Consultant Editors Professor Joseph Aoun, University of Southern California Professor Bernard Comrie, University of Southern California Dr Teun Hoekstra, University of Leiden Dr Richard Hudson, University College London Professor James Hurford, University of Edinburgh Professor Douglas Pulleyblank, University of Ottawa This series does not specialize in any one area of language study, nor does it limit itself to any one theoretical approach. Synchronic and diachronic descriptive studies, either syntactic, semantic, phonological or morphological, are welcomed, as are more theoretical ‘model-building’ studies, and studies in sociolinguistics or psycholinguistics. The criterion for a work’s acceptance is the quality of its contribution to the relevant field. All texts published advance our understanding of the nature of language in areas of substantial interest to major sectors of the linguistic research community. Traditional scholarly standards, such as clarity of presentation, factual and logical soundness of argumentation and a thorough and reasoned orientation to other relevant work, are also required.
< previous page
page_iii
next page >
< previous page
page_iv
next page >
Page iv Titles in the Theoretical Linguistics series: QUESTIONS OF INTONATION Gillian Brown, Karen L.Currie and Joanne Kenworthy THE SEMANTICS OF DETERMINERS Edited by Johan Van der Auwera THIRTY MILLION THEORIES OF GRAMMAR James D.McCawley ANAPHORA AND SEMANTIC INTERPRETATION Tanya Reinhart CAUSALITY IN LINGUISTIC THEORY Esa Itkonen THE PASSIVE: A COMPARATIVE LINGUISTIC ANALYSIS Anna Siewierska DEPENDENCY AND NON-LINEAR PHONOLOGY Edited by Jacques Durand GRAMMATICAL RELATIONS D.N.S.Bhat BASQUE PHONOLOGY José Ignacio Hualde FUNCTIONAL CATEGORIES AND PARAMETRIC VARIATION Jamal Ouhalla CLEFT AND PSEUDO-CLEFT CONSTRUCTIONS IN ENGLISH Peter C.Collins ANAPHORIC RELATIONS IN ENGLISH AND FRENCH Francis Cornish STYLISTICS AND PSYCHOLOGY: INVESTIGATIONS OF FOREGROUNDING Willie Van Peer ALLOMORPHY IN INFLEXION Andrew Carstairs WELSH SYNTAX: A GOVERNMENT-BINDING APPROACH Louisa Sadler EXISTENTIAL SENTENCES: THEIR STRUCTURE AND MEANING Michael Lumsden THE PHONOLOGY-MORPHOLOGY INTERFACE Jolanta Szpyra ACCESSING NOUN-PHRASE ANTECEDENTS Mira Ariel OLD HITTITE SENTENCE STRUCTURE Silvia Luraghi
< previous page
page_iv
next page >
< previous page
page_v
next page >
page_v
next page >
Page v Nominalizations Maria Koptjevskaja-Tamm
London and New York
< previous page
< previous page
page_vi
next page >
Page vi First published in 1993 by Routledge 11 New Fetter Lane, London EC4P 4EE This edition published in the Taylor & Francis e-Library, 2003. Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 © 1993 Maria Koptjevskaja-Tamm All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data Koptjevskaja-Tamm, Maria Nominalizations.—(Theoretical Linguistics Series) I. Title II. Series 410 Library of Congress Cataloging in Publication Data Koptjevskaja-Tamm, Maria. Nominalizations/Maria Koptjevskaja-Tamm. p. cm.—(Theoretical linguistics) Includes bibliographical references and index. I. Title. II. Series: Theoretical linguistics (Routledge (Firm)) P271.K67 1993 415–dc20 92–7229 ISBN 0-203-21464-1 Master e-book ISBN ISBN 0-203-31666-5 (OEB Format) ISBN 0 415 06020 6 (Print Edition)
< previous page
page_vi
next page >
< previous page
page_vii
next page >
page_vii
next page >
Page vii To Martin, Niklas, Sandra and Katarina Tamm
< previous page
< previous page
page_viii
next page >
page_viii
next page >
Page viii This page intentionally left blank.
< previous page
< previous page
page_ix
Page ix Contents Acknowledgements Abbreviations Part 1 Theoretical background 1 Preliminaries: the aim of this book 1.1. Goals of the study 1.2. Stating the problem 1.3. Illustrating the problem: internal syntax of Russian ANCs 1.4. The problem of arguments 1.5. Semantics of ANCs 2 ANCs as non-finite noun clauses 2.0. Introduction 2.1. Subordination, ‘balancing’ and ‘deranking’ 2.2. Types of noun clauses 2.3. Language specific variations in the semantic range of ANCs, some examples 2.4. Lexical nominalization vs. clausal nominalization 3 Noun phrases vs. finite clauses in different languages 3.0. Introduction 3.1. Noun phrases vs. finite clauses—a universal opposition? 3.2. Discussion of the basis for a classification of ANCs across languages 3.3. The typology of ANCs 3.4. Isomorphism between NPs and finite clauses: formulating the problem
< previous page
page_ix
next page > xiii xv 1 3 3 5 8 9 15 22 22 22 24 45 49 53 53 54 58 60 65
next page >
< previous page
page_x
Page x 3.5. Affinities between noun phrases and finite clauses: more examples Part 2 The data 4 The data base 4.0. Introduction 4.1. Obtaining data 4.2. Characteristics of the sample 5 Type 1: ‘Sentential’ 5.0. Introduction 5.1. ‘Ideal’ representatives: Korean and Archi 5.2. Problems in defining the pattern 5.3. Head-marking in the SENT type 5.4. Further characteristics of the SENT type 5.5. Intra-lexemic syntactic variation 5.6. Summary 6 Type 2: ‘Possessive-Accusative’ 6.0. Introduction 6.1. ‘Ideal’ representatives: Selkup and West Greenlandic 6.2. Problems in defining the pattern 6.3. Head-marking in the POSS-ACC type 6.4. Further characteristics of the POSS-ACC type 6.5. Intra-lexemic syntactic variation 7 Type 3: ‘Ergative-Possessive’ 7.0. Introduction 7.1. Languages with agentive passives 7.2. Pure OBL-POSS nominalizations vs. passives: possible connections 7.3. Languages without agentive passives 7.4. Head-marking in the ERG-POSS type 7.5. Further characteristics of the pure ERG-POSS type 7.6. Intra-lexemic syntactic variation 7.7. Summary 8 Type 4: ‘Nominal’ 8.0. Introduction 8.1. The DBL-POSS subtype
< previous page
page_x
next page > 69 77 79 79 79 82 90 90 90 94 98 102 103 108 110 110 110 116 120 122 126 129 129 130 137 156 159 159 161 161 165 165 165
next page >
< previous page
page_xi
Page xi 8.2. The POSS-ADN subtype 8.3. Summary 9 Less frequent types 9.0. Introduction 9.1. Type 5: ‘Mixed’ 9.2. Type 6: ‘Incorporating’ 9.3. Type 7: ‘Relative’ 9.4. Type 8: ‘Argument-Reducing’ 10 Splits and variation 10.0. Introduction 10.1. Lexical, semantic and functional splits 10.2. Referentiality/animacy split 10.3. Ambiguities 10.4. Alienable vs. inalienable possession and ANCs 10.5. Pattern shifts and intra-lexemic variation Part 3 Discussion 11 Explanation of type-membership 11.0. Introduction 11.1. Genetic and areal distribution of the nominalization types 11.2. Complement systems 11.3. Word order 11.4. The structure of non-derived NPs 11.5. Head-marking 11.6. Nominativity vs. ergativity 11.7. Morphological type 12 Explanation of the attested types 12.0. Introduction 12.1. Finite clauses vs. NPs vs. ANCs: direct analogy or modelling? 12.2. Sentence-like vs. NP-like properties of ANCs 12.3. ANCs as products of competing motivations 12.4. Inflection vs. derivation and ANCs 12.5. Why an action nominal construction? Summary
< previous page
page_xi
next page > 176 180 181 181 181 184 191 194 197 197 197 201 203 210 214 219 221 221 222 223 229 234 236 241 245 247 247 248 254 257 263 266 271
next page >
< previous page Page xii Appendix A: Appendix B:
< previous page
page_xii Alphabetical language index ANC types with examples Notes Bibliography Author index Language index Subject index
page_xii
next page > 273 280 300 307 324 327 330
next page >
< previous page
page_xiii
next page >
Page xiii Acknowledgements This book is a revised version of my doctoral dissertation ‘A Typology of Action Nominal Constructions’ presented in 1988 at the Department of Linguistics at the University of Stockholm and published in a limited number of copies. Without all the support and encouragement from my colleagues at the university, I would never have been able to carry my project through. The first part of the book is substantially modified and restructured. Most of the other major modifications and, hopefully, improvements concern the data. Thus, the original sample has been enlarged by six languages (from sixty-four to seventy languages). I have also done my best to recheck all the data in my dissertation, and I hope that the number of typographical errors in the examples, as well as other, more essential errors is considerably reduced in the present book. In doing this, I have been greatly assisted by numerous comments from various kind people. Some of the terms, in particular the names of the nominalization patterns, have been changed. Thus, the Possessive-Ergative, Possessive-Oblique and Possessive-Relative types have been changed into the Ergative-Possessive, Possessive-Adnominal and Relative types respectively. I have also added a more or less hypothetical Argument-Reducing type. There is also a new appendix, which lists relevant examples for all the sample languages. I am especially indebted to Östen Dahl, Jan Anward, Johanna Nichols, Martin Haspelmath and Leon Stassen for their valuable comments on the book at different stages of its preparation, and for their inspiration, advice, support and enthusiasm. Next I want to thank Anders Holmerg, Sven Malkior, Elena Maslova, Arnfinn Muruvik Vonen, Vladimir Nedjalkov, Jan-Olof Svantesson and John W.M.Verhaar who all, without sparing their own effort, gave me invaluable help. Thanks also to Björn Hammarberg, Gunnel Källgren, Birgit Nilsson and Åke Viberg who all read and commented upon
< previous page
page_xiii
next page >
< previous page
page_xiv
next page >
Page xiv various versions of the manuscript, and to Elena Skribnik who sent me publications on the Siberian languages. I am also grateful to the two anonymous reviewers from the Routledge editorial board for their criticism and suggestions. A typological investigation of this kind can hardly be carried out without the help of a large number of people with knowledge of various languages. I want to thank the following people who provided me with the necessary material: Ol’ga Boguslavskaja (Daghestanian languages), Amsalu Aklilu (Amharic), Ahmed Elgendi (Egyptian Arabic), Joakim Engwall, Oskar Aranzabal and Janna Gegus (Basque), Ivan Derzhanski and Mila (Bulgarian), Kees Hengeveld (Dutch), Diana Krull (Estonian), Vladimir Nedjalkov (Evenki, Nivkh, Chukchee), Danielle Cyr (Montagnais, Inuktitut, French), Karina Vamling (Georgian), Hartmut Traunmüller (German), Irene Fotopoulou and Hans Ruge (Greek), Alexandra Ajxenval’d, David Gil and Sven Malkior (Modern Israeli Hebrew), Ferenc Kiefer (Hungarian), Kjartan Ottósson (Icelandic), Hélène Boëthius (Ifè), Johanna Nichols (Ingush), Juanasie Koperkualuk (Inuktitut), Yukiyo Wallenström and Bengt Ek (Japanese), Nick Evans (Kayardild), Hang-seok Choi and Lev Koncevič (Korean), Farhad Shakely (Kurdish), Laine Lasmanis (Latvian), Debora Mulindwa and Gunnar Eriksson (Luganda), Jan-Olof Svantesson (Modern Mongolian), Ravinder Singh (Panjabi), Hassan Djamshidpey (Persian), Abdulaziz Lodhi and Sune Magnberg (Swahili), Anthony Diller and Preecha Juntunamalaga (Thai), Kari Fraurud and Birgit Nilsson (Turkish) and Elena Maslova (Yukagir). Many thanks also to Una Cunningham-Andersson for her careful reading of the English text and to Johan Stark for his technical help. And thanks to all other people who helped me in various ways and who are too many to be enumerated.
< previous page
page_xiv
next page >
< previous page
page_xv
next page >
Page xv Abbreviations A see section 1.4.1. ABL ablative ABS absolutive ACC accusative ACT active ADJ adjective ADMON admonitive, potential ADV adverb AFF affix AGR agreement AGT passive agent ALL allative AN action nominal ANC action nominal construction ANTI-PAS anti-passive variant of the Possessive-Accusative type AOR aorist ARG-RED Argument-Reducing type ART article ASP aspect AUX auxiliary CAUS causative CAY Central Alaskan Yupik Cl. Classical CL class COMIT comitative COMP complementizer CONSTR Construct State CONV converb COP copula CRD coordination
< previous page
page_xv
next page >
< previous page Page xvi DAT DBL-POSS DEF DEF.OBJ DET DIR DO DOM.POSS DP DU DYN ENCL ERG ERG-POSS EUPH EZ F FC FIN FOC FUT G GEN H HAB IE IMP IMPRF INC INDEF INDIC INESS INF INSTR INTR LNK LOC M MIX MN N NECESS
< previous page
page_xvi
next page >
dative Double-Possessive subtype definite definite object determiner directional direct object Dominant possession determiner phrase dual dynamic enclitic ergative Ergative-Possessive type euphonic extender ezafet feminine finite clause finite focus, focal future genitive (see section 3.5.1.) genitive see section 3.5.1. habitative Indo-European imperative imperfect Incorporating type indefinite indicative inessive infinitive instrumental intransitive linker locative masculine Mixed type mode nominal noun/neutral necessity
page_xvi
next page >
< previous page Page xvii NEG NMLZ NOM NOMN NONFIN NONFOC NONFUT NONM NONPAST NPOST O OB OBJ OBJC OBL OBL-INC OBL-POSS P PART PAS PASS PAST PAST.POSS PIE PL POL POSS POSS-ACC POSS-ADN POSS-INC PR PRED PREP PRES PRES.POSS PRET PRF PRFV PROG PROP PTCP Q
page_xvii
next page >
negation nominalization nominative Nominal type non-finite non-focal non-future non-marked non-past nominal postposition object (see section 1.4.1.) objective prefixes object marker Objective conjugation oblique Oblique-Incorporating subtype Oblique-Possessive subtype see section 1.4.1 partitive passive variant of the Oblique-Possessive subtype passive past past possession Proto-Indo-European plural politeness possessive Possessive-Accusative type Possessive-Adnominal subtype Possessive-Incorporating subtype pronoun predicate preposition present present possession preterite perfect perfective progressive proper article participle question particle
< previous page
page_xvii
next page >
< previous page Page xviii REAL REL RFL S SBJ SENT SENT-INC SENT-POSS SG SPEC SPL STAT SUBJ SUBJC SUB.POSS SUFF TNS TOP TRANS 1/2TRANS UNREAL UNS V V-l VCE VLD VN VPOSTP VPTL
< previous page
page_xviii
next page >
real Relative type reflexive subject (see section 1.4.1.) subjunctive Sentential type Sentential-Incorporating subtype Sentential-Possessive subtype singular ‘special’ systems split ergativity stative subject marker Subjective conjugation Subordinate possession suffix tense topic transitive anti-passive unreal unspecified tense-aspect-mood verb verb-initial languages voice validator verbal noun verbal postposition verbal particle
page_xviii
next page >
< previous page
page_1
next page >
page_1
next page >
Page 1 Part 1 Theoretical background
< previous page
< previous page
page_2
next page >
page_2
next page >
Page 2 This page intentionally left blank.
< previous page
< previous page
page_3
next page >
Page 3 1 Preliminaries: the aim of this book 1.1. GOALS OF THE STUDY This book is mainly concerned with a single type of nominalizations, action nominal constructions. Its aim is to make a cross-linguistic comparison between the internal syntax of action nominal constructions (ANCs) and that of two other types of constructions: the corresponding finite clauses and ordinary noun phrases. The main question raised by such a comparison can be formulated as follows: what are the means of signalling the relations between an action nominal and its dependents (subject and objects) as compared to the relationships in corresponding finite clauses and in ordinary noun phrases? Taking one easy example, what is the internal syntax of the construction the enemy’s destruction of the city as related both to the finite clause The enemy destroyed the city and to the NP the enemy’s map of the city? In the linguistics of the last three decades, it is possible to discern two different approaches to the internal syntax of nominalizations, or, more specifically, of action nominal constructions. The first approach concentrates on how to derive nominalizations in transformational grammar. In early work (Chomsky 1957, Lees 1960), nominalizations were analysed as results of transformations. Chomsky (1970) argues against the transformationalist hypothesis for the derivation of action nominal constructions in English, and proposes the lexicalist hypothesis, according to which action nominals constitute separate items in the lexicon. Chomsky’s argumentation is based on the regular similarities in the syntax of action nominals and of non-derived nouns, as well as on the idiosyncratic differences in the morphology, syntax and semantics of action nominals and the corresponding verbs. The derivation of action nominal constructions in other languages has also been the object of attention for several syntactic theories. To cite a few examples, action nominal constructions in German (Esau 1973),
< previous page
page_3
next page >
< previous page
page_4
next page >
Page 4 Modern Israeli Hebrew (Berman 1976, 1978) and Modern Mongolian (Binnick 1979) have been described from the point of view of the Standard Theory, those in Dutch (Hoekstra 1986) and Quechua (Lefebvre and Muysken 1988) have been treated within the theory of Government and Binding, while those in Georgian (Harris 1981) have been analysed within Relational Grammar. The second approach (Comrie 1976, Comrie and Thompson 1985) discusses the internal syntax of ANCs from the cross-linguistic and typological points of view. Using data from several languages, Comrie (1976) shows that there is no simple dichotomy between sentence-like and noun phrase-like action nominal constructions across languages, because such constructions vary greatly with respect to the ‘extent to which their internal structure corresponds to that of a nonderived noun phrase, rather than to the internal structure of a sentence with a finite verb’ (1976:200). This claim puts the transformationalist —lexicalist controversy in a different light. Since action nominal constructions in different languages differ as to their degree of nouniness, this variety ‘precludes us from using internal structure even as a heuristic test in universal grammar for distinguishing noun-headed noun phrases from verb-headed constructions’ (ibid.: 178). The present book follows the second of the two approaches mentioned above. Like Comrie and Thompson, I focus on the internal syntax of action nominal constructions across languages. However, the present study is not restricted to just a few languages, but is based on the material of seventy languages representing the majority of the world’s linguistic genetic stocks and linguistic areas. The first, most concrete goal of the study is to construct a syntactic typology of action nominal constructions, trying to answer at least the following questions: What are the means of signalling the relations between action nominals and their arguments as compared to the situation in finite clauses? If they are not the same, what other types of relations do they normally signal? How close are action nominal constructions to ordinary noun phrases from the point of view of their syntax? What happens to the opposition between the different arguments of finite verbs when the latter become nominalized, that is, what oppositions between the arguments tend to be retained or neglected in action nominal constructions? Having answered these questions, the following step is to explain
< previous page
page_4
next page >
< previous page
page_5
next page >
Page 5 the patterns that emerge. An explanation aims at answering the two questions: 1 ‘Why the typology contains just these attested types’ (Stassen 1985:6), and 2 Whether the choice of a particular nominalization pattern or of particular nominalization patterns in a particular language is accidental. If it is not, what are the determining factors? The rest of this chapter will be devoted to clarifying what is meant by action nominal constructions and by their internal syntax. 1.2. STATING THE PROBLEM 1.2.1. Action nominal constructions (ANCs), a definition Following Comrie (1976:178), I define action nominals (ANs) as ‘nouns derived from verbs (verbal nouns) with the general meaning of an action or process’, capable of declining or taking prepositions or postpositions in the same way as non-derived nouns, and showing ‘reasonable’ productivity. Action nominal constructions (ANCs), then, are nominalizations with ANs as their head (Comrie 1976:178). As such, they constitute a type of lexical nominalization (on the distinction between lexical and clausal nominalizations, see section 2.4.). The following sentences offer examples of ANCs which refer to a proposition (1.1), a fact (1.2), an event (1.3) and a manner of doing something (1.4). These are opposed to example 1.5, in which the verbal noun criticism is used in a concrete way, to refer to the result of the corresponding action. (1.1) The collapse of the Germans is unlikely. (Vendler 1967:132) (1.2) John’s singing the Marseillaise caused the riot. (ibid.: 135) (1.3) I heard the singing of the Marseillaise. (ibid.: 138) (1.4) John’s playing of Ravel is wonderful. (1.5) His criticism of the book is to be found on page 15. (Chomsky 1970:194). All the conditions in the definition of action nominals given above present certain problems. Here I will only touch upon two of them. More will be said in sections 1.4 and 1.5 and in chapter 2.
< previous page
page_5
next page >
< previous page
page_6
next page >
Page 6 Firstly, it can be difficult to ascertain whether action nominals in a language have the same possibilities to decline or take adpositions as non-derived nouns, since such information is not always to be found in available grammatical descriptions for many languages. For languages with morphological cases, the fact that some nouns do not follow the common case pattern seems to be most frequently registered. In Turkish, for example, infinitives with the suffix -mAG do not combine with possessive and genitive suffixes, although they may take any of the other nominal suffixes. With respect to languages without morphological cases, relevant information can be absent from descriptions. Comrie (1976) excludes the infinitive in French from the set of action nominals because it ‘may have a different preposition from a noun phrase, as in je commence le travail—‘‘I start work”, but je commence à travailler —“I begin to work”’. Such details are quite unlikely to appear in grammars of more ‘exotic’ languages. Secondly, there are no obvious criteria for whether or not certain action nominals show ‘reasonable’ productivity. Even the exact categorial status of action nominals can vary greatly in different languages. In some of them, action nominals are treated morphologically as regular verb forms, in others they constitute a group of derived nouns with a number of idiosyncratic features. More about this issue will be said in chapter 2. 1.2.2. Nouns, verbs and action nominals As should be clear from the definition in 1.2.1., action nominals in certain respects occupy an intermediate position between typical verbs and typical nouns. Thus, typical nouns include names for things, persons, places, while typical verbs denote actions and processes. Or, seen from the discourse point of view, as launched by Hopper and Thompson (1984), nouns prototypically introduce participants and ‘props’ and deploy them, while verbs ‘assert the occurrence of an event of the discourse’ (Hopper and Thompson, 1984:708). Action nominals, like discovery, shooting, etc., clearly refer to events, like verbs do, although not by asserting the occurrence of the events of the discourse, but by giving them a name. In other words, they combine semantic and discourse features of both verbs and nouns. In their morphology they also combine verbal and nominal features and different languages treat them as being closer to one or other of these word classes.1 The main hypothesis of this book is that the ‘intermediate nature’ of action nominals is reflected in their syntactic behaviour. More
< previous page
page_6
next page >
< previous page
page_7
next page >
Page 7 specifically, we may expect that the internal syntax of ANCs across languages will be more or less like that of finite clauses and/or that of NPs. Among the problems which are covered by the notion ‘internal syntax of ANCs’, I will mainly restrict myself to that of the means of signalling the relations between an action nominal and its dependents (subject and objects). 1.2.3. Means of signalling syntactic relations Before illustrating the problem, let us consider the different means of signalling syntactic relations in a phrase and in a clause (which I will abbreviate as ‘syntactic means’). The syntactic relations which are relevant for the present study include those between a finite verb and its arguments (in a finite clause), those between an action nominal and its arguments (in an action nominal construction), and those between a head-noun and its NP-dependents (in a noun phrase). Following Nichols (1986), I will use here the distinction between two types of marking built on the opposition between the head and the dependent of a construction. According to Nichols, ‘the head is the word which governs, or is subcategorized for—or otherwise determines the occurrence of—the other word. It determines the category of its phrase’ (Nichols 1986:57). The question of what is the head in a clause is not uncontroversial, but for the purpose of the present study it is sufficient to consider predicates (or finite verbs) to be such heads. Syntactic relations can be signalled in the following three main ways: 1 by head-marking, which means that the head of a construction is marked in such a way as to identify its relation to its dependents, for example, agreement between the verb and some of its arguments in a clause, possessive affixes or the Semitic Construct State (‘status constructus’) in a noun phrase; 2 by dependent-marking, which means that the dependents in a construction are marked in such a way as to identify their syntactic relations to the head of the construction, for example, different cases or adpositions for the subject and objects in a clause, a special genitive case or a special adposition for the possessor in a noun phrase; 3 by word order, for example, a more or less fixed word order between the verb and its arguments in a clause, a more or less fixed word order between the possessed noun and the possessor-noun in a NP.
< previous page
page_7
next page >
< previous page
page_8
next page >
Page 8 1.3. ILLUSTRATING THE PROBLEM: INTERNAL SYNTAX OF RUSSIAN ANCs ANCs can resemble more or less closely the corresponding finite clauses or NPs with non-derived nouns (possessive NPs) in either of the three respects mentioned in the previous section. Comrie and Thompson (Comrie 1976, Comrie and Thompson 1985) concentrate on dependent-marking and only mention word order in some cases. I would like to discuss all three syntactic means. To illustrate, consider the two Russian verbs razmyšljat’—‘to meditate’ and zavoevat’ —‘to conquer’ with the corresponding action nominals razmyšlenie and zavoevanie. Ignoring the pronouns, the following comparison can be made between the internal structure of finite clauses with these verbs and of constructions with these verbal nouns respectively, starting with sentences (1.6–1.9): (1.6) Mal’čik razmyšlja-et (o prirod-e). boy:NOM meditate-PRES.3SG (about nature-LOC) ‘The boy is meditating (about the nature).’ (1.7) Sil’nyj grom prerval [razmyšl -enij-a heavy thunder interrupted [meditate-AN-PL.ACC mal’čik-a (o prirod-e)]ANC boy-GEN (about nature-LOC)]ANC ‘Heavy thunder interrupted the boy’s meditation (about nature).’ (1.8) Aleksandr zavoeva-l Egipet. Alexander:NOM conquered-PAST.PERF Egypt:ACC ‘Alexander conquered Egypt.’ (1.9a) Aleksandrija byla osnovana v čest’ Alexandria was founded in glory [zavoeva-nij-a Egipt-a (Aleksandr-om)]ANC conquer-AN-GEN Egypt-GEN Alexander-INSTR]ANC ‘The city of Alexandria was founded as a monument over the conquest of Egypt (by Alexander).’ (1.9b) [Zanoeva-ni-e Aleksandr-a]ANC izmenilo [conquer-AN-NOM Alexander-GEN]ANC changed žizn’ Egipt-a life:ACC Egypt-GEN ‘Alexander’s conquest changed the life of Egypt.’ In Russian, the main means of signalling the relations between verbs
< previous page
page_8
next page >
< previous page
page_9
next page >
Page 9 and their arguments in finite clauses is the category of case in NPs. Subjects take the nominative, direct objects take the accusative. In passive sentences, the agent is in the instrumental case. Finite verbs agree with their subjects as regards number and person (in the present, as in example (1.6), or future) or gender (in the past, as in examples (1.7–1.9)). Word order is fairly free, although the most neutral is SVO. Full sentences thus clearly follow the nominative-accusative pattern. In noun phrases only dependent-marking is used. Possessors take the genitive, while nominative and accusative dependents are excluded. The instrumental occurs normally only in combination with nouns derived from verbs. Heads normally precede their dependents. The syntactic relations within ANCs are signalled in a quite different way than in the corresponding finite clauses. First of all, as in all types of noun phrases (with marginal exceptions) both the nominative and accusative are forbidden in ANCs. Additionally, where only one argument is present, the opposition between subject and objects is neutralized, this single argument taking the genitive. Where both the subject and direct object of a transitive verb are retained in the ANC, it is the direct object that takes the genitive, while the subject stands in the instrumental. Head-marking does not exist within ANCs, as within NPs in general. Action nominals precede their arguments, the order between the latter is rather free, though subjects tend to precede objects. Thus, ANCs in Russian stand much closer to noun phrases with non-derived nouns than to finite clauses in head-marking, dependent-marking and word order. The only difference from ordinary NPs is the possible use of the instrumental. ANCs follow the ergative-absolutive pattern rather than the nominative-accusative pattern of finite clauses. 1.4. THE PROBLEM OF ARGUMENTS 1.4.1. Grammatical relations or thematic roles? The main question addressed in this study (as it is formulated in section 1.1.), involves the grammatical relations ‘subject of’ and ‘object of’. Leaving aside whether it is appropriate to talk about the subject and object of ANCs, let us now consider another aspect of the problem with these terms. In many cases, a single language has several nominalization patterns, the choice among which seems to depend not on the grammatical relations of the arguments in the finite clause, but on their thematic roles.
< previous page
page_9
next page >
< previous page
page_10
next page >
Page 10 For example, in Russian, both the verbs oskorbit’ —‘to insult’ and uvažat’—‘to respect’ combine with a subject in the nominative case and a direct object in the accusative case, as in the sentences (1.10a) and (1.10b) below. (1.10a) Petrov oskorbil Ivanov-a. Petrov: NOM insulted Ivanov-ACC ‘Petrov insulted Ivanov.’ (1.10b) Ivanov uvažal Petrov-a. Ivanov:NOM respected Petrov-ACC ‘Ivanov respected Petrov.’ From both verbs it is possible to derive action nominals, oskorblenie and uvaženie respectively. However, these verbal nouns combine with their arguments in different ways, i.e., the former (‘insult’) shows the usual Russian nominalization pattern, where the subject is in the instrumental and the object is in the genitive, while the latter (‘respect’) comes with a much more infrequent solution, where the subject is in the genitive and the object is referred to by the prepositional phrase in the dative with the preposition k—‘towards’. This difference is exemplified in examples (1.11a) and (1.11b). (1.11a) oskorbl-enie Ivanov-a Petrov-ym insult-AN Ivanov-GEN Petrov-INSTR ‘Petrov’s insulting Ivanov’ (1.11b) uvaž-enie Petrov-a k Ivanov-u respect-AN Petrov-GEN towards Ivanov-DAT ‘Petrov’s respect towards Ivanov’ Besides uvaženie, there exist some other action nominals with the same peculiarities as regards the expression of their arguments. Their common feature is the thematic roles which they ascribe to their arguments. The most important point is that the subject of the corresponding verb is not an agent, as it is with most dynamic verbs (e.g., with oskorbit’ —‘to insult’), but an experiencer of an attitude or emotion to someone/something. Hoekstra (1986), in a paper on derivation of nominalizations, discusses two views on what he calls ‘complement inheritance’ in nominalizations, by ‘inheritance’ meaning ‘a particular conception of the phenomenon that derivationally related words show up in environments that are similar in relevant respects’ (Hoekstra 1986:549). According to the categorial view, ‘nominalizations have syntactic structures which are isomorphic to those of their corresponding verbal
< previous page
page_10
next page >
< previous page
page_11
next page >
Page 11 constructions’, which means that inheritance involves (sub)categorial information. According to the thematic view, it is the thematic grid or argument structure of the verbal base which is inherited in nominalizations (ibid.: 550). Hoekstra argues for the latter view, at least for the derivation of action nominals in English and Dutch, showing several discrepancies between verbal and nominal constructions which are unexpected under the categorial approach, but are quite systematic. They can all be explained by Hoekstra’s generalization that ‘in nominal constructions the formal realization of arguments is determined by their thematic role with respect to the head’ (ibid.: 568). Though it is true that derived action nominals in a number of languages may be sensitive to the thematic roles of their arguments, I still believe that the categorial structure of the underlying verb must be taken into account as well, otherwise it would be too difficult to explain the striking similarities between the expression of the arguments in ANCs and in the corresponding finite clauses. As regards those action nominals which should probably be considered regular verb forms, as far as I can judge from the available sources of information, the correspondence between ANCs and the underlying finite clauses seems perfectly straightforward and is based only on the grammatical relations between the head and its arguments. It is another issue, of course, that in different languages the grammatical relations themselves can be more or less sensitive to the thematic roles of verbal arguments. In talking about the arguments of verbs and action nominals, I will use the terminology launched by Dixon (1972), though with slight modifications. The same terms will apply to both types of constructions, to finite clauses and to ANCs, which does not imply that the arguments in both types of constructions are treated in the same way, either syntactically or morphologically. However, this will simplify the comparison between finite clauses and ANCs. Action nominals derived from intransitive or transitive verbs will be called intransitive or transitive action nominals respectively. The single argument of an intransitive verb and an intransitive AN will be symbolized as S (Si by Dixon). As we have already seen in this section, in a number of languages, transitive ANCs can follow one of several patterns where the choice between them depends on the exact thematic role of the arguments. In such cases, however, there always seems to exist a ‘standard pattern’ or ‘standard patterns’, which the majority of the action nominals in the language follow. This majority, among other things, includes action nominals derived from ‘prototypical’,
< previous page
page_11
next page >
< previous page
page_12
next page >
Page 12 ‘highly transitive’ verbs (Hopper and Thompson 1980), i.e., those which refer to actions and whose participants are an agent and a patient (such as to hit or to put). These arguments will be symbolized as A and P respectively. The present study is mainly restricted to such ‘standard’ ANCs, since this is the sort of information which is easiest to obtain for quite a number of languages. While I hope to have collected enough information on the ‘standard’ patterns in various languages, I do not believe it to be a realistic task to describe all the existing deviations from them. In many cases, though, the examples will contain other types of transitive verbs. This can be understood in two ways: either the coding of the arguments in ANCs in the language is not sensitive at all to the thematic roles, i.e., there are only ‘standard’ patterns in the language, or such deviations do exist, but in the case of the particular action nominal which appears in the example quoted these deviations are not relevant. Due to the gaps in the data, it is generally impossible to distinguish between the two situations. The categories S, A and P are a sort of a compromise between thematic roles and grammatical relations.2 On the one hand, I try to choose those verbs where there is no conflict between the thematic roles of the arguments and their grammatical relations to the head verb. On the other hand, S, A and P are quite neutral with respect to the notions ‘subject’, ‘direct object’, ‘absolutive’, ‘ergative’, etc. However, sometimes the latter terms will be used in the discussion of finite clauses for those languages, where the corresponding syntactic functions are uncontroversial. Basic word order in a language will be designated in terms of SVO, SOV, etc. 1.4.2. Reduction of arguments Up to now I have tacitly assumed that it is always possible to compare the expression of the arguments in ANCs with those in the corresponding finite clauses. However, in many cases, some of the possible syntactic arguments of a certain verb are not expressed or even cannot be expressed in the corresponding ANCs, even if they should obligatorily be present in combination with finite verbs. In this connection, it is appropriate to talk about reduction of arguments. Let us assume that action nominals have the same semantic arguments as the corresponding verbs. When a semantic argument ai of an action nominal X in a given sentence is not expressed, this can be
< previous page
page_12
next page >
< previous page
page_13
next page >
Page 13 the result of one of three facts: 1 the semantic argument ai of X (or any other action nominal) can never be expressed, at least if some other semantic argument, aj, is to be expressed; 2 the action nominal X is used in such a way that the semantic argument ai cannot be expressed in this particular situation; 3 the semantic argument ai of the action nominal X could in principle be expressed in the given sentence, but is omitted in accordance with the communicative goal of the sentence. The following examples illustrate these three points. (1) The semantic argument ai of X (or any other action nominal) can never be expressed, at least if some other semantic argument, aj, is to be expressed. The Russian verb napolnjat’ —‘to fill’ in one of its uses has three semantic arguments as in the sentence (1.12): (1.12) Raboči-e napolnjaj-ut bassejn worker-PL.NOM fill-PRES.3PL swimming.pool:ACC vod-oj. water-INSTR ‘The workers are filling the swimming-pool with water.’ In this sentence, voda —‘water’ is used in the instrumental, which is also the usual case for the A of ANCs in the presence of the P (in the genitive). However, it is impossible to have both ‘the workers’ and ‘water’ in ANCs corresponding to (1.12), the option being either (1.13) or (1.14): (1.13) napoln-enie bassejn-a vod-oj fill-AN swimming.pool-GEN water-INSTR ‘the filling of the swimming-pool with water’ (1.14) napoln-enie bassejn-a raboči-mi fill-AN swimming.pool-GEN worker-INSTR.PL ‘the filling of the swimming-pool by the workers’ Russian cases like the one mentioned above are lexically determined and relatively rare. However, this is not the case in some other languages, like Hungarian, Hausa or Chukchee. Thus, according to Ferenc Kiefer (personal communication) in Hungarian it is impossible to have constructions with both the A and the P which are directly dependent on one and the same transitive action nominal (derived
< previous page
page_13
next page >
< previous page
page_14
next page >
Page 14 with the suffix -ás/-és). If either of the arguments is present, Hungarian uses a possessive construction. However, when both are present it is necessary to put a type of dummy element, most often the present or the past participle of the verb történik—‘to happen’, between the action nominal and one of its arguments, which means that the whole construction looks like a relative clause (see example (1.15)). (1.15) Norvégia Németország által Norway Germany of történ-ő/történ-t elfoglal-ás-a happen-PRES PTCP/PAST PTCP occupy-AN-POSS ‘Germany’s occupation of Norway’ The same strategy, which results in the Relative type of nominalizations (see chapter 9), is used in Hausa (Ščeglov 1970:253), while in Chukchee (a Paleo-Siberian language) there is no possibility whatsoever of expressing both arguments of transitive ANs at the same time (Vladimir Nedjalkov personal communication). (2) The action nominal X is used in such a way that the semantic argument ai cannot be expressed in this particular situation. In other words, the restriction on the presence of any syntactic argument of an AN corresponding to its semantic argument ai is connected to the special syntactic position or use of this AN in the given sentence. Thus, in the Russian example (1.16), the A of the noun čtenie—‘reading’ coreferential with the subject of prodolžal—‘went on’ is deleted: (1.16) Ivan prodolžil čtenie Aivengo posle obeda. Ivan went.on read:AN Ivanhoe after lunch ‘Ivan went on reading Ivanhoe after lunch.’ Contexts that trigger some sort of ‘equi-NP-deletion’ in constructions with action nominals have been investigated, for example, by Rothstein (1974) for Polish and Padučeva (1977) for Russian. It is worth noting that the rules of equi-NP-deletion for constructions with infinitives and action nominals in a single language can be quite different.3 A number of languages do not distinguish between action nominals and participles (e.g., Turkish, Quechua). Used in the latter function, such words are syntactically dependent on some of their semantic arguments (or even other kinds of semantic dependents), instead of having the latter as syntactic arguments. See example (1.17) from
< previous page
page_14
next page >
< previous page
page_15
next page >
Page 15 Turkish: (1.17) [Orhan-ın gör-düğ-ü] adam cık-tı Orhan-GEN see-AN-3.POSS man leave-PAST ‘The man, whom Orhan saw, left.’ (Lehmann 1984:54 ex. 11) In this example the action nominal from the verb ‘see’ is not allowed to have an overt object. Its semantic ‘object’ appears as the head of a NP, in which the action nominal/participle is a dependent. (3) The semantic argument ai of the action nominal X could, in principle, be expressed in the given sentence, but is omitted in accordance with the communicative goal of the sentence. This is, in fact, a very common situation. In many languages, all of the semantic arguments of an action nominal are seldom present in ANCs; most often they are either clear from the context or irrelevant. This peculiarity of ANCs, and nominalizations in general, is, presumably, closely connected with their functions in discourse. All three cases discussed above are interrelated. Since ANCs in actual use very often involve valency reduction (Mackenzie 1985), a number of languages do not have any way of expressing all the verbal arguments within the corresponding ANC. Thus, ANCs with all the possible arguments present, and in particular with both the A and the P, are rather rare, but do, in fact, occur. The syntactic typological classification, which is presented in the later chapters, is intended to cover all possible nominalization patterns and means of signalling the arguments of action nominals, more or less regardless of their frequency in actual usage. Deletion of arguments in ANCs may be caused by fundamentally different reasons and has therefore essentially different consequences for the resulting typology. Some of the languages where action nominals cannot have two arguments directly dependent on them constitute special syntactic types (the Relative and the Argument-Reducing types, see chapter 9). On the other hand, if deletion of a certain argument in a particular situation is triggered by a special syntactic or discourse context, there may exist other contexts which do not call for such a deletion and which should be found in order to establish the marking of the arguments in the corresponding ANCs. 1.5. SEMANTICS OF ANCs The definition of ANCs given in 1.2.1. states that their head-nouns are characterized by the general idea of an action or a process. The exact
< previous page
page_15
next page >
< previous page
page_16
next page >
Page 16 range of the meanings and functions typical for action nominals across languages can vary considerably depending on several factors, the most important being the overall system of noun clauses (finite and non-finite) available in a particular language (see chapter 2). The purpose of the present section is to clarify what exactly is meant by ‘the general idea of an action or a process’. 1.5.1. General questions In discussing the semantics of action nominals and action nominal constructions across languages, we should distinguish at least the following four problems: 1 What is the whole range of meanings that ANs and ANCs can have across languages? 2 What is the range of meanings that ANs and ANCs can have in a particular language? 3 What is the meaning of an AN or an ANC in a particular context? 4 In those contexts where ANs or ANCs alternate with other types of constructions, which factors favour the choice of the former? It is clear that the exact meaning of an action nominal depends on the meaning of the underlying verb stem. Hence, the expressions ‘the range of meanings’ and ‘the meaning’ of an AN should be understood as referring to those types of meanings, or meaning categories that are characteristic for ANs (which can be visualized as some kind of semantic components or similar in their semantic description). The following subsections will be devoted to the first and the third of the problems formulated above. In most cases, I shall not distinguish between the meaning of an AN and the meaning of the ANC of which it is the head. The second problem will be discussed in section 2.3. The last problem will only be touched upon in chapter 12. As we will see in chapter 2, in at least some languages (of the complement-deranking type, defined in 2.2.), ANCs constitute the only type of noun clauses and, consequently, must be capable of expressing the whole range of meanings which are associated with noun clauses. Thus, the issue of the whole semantic range of ANCs across languages can be reduced to the question of the semantic range that noun clauses as a whole can have. Here, as in chapter 2, the discussion is confined to complements (i.e., subordinate clauses which function as arguments of verbs, adjectives, nouns, etc.). In these cases, the semantic type of a noun clause is predetermined by the meaning of the complement-taking predicate which combines with
< previous page
page_16
next page >
< previous page
page_17
next page >
Page 17 this noun clause. Therefore, the task of describing the meaning types of complements goes hand in hand with the task of classifying complement-taking predicates into semantic types. 1.5.2. Propositions, facts and events (Vendler’s theory of nominalizations) The most widespread semantic classification of noun clauses into propositions, facts and events is primarily due to Vendler (1967, 1970), who showed that noun clauses (‘nominalizations’, to use his term), at least in English, can express few and clearly defined meaning categories, predetermined by the semantic type of complement-taking predicates.4 Thus, the verbs to assert and to believe in take propositional complements, the verbs to know and to regret take factive complements, while the verbs to hear and to continue take complements referring to events (actions and processes). Vendler showed also that the meaning category of a noun clause is also, at least partly, predetermined by its form, for example whether it is a finite subordinate clause or a construction with a derived nominal as a head. In his early work, Vendler (1967) worked only with a binary opposition between facts and events (which had been used even earlier, by Lees 1960). However, in his later papers he is, in addition, forced to distinguish between facts and propositions: true propositions merely correspond to or fit the facts, a proposition is ‘the subjective appearance of an objective possibility, and, if true, of a fact’ (Vendler 1970:95–6). Vendler himself identifies the distinction between predicates which take complements referring to facts, and other predicates with the Kiparskys’ opposition between ‘non-factive’ and ‘factive’ predicates (Kiparsky and Kiparsky 1970). According to them, ‘factive’ predicates, in contrast to ‘nonfactive’ ones, presuppose the truth of the subordinate noun clause (which, thus, corresponds to a fact). Predicates which take complements of the same semantic type can co-occur. Thus, events, actions and processes occur, last, end, can be watched, observed, can be gradual or sudden. That is because events are primarily temporal objects. Facts can be mentioned and denied, can surprise us, can cause other things (Vendler 1967:141). While facts and so on are objectively given, not formed by anyone, propositions (opinions, predictions, beliefs) are formed or made by people. Propositions must have a representative reality, which enables them to represent a reality, to fit or fail to fit the facts, to be true, false, unlikely or probable (Vendler 1970:225–6).
< previous page
page_17
next page >
< previous page
page_18
next page >
Page 18 Vendler also demonstrated connections between the meanings of noun clauses and the types of matrix predicates in English, on the one hand, and the form of the noun clauses themselves, on the other hand. Thus, he divides all the various forms of nominalized sentences into two categories, imperfect nominals and perfect nominals, ‘one in which the verb is still alive as a verb, and the other in which the verb is dead as a verb, having become a noun’ (Vendler 1967:131). In the former category, nominalizations may contain tenses, auxiliaries and adverbs, while nominalizations of the latter category may not contain such verbal satellites, but rather take articles, prenominal adjectives and so on. While ‘imperfect nominals’ may refer to propositions and facts but hardly to events, ‘perfect nominals’ may refer to events, facts and propositions. These differences may be seen in the following examples, where the collapse of the Germans is a ‘perfect nominal’, while that the Germans will collapse/collapsed is an ‘imperfect nominal’. (1.18a) The collapse of the Germans is unlikely. (1.18b) That the Germans will collapse is unlikely. (1.18c) The collapse of the Germans was gradual. (1.18d) *That the Germans collapsed was gradual. (Vendler 1967:225–6) Thus, in English, noun clauses which refer to events are further from the corresponding independent sentences than those which refer to facts and propositions. As we shall see, this situation is not at all unique and is also reflected in the internal structure of ANCs across languages. 1.5.3. Mode nominals, act nominals, result nominals and object nominals Up to now, we have discussed the three main meanings of noun clauses, including ANCs. There are four more semantic categories that may be relevant in this connection. According to certain analyses, the semantic range of ANs in a number of languages includes the ‘manner of performing an action’. As an example, consider Berman’s (1976:65) analysis of the following Hebrew sentences. (1.19a) [halixat dan] madiga et kulánu. go:AN dan worries DEF.OB all.of.us ‘Dan’s going(away)/walking worries us all.
< previous page
page_18
next page >
< previous page
page_19
next page >
Page 19 (1.19b) ki ánu adáyin zquqim lo ‘because we still need him’ (1.19c) ki hi hayta hafganatit ‘because it was demonstrative (=theatrical)’ (1.19d) ki hu xozer aléha midey šavúa ‘because he repeats it every week’ According to Berman, the AN in (1.19b) can be paraphrased by ‘the fact that’, in (1.19c) by ‘the way that’ and in (1.19d) by ‘the process or action of’. This type of polysemy is quite regular in Hebrew, though not all ANs show the whole set of interpretations. Berman suggests that in the underlying structure, there must be a specific set of PRO-form heads associated with any derived verbal nominal, which will help to provide a unique interpretation for every surface occurrence of such a nominal. In contrast to Hebrew, many languages have a special type of verbal noun which refers to the way of performing the action denoted by the underlying verb. Thus, these nouns differ from typical ANs and can be called nomina modi —‘mode nominals’. For example, according to Harrison (1976), Mokilese (a Micronesian language) has a class of mode nominals derived from verbs by means of a regular process (sometimes involving reduplication), while for fact and event complements, it uses verb phrases (see chapter 3, section 3.1.) and clauses which are exactly like independent ones, except for the marking of the subject. Sentence (1.20) below contains a mode nominal (cf. example (3.3)). (1.20) Loakjidin joangoan mwumwwoawe apwal. fish:MN ‘Fishing for that kind of fish is difficult.’ (Harrison 1976:282 ex. 8) Other languages have formally distinct classes of mode nominals and action nominals. In Amharic, for example, there is a clear morphological opposition between mode nominals, like akkiyahed —‘the way of going’, astoyayot—‘the way of seeing’ vs. action nominals (infinitives), like mohed —‘to go, going’, mayot—‘to see, seeing’ (Obolensky et al. 1964:60, 455). In some cases, however, it is not possible to have an opposition between an action itself and the manner of performing it, or rather, some verbal stems can be compatible with only one of these interpretations. Thus, it is difficult to imagine event (action, process) nominalizations from stative verbs such as to know, which makes it possible or even probable for the corresponding AN
< previous page
page_19
next page >
< previous page
page_20
next page >
Page 20 to be interpretated as a mode noun. In other cases, ANs can be understood as referring to the action itself, but this is not considered as something very interesting or important, and the noun tends to get a manner interpretation as well. For example, in Russian, for some verb stems, the verbal nouns with the suffixes -nie/-tie (normally used for deriving ANs) have both the manner and the action (event, process) readings, the former being preferable, e.g., povedenie —‘behaviour’, ponimanie —‘understanding’, vosprijatie—‘perception’, tolkovanie—‘interpretation’. ANs can be opposed to still another class of verbal nouns, ‘act nominals’ (nomina acti). This distinction is illustrated by the sentences (1.21a) and (1.21b) (Schwartz 1970:303), where announcement expresses an act and not performance of an act, while leaving has an event or a fact reading. (1.21a) *His announcement when Congress had just convened, surprised everyone. (1.21b) His leaving when Congress had just convened, surprised everyone. The two oppositions, ANs vs. mode nominals, and ANs vs. act nominals, are far from being obvious, but both mode nominals and act nominals when they can be distinguished from action nominals in the same language, are only marginal to present purposes. In many cases, however, a single verbal noun can combine all these meanings. As discussed in section 10.1., such verbal nouns may sometimes have different syntactic behaviour in their different uses. Many languages have verbal nouns that combine both more abstract and more concrete meanings, as ‘action’ as opposed to ‘result of action’ or ‘action’ as opposed to ‘object of action’, such instances of polysemy often being language-specific. For example, the English verbal noun agreement refers to a fact in Their agreement to our plans surprised us and to the result of the action denoted by the underlying verb, as in The agreement they signed was submitted in four copies. The Hebrew equivalents of this noun in the two uses are haskama and heskem respectively. On the other hand, English entering, entrance, entry are all translated as knisa in Hebrew (Berman 1976:64). In the next example, the verbal noun štiya can refer both to the fact, process, or the way of drinking, on the one hand, or to the object of this activity, the drinks themselves.
< previous page
page_20
next page >
< previous page
page_21
next page >
Page 21 (1.22) Hebrew nehenéynu min haštiya bamsiba enjoyed.IPL (from) ART.drink.AN at.ART.party ‘We enjoyed the drinking/the drinks at the party’ (Berman 1976:65) Even if such a polysemy is relatively usual both in English and Hebrew, it is by no means regular, in contrast to the fact/proposition readings of, for example, -ing nominals in English. We could imagine a language where there is a systematic polysemy between such abstract and concrete meanings of verbal nouns, so that all of them or the overwhelming majority of them are characterized by this polysemy. They will be considered as ANs, if they satisfy the other criteria for such nouns. The fact that they can be used in the other way is of less importance for the present study.
< previous page
page_21
next page >
< previous page
page_22
next page >
Page 22 2 ANCs as non-finite noun clauses 2.0. INTRODUCTION Although action nominals in different languages in prototypical cases have certain common features, they show great diversity with respect to their syntactic behaviour, semantics and their exact morphological relation to the verbs they are derived from. Thus, they range from being separate lexical items—derived nominals, which can have rather idiosyncratic connections with the corresponding verbs, semantically, syntactically and morphologically—to being regular (inflectional?) verb forms, formed for any verb in a straightforward way, without any morphological, semantic or syntactic idiosyncratic features. Nominalizations with ANs as their head—action nominal constructions—are involved in the subordination system of languages, and many of their properties derive from their exact status in the overall system of noun clause types existing in a particular language. Thus, it is necessary to see ANCs in this broader perspective against the background of various noun clause systems. In this chapter I will clarify some of the terms which are used in talking about ANCs as a type of non-finite noun clause and give a sketch of the different noun clause types which exist in various languages. 2.1. SUBORDINATION, ‘BALANCING’ AND ‘DERANKING’ Let us start with some general remarks on subordination. Subordination refers to the syntactic relation which exists between a matrix clause and a subordinate clause, as well as to the process of creating such a relation. In the simplest case, a matrix clause together with a subordinate clause constitute a complex sentence. Although the exact relationship between subordination and coordination can in many cases be rather complicated (see, e.g., Foley and Van Valin 1984:
< previous page
page_22
next page >
< previous page
page_23
next page >
Page 23 ch. 6), here it is sufficient to assume that in most situations we can recognize a construction as involving subordination and, in addition, determine which is the matrix clause and which is the subordinate one. The terms ‘matrix’ and ‘subordinate’, as understood here, are relative. In other words, a clause A, which is the matrix clause with respect to a clause B, may in turn be subordinate to another clause, C. I will use the term independent clause to refer to both simplex sentences (i.e., those which only consist of one clause), and to clauses which are not subordinate to any other clause within complex sentences. In most cases, subordination involves embedding of the subordinate clause in the matrix clause. As constituents of matrix clauses, subordinate clauses fall into three functional types: complements, i.e., subordinate clauses which function as arguments of verbs, nouns, etc.; relative clauses, i.e., subordinate clauses which function as modifiers of nouns; adverbial clauses, i.e., subordinate clauses which function as modifiers of verb phrases or entire propositions. Subordination is usually signalled by the following markers of subordination which can be used separately or in any combination with each other: 1 word order—both inter-clause word order (i.e., the mutual order of the matrix and the subordinate clauses) and intra-clause word order (i.e., the order of the constituents of a clause which is typical for subordinate clauses as opposed to independent or matrix ones); 2 subordinating morphemes (e.g., the complementizer that in English); and 3 dependent verb forms, including non-finite verb forms. The concept of dependent verb forms needs some clarification. We shall say that a verb form in a language is dependent if it cannot normally be used as the predicate of an independent clause (see p. 31–2 for some problems with this definition). Note that ‘verb form’ will be understood here in a somewhat extended sense, including also verbal derivatives. When necessary, I will distingish between inflectional verb forms and derived verb forms. The distinction between the use of word order and subordinating morphemes as opposed to the use of special dependent verb forms is extremely important for the present study, and can be described in terms of an opposition between the strategies of balancing and deranking as stated by Stassen (1985:
< previous page
page_23
next page >
< previous page
page_24
next page >
Page 24 76–83) in connection with different types of constructions used for codifying temporal chaining. In balanced constructions, both predicates remain structurally of the same rank, whereas in deranked constructions, one of the predicates is deranked, that is, the form of the predicate itself signals the subordination of that clause, which, however, does not preclude the use of other markers of subordination. Both coordination and subordination can involve either embedding of the whole subordinate clause without any predicate deranking, or deranking of the predicate in the subordinate clause. A dependent verb form, then, is a verb form that appears in a deranked clause. Some languages (such as Chinese) do not have special dependent verb forms. The precise difference between the non-dependent and dependent verb forms in languages that do distinguish between them can vary considerably. Stassen presents three ways in which predicates can be deranked. Firstly, in many languages, dependent verb forms are characterized by a reduction of the verbal morphology as compared to non-dependent verb forms, such as the loss of personal flection, normal tense affixes, many distinctions of aspect and mood. Secondly, dependent verb forms can have different category status from non-dependent ones. For example, dependent verb forms can turn up as verbal nouns, participles, or gerunds. Thirdly, dependent verb forms can be characterized by the addition of some specific marker, or by the application of a special dependent mood. Non-finite verb forms, discussed in sections 2.2.2–2.2.6, constitute a subtype of dependent verb forms and include, among other things, action nominals. 2.2. TYPES OF NOUN CLAUSES Among subordinate clauses, we can distinguish a class of noun clauses, which have a distribution comparable to that of noun phrases in matrix clauses. The fact that a clause is characterized as a noun clause does not say much about its functional type. This is quite similar to the situation for noun phrases which can have a variety of functions within a simplex clause. Noun clauses are most often used as complements and as (constituents of) adverbial clauses, though the latter need not contain noun clauses. By far the most typical function of noun clauses is as arguments of the matrix predicate, or as sentential complements. Let us, therefore, look at the different varieties of sentential complements in order to get a general idea about the range of structurally different noun clauses in different languages. While some languages can have both balanced and deranked noun clauses,
< previous page
page_24
next page >
< previous page
page_25
next page >
Page 25 Table 2.1 Summary of complement types (after Noonan 1985:65) Complement Part of Syntactic relation of Range of inflectional Other characteristics type speech subject to predicate categories of predicate Indicative verb same as main clause same as main clause s-like form (nearly) identical to declarative main clause Subjunctive verb same as main clause same as main clause s-like form that differs from declarative main clause—when main clause, often used in hortative or imperative senses Paratactic verb predicate may agree typically reduced interpreted as separate assertion; with subject but does syntactically not a subordinate not form constituent clause; cannot take complementizer with it Infinitive verb predicate cannot form reduced; cannot relations with object same as constituent with take subject—verb indicative subject agreement Nominalizationnoun associative relation reduced; may take may have internal structure of NP; between subject and nominal categories frequent gradation between predicate such as case and nominalizations and infinitives number Participle adjective subject is head, rest reduced; may take syntactically may conform to or of predication is adjectival concord principles governing adjectives adverb modifier when agreeing with subject
< previous page
page_25
next page >
< previous page
page_26
next page >
Page 26 others obligatorily derank the predicates in their noun clauses. The former languages will be called complement-balancing languages and the latter complement-deranking languages (see section 2.2.1. concerning the difficulties with this distinction). Although I am aware of the asymmetry in the use of these terms, I do not believe it to be confusing. Languages like Chinese, which only have balanced noun clauses, may be called strictly complement-balancing languages. As a point of departure I will use Noonan’s (1985:65) survey of complement types, which is summarized in Table 2.1. Paratactic constructions seem to fall outside the main topic of this chapter, since they do not constitute a clear case of either subordination or coordination. The other complementation types rather neatly cover those clauses which contain both dependent and non-dependent verb forms, as well as those that contain finite and non-finite verb forms. Below I shall touch upon Noonan’s classification in more detail and show that some of his definitions are not universally applicable. 2.2.1. S-like noun clauses The first major distinction within Noonan’s complement types is between ‘s-like’ (i.e. sentence-like) complements and all other complements. This distinction does not correspond to Stassen’s opposition between balanced and deranked clauses, because s-like complements may be either balanced or deranked (see section 2.2.2.), but coincides instead with the distinction between finite and non-finite clauses as I define it.1 Characterizing s-like complements, Noonan says: 1 an s-like complement type is ‘one that without its complementizers has roughly the same syntactic form as a main clause’; 2 their predicates remain syntactically and morphologically verbs; 3 ‘any case marking in subject or objects will have the same form as in main clauses’; 4 ‘if the verb in the main clauses is inflected for subject or object agreement in some language, then the verb in any s-like complement in that language will also be inflected for subject or object agreement’; 5 ‘that a complement type is s-like does not preclude the possibility that its syntax may differ in certain respects from that of main clauses’; 6 ‘all languages have some sort of s-like complement type’ (Noonan 1985:49).
< previous page
page_26
next page >
< previous page
page_27
next page >
Page 27 Within the category of s-like complements, Noonan distinguishes between indicative and subjunctive s-like complements in the languages that have more than one s-like complement type. Indicative refers to the form that most closely resembles declarative main clauses. Non-indicative, or subjunctive s-like complement types ‘usually have a special non-indicative stem or conjugation; they may also differ from indicatives in occurring with modal particles or special complementizers’ (ibid.: 51). In prototypical cases, features (1)–(4) do in fact appear together. However, they do not imply each other and for a number of languages the complement types that are most similar to independent assertions are characterized only by some of the above mentioned features. In other words, the choice of one or another of these characteristics as a defining property of s-like complements can lead to different results. Thus, it is not, for example, quite clear which differences between s-like complements and independent clauses are possible without impinging on the requirement that they have ‘roughly the same syntactic form’. As an example, Noonan mentions word order differences between main and s-like subordinate clauses in German and special ‘dependent’ forms in Irish which occur only in subordinate clauses and after some verbal particles. Wappo (a North-American Yuki language), where the subjects of independent and matrix clauses are marked with a suffix -i which is not allowed in subordinate clauses, would constitute another such case.2 In fact, it happens sometimes that the subjects in subordinate clauses appear in a form other than that of the subjects in independent clauses, even if in all other respects these types of clauses are identical (which thus contradicts Noonan’s feature (3)). Thus, in Classical and Modern Mongolian, subjects in indirect quotations may appear in the accusative.3 In Mokilese, a Micronesian language, the subjects of complements that ‘describe the action or state, rather than simply asserting that it is a fact’, look like possessors in NPs (Harrison 1976:270). Though Harrison calls such clauses ‘nominalized sentences’, they are otherwise similar to independent clauses. The claim made by Noonan that all languages have some form of s-like complements is very strong, but should be seen in the light of his statement that ‘since all languages have ways of presenting direct quotes, all languages use s-like complements with utterance predicates, though other complement types can occur with predicates in this class for indirect discourse’ (Noonan 1985:111). However, as Munro (1982) and McCawley (1968) (both quoted in Haiman 1985:223–4) showed, direct quotations can hardly be considered as syntactically
< previous page
page_27
next page >
< previous page
page_28
next page >
Page 28 embedded sentences. Among other things, direct quotations need not be in the same language as their matrix clauses (example 2.1a), or need not even be linguistic signs at all (example 2.1b). (2.1a) Ceasar said ‘veni, vidi, vici’, (2.1b) The car said ‘br-r-m-m’. Thus, the fact that a language uses s-like clauses for direct quotations does not say anything about its complementation system in general. Noonan himself mentions languages that restrict their use of s-like clauses to direct quotations. Two additional observations should be mentioned here. Firstly, some complement constructions may be formally indistinguishable from direct quotations, for example, in the case of other predicates that take complements referring to propositions (see section 1.5.1.). An extreme case of such an extension is presented by Teleéfoól, a New Guinean language, in which direct quotations cover the semantic areas of saying, desiderative, direct imperative, naming and perceptive (Healey 1964). Secondly, languages may vary with respect to their strategies of presenting indirect quotations. Thus, Abkhaz, a Northwest Caucasian language (Hewitt 1979:6) obligatorily deranks such clauses, while Tamil (Asher 1982) and Imbabura Quechua (Cole 1982) allow for indicative clauses in indirect quotations, even though they derank all other types of complements. The difference between direct and indirect quotations may be very subtle. Thus, in Tamil, the quotative particle ŋŋu follows the embedded sentence, and the latter may either ‘consist of the actual words used by the speaker whose speech is reported, or it may involve some changes, such as a change of person, from the words uttered. The result is that reported speech is sometimes ambiguous’ (Asher 1982:1). In Ingush, quotative clauses with the marker eanna—‘having said’ after verbs of saying and asking are similar to direct quotations in many respects, including the use of shifters, but are different from them in allowing inter-clausal reflexivization (Nichols forthcoming). Even in languages that have a clear opposition between direct and indirect quotations, there may exist intermediate forms (see Haiman 1985:225–7 on degrees of incorporation of quotation complements in English and German). The range of indirect quotations also shows great cross-linguistic variation. In a number of languages, constructions which obviously started as indirect quotations have gradually become grammaticalized as complements to various types of predicates (Lehmann 1982:64, Lord
< previous page
page_28
next page >
< previous page
page_29
next page >
Page 29 1976). In Buryat, a Mongolian language, finite clauses with the quotative marker (a gerundive from the verb ge- —‘to say’), besides coding direct and indirect quotations in combination with utterance predicates, may code ‘complements’ to manipulative verbs (as ‘to invite’), to emotional verbs and verbs of evaluation (as ‘to be better‘) (Skribnik 1983). This marker itself, together with some other forms of the same verb have presumably become grammaticalized as complementizers (Pjurbeev 1979:56–61). In conclusion, languages may vary as to which predicates, besides utterance predicates, have the form of direct quotations, and to what strategies they use in coding indirect quotations. This means that languages may be complement-deranking to different degrees. However, I will still retain the distinction between complement-deranking and complement-balancing languages, since this distinction has a direct bearing on the typology of action nominal constructions. Assuming that balanced complements are extremely restricted in complement-deranking languages, if they are possible at all, chapters 5–9 give detailed information on their uses for each language which is classified as complement-deranking. 2.2.2. Deranked vs. non-finite noun clauses S-like, or finite, complements are opposed to various types of non-finite clauses, i.e., those involving such traditionally distinguished non-finite verb forms as infinitives, verbal nouns and participles. The notion of finiteness is far from being well defined (see 1981b, Joseph 1983, KoptjevskajaTamm forthcoming). In the European linguistic tradition, non-finite verb forms are recognized by their morphological and categorial features, i.e., they are characterized by a reduction of verbal morphology, which primarily involves personal flection, and/or a change of category status. Verbal nouns, including action nominals, are non-finite verb forms which are primarily characterized by a change of category status. Although traditionally, the notion of finiteness applies only to inflectional verb forms, I shall classify all action nominals as non-finites. Non-finite verb forms are dependent verb forms par excellence, because they can never function as the predicate of independent clauses. The opposite implication, however, does not hold, i.e. dependent verb forms need not be non-finite. Thus, in a number of complement-deranking languages, independent clauses and complements are structurally quite similar to each other, differing only in their use of independent or dependent verb moods respectively. Take, for example, West Greenlandic, in
< previous page
page_29
next page >
< previous page
page_30
next page >
Page 30 which ‘subordination is marked by verbal inflection alone: mood inflections are divided neatly into superordinate ones (indicative, interrogative and imperative/optative) and subordinate ones (causative, conditional, participial and contemporative)’ (Fortescue 1984:34). Complements with subordinate moods in Greenlandic satisfy all of Noonan’s criteria for s-like complements, since they have the same syntactic structure as independent clauses, their predicates remain syntactically and morphologically verbs, all the syntactic relations between the verb and its arguments are signalled in the same way as in independent clauses (by means of cases and verbal agreement). I suggest qualifying such cases as deranked s-like complements. There is not always consensus on the terminology used for referring to dependent moods. Some grammars consider them as finite forms which are called subordinate, dependent, participial, etc., while others regard them as non-finite verb forms. To illustrate the two positions, let us compare Guma’s (1971:157) description of Southern Sotho (a Bantu language) and Hewitt’s (1979, 1987) description of Abkhaz (a Northwest Caucasian language). Although the two languages have somewhat similar oppositions in their verb systems, these are treated differently. Verbal conjugation in Southern Sotho, according to Guma, consists of two non-finite moods (the infinitive and the imperative), and finite moods ‘which must always have a subject concord’ and, in their turn, are subdivided into primary moods and a secondary mood (the subjunctive, used in ‘subordinate clauses to express polite commands, requests, wishes, exhortations and habitual actions’). Primary moods include the indicative and conditional, each of which distinguishes between ‘principal’ positive and negative (used in main clauses) and ‘participial’ positive and negative (used in subordinate clauses). Thus, Guma clearly distinguishes between non-finite and finite verb forms, the latter comprising independent (‘principal’) and dependent (‘participial’ and subjunctive) forms in Southern Sotho. For Abkhaz, Hewitt, in contrast to Guma, does not distinguish between non-finite and finite dependent verb forms. Thus, characterizing various nonfinite forms, he writes: ‘Also, in addition to possessing the archetypal non-finite form, the verbal noun (masdar//infinitive)… together with a past absolutive in -n (equivalent in meaning to an English perfect participle) and (in the case of dynamic verbs only) a present absolutive, which is formally identical to the non-finite Present and is equivalent in meaning to an English present participle…Abkhaz makes a most important distinction between finite and non-finite forms of each
< previous page
page_30
next page >
< previous page
page_31
next page >
Page 31 main indicative tense. The function of the non-finite forms is, by incorporating a variety of suffixes, to represent the entire range of subordinate constructions. (Hewitt 1979:201) West Greenlandic, Southern Sotho and Abkhaz are, thus, complement-deranking languages which distinguish between dependent finite and non-finite verb forms and, consequently, between deranked s-like and non-finite complements. Other complement-deranking languages, such as Tamil and Imbabura Quechua, do not have any s-like complements at all, because their only dependent verb forms are non-finite. On the other hand, even in complement-balancing languages it may be possible to distinguish between independent and finite dependent verb forms and, consequently, between balanced and deranked s-like complements. Swahili, which, like Southern Sotho, has a complicated system of independent and dependent moods, seems to allow for balanced s-like complements (Sune Magnusson personal communication). Subjunctive forms in some European languages may also be considered finite dependent verb forms. There are also languages with dependent verb forms, but without non-finite ones. Verbs in the Algonquian languages, like Cree (Ellis 1971), distinguish among several ‘orders’, first of all, between those which are used in independent and matrix clauses and those which are used in subordinate clauses. Clauses headed by verbs of different orders are structurally similar, and there are no reasons to consider some of these orders as non-finites. The distinction between independent and dependent verb forms is not always clear-cut. First, erstwhile dependent verb forms are sometimes used in independent clauses, most often with certain modal meanings. Thus, Noonan (1985:54) writes that all languages with subjunctive complements seem to be able to use subjunctives as independent clauses. Some of these cases are clearly connected with the uses of subjunctives in dependent clauses, which might sometimes be seen in the possibility of retaining a complementizer, like in the French example Qu’il vienne!—‘Let him come!’4 There is also a general tendency to finitivize non-finite verb forms, which thus start being used as matrix predicates. Infinitives are used in a number of languages with the imperative meaning, as in the French example Ne pas se pencher dehors—‘Do not lean out’. Erstwhile non-finite verb forms, like participles and action nominals, may even enter the finite paradigm of the corresponding verbs to express certain tense and aspect meanings, and ultimately acquire a distribution which is
< previous page
page_31
next page >
< previous page
page_32
next page >
Page 32 comparable to that of finite verb forms. Thus, in several Russian dialects, adverbial gerunds (converbs) are used in a predicative function with the perfect meaning, as in Syn uže tri goda ženivši—‘The son has been married for three years’. Another example is presented by Buryat (Skribnik 1988:14–18), in which what are called ‘participles’ combine with two different sets of agreement suffixes depending on their syntactic function. As dependent predicates (in complements and adverbial clauses), they take nominal possessive affixes, while as matrix predicates they take verbal personal agreement. There is, however, one participle (with the suffix -xa- ), which always takes possessive suffixes in the meaning of the near future tense, even when used as the matrix predicate. It is somewhat unclear to what degree the distinction between finite and non-finite verb forms is appropriate in such cases. A summary of the material in this section is given in Table 2.2. According to Joseph (1983), Bulgarian is in the process of losing the distinction between finite and nonfinite verb forms and, thus, would be a possible candidate for a complement-balancing language with deranked finite complements. However, Joseph’s notion of finiteness differs from mine. Since Bulgarian has a productive way of deriving action nominals, which are qualified as non-finites in this book, it could not be counted as a representative of the missing category in the table. Table 2.2 Balanced vs. deranked and finite vs. non-finite complements Language Finite (s-like) Balanced complements Deranked Non-finite complements * − − − Chinese + − − ? + + − Russian + − + Abkhaz, Sotho, Eskimo − + + Imbabura Quechua, Tamil − − + Swahili + + + Cree − + − Key: * an impossible language ? no clear examples of such languages
< previous page
page_32
next page >
< previous page
page_33
next page >
Page 33 2.2.3. Infinitives vs. verbal nouns. Examples 2.2.3.1. Russian In the present context the most important distinction is that between infinitives and verbal nouns, and in particular, action nominals (or nominalizations, in Noonan’s terminology). Russian serves as an example of a language with a clear distinction between infinitives and action nominals, corresponding neatly to Noonan’s criteria. Compare sentences (2.2a) and (2.2b) which have an infinitive and an action nominal respectively. (2.2a) Ja ne ljublju gromk-o čita-t’ stix-i I not like loud-ADV read-INF poem-ACC.PL ‘I do not like to read poems loudly.’ (2.2b) Ja ne ljublju Pet-in-o gromk-oe I not like Peter-POSS-N.ACC loud-ADJ.N.ACC čt-eni-e stix-ov. read-AN-ACC poem-GEN.PL ‘I do not like Peter’s loud reading of poems.’ The infinitive čitat’ cannot form a constituent with its ‘notional’ subject and cannot be interpreted as having a subject different from the subject or the object of the matrix clause. It combines with objects and adverbs in the same way as finite verbs do. All verbs in Russian have infinitives which are built in regular ways. Infinitives, thus, are traditionally viewed as inflectional verb forms. They do not retain verbal agreement and tenses, but show oppositions in aspect and voice. The action nominal čtenie has a subject (Petja —‘Peter’) different from that of the matrix clause. In the example above, the subject of the AN has an ‘associative relation’5 with the AN itself, which means that the proper name stands in the special possessive form. Action nominals cannot combine with objects in the accusative and adverbs, as do finite verbs, explaining why the adverb loudly has turned into the adjective loud and the object of the verbal noun is in the genitive. This means that the internal structure of the action nominal construction in (2.2b) is the same as that of NPs. Action nominals are clearly derivational formations and cannot be formed from all verbs. They do not manifest any oppositions either in tense, aspect or voice. On the other hand, they can decline exactly as non-derived nouns. The boundary between infinitives and action nominals is, however, rather vague, and even in languages that do distinguish between
< previous page
page_33
next page >
< previous page
page_34
next page >
Page 34 infinitives and verbal nouns, the differences between them may be very small. Some examples from languages in which attribution of verbal derivatives to infinitives or action nominals is much less straightforward than in Russian are given below. 2.2.3.2. Selkup and Finnish In a number of languages, the problem of distinguishing infinitives from verbal nouns is rather difficult, since the infinitives have many morphological characteristics typical for nouns and thus come quite close to the class of verbal nouns which also exist in these languages. In Selkup (a Samoyedic language), for example, infinitives (consisting of a verbal stem and the suffix -qo, which is formally the same as the suffix of the translative case in nouns) take nominal possessive suffixes to indicate their subjects. However, they are distinguished from verbal nouns in that they do not decline and cannot combine with postpositions in the same way as the latter (Kuznecova et al. 1980:249–51). Finnish presents another interesting example of infinitives with clear nominal features. The language possesses three types of infinitives which not only combine with possessive suffixes to refer to the subject of the action, but can decline as well. Each infinitive can take several cases, though not all. These case forms of the infinitives have very specialized meanings. Thus, the inessive of the second infinitive is used to express an action which is simultaneous with the main action, while the inessive of the third infinitive is most often used in combination with the verb olla—‘to be’ to express an action in progress: (2.3) Ol-le-ssa-ni Ruotsi-ssa, tapasin useita be-INF2-INESS-1 SG.POSS Sweden-INESS met:1SG many ystäviä. friends ‘When I was in Sweden, I met many friends.’ (Karlsson 1978:185) (2.4) Mari on kirjasto-ssa luke-ma-ssa. Mari is library-INESS read-INF3-INESS ‘Mari is in the library and is reading.’ (ibid.: 188) Besides these three infinitives, there are two rather uncommon verb forms which are traditionally called the fourth and the fifth infinitives and which have a very restricted syntactic usage (only in combination with the verb olla—‘to be’). The fourth infinitive has the suffix -minen
< previous page
page_34
next page >
< previous page
page_35
next page >
Page 35 in the nominative, its only other case form is the partitive. The exact status of these five infinitives has long been a controversial matter in Finnish linguistics. Some authors regard them as infinitives, some consider them as verbal nouns. However, in Finnish, there are also other regular forms which are derived from verbs and have the suffix -minen in the nominative and which are called verbal nouns in minen. Dubrovina (1972:11) mentions, among other things, the following differences between infinitives and verbal nouns: 1 verbal nouns have the full paradigm of cases and numbers; 2 verbal nouns freely combine with various pre- and postpositions, while infinitives do so in exceptional situations only; 3 verbal nouns combine with adjectives as attributes, while infinitives combine with adverbs; 4 the object of verbal nouns takes the genitive, while the object of infinitives appears in one of the cases allowed for the object of finite verbs (the nominative, the partitive and the accusative). As will become apparent from the description of my sample (especially of the SENT and the POSS-ACC nominalization types), the ability to combine with adverbs and direct objects in their sentential form does not constitute the major criterion for distinguishing between infinitives and action nominals. However, since Finnish infinitives do not decline in all the cases and cannot combine with pre-and postpositions in the way nouns do, they cannot be regarded as action nominals. On the other hand, they are not typical infinitives—not only can they decline, at least in some cases, but they can also combine with their ‘notional’ subjects, which is rather unusual for infinitives. Thus, the subject of the second infinitive in the inessive stands in the genitive if it is not the same as the subject of the matrix clause: (2.5) Peka-n herät-e-ssä Liisa lähtee Peka-GEN wake.up-INF2-INESS Liisa goes tö-i-hin. work-PL-ILL ‘When Pekka wakes up, Liisa goes to work.’ (Karlsson 1978:186) From the typological point of view, at least some of the Finnish infinitives should rather be classified as converbs—non-finite verb forms which are primarily used with adverbial functions (see section 2.2.6.).
< previous page
page_35
next page >
< previous page
page_36
next page >
Page 36 2.2.3.3. Altaic languages Finnish and Selkup have been presented as examples of those languages where the infinitives are characterized by unusually many nominal features and thus resemble the verbal nouns which also exist in these languages as a separate category. However, in a number of languages, infinitives and action nominals are not opposed to each other, since there is only one ‘intermediate’ class of non-finite verb forms which combine certain features of typical infinitives and typical action nominals. The treatment of such categories is notoriously controversial and tends to vary even among descriptions of a single language, particularly if they are made within different linguistic traditions. This situation is exemplified by some of the Altaic languages (broadly defined), such as the Turkic, Tungus and Mongol languages, as well as Korean (for examples, see sections 2.3. and 5.1.). These languages are complementderanking with a rich system of non-finite verb forms, among which one finds those variously called ‘infinitives’, ‘masdars’, or ‘verbal nouns’ serving as the head of non-finite noun clauses and characterized by the following features: 1 they take all the nominal case endings, combine with the same range of adpositions and can be used in the same syntactic contexts as non-derived nouns (see p. 37 for the possible exception, the lack of the genitive in the case paradigm of some of these forms); 2 they can form a constituent with their subject which is in the nominative, genitive or accusative; 3 they can be formed from any verb in a very regular way; 4 their objects are in the same form as those of the corresponding finite verbs; 5 they can normally combine with the same range of adverbial modifiers as the corresponding finite verbs. Features 1 and 2 qualify these words as action nominals, while features 3–5 make them look very much like infinitives. Let us consider both facets of these words. These words satisfy all the semantic, syntactic and morphological criteria in the definition of action nominals. What makes them rather special is the fact that they are regularly formed from any verb and should thus be considered as regular verb forms and not derived nominals, and also the fact that they are extremely widely used, which implies that the range of their meanings and different uses is much greater than in, say, Russian or English. Both of these special features of the action nominals in Turkic, Tungus, Mongol6 and Korean are
< previous page
page_36
next page >
< previous page
page_37
next page >
Page 37 connected with the status of these languages as complement-deranking (more on the uses and meanings of action nominals across languages will be said in section 2.3). Is it justified to call these verb forms infinitives as well? We are left here with two options: either to admit that in some languages the classes of infinitives and action nominals coincide, or to define infinitives in such a way that they will always be distinguished from action nominals. I prefer the latter solution. In my view, infinitives are verb forms which are regularly formed from any verb, combine with their objects in the same way as the corresponding finite verbs, can be used in some contexts in which ordinary NPs are used, but do not have the same range of inflectional and distributional characteristics as non-derived nouns. Thus, Turkish, Mongolian and Korean, according to this definition, have action nominals (verbal nouns), but lack infinitives. However, there are sometimes problems in establishing whether a certain candidate for an action nominal in a particular Altaic language can take the genitive case, and the available sources may offer contradictory information. A couple of examples will illustrate this point. For Korean, Ramstedt (1968) states that verbal nouns can be used in all case formations. However, in his list of possible uses of the verbal nouns, the genitive is lacking (Ramstedt 1968:123–4). According to Vasil’ev (1961), Korean ki-infinitives have not yet become bona fide nouns in that they lack some of the cases—genitive, comitative and some others. However, recent grammars of Korean mention that these infinitives can, though only rarely, take the genitive suffixes (Lev Koncevič personal communication). In Classical Mongolian, according to Grønbech and Krueger (1955:33), verbal nouns can be declined in all the cases, which is illustrated by examples including the genitive. For Modern Mongolian, Binnick (1979:83) states that verbal nouns lack the genitive. However, according to Jan-Olof Svantesson (personal communication), verbal nouns may stand in the genitive case, as well as in all the other cases. Presumably, it is the nature of the genitive case marking itself in these languages which accounts for the total lack or infrequency of a genitive form in the paradigm of these verb forms. Firstly, on the whole, the use of the genitive in these languages is very special and is, in many respects, determined by the referential or semantic features of the particular nominal. Thus, in Turkish possessive noun phrases, the first noun (the modifier) must be marked with the genitive if it is ‘referential’, regardless of whether it is definite or indefinite. When ‘non-referential’, the modifier is in the non-marked form (Nilsson
< previous page
page_37
next page >
< previous page
page_38
next page >
Page 38 1985:69). Thus, it may be the semantics of these verb forms (denotation of the action) which does not satisfy the referential demands of the genitive case. However, this can hardly be the whole truth, since in the function of direct objects such words may take the accusative case, which, as for the genitive case, is allowed only for referential NPs. Secondly, syntactically the genitive case marks dependents of nouns, postpositions and certain non-finite verb forms, and the corresponding functions are, perhaps, not as usual for verbal nouns, which most often occur as complements to verbs. As free dependents of nouns (i.e., as dependents of nouns which are not subcategorized for such dependents), verbal nouns normally occur in the non-marked form. The genitive case occurs relatively infrequently, in combination with ‘relational’ nouns (like ‘reason’, ‘cause’, etc.) and with nouns, which are on their way to being grammaticalized as postpositions (Ubrjatova and Litvin 1984:164–72). The fact that a language has only action nominals and does not have any real infinitives, does not imply that its action nominals must be derived from the corresponding verbs in a strictly regular way. Thus, the Balkan languages such as Modern Greek and Bulgarian lack infinitives, but their action nominals are formed rather irregularly and not from all verbs. As mentioned above, some of the peculiarities of the action nominals in Turkish, Classical and Modern Mongolian, Tungus (Evenki) and Korean are connected with their status as complement-deranking languages. Support for this statement is offered by Gagauz, a Turkic language which has long been in contact with languages from other families, i.e. Bulgarian, Moldavian (Rumanian) and Russian. The verbal nouns of the other Turkic languages, which in most respects show the same features as those of Turkish, do not have any exact correspondence in Gagauz. Instead, they gave rise to two different categories in Gagauz—the infinitives in -maa/-mää and the verbal nouns in -dyk. These latter are very rarely used, and then only in restricted contexts, due to the fact that Gagauz has developed a system of finite subordinate clauses and has thus become a balancing language (Pokrovskaja 1978:126). Pure infinitives in Gagauz have originated from the dative case of the verbal noun in -mek. They do not decline and, though used in many contexts where the other case forms of the verbal nouns are used in, for example, Turkish, they are considerably more restricted in their usage. It is interesting to note that, in many contexts, the -maa-infinitive can be replaced by the subjunctive form of the same verb. This coexistence of such synonymous constructions in Gagauz shows the different directions of
< previous page
page_38
next page >
< previous page
page_39
next page >
Page 39 influence on Gagauz from the contacting languages. Thus, constructions with subjunctives reflect the common Balkan feature of loss of the primary verbal nouns or infinitives. Constructions with -maainfinitives reflect the later Balkan development of the appearance of the ‘secondary’ infinitive (as in Rumanian), which in the Gagauz case has been strengthened by contact with Russian (Pokrovskaja 1978:116). The same development characterizes Karaim, another Turkic language (ibid.: 117). Similar tendencies—namely the loss of verbal nouns or the development of severe restrictions on their use, as well as the decrease of their verbal features as connected to the appearance and growth of finite subordinate clauses—are typical for the Uralic languages (Tauli 1966:87–97). For example, of the Baltic-Finnic and Lapp languages, the verbal noun constructions have been ‘most extensively replaced by the SC [subordinate clauses] in F[=Finnish], Lapp and Hungarian, i.e. in the languages where development has been influenced by the western IE languages’ (ibid.: 88). 2.2.3.4. More examples: Irish, Warlpiri, Vedic Sanskrit, Tzutujil Irish presents a rare case of a language in which verbal derivatives (traditionally called verbal nouns) have nominal inflectional features and combine with the genitive case of the object (which, thus, distinguishes them from typical infinitives), but cannot form a constituent with the subject, which is either equi-deleted or raised (Noonan 1985:61). (2.6) Is ionadh Horn Seán a bhualadh COP surprise with:me John COMP hit:VN Thomáis Thomas:GEN ‘I’m surprised that John hit Thomas.’ (Noonan 1985:61, ex. 89) In Warlpiri, an Australian aboriginal language, ‘nominalized’ verbs are quite close to typical infinitives in a language such as Russian, in that their objects have the same form as in combination with finite verbs; and in their inability to have overt subjects (which are, in most cases, coreferential with some NP in the matrix clause). However, unlike Russian, but similar to Finnish, nominalized verbs in Warlpiri may take some, though not all, case suffixes (Simpson 1983:426–49).
< previous page
page_39
next page >
< previous page
page_40
next page >
Page 40 (2.7) Nyuntulu-rlu+ju kalaka-npa-nyanu marda-rni you-ERG+EUPH ADMON-2SG-RFL have-NONPAST witiniji+yijala yimi ngarri-rninja-ku. witness:ABS+also story:ABS tell-INF-DAT ‘You should also have witnesses to prove your story.’ (Simpson 1983:437, ex. 24) (In example (2.7) EUPH indicates a stylistic phonological extender and an old information marker; and ADMON indicates potential.) Vedic Sanskrit has a large number of verbal derivatives traditionally called ‘infinitives’. Some of these can take different cases, though not all. They take objects either in the same form as finite verbs (in the accusative) or in the dative or genitive. In certain situations, for example, when used in purpose clauses, ‘infinitives’ may have their own subjects in the dative case (Disterheft 1980:27–64). In the absence of indications to the contrary, I assume that Vedic Sanskrit infinitives are capable of building constituents with their overt subjects. (2.8) té no hinvantu [sātáye dhiyé these:NOM us(ENCL) they:drive prize:DAT thought-DAT ] win:INF ‘They should drive us in order that the thought win the prize.’ (Disterheft 1980:64, ex. 42b). In Tzutujil, a Mayan language (Dayley 1985), most verbs have infinitives or verbal nouns. ‘Root transitive verbs’ (monosyllabic transitive roots) have active infinitives which may occur freely without objects. ‘Derived transitive verbs’ (formed with a root plus derivational suffixes) have active infinitives that require the presence of an overt object. In both cases, however, active infinitives combine only with indefinite or referentially non-specific objects (Dayley 1985:105, 396). They do not take overt subjects and are not inflected for possession, which is otherwise an important flectional characteristic of common nouns in this head-marking language (ibid.: 179–81, 141). In other words, active infinitives in Tzutujil are relatively close to typical infinitives (such as those in Russian), except for their peculiar restrictions on the presence and form of objects. On the other hand, definite or referentially specific patients are allowed only in combination with passive infinitives. These take possessive prefixes referring to the patient (ibid.: 396). This means that passive infinitives form a
< previous page
page_40
next page >
< previous page
page_41
next page >
Page 41 constituent with their subjects and have acquired (at least one) nominal inflectional characteristics. In other words, passive infinitives are close to intransitive action nominals in many languages. (2.9a) Xqaamaj [choyoj chee ]. B3:A1PL:began cut:ACT.INF tree ‘We began to cut trees.’ (Dayley 1985:393, ex. 129b) (2.9b) Xqaamaj [rchojyiik (ja chee )]. B3:A1PL:began its:cut:PASS.INF the tree ‘We began to cut it (the tree).’ (ibid.: 393, ex. 129c) (In examples (2.9a,b) above, A and B refer to ergative and absolutive person prefixes, respectively.) All the cases quoted in section 2.2.3. are summarized in Table 2.3. The columns in the table correspond to the three properties which characterize infinitives vs. nominalizations in Noonan’s survey (see Table 2.1.), more specifically, ability to build a constituent with Table 2.3 Infinitives (INF), action nominals (AN) and intermediate forms: some examples Language Ability to build constituent Nominal inflectional Object in the same form as with with subject categories finite verbs Russian INF − − + AN + + − Selkup INF +p − + AN + + + Finnish INF + +/− + AN + + − Turkish + + + Vedic + +/− +/− Sanskrit Warlpiri − +/− + Irish − + − Tzutujil ACT.INF − − + PASS.INF+ + 0 ? + − − ? − − − Note: ‘p’ means that verbal derivatives form constituents only with pronominal subjects, ‘0’ means that the issue is not relevant Key: ? no clear example of such languages.
< previous page
page_41
next page >
< previous page
page_42
next page >
Page 42 subject, range of nominal inflectional categories, and relation with object. In this connection, it is appropriate to mention that typical infinitives, such as those in Russian, have developed as some case (locative, accusative, dative, etc.) of the action nominal. This is true for modern Indo-European languages and, as shown by Haspelmath (1989), for a large number of genetically non-related languages. Haspelmath suggests that it is also possible to identify a crosslinguistic category of infinitives by a combination of their syntactic features (like the ones mentioned by Noonan), their semantics (expression of purpose and so on) and their typical origin (a directional/purposive form of a verbal noun). On the other hand, infinitives may become substantivized (as in German7 and Persian) and start being used as ANs. To quote Disterheft (1980:198), ‘the infinitive can then be used throughout the nominal paradigm as was the original nominal verbal abstract, thus completing a full cycle of noun-to-infinitive-to-noun’. 2.2.4. Masdars Some words must be said about the term ‘masdar’ (Cl. Arabic, ‘the place whence anything goes forth’). In Arabic linguistics, this word applies to what we here call action nominals. In Classical Arabic, there exist more than forty patterns for deriving masdars from the underlying verb. According to Wright (1971:112), ‘the majority of verbs admit of but one form, very few of more than two or three. What these are, must be learned from the Lexicon.’ The term ‘masdar’ is used in descriptions of quite a number of different languages, especially those which have been influenced by the Arabic linguistic tradition. The heterogeneity of the cases covered by this term is quite comparable to the situation with the term ‘action nominal’. In some cases, masdars apply to regular verb forms, as in Abkhaz or in Turkish (where the Turkish tradition itself uses this term for what we have agreed to call ANs in section 2.2.3). In other cases, masdars are derived nominals, as in Georgian. 2.2.5. Participles Participles are defined as adjectival verb forms, i.e. verb forms which primarily function as an attribute to a noun in noun phrases. Thus, they are opposed to ANs which primarily function as the head in noun phrases. However, many languages do not distinguish between these two functions, or rather, between these two types of words.
< previous page
page_42
next page >
< previous page
page_43
next page >
Page 43 Nominalizations on the whole, and ANCs in particular, are broadly used in various languages for building relative clauses (Lehmann 1984). Thus, in the Altaic languages, relative clauses are based on certain non-finite verb forms, called participles, most of which, in different cases, can also be the head of ANCs. Such ANCs, in their turn, function as complements and adverbial clauses (a typology of such uses is presented in Ubrjatova and Litvin 1984, 1986; Skribnik 1988). In these languages, the grammatical tradition distinguishes between participles, which may be used as attributes to nouns, and ‘pure’ ANs, which never have the attributive function. (On the different terms used for referring to these words see sections 2.2.3.3. and 2.2.4.) Quechua does not distinguish between adjectives and nouns. It is quite natural that there does not exist any opposition between ANs and participles in that language either. Languages where a single verb form can be used both as the head of a complement and as the head of a relative clause may present additional problems for determining whether these forms can be characterized as action nominals or not. Such is the case with nominals which are derived by a totally productive and regular nominalization process in Kayardild (a non-Pama-Nyungan Australian language; Nick Evans personal communication). These nominals can refer either to the action (or the act) itself or to its agent and can be used either in typical nominal functions or as the head of relative clauses (as participles). Thus, the form wirrka-n-da (‘dance’-NMLZ-NOM) means ‘a dance’, ‘a dancer’ and ‘dancing’. Although these nominals have the whole nominal inflectional paradigm, they are highly restricted in the range of their functions when referring to actions and, thus, cannot qualify as action nominals. In other words, nominalized forms in Kayardild take most of their case suffixes when used as participles and not as action nominals. Nominalizations in Kayardild, thus, demonstrate the insufficiency of the criteria for identifying ANCs as they are stated in 1.2.1. It should be mentioned here that ANCs are not always easily distinguishable from relative clauses, for example, as complements to immediate perception predicates. Thus, in the following example from Kayardild, the complement to the matrix verb ‘to see’ could, in principle, be analysed either as an ANC with kurdama-n-kina —‘drinking’ as the head, or as a relative clause with niwan-jina—‘him’ as the
< previous page
page_43
next page >
< previous page
page_44
next page >
Page 44 antecedent: (2.10) ngada kurri-jarra niwan-jina kurdama-n-kina I:NOM see-PAST him-M.ABL drink-NMLZ-M.ABL nguku-naa-ntha wuruman-urru-naa-nth. water-M.ABL-A.OBL billycan-ASSOC-M.ABL-A.OBL ‘I saw him drinking the water in the billycan.’ (Dench and Evans 1988:34, ex. 68) (Non-standard abbreviations in 2.10 include M.ABL—Modal Ablative, the case on NPs when the governing verb is in the past; and A.OBL—Associating Oblique, the case on non-subject arguments of nominalized verbs.) 2.2.6. Converbs Converbs (adverbial participles) are non-finite verb forms used in adverbial functions. While infinitives across languages may normally occur as complements to such verbs as desiderative (‘want’) and manipulative (‘order’), converbs are not used as complements, but express instead a variety of adverbial meanings, such as temporal relations between the main and the subordinate actions, manner, cause, condition of the main action, and many others (Nedjalkov 1990). In a number of languages, converbs have originated as certain cases of verbal nouns. The exact attribution of such ‘case-inflected’ forms can be especially problematic in those languages where they form a whole class. However, while case inflection of action nominals is a living process and reflects their different syntactic and semantic uses, the cases of prototypical converbs are fossilized and are interpreted rather as a part of the whole converb marker. In other words, the marker of the verbal noun and the case marker together are conceived as an indivisible converb marker. An illuminating example of such a differentiation between action nominals and converbs is offered by non-finite verb forms in -ganča in Altay (a Siberian Turkic language), originally an action nominal with the suffix -gan in the directional case (Ubrjatova and Litvin 1986:56). These forms are used in two different functions, which are also distinguished morphologically, by the placement of possessive suffixes, which in normal nouns precede case suffixes. When -ganča forms are used with the meaning ‘according to what X V-ed’, possessive suffixes appear before the case marker, e.g., ajt-kan-ym-ča (say-AN-1SG.POSS-DIR)—‘according to what I said’. However, used with the meaning ‘until V’, the -ganča marker
< previous page
page_44
next page >
< previous page
page_45
next page >
Page 45 as a whole precedes possessive suffixes, e.g., men kel-genče-m (I come-CONV-1SG.POSS)—‘until I come’. Thus, in the latter construction -ča is obviously no longer interpreted as a real case marker. 2.3. LANGUAGE SPECIFIC VARIATIONS IN THE SEMANTIC RANGE OF ANCs, SOME EXAMPLES In the first chapter, we discussed the total range of meanings that ANs and ANCs have across languages. However, the second of the questions formulated in section 1.5.1., concerning the exact range of meanings peculiar to ANs and ANCs in a particular language, cannot be answered in a general way. The problem may be illustrated by taking examples from a few languages. In complement-balancing languages, ANCs compete with finite noun clauses (and, perhaps, with some other types of noun clauses) for meanings and functions. Thus, both in English and Russian, ANCs are capable of expressing events (action, processes), facts and propositions, even if the last two categories are normally denoted by finite noun clauses. However, for determining the possible syntactic type of a complement, it is not always sufficient to know its semantic category. In other words, even predicates within a single semantic type may show idiosyncrasies in their choice of complement type, such idiosyncrasies being either common to a group of predicates or specific to one or two of them. The first situation is illustrated by perception verbs in Russian. They take complements referring to events, actions, processes. Such complements are normally expressed by ANCs. However, there exists a class of finite subordinate clauses with the complementizer kak (‘how’) which can be used after perception verbs, as in Ja nabljudal za tem, kak deti guljali —‘I was watching the children taking their walk’, which is (almost) equivalent to Ja nabljudal za progulkoj detej—‘I was watching the children’s walk’. Most often the complementizer kak introduces adverbial manner clauses, that is, adverbial clauses which describe the way of performing an action, and not the action itself. However, in the example above, read with normal intonation, the complement clearly refers to the action. The ‘manner’ interpretation of the same clause is also possible, if it is read with extra emphasis on the complementizer kak. Recall that complements of perception verbs in many languages tend to show a special behaviour (cf. constructions accusativum cum infinitivo). Examples of lexically restricted idiosyncrasies in the choice of complements are also usual. I shall mention one example. In Russian, almost all the phasal verbs can combine with action nominals, as
< previous page
page_45
next page >
< previous page
page_46
next page >
Page 46 načat’, prodolžit’, prervat’, prekratit’, vozobnovit’, končit čtenie—‘to start, go on with, interrupt, stop, take up, finish reading’ (the majority, though not all, can combine with infinitives as well). However, the verb perestat’ —‘to stop’ cannot take action nominals as complements, the only option for the latter being infinitives. Thus, perestat’ čtenie is absolutely impossible and must be substituted by perestat’ čitat’. Many languages have several types of verbal nouns with different meanings and functions. The resulting system of verbal nouns is especially intricate in complement-deranking languages. Turkish illustrates this case. As other Altaic languages, Turkish has a number of verbal nouns and participles, the latter being used both as attributes to nouns and as nouns themselves, while the former can only be used as nouns (see sections 2.2.3.3. and 2.2.5). Noun clauses with participles as their head refer to facts and propositions, as in sentences (2.11) and (2.12) respectively. (Note that the two most often quoted grammars of Turkish, Underhill 1976:321, and Lewis 1967:165 use the term ‘fact’ to refer to both facts and propositions.) (2.11) [Halil’in gel-eceğ-in-i/ Halil:GEN come-PTCP.FUT-3SG.POSS-ACC/ gel-diğ-in-i] bili-yor-um come-PTCP.NONFUT-3SG.POSS-ACC know-PRES-1SG ‘I know that Halil will come/came, comes.’ (Underhill 1976:322) (2.12) Ahmet Bey, [kız-ın-ın saat ondan Ahmet Bey daughter-3SG.POSS-GEN clock ten:ABL evvel eve dön-düğ-ün-ü] before home return-PTCP.NONFUT-3SG.POSS-ACC say-PAST-3SG ‘Ahmet Bey said that this daughter returned home before ten o’clock.’ (ibid.: 323) In the class of verbal nouns, those with the suffixes -mEG (‘the infinitive’) and -me (‘the short infinitive’), which are theoretically interchangeable with each other, are, according to Underhill, ‘used in verbal noun constructions where only the action of the given verb, or verb phrase, is indicated’ (Underhill 1976:310).
< previous page
page_46
next page >
< previous page
page_47
next page >
Page 47 (2.13) Çocuk [ağla-mağ-a] child cry-INF-DAT begin-PAST-3SG ‘The child began to cry.’ (2.14) En sev-me-dğ-im , [ders most like-NEG-PTCP-1SG.POSS thing lesson ] study-INF-PRED ‘The thing which I dislike the most is studying.’ (2.15) Ahmet [kız-ma-ma-yı] Ahmet get.angry-NEG-INF-ACC learn-PRES-3SG ‘Ahmet is learning not to get angry.’ (Underhill 1976:310–11) The sentences above can be considered as examples of event (action, process) complements. However, the verbal noun constructions in examples (2.16) and (2.17) should rather be interpreted as corresponding to facts: (2.16) [Türkiye-nin harbe gir-me-sin-in] sebebi Turkey-GEN war:DAT enter-INF-3SG.POSS-GEN reason ‘the reason for Turkey’s entering the war’ (ibid.: 324) (2.17) [Ahmed-in er ken yat-ma-sın-a] Ahmed-GEN early go.to.bed-INF-3SG.POSS-DAT get.used-PRES-1PL ‘We are getting used to Ahmet’s going to bed early.’ (ibid.: 311) Some predicates combine both with participle and verbal noun constructions, resulting in different meanings. These predicates include the verbs bilmek (‘to know’), yazmak (‘to write’) and so on. However, there is still another small class of verbs that appear with either type of construction without any difference. These include sevinmek (‘to enjoy’) and memnun olmak (‘to be pleased’). Examples (2.18a–b) illustrate this. (2.18a) [Köpeğ-in-i ev-de dog-2SG.POSS-ACC house-LOC birak-tığ-ın-a mnun ol-du-m. leave-PTCP-2SG.POSS-DAT pleased be-PAST-1SG ‘I was pleased that you left your dog at home.’
< previous page
page_47
next page >
< previous page
page_48
next page >
Page 48 (2.18b) [Köpeğ-in-i ev-de dog-2SG.POSS-ACC house-LOC bırak-ma-n-a] memnun ol-du-m. leave-INF-2SG.POSS-DAT pleased be-PAST-1SG ‘I was pleased at your leaving your dog at home.’ (Underhill 1976:324) Besides the above mentioned verbal nouns and participles, Turkish has a class of verbal nouns derived with the suffix . They are used, primarily, to refer to the manner of performing the action involved, as in (2.19) and (2.20), and thus could be termed ‘mode nominals’ (nomina modi): (2.19) kasabaya kadar this walk-MN-INSTR town:DAT evening:DAT until olacağiz reach:PTCP become:FUT:1 PL ‘With this way of walking, at this rate, we shall have reached the town by evening.’ (Lewis 1967:172) (2.20) Bu adam bize this man we:DAT towards much under stand-MN gösterdi show:PRET.3SG ‘This man showed great understanding towards us.’ (Underhill 1976:313) However, the nouns can have other functions as well, denoting ‘the fact of the action’ (Lewis 1967:172). This is illustrated by examples (2.21) and (2.22). (2.21) Her gün bir saat yapmalısınız. every day one hour walk-MN do:NECESS:2PL ‘You ought to do an hour’s walking every day.’ (Lewis 1967:172) (2.22) hemen return-MN-1PL.POSS-LOC immediately work . start:FUT:1SG ‘Upon our return I will start work at once.’ (Underhill 1976:313) Which of the Turkish forms presented in this section should, from
< previous page
page_48
next page >
< previous page
page_49
next page >
Page 49 the point of view of their semantics, be considered as ANs (and why)? I suggest that we consider all of them as ANs in so far as they satisfy all the criteria for such nouns according to the definition in section 1.2.1. This is justified for Turkish in particular, which treats them in the same way even syntactically, i.e., all of these verbal forms show the same nominalization pattern (the Possessive-Accusative pattern, see chapter 6), independently of their meaning. An interesting fact is that Turkish ‘participles’, which code both propositions and facts, show oppositions between different tense forms, in contrast to verbal nouns. Thus, the former are in some respect closer to finite verb forms than the latter. Recall that a similar dependency between the meaning of a complement and the degree of its formal similarity to independent clauses exists in English (see section 1.5.1.). The Turkish examples show that one and the same language can have several types of ANs. Sometimes they follow the same nominalization pattern, sometimes they behave in different ways depending on their exact meaning. 2.4. LEXICAL NOMINALIZATION VS. CLAUSAL NOMINALIZATION In ANCs, by definition, the process of nominalization involves the head, causing a change of its categorial status as compared to independent clauses. Hence this type of nominalization is characterized as a type of lexical nominalization. This is opposed to clausal nominalization, the characteristic feature of which is the absence of any evidence in favour of viewing its head as a lexical noun. That is, the verb in such a clause typically has no nominal characteristics and often has such verbal characteristics as person and number, though it may be lacking in tense-aspect marking. (Comrie and Thompson 1985:392). Note that lexical and clausal nominalizations are not merely opposed as being products of complementderanking vs. complement-balancing strategies respectively. More specifically, complement-deranking does not necessarily create lexical nouns, as in the case of subjunctives or infinitives. Thus, clausal nominalization may involve both complement-balancing and complement-deranking. Nominalized clauses can, in turn, be more or less similar to noun phrases. Thus, in a number of languages, clauses are nominalized by
< previous page
page_49
next page >
< previous page
page_50
next page >
Page 50 attaching an article, as exemplified by Lakhota: (2.23a) Unglapi ‘We are going home.’ (2.23b) [Unglapi kin] iyonicip’ipi we.are.going.home the has.pleased.you ‘Our going home has pleased you.’ (Comrie and Thompson 1985:393, ex. 227) In nominalized clauses in Lakhota, finite verbs followed by the same article that is used with nouns retain all their head-marking affixes. Similar nominalization strategy is attested in Ancient (quoted by Comrie and Thompson, ibid.) and Modern Greek (Hans Ruge personal communication) and in Nama Hottentot, to mention but a few examples. In the latter language, clauses are nominalized by means of ‘person-gender-number’ suffixes -s (definite) or -ˆi (indefinite), which otherwise are obligatorily taken by ordinary nouns in the singular (Hagman 1977:126). A number of verb-final languages nominalize clauses by attaching nominal inflectional suffixes to finite verbs. In the resulting constructions, the verbs decline in the same way (or in a similar way) as non-derived nominals. However, there may still be some arguments against considering them as ANs, in particular, the presence of personal affixes. Examples (2.24a–2.24b) from Imbabura Quechua show an interesting opposition between action nominal constructions and clausal nominalizations in a single language (though the latter occur only in one subdialect). In ANCs (2.24a), the head has a nominalizing suffix instead of verbal subject suffixes, followed by the accusative marker. In clausal nominalizations (2.24b), the head retains the verbal subject affix, which is immediately followed by the accusative marker (Cole 1982:34). Note that the dependents in both types of nominalizations have the same form as in independent clauses (ANCs in Quechua follow the Sentential pattern described in chapter 5). (2.24a) Juan-ka [ñuka Marya-ta juya-j]-ta ya-n Juan-TOP I María-ACC love-AN-ACC think-3 ‘Juan thinks that I love María.’ (Cole 1982:34, ex. 107b) (2.24b) Pedro ya-n [ñuka Agatu-pi kawsa-ni-ta] Pedro think-3 I Agato-in live-1-ACC ‘Pedro thinks that I live in Agato.’ (ibid.: 34, ex. 108)
< previous page
page_50
next page >
< previous page
page_51
next page >
Page 51 Clausal nominalizations similar to those in Imbabura Quechua occur, for example, in Turkish (Lewis 1967:258), Ket, a Siberian isolate (Grišina 1979) and Diegueño, a Yuman language (Langdon 1970:177). So far, we have seen languages with rich systems of nominal and verbal morphology where the presence of the typical verbal features in the predicates of nominalized clauses can be used as a reason for considering them verbs and not ANs. However, in the case of isolating or predominantly analytic languages, this argument loses much of its weight. In languages with a low degree of morphological complexity, it may be difficult to determine whether the head has or has not retained its verbal characteristics. This results in difficulties in determining the exact type of nominalizations in such cases. Stassen (1985:80–2) quotes consecutive constructions (i.e., those expressing ‘a successive temporal ordering between events’) in Tibetan as an example of indeterminacy between predicate deranking and clause-embedding. Luckily, he has succeeded in finding some arguments in favour of the former process. For example, predicates in independent clauses can be morphologically marked for tense, which is impossible for predicates in anterior clauses (i.e., those referring to the earlier event in the chain). The same type of argumentation is used in the case of Korean nominalizations which can be classified as ANCs of the Sentential type, rather than clausal nominalizations. In the case of Tibeto-Burman languages, such as Lahu, Tibetan, Burmese (Matisoff 1972), I am inclined to take the opposite position. According to Matisoff, action nominalizations in Lahu involve adding the ‘indicative’ nominalizer ve after any clause no matter how long and complicated it is. This suggests that we should regard the whole process of nominalization as clausal. On the other hand, negation in such clauses can be expressed in two different ways, either as in independent clauses, or as in combination with ordinary NPs. This can probably be viewed as evidence for treating the head of nominalized clauses as a noun. Matisoff states that Tibeto-Burman languages on the whole tend to ‘nominalize sentences’ and quotes examples from Tibetan, Burmese and Jinghpaw. For Newari, another Tibeto-Burman language, Kölver (1977:5–8) writes that the suffix -gu nominalizes verbal phrases, clauses and sentences. Unfortunately, the examples in her description are too scarce for any generalizations to be made, but, presumably, Newari nominalizations are similar to those in Lahu. In cases of indeterminacy between clausal and lexical nominalization, I have tried to be restrictive and not to include dubious examples in my sample. It is worth mentioning that it is not always advisable
< previous page
page_51
next page >
< previous page
page_52
next page >
Page 52 to trust the terminology used in descriptions of languages. Thus, for example, for West Greenlandic, Fortescue (1984:44) states that the ‘nominalizing affix niq can convert a clause of any degree of complexity to an NP acting as object or subject of a superordinate verb’, and the same is stated for Kobon by Davies (1981:26), namely, that non-finite noun clauses represent ‘a structure derived from a clause or sentence by the attachment of the suffix -ep/-eb to the stem of the (last) verb’. However, in both cases I view such nominalizations as ANCs. The examples above suggest that there is probably no sharp boundary between clausal nominalizations and ANCs. Some languages have clausal nominalizations, some have lexical nominalizations, some have both types, and, finally, some do not distinguish between the two types.
< previous page
page_52
next page >
< previous page
page_53
next page >
Page 53 3 Noun phrases vs. finite clauses in different languages 3.0. INTRODUCTION The primary goal of this chapter is to present the basis for a classification of ANCs across languages and to give a general overview of the resulting types. However, the following two questions will also be discussed: 1 Do all languages have an opposition between finite clauses and possessive noun phrases? 2 Are there any natural ways of mapping the relations between a verb and its arguments on to possessive noun phrases, or in other words, is it justified to talk about some kind of isomorphism between finite clauses and noun phrases? Both questions are important for classifying and explaining nominalization patterns across languages. Thus, a classification which regards ANCs from the point of view of their relative closeness to the corresponding finite clauses or to non-derived possessive NPs presupposes that there is some difference between these two types of constructions. The exact nature of this formal opposition should influence the choice of possible classification parameters. In other words, the choice of the basis for a classification of nominalization patterns is dependent on the answer to the first question. The second question, on the other hand, has to do with a possible explanation of these different nominalization types. The structure of the discussion is as follows. First, some typological facts and theories relevant for the first question stated above will be presented. After that, possible parameters for a classification of nominalization patterns across languages and the resulting types will be discussed. Finally, the second question will be addressed.
< previous page
page_53
next page >
< previous page
page_54
next page >
Page 54 3.1. NOUN PHRASES VS. FINITE CLAUSES—A UNIVERSAL OPPOSITION? The question whether all languages differentiate between NPs and finite clauses has at least two sides. Firstly, since the most important constituents of the two types of constructions are nouns and verbs respectively, it is reasonable to ask to what degree these lexical categories are universal. Secondly, even if we assume that all languages differentiate between nouns and verbs, it is still possible that noun phrases and finite clauses do not exist as two structurally different constructions in some languages. Starting with the first problem, it seems that all languages make some kind of grammatical distinction between nouns and verbs, even if the details of this opposition can vary greatly. Thus, nouns and verbs can be distinguished on one or more of the following levels: morphology (i.e., with respect to the morphemes that can be included within them); immediate syntax (i.e., with respect to the units that can combine with them within phrases); and ‘prepositional syntax’ (i.e., with respect to their functions in clauses, as predicates or arguments) (these terms and their definitions are taken from Lazard 1984:30– 1). In a number of languages, the noun-verb opposition is very subtle. Thus, some analytic languages, such as the Polynesian languages or Chinese, possess a class of words that can function as the head of a noun phrase or a verb phrase depending on the other constituents of the same phrase (for example, depending on the type of particles, in the case of the Polynesian languages). Consider the following examples from Tahitian: (3.1a) ’ua ’amu te ’urī i te mā’a PRF eat DEF dog ACC DEF food ‘The dog ate.’ (3.1b) ’ua hina’aro te ’urī i te ’amu PRF want DEF dog ACC DEF way.of.eating a te mīmī POSS DEF cat ‘The dog wanted the cat’s way of eating.’ (=‘The dog wanted to imitate the cat’s way of eating.’) (Moyse-Faurie 1984:119) In these sentences, the same word ’amu combines with the verbal aspect particle ’ua—‘perfective’ and the nominal particle te —‘definite article’, constituting in this way the head of the VP and the NP respectively. In such cases, it is not obvious whether it is possible to talk
< previous page
page_54
next page >
< previous page
page_55
next page >
Page 55 about derivation of nouns from verbs, i.e., nominalization, or of verbs from nouns, i.e., verbalization (for example, see the discussion in Vonen 1988:80–4). However, nominalization as a separate process involving special derivational markers exists in most of the Polynesian languages (see section 7.2.2.).1 Thus, in Lazard’s terms (1984:37), the noun-verb opposition in these languages is somewhat blurred on the morphological level, but is manifested on both the level of ‘immediate syntax’ and on the clause level (NPs and VPs have different internal structure and have different functions within clauses). Another situation is found in certain Amerindian languages, such as Nootkan, and Austronesian languages, such as Tagalog and Malagasy, where many semantically noun-like and verb-like stems appear in exactly the same functions, both as arguments and predicates (see ex. 3.2a–b below). Tagalog (3.2a) Nagtatrabaho ang lalaki is.working TOP man ‘The man is working.’ (3.2b) Lalaki ang nagtatrabaho man TOP is.working ‘The one who is working is a man.’ (Schachter 1985:12, ex. 3.21–3.22) However, it has been shown that there are functional and distributional differences between the two word classes in these morphologically rich languages, even if the classes have many similarities (Schachter 1985, Hopper and Thompson 1984). Thus, it is mostly on the clause level that the opposition is neutralized (it is also blurred on the level of ‘immediate syntax’). On the morphological level, there is a pronounced distinction between nouns and verbs. Now, nominalization, as a process by which a predication is converted into a noun phrase, presupposes functional distinctions between noun phrases and verb phrases. In languages where one and the same verb form regularly appears as both a predicate and an argument, is it possible to talk about nominalization, and if so, in which cases? There is no consensus on this matter in language descriptions. Thus, concerning Tagalog, Schachter and Otanes (1972:62) write that there is no context in which a noun cannot be replaced by a verb or a verb phrase, but they do not call verbs in these functions nominalizations. Nominalization, however, exists in Tagalog as separate processes which result in nominalized basic forms and gerunds
< previous page
page_55
next page >
< previous page
page_56
next page >
Page 56 (ibid.: 154–64). Harrison (1976:281) in his grammar of Mokilese (Micronesia) writes the following: ‘Any intransitive verb phrase (intransitive verb, incorporated object construction, intransitive verb and following noun phrase) may be used as a noun to name an event. Such nominalizations may appear as sentence subjects, in possessive constructions, and so forth.’ In the examples below note the same verb form loakjid used either as a predicate or as an argument: (3.3a) Joamoaio loakjid rehnnoawe. my:father fish today ‘My father fished today.’ (Harrison 1976:153, ex. 76a) (3.3b) Loakjid inenin kaperen fishing ‘Fishing is a lot of fun.’ (ibid.: 281, ex. 2) Obviously, such differences in the interpretation of relatively similar cases might influence the selection of data used in this study. However, it would be an impossible task to check what exactly constitutes the noun-verb opposition in every language. For this reason, I prefer a more or less unified and, perhaps, somewhat restricting approach: if a verb form which normally occurs as a predicate, can be used as an argument without any changes at all, it is not considered as nominalized. Let us turn to the second problem, which should be understood as follows: do finite clauses and noun phrases have different internal structures in all languages? That is, do all languages have distinct ways of expressing the relations between the head and its dependents in finite clauses, as opposed to NPs? The comparison between the internal structure of ANCs with that of the corresponding finite clauses and non-derived NPs, as formulated in chapter 1, rests, to some degree, on the assumption that nonderived NPs differ structurally from finite clauses and that ANCs, consequently, can be closer to the former or to the latter. Otherwise the comparison is not reasonable. In other words, if finite clauses and NPs exploit exactly the same syntactic means, it does not make sense to oppose ANCs to one of them. The statement that NPs and finite clauses in a language have different structures does not mean that the head-dependent relations in any NP and in any clause should be signalled in totally different ways. Thus, it is easy to find cases for which this is not true, especially
< previous page
page_56
next page >
< previous page
page_57
next page >
Page 57 in the absence of any overt marking on the constituents. Let us take Korean, an SOV/GN language. Head-marking is non-existent and both subjects in clauses and characterizing attributes within NPs can be in the nominative. Thus, the clause ‘Fish swim’ and the noun phrase ‘a fish tail’ will have exactly the same structure. However, this is by no means the general situation in Korean, where noun phrases and clauses manifest a large number of structural differences (see section 5.1.). Let us start with the hypothesis that NPs and clauses exist as two structurally different units in all languages. What kind of data could constitute a counter-example? Eskimo languages with their rich morphology provide us with a well-known example of languages with striking similarities between NPs and clauses, these similarities being manifested both in the realm of word order, dependent-marking and head-marking. Examples (3.4–3.5) from Central Alaskan Yupik (CAY) illustrate this parallelism. noun phrase (3.4a) Nuka-m qetunra-a Nukaq-REL son-3SG:SG ‘Nukaq’s son’ finite clause (3.4b) Nuka-m ikayur-a-a Nukaq-REL help-INDIC-3SG:3SG ‘Nukaq helped him.’ (Woodbury 1985:63, ex. 1) noun phrase (3.5a) angya-t muragarr-lainaa-t boat-PL wood-only-PL ‘wooden boats’ finite clause (3.5b) Angya-t kit-u-t boat-PL sink-INDIC-3PL ‘The boats sink.’ (ibid.: 64, ex. 3) In (3.4a), the relative case marker -m indicates that Nukaq is the possessor-NP, while in (3.4b), the same marker indicates that Nukaq is the agent of the clause (which follows the ergative-absolutive
< previous page
page_57
next page >
< previous page
page_58
next page >
Page 58 pattern). The heads of both constructions appear with the same marker -a, which, in the case of the NP, agrees with the possessor in person and number (the third person singular) and shows the singularity of the possessum, while in the case of the clause, it registers the person and number of the A (the third person singular) and of the P (also the third person singular). In (3.5a), number is marked on both constituents of the NP, and in (3.5b) the same number suffixes mark both the intransitive subject (S) and the intransitive verb. This kind of similarities has given rise to the idea of ‘lumping’ noun phrases and clauses, i.e. treating NPs and clauses as a single category. According to different hypotheses, clauses are (sets of) NPs, or possessive NPs are clauses, or both clauses and NPs are derived from the same type of underlying structure (for the survey of the different theories see Woodbury 1985, Walter 1981 and Kalmár 1979). Now, nouns and verbs, despite the extensive parallelism, are distinct word classes in Eskimo, or rather, stems are rigidly divided into verb bases and noun bases. Special derivational suffixes serve to derive nominal bases from verbal ones, or vice versa. Still, it may be suggested that the internal structure of NPs is the same as that of finite clauses, in other words, that the head-dependent relations in these two types of constructions do not differ. Woodbury’s point is that despite the obvious similarities, NPs and finite clauses are not exactly alike and that there are distinct intermediate construction types in between them. Thus, in the absence of clear counter-examples, the above formulated hypothesis that noun phrases and finite clauses are structurally distinct in all languages, is assumed to be true. 3.2. DISCUSSION OF THE BASIS FOR A CLASSIFICATION OF ANCs ACROSS LANGUAGES If we want to classify ANCs with respect to their sentence-likeness or NP-likeness, we should do this on the basis of some parameters which have different values for noun phrases and finite clauses in the majority of languages, if not all. Apart from such exceptional cases as the Eskimo languages, noun phrases and finite clauses have different structures in most languages and can be distinguished using some or all of their syntactic means (dependent-marking, head-marking or word order). The three syntactic means are relatively independent of each other, and a combination of all of them as parameters would lead to an extremely detailed classification. For a relatively small language sample, such as the one used in this study, this might be especially dangerous, since the resulting groups would
< previous page
page_58
next page >
< previous page
page_59
next page >
Page 59 be too small for possible typological generalizations. Of the three syntactic means, dependent-marking is chosen as the main parameter for a classification of ANCs. In some cases, when this parameter alone does not offer sufficient information, head-marking is used as an additional parameter. Let us consider the reasons for this choice, taking word order, head-marking and dependent-marking one by one and considering the advantages and disadvantages associated with each of the three syntactic means as a possible parameter for the classification. 3.2.1. Word order In many languages, there is no word order difference between clauses and noun phrases, since the head-dependent word order is the same in both finite clauses and NPs. For these languages, it would be impossible to say on the basis of word order alone whether ANCs are more sentence-like or more NPlike. Such is the situation in V-1/NG and SOV/GN languages, which together constitute the absolute majority in Hawkins’ Expanded Sample (Hawkins 1983:288). For the languages with different headdependent word order, my sample shows an interesting generalization, namely, that ANCs always follow the nominal pattern. This fact does not, of course, lead to any taxonomy either. Thus, word order could hardly serve as the basis for a classification of ANCs. 3.2.2. Head-marking Since many languages in the sample do not have head-marking patterns at all, either within finite clauses or NPs, problems with head-marking as a parameter are similar to those with word order. For the languages with different head-marking patterns in finite clauses and NPs, the sample shows that ANCs, with some exceptions, tend to be similar to non-derived NPs in this respect. This issue will be discussed in more detail in sections 11.5. and 12.2. Thus, head-marking alone could hardly serve as the basis for a classification of action nominal constructions. However, as will be shown below, it may be used as an additional parameter in some cases. 3.2.3. Dependent-marking It is with respect to dependent-marking that most languages in the sample manifest the greatest distinction between noun phrases and finite clauses. In most cases, the form of adnominal dependents is
< previous page
page_59
next page >
< previous page
page_60
next page >
Page 60 distinct from that of the S, P and A in finite clauses, which makes it relatively easy to qualify the arguments in ANCs as sentential or nominal dependents. In a few cases, such as Egyptian Arabic (see section 6.2.) or Abkhaz (see section 7.1.), which almost exclusively use head-marking and in which the possessor-nominal, A, S and P appear in the same non-marked form, it can be difficult to base classification on dependent-marking alone. With respect to head-marking, however, the S, A and P in finite clauses are opposed both to each other and to the possessor-nominal, so head-marking can serve as an additional criterion for the classification of ANCs in these languages. 3.3. THE TYPOLOGY OF ANCs From the point of view of logical possibilities, we can expect to find the four major types of ANCs given below (‘sentential’ as opposed to ‘nominal’ syntactic means is to be understood as signalling syntactic relations in finite clauses as opposed to those in NPs): 1 the relations between an AN and all of its arguments are signalled by the sentential syntactic means; 2 the relations between an AN and all of its arguments are signalled by the nominal syntactic means; 3 the relations between an AN and some of its arguments are signalled by the sentential syntactic means, while its relations with the other arguments are signalled by the nominal syntactic means; 4 the relations between an AN and some or all of its arguments are signalled by special means, used in neither finite clauses nor in NPs. It turns out that this last logical possibility is not realized across languages; that is, ANCs, in a sense, do not exist as autonomous, independent constructions, but always ‘borrow’ syntactic means from finite clauses and/or NPs. This corroborates Comrie’s (1976:200) conclusion that ANCs in different languages can be classified ‘according to the extent to which their internal structure corresponds to that of a nonderived noun phrase, rather than to the internal structure of a sentence with a finite verb’. However, while Comrie suggests that ‘this classification principle leads ultimately to a continuum, rather than a dichotomy between noun phrase-like and sentence-like action nominals’, the present discussion shows that the description ‘continuum’ is hardly appropriate in this connection. Though ANCs can be placed on a scale according to the extent of their ‘sentence-likeness’ and ‘NP-likeness’, this scale consists of discrete points, corresponding to well-defined types of ANCs. Dynamic metaphors will
< previous page
page_60
next page >
< previous page
page_61
next page >
Page 61 be used to define nominalization patterns, although the question of ANC derivation will be left open. Thus, dependents retain sentential marking, but genitivize (=assimilate to the possessor-nominal of non-derived NPs). The resulting classification consists of the following nominalization types:2 1 Sentential type (SENT)—all the arguments in ANCs retain their sentential dependent-marking; 2 Possessive-Accusative type (POSS-ACC)—the S and the A in ANCs genitivize, while the P retains the sentential dependent-marking; 3 Ergative-Possessive type (ERG-POSS)—the P and the S in ANCs genitivize, while the A gets another dependent-marking. Within this type it is, perhaps, reasonable to distinguish between (a) the Sentential-Possessive subtype (SENT-POSS), in which the A retains its sentential marking, and (b) the Oblique-Possessive subtype (OBL-POSS), in which the A turns up as an oblique NP; 4 Nominal type (NOMN)—all the arguments in ANCs assimilate to dependents in non-derived NPs. Here it is also possible to distinguish between (a) the Double-Possessive subtype (DBL-POSS), in which both the S, P and A genitivize, and (b) the Possessive-Adnominal subtype (POSS-ADN), in which both the S and the A genitivize, while the P gets the same dependent-marking as some oblique NPs; 5 Mixed type (MIX)—the S genitivizes, the A turns up as an oblique NP (perhaps in the same form as the Agent in passive clauses), while the P retains its sentential dependent-marking; 6 Incorporating type (INC)—the P forms a part of the complex AN, the S retains its sentential dependent-marking or genitivizes, while the A is either similar to the S or turns up as an oblique NP; 7 Relative type (REL)—the S and P genitivize or, at least, appear as adnominal dependents, while the A is expressed within the relative clause referring to the AN. Note that there is a crucial difference between the two types of oblique NPs mentioned in the definition of the OBL-POSS (3b) and POSS-ADN (4b) above. The former are normally restricted to clauses, in which they are used for referring to ‘chômeurs’, or ‘retired terms’, in the terminology of Relational Grammar, that is, to nominals which can be ascribed the role of subjects or objects at a certain level of analysis, but not in the output structure. These nominals can hardly appear at all within non-derived NPs. The latter type of oblique NPs
< previous page
page_61
next page >
< previous page
page_62
next page >
Page 62 are not restricted in such a way and can freely serve as dependents within non-derived NPs. The INC and REL types are ‘valency-lowering’ as compared to the other nominalization patterns, since even ANs derived from transitive verbs have only one external argument. However, in the case of transitive ANs, both the A and the P may be expressed at the same time, although one of them either builds a part of the compound AN or constitutes a part of the relative clause referring to the AN. Chukchee seems to offer an extreme example of the same tendency, since its transitive ANs never combine with both the A and the P at the same time. It might be appropriate to attribute ANCs in such languages to a special type, the Argument-Reducing type (ARG-RED). However, my information on Chukchee and other candidates for this type is currently insufficient to support such a claim. Within the POSS-ACC and the ERG-POSS types, it is possible to distinguish between pure and derived ANCs (including anti-passive and passive ANCs respectively), where derived ANCs involve ANs with overt voice markers. In derived constructions, the action nominals are correlated with verbs detransitivized by means of passivization or anti-passivization. Thus, the only effect of nominalization consists in assimilating the ‘subject’ of such intransitive verbs to the possessor-nominal in non-derived NPs. Strictly speaking, since no transitive verbs are involved in the derivation of such constructions, they are outside the general classification. Both the ‘valency-lowering’ types INC, REL and ARG-RED, and the derived types are manifestations of one and the same tendency to reduce the valency of ANs as compared to the corresponding transitive verbs. On the basis of ANCs involving overt anti-passive and passive markers, Mackenzie (1985) has suggested that valency reduction, which is given regular morphological expression in such ANCs, always underlies the process of nominalization. The defining features of the types are summarized in Table 3.1. The columns headed ‘S’, ‘A’ and ‘P’ show whether these arguments in ANCs have retained their sentential dependent-marking (FC, the abbreviation for ‘finite clause’), have genitivized (G), are obligatorily deleted or have the same dependent-marking as oblique nominals either restricted to clauses (OBLFC) or compatible with nonderived nouns (OBLNP). Where the P builds a compound with an AN, it is symbolized by ‘p-’. In chapters 5–9 (and partly chapter 10) the nominalization patterns defined above will be described in great detail. Note that a language may have several nominalization patterns and the resulting
< previous page
page_62
next page >
< previous page
next page >
page_63
Page 63 Table 3.1 Syntactic types of ANCs (nominalization patterns) in the sample Type of ANC S A P 1 Sentential (SENT) FC FC FC 2 Possessive-Accusative (POSS-ACC) Pure G G FC Anti-passive 3 Ergative-Possessive (ERG-POSS) Pure: a Sentential-Possessive (SENT-POSS) G FC G b Oblique-Possessive (OBL-POSS) G OBLFC G Passive 4 Nominal (NOMN) a Double-Possessive (DBL-POSS) G G G b Possessive-Adnominal (POSS-ADN) G G OBLNP 5 Mixed (MIX) G OBLFC FC 6 Incorporating (INC) a Sentential-Incorporating (SENT-INC) FC FC pb Possessive-Incorporating (POSS-INC) G G pc Oblique-Incorporating (OBL-INC) G OBLFC p7 Relative (REL) G REL G? ?8 Argument-Reducing (ARG-RED) G
Number of languages 15+?2 24+?1+2* 1 3 22 6 8+?3+7* 3 2+?2+1* 2 8 1 2 ?1
Note: In the last column, ? refers to those languages in which the nominalization pattern is only marginal or questionable, * refers to patterns restricted to highly referential or pronominal arguments.
< previous page
page_63
next page >
< previous page
page_64
next page >
Page 64 classification involves types of ANCs rather than the languages in which they occur. Although in a number of cases, one of the ANC types existing in a language is central or basic, while all the others are more or less marginal, this is by no means usual. Still, such qualifications as ‘SENT languages’, which should be read as ‘languages which have ANCs of the SENT type (and, perhaps, of other types)’ will often be found. Chapters 5–9 are structured in the following way. Firstly, each type is illustrated by means of one or several ‘ideal’ representatives, for which a detailed description of syntactic means used in finite clauses, non-derived NPs and ANCs is given with relevant examples, and some other features of ANCs are mentioned. Then, those languages, which are classified as having ANCs of the type under consideration, but in which the exact affiliation of such constructions is connected with some problems, are discussed. In most cases, these problems come from a weak differentiation of the syntactic means used in finite clauses and in non-derived NPs. Next comes head-marking used in ANCs of the type in question. After that, typological, genetic and areal characteristics of the languages with such constructions, as well as other sentential/verbal and nominal features of these ANCs are discussed. Finally, the attested cases of intralexemic syntactic variation are mentioned, that is, those cases where the head of ANCs of the type under discussion can combine with its dependents according to other patterns (at least in some languages). For each nominalization pattern the data are summarized in a table. Word order in nonderived NPs will be discussed in G-N terms, in contrast to the otherwise consistent use of the abbreviations ‘G’ and ‘H’ introduced in this chapter (section 3.5.1.). It will be useful here to state the following generalizations concerning ANCs across languages: 1 In all languages, ANCs involve marking (both dependent-marking and head-marking) used in either finite clauses or in possessive NPs. 2 Of the two arguments in transitive ANCs, the one most likely to lose its sentential dependent-marking and head-marking is the A (except for the marginal SENT-POSS type). 3 Either all the arguments in a transitive ANC retain their sentential dependent-marking or at least one of them genitivizes (except for the marginal MIX type). 4 Head-dependent word order is the same in ANCs and non-derived NPs (with the possible exception of some Kwa languages (see section 9.2.)).
< previous page
page_64
next page >
< previous page
page_65
next page >
Page 65 5 If an ANC retains its sentential head-marking for some argument, it also retains its sentential dependent-marking for the same argument (except for one type of intransitive ANCs in Tabassaran). These generalizations are discussed in chapter 12. 3.4. ISOMORPHISM BETWEEN NPs AND FINITE CLAUSES: FORMULATING THE PROBLEM As we have seen in chapter 2, nominalization involves different processes by which finite clauses are turned into NPs (at least, with respect to their functions). In other words, nominalization creates structurally different types of noun clauses, ranging from those which are practically identical to the corresponding finite clauses to those which have totally assimilated to non-derived NPs. The closer a noun clause is to a non-derived NP, the more dramatic are the changes in its structure as compared to the corresponding finite clause. In this connection, it is reasonable to ask whether there are any natural ways of mapping finite clauses onto non-derived NPs. In other words, is it justified to talk about total or partial isomorphism between NPs and finite clauses, even in languages where these two constructions manifest many dissimilarities? If this question receives a positive answer, we can explain the existence of at least some of the nominalization patterns across languages. A negative answer leads to the conclusion that these patterns are totally accidental. The problem of parallelism between finite clauses and NPs has received much attention within Generative Grammar during the different stages of its development. Analyses of English NPs have distinguished several types of possessive constructions (in the wide sense of the term ‘possessive’), for example, those with prenominal and postnominal modifiers, or those with alienable and inalienable possession. The X-bar approach, as it was formulated in the 1970s by Chomsky (1970), Jackendoff (1977) and others, suggests cross-category generalizations grounded on the syntactic parallels between the different categories. Following this line, Giorgi and Longobardi (1988) propose a typology of noun phrases grounded on the distinction between head-subject and head-complement combinations within them. An analysis of noun phrases as determiner phrases (DP), elaborated in Abney (1986) and a number of other publications, achieves an almost complete parallelism between the structure of noun phrases and clauses. However, as shown in Haspelmath (1990) and Payne (1991), typological facts argue against at least the crucial claim
< previous page
page_65
next page >
< previous page
page_66
next page >
Page 66 of the DP analysis, namely, that the noun phrase is headed by a functional category DET (determiner). Let us look at two totally different approaches to the problem of parallelism between finite clauses and noun phrases. Keenan (1974) observes several similarities between possessors and subjects, mainly with respect to pronominalization and quantification. More specifically, subjects of clauses cannot be pronominalized by any NPs in the predicate (even if the latter precede them), while possessor-NPs cannot be pronominalized by the head of the larger NP (even if the latter precedes and dominates them). This is seen in the following examples from English (3.6) and Malagasy (3.7). (3.6a) Johni hit himselfi. *Hei/he-selfi hit Johni (3.6b) *hisi lawyeri (3.6c) Johni’s picture/analysis of himselfi *hisi picture/analysis of Johni (3.7a) Namono tenai Rabei killed self Rabe *Namono an-dRabei izyi/ny tena-nyi. killed Rabe he/the self-his ‘Rabe killed himself.’ (3.7b) *ny namani’-nyi the friendi (of)-hisi Secondly, quantified subjects of clauses can have wider scope than any NPs within the predicate, while universal quantifiers in a possessor position can have wider scope than NPs in the head of the larger (possessive) NP. Thus, in the sentence Every girl kissed some boy, every girl has wider scope than some boy, which means that the choice of the subject referent does not depend on anything else in the sentence, while that of the object referent can depend on the referent of the subject. In the example a representative of every country attended the meeting, the referent of representative in the preferable reading depends on the referent of every country . Keenan suggests that these similar properties of subjects and possessors follow from their logical similarity. More specifically, both types of NPs constitute the argument expression of the corresponding functional expression, namely, simplex clauses and possessive NPs respectively. To determine the truth of a functional expression, or to evaluate it, we must first of all identify the referent of its argument
< previous page
page_66
next page >
< previous page
page_67
next page >
Page 67 expression and the function which applies to it. However, the reference of the argument is not dependent on the truth of the function which applies to it. Thus, ‘to evaluate the truth of a simple sentence we must mentally identify the referent of the subject and then determine whether the predicate holds of it or not’ (Keenan 1974:299). In possessive constructions (especially inalienable and ‘abstract’ possessive constructions, like the inside of the bottle ), the possessor (genitive) NP is the argument expression and the head NP represents the function… To evaluate e.g. the inside of the bottle we must identify the referent of the bottle and then determine its inside. We do not in any sense first conceive of the set of all ‘insides’ and then pick the one which the bottle in question just happens to have. (Keenan 1974:299) While Keenan considers only similarities between the subject of finite clauses and the possessor without mentioning objects, Seiler (1983) has investigated the interrelation between possessive phrases and both verb-subject and verb-object structures. He starts with their morphological and syntactic similarities and then proposes a semantic explanation for them. The morphological facts include ‘affinities or even identities in form and meaning between pronominal elements in a possessive function on the one hand, and personal pronominal elements in subject and/or in object function on the other hand’ (Seiler 1983:91). The North American Indian languages considered by Seiler are categorized into five types, in which possessive elements are similar either exclusively to subject elements, exclusively to object elements, or partly to subject and partly to object elements. In those languages which distinguish between inalienable and alienable possession, pronominal object elements tend to be similar to markers of inalienable possession (however, in Tunica, a Gulf language, there is an exact correlation between alienable and inalienable possessives, on the one hand, and the object markers of dynamic verbs and the subject markers of static verbs, on the other hand). The data from Zuni illustrate what is meant by similarities between possessive and object pronominal elements.
< previous page
page_67
next page >
< previous page
page_68
next page >
Page 68 Pronominal elements in Zuni (after Seiler 1983:94) SUBJ OBJ DAT POSS SG1 ho’o hom homan hom 2 t’o’o t’om t’oman t’om 3 —— —— an an DU1 hon ho’na ho’nan ho’na 2 t’on t’o’na t’o’nan t’o’na 3 a·tci a·tcia a·tcinaiyan a·tcia PL1 hon ho’na ho’na·wan ho’na·wan 2 t’on t’o’na t’o’na·wan t’o’na·wan 3 —— —— a·wan a·wan The syntactic facts discussed by Seiler include possessor-to-object and possessor-to-subject promotion and subject/object interpretation in nominalizations. The former is illustrated by the following examples from Haya (a Bantu language). Possessor-to-object promotion (3.8) ŋ-ka-hénd’ ómwáán’ ómukôno I-P3-break child arm ‘I broke the child’s arm.’ (lit. ‘I broke the child the arm.’) (Seiler 1983:109, ex. 28) Possessor-to-subject promotion (3.9) omwáána n-aa-núúk’ ómukôno child PR-he-smell arm ‘The hand of the child smells.’ (lit. ‘The child smells (with respect to) the hand.’) (ibid.: 110, ex. 36) For describing and explaining the above-mentioned morphological and syntactic facts, Seiler operates with several semantic-conceptual categories (represented in capital letters), among other things, POSSESSION, which may be either inherent or established. According to Seiler, both inherent and established POSSESSION are co-present in every possessive structure, though in different proportions. Established POSSESSION has to be established by explicit means and calls for a POSSESSOR-AGENT that does the acquiring. Inherent, intimate POSSESSION does not need to be acquired. Rather, the POSSESSOR appears as being qualified by his POSSESSUM,
< previous page
page_68
next page >
< previous page
page_69
next page >
Page 69 or as being EXPERIENCER when something happens to his POSSESSUM. In any case, he is INACTIVUS. (Seiler 1983:113) Now, if we try to correlate some of these semantic-conceptual categories with morphosyntactic categories operating within verb structures, we see that the ‘object represents the GOAL or EXPERIENCER or INACTIVUS roles. The subject of transitive verbs may stand for AGENT or EXPERIENCER, the subject of intransitive verbs for EXPERIENCER and INACTIVUS’ (Seiler 1983:114). Thus, the affinities between possessive and objective are connected to inherent POSSESSION, which explains, among other things, why the objective pronominal elements are similar to the ‘inalienable’ possessive elements in the languages with two sets of possessive elements. The affinities between possessive and subjective have establishing POSSESSION as their basis. To conclude, Keenan and Seiler approach the problem of affinities between possessive NPs and clauses (or verb structures, in Seiler’s case) from different points of view. Keenan states the logical similarity between the two in the sense that they are both functional expressions, whose reference crucially depends on the reference of their argument expressions (the possessor and the subject respectively). Seiler claims that the different ways by which possession is established (i.e., whether it is inherent or acquired) determine the different roles of the possessor in the situation of possession and, consequently, its interpretation as being more or less active and, correspondingly, more subject-like or object-like.3 3.5. AFFINITIES BETWEEN NOUN PHRASES AND FINITE CLAUSES: MORE EXAMPLES 3.5.1. Possessivity and referentiality In the preceding section Keenan’s idea about possessive NPs as functional expressions in which the reference depends on that of the possessor-nominal (the argument expression) within them was presented. In fact, many languages demonstrate formal connections of this kind. Two examples of such connections will be given below. The abbreviations ‘G’ and ‘H’ refer henceforth to the possessor-phrase and the possessum-phrase of NPs.
< previous page
page_69
next page >
< previous page
page_70
next page >
Page 70 3.5.1.1. Referentiality of possessors with genitive marking In some languages, for example, the Altaic languages, adnominal modifiers in possessive NPs may choose between the genitive and the non-marked form depending on the referential features of the nominal in question. For example, in Turkish, the alternation between the genitive case and the nonmarked form within NPs is used for expressing the difference between more and less referential attributes. The genitive marking in the following examples is chosen when the possessor-noun is characterized by the feature ‘referential’ (Nilsson 1985:69). (3.10a) bir a fish-GEN tail-3SG.POSS ‘the tail of a fish’ (3.10b) bir balık a fish tail-3SG.POSS ‘a fish tail’ Although in both examples above we describe the object as being a tail of some fish, it is hardly possible to talk about real possessivity in the latter case, which rather gives a categorial characterization to this object. In Turkish, it is the genitive which is used specifically for the marking of the G in possessive phrases, while the non-marked form has a much wider usage, both within NPs and sentences. Thus, NPs with genitive attributes in Turkish constitute a good example for Keenan’s idea. In such NPs, it is necessary to identify the possessor in order to understand the reference of the whole possessive NP. In NPs with non-marked attributes, such initial identification is not needed. It should be noted here that languages which show alternation between a special possessive form and a non-marked form for adnominal attributes may differ considerably in the exact details of such an alternation. Thus, while proper nouns in Korean are always put in the genitive case when they appear as possessors (due to their high degree of referentiality), the corresponding possessor nominals in Selkup tend to appear in the non-marked form (due to the economy principle, since it is clear that they should be interpreted as real possessors in the majority of situations) (Kuznecova et al. 1980:378). I shall return to the question of Gs as highly referential NPs in a discussion of possessive pronouns (section 3.5.3.).
< previous page
page_70
next page >
< previous page
page_71
next page >
Page 71 3.5.1.2. Possessors as determiners A number of languages in which nouns normally show a formal opposition in definiteness and so on, neutralize this opposition when a noun appears as the possessum (H) within possessive NPs. S-genitives in English constitute a well-known example. Another well-known example is found in the Semitic languages, in which the H appears in a special form (called the Construct State) and cannot combine with articles. In these cases, it is possible to consider Gs as a sort of determiner within the whole possessive NPs, since they are incompatible with other determiners (e.g., articles) and have a similar semantic impact on the reference of the whole NP as the latter (see also Lyons 1985, 1986 on the distinction between ‘determiner-genitive’ and ‘adjectival-genitive’ languages). In languages without articles, it can be difficult or even meaningless to answer the question whether the Gs might be viewed as determiners in possessive phrases. Even in languages which consistently signal the opposition in definiteness, Gs need not always be determiners. Thus, genitives in German, de-nominals in French and of -genitives in English do not limit possible combinations of the H with the definite and indefinite articles, e.g., un livre de la fille—‘a book of the girl’ vs. le livre de la fille—‘the book of the girl’. Sometimes the exact relation between Gs and other types of determiners within a language can be intricate. Thus, in Swedish, demonstrative pronouns such as dessa—‘these’ are normally incompatible with the definite article de—‘this:PL’. Prenominal genitives, on the other hand, are incompatible with articles,4 but can, in certain situations, be preceded by demonstrative pronouns, which leads to such phrases as dessa pojkens böcker—‘these books of the boy’. Of the two connections between referentiality and possessivity, the former is directly related to Keenan’s claim, while the exact status of the latter is far from obvious. I feel that Gs included in the class of determiners are more subject-like than Gs of the non-determiner type, but at the present moment I lack any satisfactory explanation for this intuition. 3.5.2. Morphological affinities between possessive phrases and verb-subject vs. verb-object structures Morphological affinities between possessive phrases and verb-subject or verb-object structures are found not only within head-marking (see section 3.4. for Seiler’s observations), but also in the domain of
< previous page
page_71
next page >
< previous page
page_72
next page >
Page 72 dependent-marking, since a single form can function as both the possessor and as the subject or object. If a language signals possession within NPs by means of dependent-marking, such dependent-marked Gs will be called ‘genitives’ (even where the marking is done by means of adpositions).5 Genitives across languages either are only adnominal or can have a wider domain of application, being, thus, heterogeneous with respect to their structural environments. English s-genitives or Turkish genitives are only used within NPs as adnominal modifiers. Genitives which are not exclusively adnominal may have either subjective, or objective functions, or combine both of them. Several examples below will illustrate this cross-linguistic variation. In the Eskimo languages (e.g., in Central Alaskan Yupik Eskimo, see section 3.1.), a single case, the relative case, is used for both marking the Gs within possessive NPs, and for marking the agents (or transitive subjects) in finite clauses with the ergative pattern. In Kabardian (a Northwest Caucasian language), there is a system of two cases, one of which (‘ergative’) is used for both the transitive subject (the agent of ergative sentences) and for the G, and the same two functions characterize the genitive case in Lakk (a Northeast Caucasian language), which has ten ‘grammatical’ cases (Allen 1964:340). It is even more common for genitives to appear as objects of finite verbs. This is trivially true for nominative-accusative languages with only two distinct case forms for their nominals—the direct case, used for the subjects of finite clauses, and the oblique case, used otherwise. More interesting, though, is the use of the same form for the G in possessive NPs and the object in finite clauses by languages which distinguish among several cases. Usually, this form is not the only one reserved for marking objects, but alternates with some other(s), which are probably more often used in objective function. In German, for example, the object usually appears in the accusative case. However, some verbs are subcategorized for object NPs in the genitive case, the same case which marks Gs. Russian (and other Slavonic languages) presents an interesting example of the morphological affinities between the G in possessive NPs and the subject vs. object of finite clauses. Traditional grammar states that direct objects are usually put in the accusative case. However, this case as distinct from some other cases, exists for a restricted number of nouns, namely in the singular for nouns with the ending -a (mainly feminine). For all the other nouns, the accusative coincides either with the nominative (in the case of inanimate nouns) or with the genitive (in the case of animate nouns), feminine nouns manifesting the opposition in animacy only in the plural. Nouns that
< previous page
page_72
next page >
< previous page
page_73
next page >
Page 73 distinguish between the genitive and the accusative can alternate between them in negated sentences and also in combination with certain verbs. In other words, the choice of the genitive objective marking is governed by both lexical and syntactic conditions and serves as a signal of low transitivity of the clause (Hopper and Thompson 1980). The genitive case is sometimes used for marking the subjects of finite clauses, but only those subjects ‘which are lowest on the subject property hierarchies’ (Dahl 1980), that is, subjects of existential verbs under certain conditions. Thus, the genitive case in Russian is involved in both the marking of Gs and of objects and subjects, though the latter should perhaps be considered as ‘demoted unspecified complements of the verb’ (ibid.). Finnish and Estonian are quite similar to Slavonic languages in this respect. The distinction between determiner-genitives and non-determiner-genitives and the distinction between purely adnominal genitives and those with a wider range of functions are independent of each other. Thus, the postnominal genitive in German, as is clear from the description above, marks nondeterminer-genitives and objects of certain verbs. The genitive in Icelandic, which is also used outside NPs for marking the objects of some verbs, is of the determiner type. 3.5.3. Word order in possessive NPs and in finite clauses With respect to word order, certain dependents within NPs can be more similar to either the subject or the object of clauses. This concerns both the head-dependent word order in these constructions, as well as the mutual word order relations among the dependents. Such similarities are especially revealing in SVO languages, where the subject and the object are on the different sides of the finite verb. Thus, the two genitives in English, the prenominal s-genitive and the postnominal of -genitive, can be correlated with the subject and the object of finite clauses respectively. Although in many situations these two forms are interchangeable, they manifest a number of differences. Thus, only s-genitives are determiners in possessive phrases and tend to be animate, referential and definite, while these restrictions are not valid for of -genitives, which can have a partitive meaning and thus function as complements to nouns (and have the same status within NPs as objects within clauses, see Jackendoff 1977:106). English is a rare type of language with two structurally different genitives, both of which are lexically almost not restricted. Much more frequent are languages, especially of the VO types, which have two types of possessive constructions with a very uneven distribution, to
< previous page
page_73
next page >
< previous page
page_74
next page >
Page 74 a great degree depending on the semantics of the G. One of these two possessive NPs, with the order NG, is used with the majority of nouns in the G function. The other one, with the order GN, involves pronominal Gs (Ultan 1978:24). Let us again turn to Russian (representative for a number of Slavonic languages) to illustrate the dependence between the word order in possessive constructions and the similarities between these constructions and verb-subject or verb-object structures. In Russian, the genitive is the usual case for postposed adnominal dependents, including Gs. However, there are a group of words which distinguish between a special possessive form and a genitive case which is used to mark object NPs (see 3.5.2. above). These words include personal pronouns and some proper names including the names for near relatives, such as mother or father.6 Possessive forms are restricted to prenominal dependents. As Russian is a SVO language, its postnominal genitives resemble objects, appearing after verbs, and its possessive adjectives resemble subjects, appearing before verbs. The formal and functional relations between the possessive and the genitive forms are not the same for all the words that have both. Table 3.2 summarizes the differences and the similarities between these forms for the different groups of words. Thus, it seems reasonable to talk about some sort of hierarchy of the relations between possessorphrases (or typical adnominal dependents) and genitive phrases as verbal objects such as that given below: 1,2 >3>PROPER NAMES>OTHER NOUNS Note that this is one of the variants of the animacy hierarchy (Silverstein 1976, see also Corbett 1987:326 on the use of possessive forms in Slavonic). All the words which have a special possessive form are highly referential and, thus, subject-like in this respect (recall Keenan’s observation that highly referential NPs can always occur as subjects (1976:319)). As shown in section 3.5.1., correlations between referentiality and possessivity are typical for many languages and provide the empirical basis for Keenan’s idea about the similarities between Gs and subjects. Russian NPs are relatively harmonic in the sense that their structure is similar to the structure of finite clauses in several respects: prenominal possessive forms are subject-like both in their position with respect to the lexical head, and in their high referentiality, while postnominal genitives are object-like both as concerns their position in the construction and the wide range of their uses (see 3.5.2.). Other
< previous page
page_74
next page >
< previous page
page_75
next page >
Page 75 Table 3.2 The use of the genitive case and of the possessive forms in Russian NPs and VPs In VPs (bojat’sja—‘to be afraid of—’) In NPs (—stol ‘—table’) PR 1/2 SG bojat’sja menja/tebja moj/tvoj stol PR 1/2 PL bojat’sja nas/vas naš/vaš stol PR 3 bojat’sja ego/ee/ix ego/ee/ix stol Proper names bojat’sja Peti stol Peti=Petin stol Other nouns (sosed —‘neighbour’) bojat’sja soseda stol soseda
< previous page
page_75
next page >
< previous page
page_76
next page >
Page 76 languages may present more complicated cases. Thus, in Modern Israeli Hebrew all nominal dependents within NPs follow the head, while the word order in clauses is SVO. Here, perhaps, it is possible to take the mutual precedence relations among dependents as the basis for a comparison with the structure of finite clauses. For instance, both the G and the subject precede all the other dependents in a NP and in a clause respectively (this parallelism is strengthened by the referential characteristics of possessive phrases, as discussed in 3.5.1.). On the other hand, H-G complexes, like VPs, are tight constituents, since no other constituent can intervene between the H and the G (at least, in one type of possessive phrases, see sections 6.2. and 7.2.2.), which suggests that Gs be considered as object-like. In fact, with regard to head-marking, this suggestion is quite reasonable historically, as possessive suffixes in Modern Israeli Hebrew are almost exactly identical to the obsolete object suffixes on verbs. To summarize, possessive NPs across languages vary as to their affinities with verb-subject and/or verbobject structures. However, if a language has two or more possessive constructions, the one in which the G syntactically resembles subjects will often have restrictions on its referentiality/animacy similar to those of subjects. One of the hypotheses proposed here is that the types of possessive phrases a language has influence or determine the choice of nominalization pattern(s) in the language.
< previous page
page_76
next page >
< previous page
page_77
next page >
page_77
next page >
Page 77 Part 2 The data
< previous page
< previous page
page_78
next page >
page_78
next page >
Page 78 This page intentionally left blank.
< previous page
< previous page
page_79
next page >
Page 79 4 The data base 4.0. INTRODUCTION Both ‘primary’ and ‘secondary’ sources were used to create the data base for this study. Data from primary sources was obtained by working with informants or analysing language use; data from secondary sources was obtained from descriptions of languages (including extant grammars and work specialising in nominalizations). (For the terms ‘primary’ vs. ‘secondary’, see Dahl 1985:36). There were two kinds of difficulties in obtaining relevant data: (1) those connected with determining the frame (Bell 1978:126) for the study, that is, connected with finding out which languages have action nominal constructions; and (2) those connected with gathering information about the internal syntax of such constructions. Below I will give some examples of these problems and show some of the advantages and disadvantages of using ‘primary’ and ‘secondary’ data. 4.1. OBTAINING DATA 4.1.1. Determining the frame of the study Since ANCs belong to a relatively complex part of syntax and are rather infrequent, they are often omitted from language descriptions. Consequently, if a description of a language does not say anything about action nominals, this cannot be interpreted as an indication of the non-existence of this category in that particular language. The extreme case is presented by those grammars which do not even mention any forms that could be plausible candidates for action nominals in the language. In such a situation, one is obliged to exclude the language or to try to find another description. However, even if a grammar mentions plausible candidates for action nominals, checking
< previous page
page_79
next page >
< previous page
page_80
next page >
Page 80 whether they actually meet the three types of criteria—derivational, semantic and morphological— formulated in section 1.2.1. is, as we have seen, a far from trivial matter and calls for a great deal of background information about the grammar of the language. For example, the ability to decline and to combine with adpositions in the same way as non-derived nouns do is crucial for determining whether a certain word can, in principle, be a candidate for being termed an action nominal. However, I have often come across grammatical descriptions which mention only one or two morphological cases of ‘verbal nouns’ or ‘infinitives’ without any comments on their ability to be used in other cases. Thus, in connection with noun clauses in Mangarayi (an Australian Aboriginal language), Merlan (1982:9–12) writes that the predicates of purpose and causal clauses are expressed by nominalized verb forms with dative-purposive or ablative case endings respectively. In the section on derivational morphology (ibid.: 173–5), she gives a list of the suffixes for deriving ‘abstractive nouns’ and ‘gerunds’ from verbs, but does not mention whether these derived forms can be used for other functions than those described above and whether they can decline in the other cases. Hence, I have not succeeded in determining the exact status of these Mangarayi forms. 4.1.2. Gathering information about internal syntax of ANCs It often happens that a description of a language states explicitly that it has action nominals (which are really action nominals), however, the information about the internal structure of phrases with them as heads is very scarce. For instance, the grammar just mentions the fact that the genitive in the language may have both the subjective and the objective reading, so that a combination of a transitive action nominal with only one dependent is ambiguous. However, nothing is said about transitive ANCs with both the A and P present—whether they are possible at all, and if so, what their structure is. An example of this kind is provided by Armbruster’s (1960) description of Nubian, which makes it clear that a combination of an AN with one argument looks like a possessive NP. It is, however, impossible to decide whether the A and the P are not allowed to co-occur within one and the same ANC (in which case Nubian could be attributed to the more or less hypothetical Argument-Reducing type, see section 3.3.), or whether information on such combinations is simply lacking in the description. Press’s grammar of Chemehuevi (1980), to quote a somewhat different example, states that the subjects of action nominals
< previous page
page_80
next page >
< previous page
page_81
next page >
Page 81 with the suffix behave like possessors, but it does not mention the objects at all. Unfortunately, for quite a number of languages, such descriptions constitute the only source of information available. Consequently, in these cases, the languages, though belonging to the frame of the study, could not be included in my sample. In some cases, however, I have succeeded in supplementing more or less insufficient ‘secondary’ data by having a questionnaire translated into several languages, viz., Amharic, Egyptian Arabic, Basque, Bulgarian, Chukchee, Estonian, Evenki, Modern Greek, Hungarian, Inuktitut, Kurdish, Luganda, Nivkh, Persian, Swahili and Thai. The work with informants has shown that it is impossible to construct a questionnaire which would be able to pick up ANCs in all languages due to the cross-linguistic differences in the range of syntactic functions and meanings typical for such constructions (see chapters 1 and 2). In many cases, ANCs alternate with other types of finite or non-finite clauses. Consequently, one runs a risk that informants will not choose ANCs in translating the sentences in the questionnaire, if they have not been provided with specific instructions to do so. It is obvious that an ordinary language user does not know what an action nominal is. Thus, in order to obtain relevant data, I needed to specify for my informants which types of constructions they should use in their translations. In other words, this method could be successfully applied when there was enough information about ANs in the language to enable me to identify and then describe them to the informant. The questionnaire contains some very simple sentences which, if their translation equivalents involve ANCs, may illuminate different types of the latter, viz., intransitive ANCs with the lexical and pronominal Ss, and transitive ANCs with only one or both arguments present, the latter being either pronominal or lexical. The questionnaire is not designed for gathering natural examples of the uses of ANCs across languages, but consists of sentences in which they may be used, even if the choice of them does not always correspond to the best translation. In some cases, the informants have indicated that the sentence is correct, though sounding unnatural, and they would have preferred to use a finite clause. For present purposes, I do not consider this to be a great problem. The actual range of uses typical for ANCs in different languages might be a fascinating topic for typological research, but is outside the scope of this study. In several cases, my informants were linguists and experts on the languages under consideration, who could provide me with additional information about the features of ANCs, such as the possibility of containing adverbs or manifesting tense oppositions.
< previous page
page_81
next page >
< previous page
page_82
next page >
Page 82 Data obtained from such informants could hardly be classified as solely ‘primary’. The work with two of the languages, Montagnais (an Algonquian language) and Chukchee (a Paleosiberian language), has demonstrated the necessity of instructing informants about the type of constructions they should use in translating the questionnaire. The Montagnais informant consistently used different types of subordinate clauses, which means either that the language does not belong to the frame of the study, or that I have missed the relevant data. Chuckchee has a productive way of building action nominals, but, in translating my questionnaire, the informants used participial constructions. When instructed about the type of constructions to be used, the Chukchee informants were either unable to build transitive ANCs, or offered clumsy word-for-word translations from Russian. This corresponds to Vladimir Nedjalkov’s (personal communication) observation that Chukchee is a ‘participle-oriented’ language, in which transitive ANCs with both arguments present are virtually nonexistent (see section 9.4.). 4.2. CHARACTERISTICS OF THE SAMPLE 4.2.1. Choice of languages In the early stages of this investigation, it was difficult to devise an efficient strategy for choosing an appropriate sample, due to the difficulties in determining the frame of the study and to the unevenness of available information. The only viable method seemed to be to collect all the kinds of information I came across, which would result in what Bell (1978) calls ‘a convenience sample’. The sample obtained by such a method manifests both the genetic and areal biases. Too many languages in the sample have a long literary tradition and are spoken in industrialized societies which also may have consequences for the validity of the generalizations.1 However, to quote Bell, ‘convenience samples are appropriate when very little is known very well. In exploratory studies the risk of sampling error may well be commensurate with other errors or with the expected level of validity’ (Bell 1978:128). The starting point for the present investigation was exactly the case described above—very little was known at all. However, at a later stage, it seemed unwise to exclude such information as could shed light on various phenomena, even though it might contribute to genetic and areal biases. The primary aim of the
< previous page
page_82
next page >
< previous page
page_83
next page >
Page 83 sample is to cover all types of ANCs, and this consideration overrides considerations of genetic and areal representation. Thus, on the whole, the study should be based on a variety sample, rather than on a probability sample (Perkins 1988, Bakker et al. forthcoming). More specifically, it has turned out that even closely related languages may differ in their choice of nominalization patterns, for example, as do German (with its Ergative-Possessive ANCs) and Swedish (with its Possessive-Adnominal ANCs). Since one of the underlying assumptions for the present research is that the nominalization patterns are distributed non-randomly, such cases offer a unique opportunity to check which factors may influence the choice of particular nominalization patterns by languages. On the other hand, genetically related languages with the same type of ANC may differ in other grammatical details, which suggests that the latter have relatively little impact on the choice of nominalization patterns. In other cases still, closely related languages were quite similar in all respects relevant to my study, which would make an inclusion of them rather meaningless. Considerations such as those mentioned above led to the following decision. The total set of languages for which I have information, is greater than the sample. To be included in the sample, languages must have ‘interesting’ properties. More specifically, for each pair of closely related languages, both are included in the sample if they meet at least one of the following conditions: (1) they choose different nominalization patterns; and (2) they differ with respect to certain important grammatical features. These include basic word order in clauses and in NPs, the dependent-marking system and the head-marking system. The resulting sample is typologically, rather than genetically (and areally) based. Some examples below illustrate this method. Both Inuktitut and West Greenlandic (SOV/GN Inuit languages) have ANCs of the anti-passive type, are both ergative and have the same dependent-marking and head-marking systems (described in sections 3.1. and 6.1.). Only one of them, West Greenlandic, for which I have more data in other respects, is chosen. Similarly, only one from each pair of languages such as Swahili and Luganda (Bantu), Uzbek and Turkish, Avar and Archi (Daghestanian languages), Italian and Rumanian, Russian and Czech, Bulgarian and Macedonian is reported on. In the following cases, however, I have chosen to include several closely related languages in the sample. In the Andean division, three Quechua languages, Imbabura Quechua, Cuzco and Huellaga Quechua, differ in their head-marking type and/or in their ANCs (see section 5.3.). In the Semitic family, four
< previous page
page_83
next page >
< previous page
page_84
next page >
Page 84 languages are included, of which two are closely related Southwest Semitic languages, Classical Arabic and Egyptian Arabic. The latter have different types of ANCs and different dependent-marking possibilities. Three varieties of Mongolian, Classical Mongolian, Modern Mongolian (Khalkha) and Buryat, differ in head-marking and in nominalization patterns (choosing between the Sentential and the Possessive-Accusative). Similar considerations are valid for the Daghestanian, Fennic, Germanic, Slavonic, Italic, Iranian, Greek, Polynesian and Turkic languages. Kwa languages, a classical example of a typologically unstable group, have especially many representatives in the sample. Thus, Ewe, Ifè and Vata/Gbadi differ in their word order, while Vata and Gbadi have different nominalization patterns (the Possessive-Accusative and the Double-Possessive respectively). 4.2.2. Genetic characteristics of the sample: general The seventy languages for which I have satisfactory data on the internal syntax of intransitive and transitive ANCs are presented in Table 4.1, together with their genetic affiliation. As it is clear from Table 4.1, the language families are represented very unevenly in the sample. Such a state of affairs is, perhaps, not as deplorable as it looks, since in some language families ANCs do not exist at all, while in others they are quite common. In other words, some language families and even larger ‘stocks’ appear to be completely or partly outside the frame of the study. An interesting question, which will not be answered here, is whether languages with and languages without ANCs constitute two natural classes. At present I am aware of two constructions, the existence of which is likely to imply the non-existence of (transitive) ANCs in a language. One involves various forms intermediate between action nominals and other non-finite verb forms, while the other is clausal nominalizations. Let us discuss these two cases in more detail. It should be mentioned, however, that in rare cases both types of constructions coexist with ANCs in a single language (for examples see chapter 2).
< previous page
page_84
next page >
< previous page
page_85
next page >
Page 85 4.2.2.1. Languages with forms intermediate between ANs and other non-finite verb forms (‘prototypical’ infinitives, participles and converbs) On Australian languages, Dixon (1976:323) writes: There are usually also one or two suffixes for deriving nominals from verbal roots—generally agentive nominals like ‘hunter’ from ‘hunt’, and sometimes other types of deverbal noun like ‘the vomit’ from ‘to vomit’ and ‘a fight’ from ‘to fight’. Processes of nominalization are not used as much as those of verbalization and may be entirely lacking from some languages. Descriptions of specific Australian languges either do not mention any verbal nouns with the appropriate meanings, or just provide a couple of examples, such as those quoted by Dixon, making it impossible to determine whether these can take any arguments at all. My general impression, however, is that Australian languages do not, on the whole, have action nominals as defined in section 1.2.1., capable of following the nominal inflectional paradigm, appearing with nominal functions and/or building ANCs of their own, with the S, A and P non-coreferential with any NPs in the matrix clause (see section 2.2.3.4 on Warlpiri and 2.2.5. on Kayardid). Mayan languages (American Indian languages) Tzutujil (Dayley 1985) and Sacapultec (DuBois 1981) have active infinitives and passive verbal nouns, none of which are capable of having overt As (see section 2.2.3.4). 4.2.2.2. Languages with clausal nominalizations These include Sino-Tibetan languages, a Khoisan language Nama Hottentot (Hagman 1977), an Indo-Pacific language Hua (Haiman 1980), a language isolate Ket (Grišina 1979) as well as a number of American Indian languages, such as Yuman languages Diegueño (Langdon 1970) and Hualapai (Watahomigie et al. 1982), an Uto-Aztecan language Yaqui (Lindenfeld 1973), a Salishan language Squamish (Kuipers 1967), a Siouan language Lakhota (Van Valin 1985), and an Athapascan language of the Na-Dene stock Navajo (Platero 1974:203 mentions /-ígíí/ and /-yęę/ as generalized nominalizers ‘appearing on any sentence dominated by the NP node’). Another Athapascan language for which I have information, Sarcee (Cook 1984), seems to have ANCs of the Possessive-Accusative type, however, the data are insufficient for including the language in the sample.
< previous page
page_85
next page >
< previous page
page_86
next page >
Page 86 Table 4.1 Languages with ANCs in the sample by genetic groups and macroareas Northern Eurasia Altaic Tungus EVENKI Turkic TURKISH, TUVA Mongol CLASSICAL MONGOLIAN, BURYAT, MODERN MONGOLIAN (KHALKHA) Japanese JAPANESE Isolated KOREAN Caucasian Northeast Caucasian Daghestanian: AGUL, ARCHI, TABASSARAN, Vejnax:INGUSH South Caucasian=Kartvelian GEORGIAN Northwest Caucasian ABKHAZ Indo-European Baltic LATVIAN Celtic OLD IRISH, WELSH Germanic DUTCH, GERMAN, ENGLISH, ICELANDIC, SWEDISH Greek ANCIENT GREEK, MODERN GREEK Indic PANJABI Iranian KURDISH, PERSIAN Italic LATIN, FRENCH, RUMANIAN Slavonic BULGARIAN (and MACEDONIAN), POLISH, RUSSIAN (and CZECH) Uralic Finno-Ugric: Baltic-Finnic ESTONIAN, FINNISH Ugric HUNGARIAN Samoyedic SELKUP Paleosiberian Chukotko-Kamchatkan CHUKCHEE (CHUKCHI)
< previous page
page_86
next page >
< previous page Page 87 Isolates Africa and Near East Afro-Asiatic Chadic Egyptian Semitic Niger-Congo Bantu Kwa: Kru West-Atlantic Southern Eurasia Dravidian Kam-Tai North America Eskimo-Aleut: Inuit Penutian: Yokuts Central and South America Carib Andean Pacific Austronesian Northwest Polynesian Papuan East New Guinea Highlands Gum
< previous page
page_87
next page >
GILYAK (NIVKH), YUKAGIR, BASQUE, BURUSHASKI HAUSA ANCIENT EGYPTIAN AMHARIC, CLASSICAL ARABIC, (COLLOQUIAL) EGYPTIAN ARABIC, MODERN ISRAELI HEBREW SWAHILI (and LUGANDA) EWE, IFÈ, GBADI, VATA FULA TAMIL THAI INUIT (INUKTITUT and WEST GREENLANDIC) WIKCHAMNI HIXKARYANA CUZCO QUECHUA, IMBABURA QUECHUA, HUALLAGA (HUÁNUCO) QUECHUA TAGALOG MAORI, PUKAPUKAN, SAMOAN (and TOKELAUAN), TAHITIAN, TONGAN KOBON AMELE
page_87
next page >
< previous page
page_88
next page >
Page 88 At least two typological features of languages favour clausal nominalizations—a high degree of analyticity and, thus, a minimum of derivation, and a high degree of head-marking. The latter fact is mentioned by Nichols (1986:113) in connection with the concept of centricity as applied to languages with different morphological marking types. She suggests that in at least some head-marking languages, clauses are endocentric in the sense that they have the same distribution as their headmarked finite verbs. She shows that this is by no means true for all head-marking languages by quoting Abkhaz, a complement-deranking Northwest Caucasian language, in which finite verbs cannot replace subordinate clauses and, thus, have another distribution than the latter. It is possible that Ket, besides having clausal nominalizations, also has ANs, but it is difficult to get clear data on this issue from the different available sources. Thus, Uspenskij (1968:206) and Živova (1983:130) mention infinitives which are either equal to the verbal stems or are derived from them by means of certain suffixes. Krejnovič (1979b:338) calls, presumably, the same words ‘action nominals’ and describes them as a separate part of speech, which has an intermediate status between nouns and verbs and is used for deriving the latter. These words decline and take possessive prefixes in the same way as ordinary nouns. The only examples I have found in the literature involve either intransitive ‘action nominals’ or transitive with only the P present. In both cases, such constructions resemble possessive NPs. On the expression of the A and combinations of both the A and the P, I have no material. 4.2.3. Other characteristics of the sample 4.2.3.1. Areal distribution of the languages The only continent whose languages are totally missing from the sample is Australia, (see section 4.2.2.). All the other major regions are represented, though rather unevenly (thus, the sample includes only two North American languages, four South American and two New Guinean languages). 4.2.3.2. Word order types If by word order we mean the basic word order within finite clauses and the N-G order within NPs, the sample represents all the combinations from Hawkins’ list (1983), viz. SOV/GN, SOV/NG,
< previous page
page_88
next page >
< previous page
page_89
next page >
Page 89 V-initial/NG, V-initial/GN, SVO/GN, SVO/NG, plus one OSV/GN-language (Hixkaryana). However, if, in addition, both the position of adpositions and that of adjectives with respect to the head nominal are taken into consideration, the following four less frequent types are missing as compared to the fifteen attested types of the twenty-four possible ones: V-1/GN/AN (represented by some Aztec-Tanoan languages in Hawkins’ sample) and SVO/GN/NA, both with pre- and postpositions. 4.2.3.3. Other typological characteristics Most languages in the sample are synthetic, both agglutinating and fusional. The predominant part of the languages are nominative-accusative, though eleven are ergative and three manifest split ergativity. There are no active languages in the sample. As suggested by Johanna Nichols (personal communication), there may be a high correlation between the property of being an active language and the property of having clausal nominalizations. Tagalog and Yukagir organize their finite clauses somewhat differently, basing them on both grammatical relations and on pragmatic functions (Andrews 1985:138–47, Maslova 1989). Only four typical head-marking languages (Abkhaz, Hixkaryana, Egyptian Arabic and Ancient Egyptian) are included in the sample. The sample also shows examples of predominantly dependent-marking, split-marking and double-marking languages.
< previous page
page_89
next page >
< previous page
page_90
next page >
Page 90 5 Type 1: ‘Sentential’ 5.0. INTRODUCTION The Sentential type (SENT) is characterized by the retention of the sentential dependent-marking in ANCs. 5.1. ‘IDEAL’ REPRESENTATIVES: KOREAN AND ARCHI Korean and Archi serve as ‘ideal’ representatives of this type, because ANCs in them retain all the devices used for expressing the syntactic relations between verbs and their arguments in independent clauses: dependent-marking, word order and, in the case of Archi, head-marking. Korean, a nominative-accusative non-rigid SOV language, uses only dependent-marking both on the clause and phrase levels. It is both subject- and topic-prominent (Li and Thompson 1976:483), which means that both the subject and the topic get marked by special postpositional particles (i/ka and ŭn/nŭn respectively, the former being called ‘nominative case’ in most grammars) and the subject of a clause need not be its topic. Not all sentences have overt topics. In spoken Korean, subjects may occur in the non-marked form (Martin and Lee 1969:48). Direct objects are either in the non-marked form or in the ‘accusative case’ (i.e., with the postpositional particle (l)ŭl) depending on their degree of referentiality; indirect objects are in the ‘dative case’ (i.e., with the postpositional particle eke ). Deletion of subjects is allowable and, in fact, very common (Kim 1987:894). (5.1) John i Mary eke chæk John NOM Mary DAT book ACC give-PAST-FIN ‘John gave a book to Mary.’ (Kim 1987:894)
< previous page
page_90
next page >
< previous page
page_91
next page >
Page 91 (5.2) Na ton i is'-o I TOP money NOM exist-FIN ‘As to me, money exists.’ (=‘I have money.’) (Ramstedt 1968:38) (5.3a) koki cap-as’-ta fish catch-PAST-FIN ‘caught fish’ (5.3b) koki cap-as’-ta fish ACC catch-PAST-FIN ‘caught the fish’ (ibid.: 45) Ordinary noun phrases make extensive use of dependent-marking patterns, but they differ from sentences with finite verbs as to what cases are allowed for nominal modifiers within them. Nominative and accusative cases are excluded from them. Usually, adnominal modifiers appear in the ‘genitive’ case, i.e., with the particle ŭi. Dependent nouns in the uninflected form are possible where they are to be understood as characterizing, non-referential terms which are not able to have their own modifiers. Dependents precede their head. (5.4a) mal Korea language ‘Korean language’ (5.4b) kŭ koŋ ŭi ilŭm i that man GEN name NOM ‘the name of that man’ (Ramstedt 1968:40) ANCs have verbal nouns with the suffixes m/ŭm and ki as their head. In the majority of cases, the former is used for factive complements, while the latter is used for non-factive ones (Kim 1987:897). They retain the sentential syntax both as regards dependent-marking and word order. The only difference between finite clauses and ANCs consists of the impossibility of using topic particles in the latter, as well as in all other types of subordinate clauses. This can be attributed to the fact that topicalization is a typical ‘root’ phenomenon (i.e., a phenomenon which is allowed only in independent, not embedded clauses), and is, thus, not allowed to occur in dependent clauses. Example (5.6) illustrates the possibility of deleting the A in ANCs without any changes in the form of the P. (5.5) [ ka motŭn ton ŭl ] she NOM all money ACC have-PAST-AN t’æmun e kŭ nŭn ka pilyoh-æs’-ta. reason in he NOM she ACC need-PAST-FIN ‘He needed her because she had all the money.’ (Hang-seok Choi, personal communication)
< previous page
page_91
next page >
< previous page
page_92
next page >
Page 92 (5.6) [Ku ton Pak eke cu-kı] that money ACC Pak mister DAT give-AN ACC palæ-yo hope-FIN ‘I hope you gave that money to Mr Pak.’ (Martin and Lee, 1969:231) Archi is also a dependent-marking SOV language. However, it differs from Korean in three respects: firstly, it has one head-marking feature (discussed immediately below), secondly, it is ergative, and thirdly, it is not topic-prominent. Major syntactic functions in finite clauses are expressed, primarily, by means of dependent-marking strategies: the S and the P (the absolutives) are put in the nominative, the A is put in the ergative. The only head-marking feature of Archi is agreement of verbs with the class of their absolutives (all the nouns in Archi are divided into eight classes). Thus, Archi is an ergative language both with respect to its dependent-marking and head-marking in clauses. (5.7) zari ši-a-b-u I:ERG letter:NOM.CL4 wrote:CL4 ‘I wrote a letter.’ (Kibrik 1977, vol. 3:185) Ordinary NPs use only dependent-marking strategies, the usual case for adnominal modifiers being the genitive. Neither the nominative nor the ergative are allowed within ordinary noun phrases. Adnominal modifiers usually precede their heads. (5.8) čaqlin muh pea:GEN corn ‘peas’ (ibid.: 91) ANCs retain both the dependent-marking and the head-marking strategies used in finite clauses. (5.9) wez sini [un ] I:DAT know you:CL1.NOM work:AN:CLI:NOM ‘I know that you work.’ (ibid., vol. 2:312)
< previous page
page_92
next page >
< previous page
page_93
next page >
Page 93 (5.10) [labu armili girman we:GEN army:ERG Germans:CL1/2PL:NOM Ralq’-b-amul] el ko defeat:AN:CL1/2PL:NOM we:NOM learned ‘We learned about the defeat of the Germans by our Army.’ (ibid., vol. 3:185) Both Korean and Archi are strongly complement-deranking languages and resort to numerous non-finite verb forms for expressing subordinate relations. Among other things, ANCs are used much more often and in many more contexts than in typical complement-balancing languages and retain certain verbal categories. Thus, Korean ANCs show oppositions in tense (cf. ex. 5.5), though not exactly the same as in sentences with finite verbs, politeness, and voice (the latter category seems to be derivational). ANCs in Archi, like any constructions with verb forms, have the following verbal features. Firstly, they are characterized by a certain aspect-tense: action nominals can be derived either from the underlying verbal stem or from any of the aspect-tense forms. In the former case they have the general meaning of an action or state, in the latter case they refer to concrete actions or states connected with some particular time. Secondly, action nominals can be negated in the same way as any verb form. In both languages, ANs can have adverbial modifiers of time, place, degree and so on. Thus, on the whole, ANs in Korean and Archi are, in many respects, close to finite verb forms. In other respects, however, they are treated as nouns. In Archi, ANs, exactly like non-derived nouns, decline in all the cases, may be put in the plural and trigger agreement (they belong to the fourth class of substantives). In Korean, since the categories of number and gender are not relevant for nouns, the only nominal feature characterizing ANs is their ability to decline. In Korean, verbal nouns with the nominalizer m/ŭm can develop more concrete senses and refer to results, objects and products of actions rather than the actions themselves. In this connection, they assume some nominal syntactical properties: their S and A take the genitive case; certain postpositions, when combined with them, are used in a special adnominal form; they are modified by participles and not by converbs as finite verbs are (Vasil’ev 1963).
< previous page
page_93
next page >
< previous page
page_94
next page >
Page 94 (5.11 a) Caju lŭl vihaj s’unda freedom-ACC for:VPOSTP fight:FIN ‘(He, she etc.) fights(s) for freedom.’ (5.lib) caju lŭl vihan s’um freedom-ACC for:NPOSTP fight:AN ‘a fight for freedom’ (ibid.) In Archi, nouns derived from one-place stative verbs and ‘ bos- verbs’ (i.e., verbs derived with the suffix bos) assimilate to the syntax of ordinary NPs. Some details remain unclear from the description concerning, first of all, the exact range of verbs that behave in such a way. If we take the clearest example, that of one-place stative verbs (with the meanings close to adjectives in the European languages), the following features characterize nominals derived from them: (1) their S is put in the genitive; (2) they do not agree with their arguments and (3) they are incompatible with adverbs, which must be transformed into adjectives, the latter agreeing with the action nominal in class (Kibrik 1977, vol 3.: 180). (5.12a) tor horōkejru she:NOM for.a.long.time:ADV is.glad:PRES ‘She has been glad for a long time.’ (5.12b) [tormin she:GEN for.a.long time:ADJ.CL4 ] ez sini being.glad:AN.CL4:NOM I know ‘I know that she has been glad for a long time.’ (ibid.) These peculiarities are connected with the low degree of ‘verbality’ typical for stative verbs. They have only one aspect-tense stem, as contrasted to dynamic verbs, and even finite forms do not agree with any of their arguments. 5.2. PROBLEMS IN DEFINING THE PATTERN As regards Korean and Archi, there is a strongly pronounced difference between their dependentmarking patterns on the clause and phrase levels, i.e., special cases are reserved for expressing the S, A and P of finite verbs and are not permitted within NPs, where the G is put in still another case, normally the genitive (this difference is
< previous page
page_94
next page >
< previous page
page_95
next page >
Page 95 greater in Archi, because Korean does allow one and the same non-inflected form as both subjects and Gs). Thus, for these two languages it is easy to say whether their ANCs have retained the sentential dependent-marking or not. However, problems with classifying nominalization patterns arise when the languages in question do not show any clear distinction between dependents in finite clauses and those in NPs, making the distinction between the SENT and the POSS-ACC types blurred or neutralizing it completely. These cases will be discussed below. 5.2.1. (Almost) identical dependent-marking of Gs in NPs and that of As in finite clauses: Burushaski The Indian language isolate Burushaski presents some problems for the classification, due partly to the insufficient data in my source (Lorimer 1935). In this language, dependent-marking on the clause level follows the ergative pattern—obligatorily in the past tense and with certain verbs in any tense, and optionally with any tenses of any transitive verb (ibid.: 64). In these cases, the A is put in the ergative case, as distinct from the non-marked form of the P. Otherwise, both the A and the P are put in the same non-marked form. Finite verbs have the same type of agreement for the A and the S (i.e., for the subject). In many cases, but not always, verbs may be head-marked for the P by means of a special set of prefixes. Within NPs, there is a set of kinship terms which take head-marked possession; in all the other cases the G is put in the genitive case and no head-marking appears on the H. Now, the problem is that the genitive case is almost identical to the ergative, the only difference appearing in words referring to human females. Lorimer himself gives no analysis for the forms of the arguments in ANCs. In the few examples I have found in his grammar, the As do not refer to human females, their form is ambiguous and may be interpreted both as the ergative and the genitive case. The P is in the nonmarked form. Thus, ANCs in Burushaski can be attributed either to the SENT type (if the A appears in the ergative), or to the POSS-ACC type (if the A is put in the genitive). I assume, though, that the former description is correct. My supposition is based on the fact that the S in ANCs may turn up in the nominative case, that is, in its sentential form. My sample suggests that retention of sentential marking for the S in ANCs of a language implies retention of sentential marking for the A. More specifically, intransitive ANCs across languages are the easiest to assimilate to non-derived NPs, and even some of the languages with the SENT type of nominalizations
< previous page
page_95
next page >
< previous page
page_96
next page >
Page 96 allow genitive Ss in intransitive ANCs (see section 5.5.). The only example of the opposite situation is found in Tahitian (see section 10.3.), where, in ANCs, the As genitivize, but the Ss may retain sentential marking. (A somewhat different exception is provided by Basque, see 10.3.) Additional data on Burushaski would shed light on the actual state of affairs concerning ANCs in this language. 5.2.2. The absence of dependent-marking for Gs in NPs, as well as for Ss, As and Ps in finite clauses In the case of Nivkh (Gilyak) and Kobon, the problems in defining the type of ANCs are of a more fundamental nature, as they arise from the system of dependent-marking in the languages. Only ANCs with pronominal arguments can be classified as belonging to the SENT type, while for other ANCs, the exact attribution to any type within the present classification does not make sense (more specifically, for these languages, there would be no opposition between the SENT, POSS-ACC and DBL-POSS types). Nivkh, a Paleosiberian SOV/GN-language, presents an interesting case in several respects. The nonmarked form of nouns serves both as the S, A and P of finite clauses (except those with causative verbs), and as the G in non-derived NPs, which means that nominal dependent-marking is not a sufficient criterion for opposing finite clauses to noun phrases. Pronouns, however, distinguish between the non-marked form and the attributive forms, the latter appearing only in NPs in two variants—either as free preposed pronouns, or as pronominal prefixes on the head-noun (as in example 5.13b). Verbs do not agree with the subject, but a number of them take special markers signalling the distinction between transitive and intransitive verbs (all the transitive verbs have initial consonants from one series, while the intransitive verbs have initial consonants from another series). A group of verbs (the most ancient ones) take special pronominal prefixes, when there is no overt object in the clause.1 (5.13a) man-dwelling ‘a man’s dwelling’ (Krejnovič 1979a:302) (5.13b) v-ytyk his-father
< previous page
page_96
next page >
< previous page
page_97
next page >
Page 97 (5.14a) Ty č’-xysk this dog your-cat pursue-FIN ‘This dog was pursuing your cat.’ (5.14b) If lu-d’. he permanently sing-FIN ‘He always sings.’ In ANCs, all the S, A and P, even when pronominal, appear in the non-marked form. The transitive verbs, distinguishing between the presence and the absence of an overt object, retain this distinction in nominalizations (as in examples 5.16a–b below). Thus, ANCs with pronominal arguments follow the SENT pattern. (5.15a) Imŋ [qan č-xysk] ama-d’. they dog your-cat pursue-AN look.at-FIN ‘They were looking at the dog pursuing your cat.’ (5.15b) [If lu-d’] ŋalzaj-gu-d’. he permanently sing-AN be.angry-CAUS-FIN ‘His perpetual singing makes (me) angry.’ (5.16a) Puškin j-uru-d’ Pushkin OBJ-read-AN ‘Pushkin’s reading (of something)’ (5.16b) Puškin uru-d’ Pushkin read-AN ‘(someone’s) reading of Pushkin’ Kobon, a New Guinean SOV language, also has very few dependent-marking patterns. On the clause level, only pronouns have distinct object forms used optionally. Nouns in the non-inflected form function as both subjects and objects. Within NPs, G-H relations are signalled by means of juxtaposition. What is interesting, though, is the different word orders in NPs with lexical and pronominal Gs. The former precede the head, while the latter follow it. In ANCs, the Ss and the As, both lexical and pronominal, precede the head. Thus, ANCs with pronominal Ss and As have the same word order as in finite clauses, as opposed to non-derived NPs, and belong to the SENT type. Other ANCs show the same headdependent word order and involve the same non-inflected form of the dependents as both finite clauses and non-derived NPs. A somewhat different situation is presented by Tuva (and some other Turkic, especially Siberian Turkic languages). In ordinary NPs,
< previous page
page_97
next page >
< previous page
page_98
next page >
Page 98 the G appears either in the genitive or in the non-marked form depending on its referential characteristics, while the head in both cases takes possessive suffixes. The subject in finite clauses is normally in the non-marked form. In ANCs, the head bears possessive suffixes, while the A and the S may choose between the non-marked form and the genitive case form. In other words, the dependentmarking of the A and S in ANCs is subject to the same variation as the dependent-marking of the G in ordinary NPs (Čeremisina 1981a, Ubrjatova and Litvin 1984). While ANCs with the A or S in the genitive case are unambiguously classified as manifestations of the POSS-ACC type (see chapter 6), those with the A or S in the non-marked form present classificatory problems. They may be characterized both as SENT ANCs (since the dependents retain their sentential form) and as POSS-ACC (since the A/S appears in the same form as the G). In the latter case, we have to admit that the POSS-ACC type in Tuva is manifested in two different ways. The latter decision, which is taken for Selkup and Buryat (see chapter 6), would be preferable if we could show that the choice between the genitive case and the non-marked form both in ordinary NPs and in ANCs is conditioned by similar factors (first of all, by the referential characteristics of the nominal in question). There are, however, indications that the distribution of the genitive and the non-marked forms in ANCs differs from that in non-derived NPs (see section 10.1. for discussion). In any case, Tuva shows that the boundary between the SENT and the POSS-ACC types is rather vague and, to a certain degree, artificial. 5.3. HEAD-MARKING IN THE SENT TYPE As mentioned above, Korean and Archi have the same head-marking patterns in ANCs and in clauses (for Korean, this means the absence of head-marking in both types of constructions). However, for the SENT type on the whole, this is by no means a characteristic feature. This section is devoted to the relation between head-marking used in ANCs of the SENT type, as compared to head-marking in finite clauses vs. noun phrases. As Nichols (1986:75) has stated, ‘if a language has major, salient, headmarking morphology anywhere, it will have it at the clause level’. Although most languages in the sample are predominantly dependent-marking, the implication quoted above can apply to them in a slightly modified formulation (some counter-examples are not relevant for the SENT type), i.e., if a language makes use of head-marking within non-derived possessive NPs, it is sure to have headmarking morphology at the clause level as well. This
< previous page
page_98
next page >
< previous page
page_99
next page >
Page 99 means that we are left with the following three possibilities as regards combinations of head-marking in finite clauses and NPs in the sample SENT-languages: (1) no head-marking, (2) head-marking only in finite clauses, and (3) head-marking both in finite clauses and in NPs. These groups will be discussed separately below. 5.3.1. No head-marking This is the case with Korean, Classical Mongolian and Tongan. Nivkh is also close to these languages, except for its lexically conditioned verbal object prefixes, appearing in the absence of overt direct objects. 5.3.2. Head-marking in finite clauses, no head-marking in NPs This is represented by Daghestanian languages, Burushaski, Kobon, Imbabura Quechua and Tamil. Ingush has some minor head-marking possibilities. The languages in this group are heterogeneous as regards head-marking in their ANCs. Most of them follow the nominal pattern, i.e., they use or avoid head-marking in their ANCs to the same extent as in non-derived NPs. In Daghestanian languages, however, ANCs use head-marking, although in non-derived NPs, head-marking is absent. It may be suggested that this difference is not accidental, but depends upon the different nature of verbal agreement in both cases. In the Daghestanian languages, verbs agree in class with their absolutives, while verbs in Imbabura Quechua, Tamil, Basque and Kobon have only personal agreement. Thus, class agreement seems to be retained more easily than personal agreement. Tabassaran is a good illustration of this statement. In the Kondik dialect of Tabassaran there are two action nominals (masdars). In finite clauses, the Ss and Ps are put in the nominative, while the As take the ergative. Verbs agree in class with their absolutives (i.e., according to the ergative pattern) and in person with semantically and communicatively salient NPs (in accordance with a strict hierarchy, see Kibrik 1985:287). In intransitive ANCs, the S can be put either in the nominative or in the genitive, while transitive ANCs always follow the SENT pattern. While action nominals retain class agreement with the S and P, independently of the overt expression of the S, personal agreement of verbs with their subjects exists only in finite clauses.
< previous page
page_99
next page >
< previous page
page_100
next page >
Page 100 (5.17a) Daši/dašijin R-uIb-az . father:CL1:NOM/GEN come-AN:CL1-DAT awaited:we ‘We waited for Father’s coming.’ (Bogatyrev and Boguslavskaja 1982:85, ex. 66–7) (5.17b) XuIni/XuIndin li-b-ch-ub cow:CL2:NOM/GEN go-CL2-AN-NOM . good I:not:see ‘I do not see the good cow’s step.’ (ibid.: 85, ex. 68–9) Bantu languages (the POSS-ACC type, see section 6.3.) offer another example of retention of class agreement in ANCs. The exact situation in Ingush with its minor head-marking patterns on the clause level, is unclear from my source of information on the language.2 5.3.3. Head-marking both in finite clauses and in NPs Classical Arabic, Cuzco Quechua and Huallaga (Huánuco) Quechua (and also Tuva, if its ANCs are of the SENT type) belong to this group. Although all these languages are double-marking, where certain syntactic relations are marked both on the head and the dependent of a constituent (Nichols 1986:72), they behave quite differently with respect to head-marking in ANCs. This probably depends on the different nature of double-marking in NPs of these languages, more specifically, on the difference between head-marking that only signals the presence and type of dependency (as in Classical Arabic), and head-marking that, additionally, indexes certain grammatical categories of the dependent, as in Cuzco and Huallaga Quechua and in Tuva (for this distinction see Nichols 1986:78). In Classical Arabic, head-marking in NPs appears in two variants - either as possessive suffixes (in the absence of an overt external G in the same NP), or as what is called the Construct State of the headnoun, when it has a dependent in the genitive case. Only the latter situation is an example of doublemarking, and in these cases the head-marking (or the absence of head-marking3) merely registers the presence of syntactic dependency as such. The Construct State of the H in NPs seems to be inseparably linked to the presence of a genitive dependent. ANCs of the SENT type do not contain such genitive dependents and, thus, do not involve any head-marking, their head appearing in the Absolute State.
< previous page
page_100
next page >
< previous page
page_101
next page >
Page 101 In Cuzco and Huallaga Quechua, the structure of ANCs shows that there is hardly any strict boundary between the SENT and the POSS-ACC types, since dependent-marking and head-marking can follow different patterns. In ordinary NPs, the H takes possessive suffixes independently of the presence or absence of an overt external G in the same NP. Thus, nominal head-marking in these Quechua languages is not as intimately linked with dependent-marking as in Classical Arabic and, in doublemarked NPs, it not only signals the presence of a G, but also indexes the person and number of the latter. The possessive suffixes are formally very similar to the verbal subject suffixes, differing only in the second person. Now, in ANCs, head-marking of the S and A in ANCs follows the nominal pattern, while dependent-marking and head-marking of the P follow the sentential pattern. Such ANCs actually occupy an intermediate position between the SENT and the POSS-ACC types. In Cuzco Quechua, one and the same action nominal with the same head-marking can, in fact, combine with the A or S both in the nominative and in the genitive. In the latter case, there is, thus, no contradiction between the dependent-marking and the head-marking within the ANC (note, however, that the P in such constructions cannot retain its sentential dependent-marking and appears in the non-marked form, see section 5.5.). For Huallaga Quechua, however, this option is not reported in my source. Note that neither the G in non-derived NPs nor the S or A of ANCs needs to be present, as their person and number is indexed within the head of the NP or the ANC respectively. This means that ANCs without an overt independent S or A look exactly like non-derived NPs without an independent G, except for the person agreement with the P. It is worth mentioning that some of the languages, closely related to Classical Arabic vs. Huallaga Quechua, have chosen quite opposite ways for avoiding double-marking in NPs. Thus, modern Arabic dialects (for example, Egyptian and Syrian Arabic) have lost dependent-marking in NPs (the G appears in the non-marked form), but have retained head-marking (the H is put in the Construct State). In the Ecuadorian Quechua languages (for example, in Imbabura Quechua), head-marking in NPs is lost, while dependent-marking is retained (the G appears in the genitive case). Tuva is superficially quite similar to Cuzco Quechua in that ANs with the nominal head-marking can combine with the A/S in both the non-marked (sentential) form and in the genitive case. However, in contrast to Cuzco Quechua, where the G of ordinary NPs is always put in the genitive, provided that the head bears possessive affixes, in
< previous page
page_101
next page >
< previous page
page_102
next page >
Page 102 Tuva, the same variation is found in ordinary NPs. Tuva demonstrates another peculiarity, which concerns head-marking within adverbial ANCs (i.e., those SENT ANCs which are used in adverbial functions). For the third person S and A, the AN does not take any overt possessive suffix, in contrast to the H in possessive NPs and to the head of ANCs used in complement functions. Thus, the structure of the head-marking paradigm for ANs in adverbial uses is similar to that of finite verbs, which also lack any overt expression for the subject in the third person. On the whole, it is questionable whether such constructions should be analysed as ANCs and not as converbial clauses (see the discussion in 10.1.). All the facts mentioned in the present section are summarized in Table 5.1 (pp. 106–7). 5.4. FURTHER CHARACTERISTICS OF THE SENT TYPE Among the languages included in this group are nominative-accusative, ergative-absolutive languages, one language with split ergativity (Burushaski) and one which does not dependent-mark the major functions in finite clauses (Nivkh). Almost all of the SENT-languages have SOV/GN word order, except for the two V-1-languages, Classical Arabic and Tongan. SENT nominalizations in the former are qualified by Wright (1971:2, 58) as very rare. Almost all of the SENT-languages are exclusively or predominantly complement-deranking, except for the same two languages, Classical Arabic and Tongan, and Basque, all of which have finite complements. In a number of cases, ANCs of the SENT type have retained some other features characteristic of sentences. It is here that the quality of available information varies extremely. However, even this information invites generalizations. Thus, Korean, Tamil, Archi, Quechua, Classical Mongolian and Tuva have some tense distinctions in their action nominals. Korean, Archi, Tamil, Quechua and Kobon retain adverbs or adverbial modifiers, not allowed in non-derived NPs. Nominalizations in Tamil, Kobon and Imbabura Quechua show the same range of reflexive relations as independent clauses. As for negation, Archi and, presumably, Tamil use the same device in ANCs and in independent sentences. Quechua (at least Imbabura and Cuzco) negates ANs and finite verbs in slightly different ways, depending on the impossibility of using one of the two parts of the sentential negation with non-matrix predicates (since it is a validator, see below). Kobon does not retain negation in ANCs.
< previous page
page_102
next page >
< previous page
page_103
next page >
Page 103 The category of voice in ANCs is heterogeneous. Thus, in Korean, there seems to exist a number of derivational suffixes that link different verb-stems, serve to express some voice oppositions and are left intact by the processes of nominalization. Imbabura Quechua uses various syntactic (periphrastic) constructions as the passive correspondences to active sentences. These cannot be rendered by nominalization. Most languages with ANCs of the SENT type do not have any voice distinctions at all. In Imbabura Quechua and Kobon, ANCs cannot be coordinated with ordinary NPs. Though, on the whole, SENT ANCs retain many properties of finite clauses, in some cases, they do not allow certain root phenomena, manifesting restrictions on outputs of rules characteristic of independent clauses. For instance, Quechua has special classes of clitics, validators and topic markers, which can only be suffixed to major constituents of the matrix clause. Korean topic particles are also restricted to the matrix clause. Modal oppositions are never retained in ANCs. Word order in nominalizations and in other types of subordinated constructions tends to be more rigid than in matrix clauses. This is true for, for example, Tamil and Imbabura Quechua. All these phenomena may be attributed to the fact that subordinate clauses do not normally have their own illocutionary force (see Foley and Van Valin 1984:239, and Lehmann 1988:193–5). Another difference between the structure of finite clauses and ANCs is manifested by Classical Mongolian, in which the accusative is often used to refer to the subject of subordinated predicates under special conditions: ‘at the beginning of a sentence or a clause, an accusative may be used to indicate that the word is not subject to the final verb but to the closest verb’ (Grønbech and Krueger 1955). A similar rule applies in Khalkha Mongolian, a POSS-ACC language (see section 6.4.). 5.5. INTRA-LEXEMIC SYNTACTIC VARIATION In a few of the languages described here, a single AN can combine with its arguments not only according to the SENT pattern, but also according to the following patterns: 1 POSS-ACC type: Tuva; some of the Daghestanian languages of the Lezghian group, such as Rutul or Agul. In the latter, the syntactic difference in the shape of nominalizations is accompanied by the semantic opposition ‘fact/process’ (see section 10.1.). In Classical
< previous page
page_103
next page >
< previous page
page_104
next page >
Page 104 Arabic, the most usual type of ANCs is the POSS-ACC, the SENT type being used very rarely and only with transitive ANs. 2 ERG-POSS type: Tongan, for which I lack explicit information on the frequency of the different nominalization patterns, and Classical Arabic. 3 INC type: Imbabura Quechua. Finite verbs do not allow incorporation of their objects, which must appear in the accusative. In ANCs, however, the Ps can be incorporated, i.e., they are not case marked and are placed immediately before the nominalized verb (when the P is not incorporated, it can be separated from the AN by adverbial modifiers). The A appears in the nominative. In other words, Imbabura Quechua is consistent in retaining the sentential marking for the S and the A in ANCs. (5.18a) Juzi-ka [uka kaya randi-na-ta] Jose-TOP I tomorrow buy-AN-ACC kri-n believe-3SG (5.18b) Juzi-ka [uka kaya randi-na-ta] Jose-TOP I tomorrow buy-AN-ACC kri-n believe-3SG ‘Jose believes I will buy a sheep tomorrow.’ (Cole 1982:37, ex. 118–19) As Cole (1982) notes, incorporation in Imbabura Quechua is restricted to surface noun-noun combinations and could perhaps be considered a special case of nominal compounding. Compare the example above with the noun-noun compounds rumi ñan ‘stone road’ and warmi wagra ‘female cow’ (ibid.: 198). The situation in Cuzco Quechua seems to be more complicated, since the P in the non-marked form may be separated from the head-noun and at least superficially does not form a compound with it. In the following example, the P is separated from the action nominal by the A: (5.19) Pi-qpa-ta-n muna-nki [ who-GEN-ACC-AFF want-2 Xwancha-q riku-na-n-ta]. Juan-GEN see-AN-3-ACC ‘Of whom do you want Juan to see the co-worker?’ (Lefebvre and Muysken 1988:130)
< previous page
page_104
next page >
< previous page
page_105
next page >
Page 105 Lefebvre and Muysken (1988:118–40) argue at length that such non-marked Ps are in fact realizations of the objective case within the [+N] domain of assignment, while Ps with the suffix -ta (in the accusative case) are realizations of the objective case within the [−N] domain of assignment. However, it is still unclear to what degree such constructions are reminiscent of non-derived NPs. For the A there is a choice between the nominative and the genitive case (see below). In all of the Daghestanian languages of the Lezghian group, except Budux, genitive dependents are allowed within ANCs. Intransitive ANCs are easiest to assimilate to non-derived NPs. Thus, in the Djubek dialect of Tabassaran, intransitive ANs with the suffix -val combine with the Ss in the genitive case (while constructions with -uv- ANs retain sentential dependent-marking). For Archi, see section 5.1. In transitive ANCs with one argument present, this argument may sometimes be put in the genitive. In this case, it is most often the A. ‘Objective’ genitives are extremely rare in Daghestanian languages and occur only in Tsaxur and Rutul (Klimov and Alekseev 1980:234). In Cuzco Quechua, the S and the A in ANCs may be put in the genitive case. The difference between the nominative and the genitive S/A is partly determined by the syntactic function of the ANC. Thus, when it appears in the subject position, only the genitive marking is possible for its S or A, while otherwise both types of marking are allowed (Lefebvre and Muysken 1988, Muysken 1989). Especially interesting is the fact that the A in the genitive case cannot combine with the P in the accusative case (i.e., Cuzco Quechua lacks ANCs of the POSS-ACC type), that is, retention of the sentential dependentmarking for the P implies retention of the sentential dependent-marking for the A. In Basque, at least in northern dialects, definite Ps of ANCs may appear in the genitive case in the absence of an overt A (N’Diaye 1970:141, Saltarelli 1988:155). In a few cases, ANs acquire more concrete senses, which are accompanied by putting their Ss or As in the genitive case, as in Korean (see section 5.1.). All the data described in the present chapter are summarized in Table 5.1. The table includes the following columns for each language: 1 basic word order in finite clauses (FC) and NPs (NP); 2 case marking system—nominative-accusative (ACC), ergative (ERG), split ergativity (SPL), ‘special’ systems (SPEC, in the case of Tagalog and Yukagir), non-distinction of the S, A and P (‘−’)
< previous page
page_105
next page >
< previous page Page 106 Table 5.1 Languages of the SENT type Language Word order Case marking system Agul SOV/GN ERG ?Classical V-1/NG ACC Arabic Archi SOV/GN ERG
Balanced complements −a +
Basque
SOV/GN
ERG
+
Burushaski
SOV/GN
SPL b
(+)c
Ingush Kobon Korean
SOV/GN ERG SOV/GNGprACCpr SOV/GN ACC
(+)g (+)i −j
Classical Mongolian Nivkh Quechua, Cuzco Quechua, Huallaga Quechua, Imbabura Tabassaran
SOV/GN
ACC
−k
SOV/GN SOV/GN
− ACC
(+)e (+)l
SOV/GN
ACC
−k
SOV/GN
ACC
−m
SOV/GN
ERG
(+)a,n
Tamil
SOV/GN
ACC
−p
−a
Head-marking Other Other features FC NP ANC patterns − − − 2 ACC- − − 2, 3A SAPpr ERG-SP − SP ASP NEG ADV RFL ERG- − − SAP ACC- −/(+) d − SAP ?h − ?h ACC-SA − − ADV −CRD − − − TNS POL VCE ADV −TOP − − − TNS ACC-P f − ACC- + SAP ACC- + SAP ACC-SA −
Pf saP 6A, 6B
TNS RFL −VLD
saP 6A
TNS
−
ERG- − SP o ACC-SA −
SP
TNS RFL −VLD −CRD
Tongan V-1/GprNG ERG/ACCpr + − − !Tuvaq SOV/GN ACC (+) ACC-SA + Key: ?—Rare pattern in the language !—pattern with problematic or indeterminate attribution Xpr—patterns relevant only for pronouns TNS, ASP—tense, aspect distinctions CRD—coordination of ANCs with non-derived NPs is allowed RFL—reflexive relations expressed as in finite clauses POL—politeness VLD—validator TOP—topic marker ADV, ADJ—adverbs, adjectives compatible with ANs NEG—negation expressed as in finite clauses −—the category absent from ANCs
< previous page
next page >
page_106
page_106
−
TNS ADV RFL NEG CRD
– 2pr, 3A sar 2 TNS
next page >
< previous page
page_107
next page >
Page 107 Notes: a On subordinate clauses in Daghestanian languages, Bokarev (1979:170) writes the following: ‘Due to the lack of special subjunctions (in Daghestanian languages), the content of subordinate clauses is expressed by means of participial, converb and other reduced clauses.’ b On split ergativity in Burushaski see section 5.3. c On noun clauses in Burushaski, Lorimer (1935:400) writes: ‘In general when one verb is dependent on another, the second is put in the Infinitive (whether in its simple form or in a case form) or in the Present Base+ar.’ d A set of kinship nouns uses special possessive affixes. e Verbs of cognition, perception and impersonal verbs of evaluation can take asyndetic finite subordinate clauses, that is, without any conjunctions. In these cases the subordinate finite verb is placed immediately before the matrix verb (Jaxontov 1981:48). f On object markers in Nivkh, see section 5.3. g In Ingush, ‘ordinary indicative verbs without overt markers of subordination can be used with verbs of saying (in alternation with subjunctive and quotative forms), with mott —‘‘seem”, and (in alternation with nominalization) lov —“want, would like”, while verbs of saying and asking take clauses followed by the quotative eanna—“having said”’ (Nichols forthcoming). h Very few verbs show agreement with the third person singular of the subject in the present (an accusative pattern), while about thirty verb roots agree in gender class with the nominative S or P (an ergative pattern) (Nichols forthcoming). I have no information about the head-marking possibilities of the corresponding ANs. i Verbs of perception can take finite clauses without conjunctions as complements, preceding or following the matrix verb (Davies 1981:27). j On subordinate clauses in Korean, Ramstedt (1968:223) writes: ‘The participles are used with all kinds of nouns but the language has a special inclination to use certain nouns with a very general sense (thing, fact, case, place, word, while, way, etc.) after the participles where European languages have subordinate clauses.’ Kim (1987:897) mentions the quotative complementizer ko which is preceded by finite verb forms. k My sources of information do not mention any finite subordinate clauses in the langauge. l Cuzco Quechua has non-nominalized (balanced) subordinate clauses followed by chay in the appropriate case or with the suffic -qa (Lefebvre and Muysken 1988:12), but ‘the primary way to form complement clauses is through morphological nominalization’ (Muysken 1989:630). m Both direct speech (quotes) and indirect speech in Imbabura Quechua can be rendered by finite clauses followed by the word nin—‘says’. The latter has acquired a function of ‘validator’, indicating the evidential status of the preceding clause, rather than being a superordinate verb of saying (Cole 1982:13–14). n At least some verbs, such as verbs of fearing and utterance, may take finite complements followed by the quotative k’urdi, such complements being either direct or indirect speech (Bergel’son et al. 1982). o On personal agreement in Tabassaran, see section 5.3. p In Tamil, finite clauses with the quotative particle— can indicate either direct or indirect speech (see section 2.2.1., also Asher 1982:3). q ANCs in Buryat and Selkup might also be classified similarly to Tuva (see the discussion in sections 5.2., 6.1. and 6.3.). r On the differences between the head-marking paradigm of ANCs in adverbial functions and of nonderived nouns in Tuva, see sections 5.3. and 10.1.
< previous page
page_107
next page >
< previous page
page_108
next page >
Page 108 and a combination of ergative and nominative-accusative patterns (A/E, in the case of Pukapukan); 3 use of balanced complements, primarily, when the arguments of complements are not coreferential with those of the matrix clauses (excluding cases of direct speech)—not possible (‘−’), restricted (‘(+)’), unrestricted (‘+’); 4 head-marking, which includes: a for finite clauses, both the type of head-marking (i.e., its absence (‘−’), nominative-accusative (ACC), ergative (ERG), splitergative (SPL)), and also which of the arguments are head-marked. Thus, ‘ACCSAP’ means ‘the same head-marking for the S and the A, as opposed to the P, which is also headmarked’; while ‘ACC-SA’ means ‘the same head-marking for the S and the A, as opposed to the P, which is not head-marked at all’; b for NPs, presence (‘+’) or absence (‘−’) of head-marking; c use of head-marking in ANCs—for each of the head-marked dependents S, A and P, the character of head-marking, whether it is sentential (symbolized by the capital letters S, A and P) or nominal (symbolized by the small letters s, a and p); 5 other nominalization patterns according to which the same AN can combine with its dependents (1— SENT, 2—POSS-ACC, 2′—antipassive, 3A—SENT-POSS, 3B—OBL-POSS, 3′—passive, 4A—DBL-POSS, 4B—POSS-ADN, 5—MIX, 6A—SENT-INC, 6B—POSS-INC, 6C—OBL-INC, 7—REL); 6 other features of ANCs. Table 5.1 and the similar ones in chapters 6–9 are followed by notes which are referred to by means of italicized letters. 5.6. SUMMARY 0 Definition of the SENT. Retention of the sentential dependent-marking for the Ss, As and Ps of ANCs. 1 Further syntactic devices: a head-marking. Almost complete absence of head-marking within NPs except for Classical Arabic, Cuzco Quechua and Huallaga Quechua. In ANCs, the tendency for personal agreement to follow the nominal pattern, and for class agreement and agreement with the P to follow the pattern of finite clauses; b word order. Predominating word order SOV/GN except for two V-1/NG-languages (Classical Arabic and Tongan);
< previous page
page_108
next page >
< previous page
page_109
next page >
Page 109 2 Retention of sentential features. Relatively common oppositions in tense, sentential negation, adverbial modifiers. 3 Nominal features. Inability of SENT-nominalizations to be coordinated with non-derived NPs as an example of the difference in the categorial status of the two constructions. 4 Other patterns. The POSS-ACC, ERG-POSS (SENT-POSS) and INC types. 5 Typological features of the languages. a both nominative-accusative and absolutive-ergative languages, one language with split ergativity (Burushaski) and one without dependent-marking within clauses (Nivkh); b predominantly complement-deranking languages, except Classical Arabic, Tongan and Basque; c preference for predominating dependent-marking: with no head-marking (Korean, Tongan and Classical Mongolian); with head-marking on the clause level (Basque, Burushaski, Daghestanian languages, Ingush, Imbabura Quechua and Tamil); four double-marking languages (Cuzco and Huallaga Quechua, Classical Arabic and, probably, Tuva), and two examples of languages without dependent-marking at all (Nivkh) or with dependent-marking restricted to pronouns (Kobon). 6 Genetic and areal affinity of the languages. Dravidian, Caucasian, Andean Equatorial, Afro-Asiatic, Austronesian, Altaic and Indo-Pacific families. Three language isolates—Basque, Burushaski and Nivkh. ‘The Indo-Asian Language Aria’ (Masica 1976), as well as other regions—South America and New Guinea.
< previous page
page_109
next page >
< previous page
page_110
next page >
Page 110 6 Type 2: ‘Possessive-Accusative’ 6.0. INTRODUCTION In ANCs of the Possessive-Accusative (POSS-ACC) type, S and A genitivize, while the P retains the sentential dependent-marking. The definition above covers two cases—pure POSS-ACC ANCs and anti-passive ANCs. The head of pure POSS-ACC nominalizations is non-marked with respect to voice. In anti-passive ANCs, the head contains an overt anti-passive marker. Anti-passive ANCs are outside the general classification of ANCs, because, in this case, nominalization involves either only intransitive verbs or lexically transitive verbs which have been intransitivized. In other words, such ANCs do not present direct correlates to finite clauses with transitive verb forms as the head. 6.1. ‘IDEAL’ REPRESENTATIVES: SELKUP AND WEST GREENLANDIC Both Selkup and West Greenlandic, SOV/GN-languages, have ANCs of the POSS-ACC type, but these constructions correspond, morphologically, to different finite clauses. In Selkup, a nominative-accusative language, verbal nouns in POSS-ACC nominalizations have no voice markers, while those in West Greenlandic, an ergative language, contain the anti-passive suffix and are, thus, correlated to (marked) anti-passive finite clauses. The former situation, or the pure POSS-ACC type, is characteristic for many languages in the sample, the latter, the anti-passive type, is found only in two varieties of Inuit, West Greenlandic and Inuktitut. In Selkup (Kuznecova et al. 1980), the subject of independent clauses is in the nominative, the nonmarked form, while the direct object is put in the accusative or in the nominative, if no misunder-
< previous page
page_110
next page >
< previous page
page_111
next page >
Page 111 standing can arise in interpreting its function (i.e., whether it is the subject or the direct object).1 Verbs agree with their subjects in person and number. In Selkup, there are two types of verbal conjugation— Subjective and Objective. Intransitive verbs are always conjugated according to the Subjective conjugation, while transitive verbs can choose between the two, the choice depending on the character of their direct object. Thus, if the direct object is a pronoun of the first or second person or if it is indefinite, the Subjective conjugation is chosen. In most other cases, transitive verbs are conjugated according to the Objective conjugation and are, thus, in a sense head-marked for their direct object as well. Word order is rather free with the basic order being S(IO)OV. (6.1) tātyŋyty big wood.grouse:NOM brought:3SG.OBJC ‘She brought a/the big wood-grouse.’ (Kuznecova et al. 1980:385) (6.2) big wood.grouse-ACC killed:2SG.OBJC ‘You killed a/the big wood-grouse.’ (ibid.) (6.3) he:NOM you:ACC see:FUT:3SG.SUBJC ‘He will see you.’ (ibid.: 235) (6.4) kana-p he:NOM dog-ACC see:FUT:3SG.OBJC ‘He will see this dog.’ (ibid.) (6.5) kana-p he:NOM dog-ACC see:FUT:3SG.SUBJC ‘He will see some dog.’ (ibid.) In non-derived NPs, both dependent-marking and head-marking are used, though not always together. As regards dependent-marking, the accusative is excluded. The usual case for adnominal dependents (including the G) is the genitive. The nominative is not, however, excluded from NPs and in some cases it appears where otherwise the genitive would be expected, especially when the G is expressed by a proper name and ‘a possibility of mixing up the attribute and the head
< previous page
page_111
next page >
< previous page
page_112
next page >
Page 112 is not great’ (Kuznecova et al. 1980:378). As regards head-marking, Selkup has a set of possessive suffixes which are very similar to the verbal personal suffixes in the Objective conjugation. When the G is the first or the second person, the possessive suffixes can be duplicated by the appropriate independent personal pronouns in the genitive, i.e., double-marking is allowed. In the third person, such a situation is avoided—that is, either the H bears a possessive suffix or the G is explicitly named by a noun or a pronoun in the genitive. Dependents precede their heads. (6.6b) (mat) I:NOM/GEN brother-(NOM)1SG.POSS ‘my brother’ (6.7a) pō-t pīry (6.7b) *pō-t pīry-ty tree-GEN height tree-GEN height-3SG.POSS ‘the height of the tree’ (Kuznecova et al. 1980:187) (6.8) salaqoj Salaqoj:NOM house ‘Salaqoj’s house’ (ibid.: 378) ANCs have verbal nouns with the suffix -ptä (and its variants) as their head. The P retains its sentential dependent-marking, appearing either in the nominative or in the accusative, while the S and the A are put in the genitive, like the G in NPs. With respect to head-marking, ANCs totally assimilate to possessive NPs, which means that their head bears possessive suffixes. Though Kuznecova et al. (1980) write that possessive suffixes appear only in the absence of an overt third person S or A in the genitive, examples (6.9b) and (6.10a) below show that there exist cases where double-marking occurs even in ANCs with the third person S/A. Also, like the G in non-derived NPs, the S/A may appear in the nonmarked form, as in example (6.10b). In the absence of any indications to the contrary, I assume that, on the whole, the variation in the form of the S/A is identical to that of the G and hence propose to subsume all such ANCs in Selkup under the POSS-ACC type. (6.9a) [Qenty-ptä-n-y] kunty cunty-s-ak go-AN-GEN-1SG.POSS during shoot-PAST-1SG ‘I was shooting when I was going.’
< previous page
page_112
next page >
< previous page
page_113
next page >
Page 113 (6.9b) [ījat-y-n pisyc-cä-n-tyt] child-PL-GEN laugh-AN-GEN-3PL.POSS for ‘for the sake of the children’s laughter’ (Kuznecova et al. 1980:251) (6.10a) Mat ašša [ I:NOM NEG know-PAST-1SG brother-GEN-1SG.POSS ima-p ] wife-ACC find-AN-NOM-3SG.POSS ‘I did not know that my brother had married’ (lit. ‘my brother’s finding a wife’). (ibid.: 173) (6.10b) [Losy ira-p tü amty-ptä-qyn-ty] devil old.man-ACC fire eat-AN-ABL-3SG.POSS ‘When fire was eating the old man devil…’ (Tereščenko 1973:301) West Greenlandic (Fortescue 1984), a consistent double-marking ergative language, resorts to several other types of constructions with transitive verbs (i.e., anti-passive, or ‘half-transitive’, according to Fortescue’s terminology, and passive), besides ergative sentences. In finite clauses of the ergative pattern, the S and the P are in the absolutive case, while the A is in the relative case. In ‘halftransitive’ (anti-passive) constructions, where ‘half-transitivizing’ affixes are attached to the verb, the A is in the absolutive and the P is in the instrumental case. In passive constructions, the P is in the absolutive and the A is in the ablative. Intransitive verbs agree with their Ss, transitive verbs in an ergative construction agree with both their As and Ps. The sets of the endings for intransitive and transitive verbs are formally very far from each other. In ‘half-transitive’ and passive constructions, the verbs are treated as intransitive and thus agree only with their Ps. The most neutral word order is SOV. (6.11a) inuit tuqup-pai people:ABS kill-3SG.3PL.INDIC ‘He killed the people.’ (6.11b) inun-nik tuqut-si-vuq people-INSTR kill-1/2TRANS-3SG.INDIC ‘He killed people.’ (Fortescue 1984:86)
< previous page
page_113
next page >
< previous page
page_114
next page >
Page 114 (6.12a) inuit nanuq taku-aat people:REL bear:NOM see-3PL.3SG.INDIC ‘The people saw the polar bear.’ (6.12b) nanuq (inun-nit) taku-niqar-puq bear:ABS (people-ABL) see-PASS-3SG.INDIC ‘The polar bear was seen (by the people).’ (ibid.: 265) In non-derived NPs, both the relative and the absolutive cases are allowed. In their head-marking, NPs are also rather close to independent clauses (see section 3.1. for discussion of the Eskimo languages). Gs are in the relative case and precede their heads. Since there is no distinction between adjectives and nouns, juxtaposition of simple nominals is possible for expression of material and some other semantic relations. The H bears possessive suffixes which agree with the G in the relative (when it is present) and are identical to the verbal suffixes referring to the A when the P is in the third person singular or plural. (6.13) (uanga) illuga I:REL house:1SG.POSS ‘my house’ (6.14) Isurtu-p Uummanna-a Isortoq-REL Uummannaq-3SG.POSS ‘Isortoq’s Uummanaq’ (lit. ‘heart-shaped mountain’) (Fortescue 1984:109) ANCs contain verbal nouns with the suffix -niq as their head. Intransitive ANCs assimilate to possessive NPs (example 6.15). If an action nominal is derived from a transitive verb, there are several possibilities. When only its P is present, the action nominal is derived from the underlying transitive verbs either without any intransitivizing affixes (example 6.16a) or with the passive suffix (example 6.16b). The P genitivizes (appearing in the relative case) and the AN bears the corresponding possessive suffix. However, if both arguments are present, the action nominal is derived from the anti-passive variant of the underlying verb. In this case, the A genitivizes, appearing in the relative, and the action nominal (with the ‘half-transitivizing’ affix) bears the corresponding possessive suffix, while the P is in the instrumental, as it is in the corresponding anti-passive finite clause (see examples 6.16c and 6.17). Thus, we see that the G in an action nominal construction always corresponds to the absolutive argument in the
< previous page
page_114
next page >
< previous page
page_115
next page >
Page 115 corresponding finite clause. Strictly speaking, West Greenlandic has no transitive ANCs. (6.15) [umiarsu-up qassi-nut tikin-ni-ssa-a] ship-REL how.many-ALL arrive-AN-FUT-3SG.POSS nalunngil-ara know-1SG.3SG:INDIC ‘I know when the ship will arrive.’ (Fortescue 1980:45) (6.16a) angutip tuqunnira man:REL kill:AN:3SG.POSS ‘the killing of the man’ (i.e., ‘he is killed’) (ibid.: 46) (6.16b) nalu-aa qinnuta-ata not.know-3SG:3SG.INDIC request-3SG.POSS.REL qanuq [naammassi-niqar-ni-ssa-a] how implement-PASS-AN-FUT-3SG.POSS ‘She didn’t know how his request would be implemented.’ (ibid.: 45) (6.16c) angutip tuqutsinira man:REL kill:1/2TRANS:AN:3SG.POSS ‘the man’s killing’ (i.e., ‘he is killed’) (ibid.: 46) (6.17) [ikinngum-mi-nik tuqut-si-nir-a] friend-his.RFL-INSTR kill-1/2TRANS-AN-3SG.POSS tusar-para hear-1SG.3PL.INDIC ‘I heard of his killing his friend.’ (ibid.: 45) ANCs in West Greenlandic have no mood oppositions typical for finite clauses. However, both tense and modality (expressed optionally by means of verbal affixes which, in finite clauses, are followed by obligatory mood affixes) can be retained in them, though these categories are not generally found in nominalizations. As regards tense (at least future), action nominals can choose between the verbal future affix -ssa which precedes the nominalizing suffix -niq, and the nominal future affix -ssaq which follows it. The latter is used in possessive noun phrases to refer to future possession. My source of information does not mention anything about the relation between the
< previous page
page_115
next page >
< previous page
page_116
next page >
Page 116 corresponding verbal and nominal past suffixes. Negation is retained in ANCs as well as all adverbials. 6.2. PROBLEMS IN DEFINING THE PATTERN The definition of the POSS-ACC type, as given above—i.e., the S and the A genitivize, while the P retains its sentential marking—presupposes that there is some difference between the marking of Gs and that of Ss, As and Ps in finite clauses. In a number of languages, however, this distinction either does not exist at all or is very inconspicuous, making their ANCs difficult to classify. Some examples of such difficult cases are presented below. 6.2.1. Languages without any dependent-marking of Gs in possessive NPs, and of Ss and As in finite clauses Modern Israeli Hebrew (in most cases referred to as ‘Hebrew’ in this book) will be described in more detail in the next chapter, see section 7.2.2. and examples (7.35–7.40). It uses the same non-marked form for the subjects of finite clauses, indefinite direct objects and Gs (in one of its three different possessive NPs). The H is put in the Construct State and is followed by the G in the non-marked form. ANCs in Hebrew manifest two (or, in certain dialects, three) patterns. According to one of them, applying only to definite Ps, the action nominal is in the Construct State and immediately precedes the S or the A in the non-marked form, while the P retains its sentential form, being introduced by the preposition et. Since the S and the A of ANCs are in the same non-marked form as in the corresponding finite clauses, it would be possible to qualify the ANC itself as belonging to the SENT type, described in the preceding chapter, or just state that the distinction between the SENT and the POSS-ACC types is not relevant for Hebrew. However, I would prefer classifying the Hebrew pattern under consideration as belonging to the POSS-ACC type. The following arguments seem to favour this decision: (1) Pronouns distinguish between the possessive form (possessive affixes) and the subject form (independent pronouns). In ANCs, pronominal Ss and As appear as possessive suffixes on the action nominal and not as independent pronouns, which means that ANCs with pronominal dependents are unambiguously classified as following the POSS-ACC pattern. However, in many languages, personal pronouns show other formal oppositions than ordinary nouns, and, moreover, ANCs with pronominal arguments can follow other patterns than those with
< previous page
page_116
next page >
< previous page
page_117
next page >
Page 117 lexical nouns (cf. section 10.2.). Therefore, the existence of the special possessive pronominal suffixes in Hebrew need not mean that ANCs with lexical nouns should be classified as belonging to the POSSACC type. This argument is, however, further strengthened by the next one. (2) Hebrew has not one, but three types of possessive phrases with lexical Gs (they are described in section 7.2.2.). In the other two types, the G is introduced by the preposition . In most cases, all three constructions are in free variation. What is relevant for the present issue is the fact that the preposition can only be used within NPs for introducing the G, the same applying to ANCs. This means that in those cases where there is a formal opposition in the dependent-marking of Gs and the subjects of finite clauses (Ss and As), ANCs treat their Ss and As as Gs. These two considerations can serve as arguments for classifying Hebrew ANCs with definite Ps as following the POSS-ACC type. The other variant of Hebrew ANCs follows the ERG-POSS type described in chapter 7 (some dialects allow also for DBL-POSS ANCs, see chapter 9). Egyptian Arabic presents a still more complicated situation. Within the class of nouns, the language does not differentiate at all among the forms functioning as subjects, direct objects and Gs. Indirect objects following direct objects are expressed by prepositional phrases with the preposition li . However, if they precede direct objects, they do not normally have li and are thus in the same form as both subjects and direct objects. In possessive NPs, the H is in the Construct State, exactly as in Hebrew, but this is the only type of possessive NPs in the language. As regards pronouns, there is an opposition between independent pronouns used as subjects of finite clauses, and two almost identical sets of bound affixes—those used as nominal possessive suffixes and those used as verbal object suffixes. In ANCs with lexical arguments, the S, A and P are in the non-marked form, the AN itself is in the Construct State and the A precedes the P.2 In constructions with pronominal arguments, the AN bears possessive suffixes which refer to the S or the A, while the P is expressed by a special form of the corresponding independent pronoun. Thus, ANCs with pronominal arguments belong to the POSS-ACC type. As for ANCs with lexical arguments, the question as to which type they belong might be considered irrelevant. Note that Classical Arabic, which has distinct cases for subjects, objects and Gs, has ANCs of both the SENT and the POSS-ACC types, though the former are very seldom used. Similar problems arise also in Evenki and Yukagir. Evenki resorts
< previous page
page_117
next page >
< previous page
page_118
next page >
Page 118 to non-marked forms for expressing the S and A in finite clauses, as well as the G in NPs. The H of NPs bears possessive suffixes, whether an overt external G is present in the same NP, or not. In ANCs, the verbal noun appears with the same possessive suffixes, while the S and A turn out in the non-marked form. In Yukagir, the exact expression of the A, S and P in finite clauses depends on their pragmatic role (focus vs. non-focus) and on their deictic characteristics (first or second person vs. all the other persons) (Maslova 1989). Most As and non-focal Ss appear in the non-marked form which is identical to the form of the G in NPs, as in (6.18), and only third person pronouns have a distinct form with the suffix -l when they function as non-focal As and non-focal Ss (cf. (6.18b) and (6.18d)). Note that in the examples below the non-marked form is glossed differently for different classes of NPs due to its different status in their case systems; thus, for third person pronouns it is glossed as ‘ergative’: (6.18a) I-NOM V.FOC-come-1SG ‘I (S NONFOC) came.’ (6.18b) I-LOC he-ERG call-SG ‘He (A FOC) called me.’ (6.18c) mat-ul you-NOM I-ACC V.FOC-see-FUT-2PL ‘You (A NONFOC) will see me (P NONFOC).’ (6.18d) he-ABS I-LOC V.FOC-see-3SG ‘He saw me.’ (Maslova 1989:10) (6.18e) kelu-j old.mani-NONM hei-ERG house-LOC come-3SG ‘The old mani came into hisi house.’ (Maslova personal communication) Most nouns have a distinct form for focal Ss and Ps, while the exact expression of non-focal Ps depends on the type of the A in the same clause. In ANCs, the Ss and the As appear in the non-marked form and look like the Gs in possessive NPs, while the P has the same form as the non-focal P in certain types of finite clauses. Note, however, that the only dependent which clearly distinguishes between finite
< previous page
page_118
next page >
< previous page
page_119
next page >
Page 119 clauses and ANCs is the S expressed by the third person pronoun (in finite clauses it appears with the suffix -l). There is, however, another feature which ANCs have in common with possessive NPs, in contrast to finite clauses, namely head-marking. Nouns may take the possessive suffix -gi referring to the third person G (normally, when the G is not coreferential to the A or S of the clause), and exactly the same suffix appears in ANCs. Thus, in (6.19a) the possessive suffix on the AN shows that its A cannot be coreferential with the S of the matrix clause, while the use of the pronoun in the ANC of (6.19b) indicates that its S is coreferential with the (deleted) S of the matrix clause. Compare this with the possessive NPs in (6.18e) and (6.19d). (6.19a) [ ] hunter-PL-NONM see-AN-PL-3i.POSS-ATTR before hej-ABS V.FOC-run-3SG ‘Before the hunters saw (him), he ran away.’ (6.19b) [ ] kelu-j he-NONM work-AN-ABL come-3SG ‘He came from his work.’ (6.19c) mit-ul lew-l-ŋin or we-ACC eat-AN-DAT be.tormented-CAUS-PRES-3SG ‘Or is he tormenting (us) in order to eat us?’ (6.19d) āmā-gi father-3SG.POSS V.FOC-come-3SG ‘His father came.’ (Maslova personal communication) Thus, those Yukagir ANCs for which the distinction between the SENT and the POSS-ACC types makes sense are unambiguously attributed to the POSS-ACC type, while for most other ANCs in Yukagir this distinction is neutralized. On problems with attribution of ANCs in Tuva see section 5.2. 6.2.2. Languages with the same dependent-marking of Gs in possessive NPs and of Ps in finite clauses Tagalog (Schachter and Otanes 1972) offers an example of this case. In finite clauses, one of the arguments of the verb is chosen as the so-called topic of the clause (although many linguists apply the term
< previous page
page_119
next page >
< previous page
page_120
next page >
Page 120 ‘focus’ to the same entity), it appears in a special form (as an ang- phrase) and the verb takes a special affix depending on the semantic role of its topic. The arguments may be in any one of at least nine different kinds of semantic relations to the verb, the most important among them being those of actor, object and directional. Both actor complements and object complements (non-topics) are expressed by ng -phrases, while directionals (non-topics) are expressed by sa- phrases. In possessive NPs, there are at least two possibilities. Either the G is expressed by a ng -phrase following the H (the more common type of possessive NPs), or the G is a sa-phrase either preceding or following the H. With pronominal Gs, both constructions are equally common. ANCs have gerunds (based on actor-focused verbs) as the head. The S and the A of these constructions genitivize,3 and the ANC may also contain any other complements allowed in the corresponding finite clause, for example, the P expressed by a ng -phrase. The problem is that both the P in actor-focused finite clauses and the G in non-derived NPs appear as ng -phrases and the type of Tagalog ANCs, whether they are of the POSS-ACC type (if the P retains its sentential dependent-marking), or whether they are of the DBL-POSS type (if the P genitivizes) is indeterminate. Luckily, however, the existence of the two types of possessive NPs in Tagalog helps to differentiate the A or S and the P in ANCs, favouring the point of view according to which ANCs in Tagalog belong to the POSS-ACC type rather than to the DBL-POSS type. While the A and S in ANCs can be expressed by sa-phrases, exactly like Gs, this is not allowed for the Ps, in accordance with their behaviour in finite clauses. This means that Ps in ANCs do not genitivize, but retain their sentential marking. Still other problems are presented by Ifè and some other SVO/NG Kwa-languages, which use the same non-inflected form for the S, A, P (both in finite clauses and in ANCs) and the G. The problem is that in ANCs, the P precedes the verbal noun, in opposition both to the word order in finite clauses (SVO) and to the head-dependent word order in NPs (NG), which causes difficulties in determining whether such ANCs follow the POSS-ACC or the INC pattern. This issue is discussed in more detail in section 9.2. 6.3. HEAD-MARKING IN THE POSS-ACC TYPE In the SENT type discussed in the preceding chapter, head-marking in NPs is used only in Cuzco and Huallaga Quechua, Classical Arabic and Tuva. Among the POSS-ACC languages, head-marking in NPs is much more frequent. This is true for eleven of the twenty-five pure
< previous page
page_120
next page >
< previous page
page_121
next page >
Page 121 POSS-ACC languages in the sample, representing Afro-Asiatic, Altaic, Gum, Niger-Congo and Uralic families as well as Yukagir, and for the only language with anti-passive ANCs, Inuit. Modern Mongolian4 has reflexive possessive suffixes. Head-marking on the clause level, that is, in finite clauses, occurs in fifteen languages representing seven language families. ANCs of the POSS-ACC type tend to use head-marking in accordance with the structure of non-derived possessive NPs. In this respect, Bantu languages constitute an exception, as action nominals in them can retain the agreement with the class of the P characteristic for finite verbs, while non-derived possessive NPs have only dependent-marking. The situation on the whole is similar to that of the Daghestanian languages discussed in section 5.3. The much greater use of head-marking and, in particular, the existence of possessive affixes (which normally implies the existence of verbal agreement in the same language, according to Nichols’ implication, see section 5.3.) in the languages of the POSS-ACC type as compared to those of the SENT type, seems to be a relevant factor in the choice of nominalization pattern. In languages with headmarking in NPs, the ability to be head-marked in a certain way is one of the nominal inflectional categories and could, probably, even be included among the language specific criteria for identifying ANs. Cuzco and Huallaga Quechua, double-marking languages described in section 5.3., are examples of languages in which the sentential dependent-marking in ANCs comes into conflict with the nominal head-marking. In their ANCs, both the S, A and P retain sentential dependent-marking, while the head of the constructions takes possessive affixes referring to the S or the A. Thus, there arises a conflict in the status of the A or S. In dependent-marking, they are treated as verbal dependents, but according to the head-marking of the verbal noun, they are nominal dependents. Evidence presented in this book actually demonstrates that action nominals can combine both verbal and nominal features, which means, among other things, that some of their arguments are treated as sentential dependents, while others are treated as nominal dependents. However, these discrepancies concern different arguments. In the case of Cuzco and Huallaga Quechua, it is the same argument that is treated in the two different ways in one and the same ANC. Probably, for these Quechua languages, this conflict is not felt to be great, because the possessive affixes are almost identical to the verbal subject affixes. However, these are the only languages in the sample that manifest such a discrepancy between the head-marking and dependent-marking of arguments in ANCs.
< previous page
page_121
next page >
< previous page
page_122
next page >
Page 122 Buryat and Selkup present another case of interdependence and even complementary distribution of head-marking and dependent-marking within ANCs. The head of ANCs may take possessive suffixes which refer to its A or S in the genitive case or in the non-marked form. However, possessive suffixes and the genitive case co-occur relatively rarely (for example, to emphasize the A/S) and, normally, ANCs resort to only one of these syntactic means. Thus, when the A/S is in the first or second person, it is usually expressed by the corresponding possessive suffixes. In the third person, the AN in Buryat is often not head-marked (Skribnik 1988:29–30), while in Selkup either the AN is not head-marked or the S/A is in the non-marked form (see section 6.1.). Since this variation in the expression of the S and A in ANCs is parallel to the variation in the expression of the G in non-derived NPs, I choose to consider all these ANCs in Buryat and Selkup as variants of the POSS-ACC type (compare this with the Tuva case discussed in section 5.3.). Some words should be said about the formal characteristics of head-marking in finite clauses and NPs. As described in section 6.1., the possessive suffixes in Selkup are very similar to the verbal subject suffixes. The same is true for Turkish, Tuva and Buryat and, with some modifications, for Inuit (the possessive suffixes are identical to the verbal suffixes referring to the A when the P is in the third person singular or plural). On the other hand, in the Semitic languages, the possessive suffixes are almost identical to the object suffixes which appear in a verb with pronominal objects (in the absence of overt external objects). Similarities of the latter sort are more frequent in the languages with nominalizations of the ERG-POSS type (see section 7.4.). 6.4. FURTHER CHARACTERISTICS OF THE POSS-ACC TYPE In the main, languages with pure POSS-ACC ANCs are nominative-accusative languages, the only exception being Agul (and other Daghestanian languages of the Lezghian group), in which POSS-ACC ANCs coexist with SENT ANCs. In Inuit, an ergative language, nominalizations correspond morphologically not to the unmarked, ergative sentential constructions, but to the derived anti-passive (‘half-transitive’) finite clauses. Pure POSS-ACC nominalizations in Inuit could not be distinguished from SENT nominalizations, as they would contain the A in the relative case (which is used both
< previous page
page_122
next page >
< previous page
page_123
next page >
Page 123 for Gs and for agents in finite ergative clauses) and the P in the absolutive case (the non-marked form). Tongan, a West Polynesian ergative language, does have a near approximation to the pure POSS-ACC nominalizations, but only with the pronominal As. (ANCs with pronominal arguments, which, in a number of languages, follow other nominalization patterns than ANCs with lexical arguments, will be dealt with in section 10.2.) Pukapukan, a Polynesian language of the Samoic Outlier group, which has POSS-ACC ANCs, builds finite clauses according to both the nominative-accusative and the ergative patterns. Thus, no ergative language has pure POSS-ACC nominalizations as its only nominalization pattern. As regards the word order in the POSS-ACC languages, there is more variation than among the SENT languages. Thus, the following word order types are represented here: V-initial/NG, SVO/NG and SOV/GN. English, an SVO-language, has both preposed and postposed Gs. The unclear status of ANCs in Ifè, an SVO/NG language, was mentioned in section 6.2.2. Among the languages of the group under consideration there are both complement-balancing and (predominantly) complement-deranking languages. For the Bantu languages, the situation is rather complicated and different languages seem to have principally different types of complement systems. Thus, while both Luganda and Swahili allow the indicative mood in their subordinate clauses, Nkore-Kiga (Taylor 1985:15) and Southern Sotho (see section 2.2.2.) are complement-deranking languages. A number of languages retain certain sentential features in ANCs of the POSS-ACC type, such as tense or aspect (Inuit, Buryat, Modern Mongolian, Turkish, Tuva and Tagalog) and the active/ passive opposition signalled by overt voice markers (Amharic, Turkish, the Bantu languages and Maori). As mentioned in section 6.1., in the case of West Greenlandic, it is not obvious whether the category of tense in ANCs is a sentential or a nominal property. Since many languages either lack any distinction between adverbs or adjectives or have a minimal one, it is not always clear whether nominalizations in their use of adverbs vs. adjectives are closer to finite clauses or to ordinary NPs. Those languages which have a regular means of deriving adverbs from adjectives or vice versa can behave in different ways in nominalizations. Thus, Turkish (Kari Fraurud personal communication) allows adverbs (where they are distinct from adjectives, being derived from them by reduplication),
< previous page
page_123
next page >
< previous page
page_124
next page >
Page 124 Table 6.1 Languages of the POSS-ACC type (see comments to Table 5.1 in section 5.5., pp. 105–8) Language Word Case marking Balanced Head-marking Other Other features order system complements FC NP ANC patterns Pure Agul SOV/GN ERG −a − − − 1 Amele SOV/GN − + ACC- −/+b P SAP Amharic SOV/GN ACC + ACC- +pr c sapr 3′ VCE SAP Arabic, Classical V-1/NG ACC + ACC- + saPpr1, 3A SAPpr Arabic, Egyptian V-1/NG − + ACC- + sa 4B ADJ ADV SAPpr NEGAN Buryat SOV/GN ACC (+) ACC- + sa ?d SA English SVO/GNGACCpr + ACC- − − 3A, 3′, ADV ASP NEG SA 4A Evenki SOV/GN ACC (+)e ACC- + sa TNS SA Fula SVO/NG − + ACC- − Ppr SAPpr Hebrew, Modern SVO/NG ACC + ACC- +f sa 3A, ?4A ADJ NEGAN SA ADV CRD Ifè SVO/NG − − − − − Maori VACC −g − − − 3A, 3′, 5 NEGNP 1/GprNG Mongolian, Modern SOV/GN ACC (+)h − + sa SPEC i ADV TNS ?Polish SVO/*NG ACC + ACC- − − 3B, 4A* RFL SA Pukapukan VA/E + − − − 3B 1/GprNG Selkup SOV/GN ACC (+)j ACC- + sa ?d SA k Swahili SVO/NG − + ACC- − P 3′ NEG VCE RFL SAP Tagalog VSPEC + − − − 1/GprNG Tahitian VACC −g − − − 3A 1/GprNG Thai SVO/NG − + − − − 5 Turkish SOV/GN ACC (+)l ACC- + sa 3′ TNS ADV NEG SA VCE −CRD Tuva SOV/GN ACC (+) ACC- + sa !1 TNS SA Vata SOV/GN − + − − − Wikchamni SOV/GN ACC − − − − Yukagir SOV/GN SPEC − SPEC + sa Restricted to pronominal As Egyptian, Ancient V-1/NG − + ACC- + sa 3, 5pr SAPpr
< previous page
page_124
next page >
< previous page
page_125
next page >
Page 125 Tongan V-1/GprNG ERG + − − − 1, 3A Anti-passive ANCs Inuit SOV/GN ERG − ERG-SAP ACC-SA + sa 3′ TNS m Key: ?—Rare pattern in the language !—pattern with problematic or indeterminate attribution Xpr—patterns relevant only for pronouns X*—patterns relevant only for highly referential nominals (in addition to pronouns) *NG—besides the usual NG-order in NPs, there are possessive NPs with preposed highly referential Gs (including possessive pronouns). TNS, ASP—tense, aspect distinctions CRD—coordination of ANCs with non-derived NPs is allowed RFL—reflexive relations expressed as in finite clauses POL—politeness VLD—validator TOP—topic market ADV, ADJ—adverbs, adjectives compatible with ANs NEG/NEGNP—negation expressed as in finite clauses/as in non-derived NPs NEGAN—special negation in ANCs −—the category absent from ANCs Notes: a See footnote a to Table 5.1., p. 107. b Only inalienable nouns are head-marked for the possessor. c In Amharic, independent possessive pronouns are in free variation with possessive suffixes on the head. d See the discussion in section 6.3. e In Evenki, finite subordinate clauses (which have developed under the influence of Russian) can be used as complements of modal verbs, verbs of perception, cognition and utterance. This use is, though, relatively rare (Nedjalkov 1983). f In Hebrew, possessive NPs show three different patterns, in one of which the H is not marked for the G. All the three are exploited in ANCs (see section 7.2.2.). g Some East Polynesian languages, among others Maori (Chung 1978:21) and Tahitian (Koptjevskaja 1984), distinguish between tense-aspect particles which are used in independent vs. subordinate clauses, so that the two types of clauses can differ as to which tense-aspect oppositions are manifested in them. h Indirect speech can be rendered by finite subordinate clauses, see section 6.5. i The S and the A may appear in the accusative case (see section 6.5.). j Selkup, under the influence of Russian, resorts now to finite subordinate clauses, which, however, do not have any ‘stable’ structure (Kuznecova et al. 1980:406). k For the special character of object agreement in Selkup, see section 6.1. l In Turkish, finite clauses can sometimes function ‘as a noun clause or adjectival clause within a longer sentence’, especially in colloquial speech (Lewis 1967:258). Such clauses can sometimes take case endings, that is, become clausal nominalizations (ibid.: 258). With verbs of perception and ‘thinking’, asyndetic subordination is usual (ibid.: 274). Turkish has borrowed the Persian subordinate conjunction ki, which ‘opened the door to the Indo-European pattern of sentence’ (ibid.: 211). m For the dual character of tense distinction in Eskimo, see section 6.1.
< previous page
page_125
next page >
< previous page
page_126
next page >
Page 126 while in Hebrew (Berman 1976, Sven Malkior personal communication) and in Egyptian Arabic (Wise 1975:81–5), different types of adverbs are treated in different ways—some are allowed, some must be substituted for by the corresponding adjectives, and some can be used interchangeably with the corresponding adjectives. Negation is sometimes expressed as in the corresponding finite clauses (in the Bantu languages, Turkish and Tuva), while, sometimes, it is identical to negation used with ordinary nouns (as in Bulgarian and East Polynesian). Hebrew has a special type of negation which applies only to verbal nouns. In Turkish (Kari Fraurud personal communication), verbal nouns cannot be coordinated with ordinary nouns, while those in Hebrew (Sven Malkior personal communication) can.5 All the data presented in this section are summarized in Table 6.1. 6.5. INTRA-LEXEMIC SYNTACTIC VARIATION In some POSS-ACC languages, a single verbal noun can be the head of other types of ANCs: SENT type: Agul and, probably, Tuva. In Tongan, ANCs with non-pronominal and pronominal Ss/As follow the SENT or the POSS-ACC pattern respectively. In Turkish, indefinite subjects in ANCs may appear in the non-marked form, as in the following example: (6.20) [kimse gir-me-me-si] için anybody:NONM enter-NEG-AN-3SG.POSS for kapıyı kilitledi… door:ACC locked:3SG ‘…he locked the door so that nobody would enter…’ (Nilsson 1985:80) The semantic distinction between the genitive case and the non-marked form in ANCs is relatively similar to (although not identical to) the corresponding distinction in non-derived NPs (Nilsson 1985:80– 6). Unfortunately, I lack detailed information on ANCs with non-marked arguments (among other things, it is not clear whether or not they are restricted to ANs derived from certain types of verbs, like existential verbs and verbs of motion). Pure ERG-POSS type: Classical Arabic and Tongan (SENT-POSS), English, Hebrew, Maori, Pukapukan and Polish. In Hebrew, the choice between the two patterns is connected to the referential characteristics of the P (see section 10.2.). In English, this variation
< previous page
page_126
next page >
< previous page
page_127
next page >
Page 127 appears as the choice between verbal gerunds, on the one hand, and nominal gerunds and verbal nouns, on the other hand. For the other languages with the same variation between the nominalization patterns, I am not aware of any factors which can influence the choice. Passive variety of the ERG-POSS type: Swahili (and Luganda), Turkish, Amharic, Inuit, Maori and English. DBL-POSS type: Modern Hebrew (substandard) and English. POSS-ADN type: Egyptian Arabic. MIX type: Maori, Thai and Ancient Egyptian (for ANCs with certain pronominal As, see section 10.2.). Modern Mongolian allows the S and A of a verbal noun to be either in the genitive or in the accusative. It is worth mentioning that the accusative can be used in other cases for marking the subject of a subordinate clause, more specifically, the subject of indirect discourse. The only difference between indirect and direct speech involves the subject, which can be pronominalized or deleted depending on coreferentiality with some NP in the matrix clause. When overt, the subject of indirect speech generally takes the accusative case. It would be tempting to try to find some difference between the genitive and accusative Ss and As of verbal nouns. Binnick quotes Bauwe, who stresses that ‘a genitive can only appear when the VN has lost its verbal character and become purely nominal’. He himself states that the discussion of the problem seems unclear to him, but ‘perhaps what is meant is that nominalized Ss (sentences) take genitival subjects, whereas VNs which do not have purely nominal functions do not’ (Binnick 1979:99). In Amele, the same nominalizing marker -ec/-oc is used both for deriving ANs and for marking indirect quotes, which in other respects have the structure of finite clauses (Roberts 1987:16). The main features of the POSS-ACC languages are summarized in Table 6.1 on p. 124. 6.6. SUMMARY 0 Definition of the POSS-ACC type. Genitivization of the S and the A, retention of the sentential marking by the P. Subdivision into the pure and the anti-passive subtypes. 1 Further syntactic devices: a head-marking. As compared to the SENT languages, more frequent head-marking within NPs. The tendency of ANCs to follow the nominal pattern, except for Bantu languages, which
< previous page
page_127
next page >
< previous page
page_128
next page >
Page 128 retain the verbal class agreement with the P; retention of pronominal object clitics by Classical Arabic and Amele; b word order. The identity of the AN-P word order in ANCs and of the H-G word order in non-derived NPs (with the possible exception of Ifè). 2 Retention of verbal features. Oppositions in voice, tense and aspect, sentential negation and compatibility with adverbial modifiers in POSS-ACC nominalizations of some languages. 3 Nominal features. Compatibility with adjectives, nominal negation and possibility of coordination with non-derived NPs in some languages, as well as special negation and non-ability to be coordinated with non-derived NPs in other languages. 4 Other patterns. SENT, ERG-POSS (SENT-POSS, OBL-POSS and passive patterns), NOMN (both DBLPOSS and POSS-ADN), MIX and INC. 5 Typological features of the languages: a for the pure POSS-ACC type, mostly nominative-accusative languages except for one ergative language and one example of a language with both nominative-accusative and ergative clauses (Pukapukan); b word orders SVO/NG, SOV/GN and V-1/NG. Some examples of preposed possessive pronouns; c both predominantly dependent-marking languages with no head-marking (Agul, Tagalog, Tahitian, Maori, Pukapukan, Wikchamni), with little head-marking (Modern Mongolian, Polish, English), with split-marking (Bantu languages), as well as double-marking (Amharic, Buryat, Classical Arabic, Selkup, Turkish and Tuva). Two examples of languages with double-marking on clause level and head-marking on phrase level (Evenki and Modern Israeli Hebrew). Two examples of head-marking languages (Egyptian Arabic and Amele) and one language without marking on clause level (Thai). 6 Genetic and areal affinity of the languages. For the pure POSS-ACC type, Afro-Asiatic, Gum, IndoEuropean, Niger-Congo, Altaic, Uralic, Austronesian and Penutian (Yokuts) language families representing ‘The Indo-Asian Language Area’ (Masica 1976) as well as other regions—Austronesia, Africa, Northern America and New Guinea. Inuit (of the Eskimo-Aleut language family) as the only language with anti-passive ANCs.
< previous page
page_128
next page >
< previous page
page_129
next page >
Page 129 7 Type 3: ‘Ergative-Possessive’ 7.0. INTRODUCTION The Ergative-Possessive type (ERG-POSS) is characterized by genitivization of the Ps and the Ss in ANCs, which are marked in the same way as opposed to the As. The definition above covers both pure ERG-POSS ANCs, in which the head does not have voice markers and which correlate morphologically to finite clauses non-marked with respect to voice, and passive ANCs, in which the head has an overt passive marker and which correlate to passive finite clauses (i.e., marked with respect to voice). The latter cases are, strictly speaking, outside the general typology of ANCs, since nominalization is not directly related to finite clauses with transitive verb forms. In other words, nominalization involves either intransitive verbs or lexically transitive verbs which have been intransitivized by passivization. These cases find a direct analogy in the anti-passive variant of the POSS-ACC type (see section 6.1.). In the majority of the pure ERG-POSS languages, the A has the same marking as the agent in passive clauses. In some of these languages, a single verbal noun can combine with its arguments both according to the ERG-POSS pattern and according to some other nominalization pattern, similar to the opposition ‘active vs. passive finite clauses’. It will be argued below that the pure ERG-POSS type in the general case is not directly correlated with the corresponding finite passive clauses. Languages which have such ANCs without having any passive clauses will be used to support this argument. Due to the relative heterogeneity of the ERG-POSS languages, more space will be devoted to this type than to the other types. The
< previous page
page_129
next page >
< previous page
page_130
next page >
Page 130 following groups of languages will be examined below: 1 languages with agentive passives: 1.1 languages with pure ERG-POSS ANCs, 1.2 languages with passive ANCs; 2 ERG-POSS languages without agentive passives. Among the languages of the last group, there are three languages in which ERG-POSS ANCs retain the sentential dependent-marking for the As. Therefore, it may be reasonable to distinguish between the Oblique-Possessive (OBL-POSS) and the Sentential-Possessive (SENT-POSS) variants of the ERG-POSS type. 7.1. LANGUAGES WITH AGENTIVE PASSIVES 7.1.1. Languages with pure OBL-POSS ANCs. Ideal representatives: Georgian, Russian and Abkhaz In Georgian, Russian and Abkhaz, the A in ANCs takes the same dependent-marking as the agent in passive clauses, which is the only option available for these three languages in forming ANCs (for Russian ANCs with pronominal arguments see section 10.2.). Georgian and Russian are predominantly dependent-marking languages. Their only head-marking feature is verbal agreement. Abkhaz is predominantly head-marking. Georgian has traditionally been considered as a language with ‘split’ ergativity since the morphological coding of the arguments in a finite clause depends on its tense. This description of Georgian will follow Harris (1981), who claims that the Georgian syntax is based on the distinction between active and inactive verbs. Dependent-marking on the clause level is determined by the type of the verb together with the tense ‘series’. In the Series 1 tenses, the S and the A are in the nominative, while both the P and the indirect object are in the dative. In the Series 2 tenses, the A and the S of active verbs are in the ergative, while the P and the S of inactive verbs are in the nominative. In the Series 3 tenses, all the Ss and Ps assume the nominative, the A takes the dative, while the indirect object appears as a postpositional phrase with the postposition -tvis . In passive clauses, the subject is in the nominative and the agent is referred to by means of a postpositional phrase with the postposition mier . Even the indirect objects of active clauses are marked differently in the corresponding passive ones, turning out as NPs with the suffix -tvis . Head-marking on the clause level consists of number and person agreement of the verb with its
< previous page
page_130
next page >
< previous page
page_131
next page >
Page 131 subject, direct or indirect object. Word order is fairly free with the basic word order SOV. Series 1 (7.1) nino ačvenebs suratebs Nino:NOM she:shows:him:it:SERIES 1 pictures:DAT gias. Gia:DAT ‘Nino is showing pictures to Gia.’ (7.2) nino amtknarebs. Nino:NOM she:yawns:SERIES 1 ‘Nino yawns.’ (7.3) ekimia. Vaxtang:NOM doctor:NOM:he:is:SERIES 1 ‘Vaxtang is a doctor.’ (Harris 1981:39–40, ex. 1a–1c) Series 2 (7.4) ninom ačvena suratebi Nino:ERG she:showed:him:it:SERIES 2 pictures:NOM gias. Gia:DAT ‘Nino showed the pictures to Gia.’ (ibid.: 40, ex. 4a) (7.5) ninom daamtknara. Nino:ERG she:yawned:SERIES 2 ‘Nino yawned.’ (7.6) ekimi iqo. Vaxtang:NOM doctor:NOM he:was:SERIES 2. ‘Vaxtang was a doctor.’ (ibid.: 40, ex. 3a–3b) Series 3 (7.7) turme rezos samajuri učukebia apparently Rezo:DAT bracelet:NOM he:gave:it:SERIES 3 dedistvis. mother:for ‘Apparently Rezo gave a bracelet to his mother.’ (ibid.: 117, ex. 2a)
< previous page
page_131
next page >
< previous page
page_132
next page >
Page 132 Passive (7.8) vašli micemulia . apple:NOM given:it:is:SERIES 1 teacher:for ‘The apple is given to the teacher.’ (ibid.: 103, ex. 1b) (7.9) bavšvi mier. child:NOM bitten:it:is:SERIES 1 dog:GEN by ‘The child is bitten by a dog.’ (ibid.: 103, ex. 2b) Ordinary noun phrases and ANCs use only dependent-marking. Dependents in the nominative and ergative cases, as well as with the postposition mier, are excluded from non-derived NPs. The G in the genitive case normally precedes the H. (7.10) gelas saxli Gela:GEN house:NOM ‘Gela’s house’ In ANCs, the S and the P are always put in the genitive, the A is expressed by means of a postpositional phrase with the postposition mier, while the indirect object takes the suffix -tvis 1. To judge from the available examples, the arguments (or at least one of them) usually precede the head. (7.11) datvis am tqeši . bear:GEN kill:AN.NOM this woods:in forbidden:it:is ‘Killing bears in these woods is forbidden.’ (Harris 1981:157, ex. 11) (7.12) mamas unda qvavilebis micema father:DAT he:wants:it flowers:GEN give:AN.NOM mziastvis vanos mier. Mzia:for Vano:GEN by ‘Father wants Vano to give flowers to Mzia.’ (ibid.: 157, ex. 13) (7.13) tamadis damtknareba supraze uzrdelobaa. toast.master:GEN yawn:AN.NOM table:on rudeness:it:is ‘It is rude for the toast master to yawn at the table.’ (ibid.: 157, ex. 14) We see that transitive ANCs and passive clauses in Georgian have the following features in common: (1) the P in transitive ANCs and in
< previous page
page_132
next page >
< previous page
page_133
next page >
Page 133 passive clauses gets the same marking as the S in intransitive ANCs or in intransitive clauses respectively; (2) the A appears as a postpositional phrase with the postposition mier; and (3) the underlying indirect object takes the suffix -tvis. This is summarized in Table 7.1. It is interesting to compare Georgian and Russian ANCs of the pure ERG-POSS types. Since Russian was discussed in section 1.3., I will confine myself here to summarizing the relevant facts in Table 7.2. Tables 7.1. and 7.2. show that there are two important differences between Georgian and Russian ANCs. First, in Russian, the indirect object in ANCs and in finite passive clauses has the same marking as in the corresponding active finite clauses, while in Georgian, its marking is influenced by the processes of passivization and nominalization. Second, in Georgian, the marking of the A in ANCs does not depend on the presence or absence of other arguments, while in Russian the A is distinct from the S and P only in the presence of the P in the same ANC. In both Georgian and Russian, action nominals correspond to ‘derived nominals’ and not gerunds (in the sense of Chomsky 1970). Thus, for example, the formal and the semantic relationship between them and the associated propositions is often idiosyncratic; some verbs do not have any corresponding verbal nouns. The internal structure of Table 7.1 Georgian: marking of S, A, P and IO in finite clauses and ANCs Sact Sinact A P IO Series 1 tenses NOM NOM NOM DAT DAT Series 2 tenses ERG NOM ERG NOM DAT Series 3 tenses NOM NOM DAT NOM -tvis Passive clauses —— —— mier NOM -tvis ANCs GEN GEN mier GEN -tvis Table 7.2 Russian: marking of S, A, P and IO in finite clauses and ANCs S A P IO Active clauses NOM NOM ACC DAT Passive clauses —— INSTR NOM DAT ANCs with one argument GEN GEN GEN DAT ANCs with two arguments —— INSTR GEN DAT
< previous page
page_133
next page >
< previous page
page_134
next page >
Page 134 ANCs differs from that of non-derived NPs in allowing certain types of dependents (for Russian, dependents in the instrumental case, which can normally only be taken by verbal nouns of various types, and for Georgian, dependents with the postposition mier and the suffix -tvis ). In Russian, only adjectives can qualify action nominals (see, however, section 10.1.), while in Georgian, adverbs are allowed in such constructions (in fact, some forms can be qualified either by adverbs or by adjectives depending on their more abstract or more concrete meaning, that is, on their more verbal or more nominal character, Hewitt 19832). Both languages are complement-balancing. Thus, in both Georgian and Russian, the dependent-marking of As in ANCs is similar to that of agents in passive clauses. However, the two languages differ in their treatment of the A in transitive ANCs with the P missing. In Abkhaz (Hewitt 1979), a typical head-marking language, major syntactic functions in independent clauses are expressed by means of verb agreement in the form of three sets of affixes in the verb complex according to the ‘ergative’ pattern, as follows: Set 1 affixes, occupying the first pre-radical slot of the verb complex, correlate with S or P (absolutives); Set 2 affixes, which follow immediately (if there are no other pre-radical constituents), correlate with the indirect object; and Set 3 affixes, following upon Set 2 affixes and any preverbs, correlate with A. Set 2 affixes are also used within the verb complex in association with, e.g., locative and comitative particles. Postpositions are used to indicate location and oblique functions of the noun. Word order is free with the basic order being A IO P V (i.e., SOV). In addition to some artificial constructions corresponding to the dynamic passives of other languages, there is a static passive construction, where the agent may be marked by the postposition la —‘by, by means of’ (the instrumental postposition) or —‘from’. (7.14) à SET1 the-girl-PL they-swim-FIN ‘The girls swim.’ (Hewitt 1979:104)
< previous page
page_134
next page >
< previous page
page_135
next page >
Page 135 (7.15) (sarà) I the-child-PL the-book-PL to- yt’ SET1- SET2-SET3 them- to.them-I- give- FIN ‘I give the books to the children.’ (ibid.: 105) In non-derived NPs, the H gets the appropriate Set 2 pronominal prefix. The non-marked G, if present at all, precedes the H. Some postpositional phrases such as ablative and benefactive PPs, are allowed within NPs, while others, such as instrumental ones, are excluded. (7.16) (sarà) SET2 I my-house ‘my house’ (ibid.: 116) (7.17) SET2 the-boy his-house ‘the boy’s house’ (ibid.) Abkhaz is a complement-deranking language with a rich system of non-finite verb forms (see section 2.2.2.). Action nominal constructions have as their head a regular verb form (‘masdar’) which is formed by suffixing the formant -ra to the verbal root and prefixing the article -a (the same formant is used to form abstract nouns from nouns and adjectives). As in non-derived NPs, both Set 1 and Set 3 pronominal affixes are prohibited in ANCs. Thus, the S and the P are expressed by means of Set 2 affixes, i.e., exclusively by means of head-marking. The A is dependent-marked as an instrumental PP (these are otherwise prohibited in non-derived NPs). The indirect object seems to be retained in nominalizations, which leads to a sequence of two Set 2 affixes, though the resulting expression is not natural Abkhaz (ex. 7.20).
< previous page
page_135
next page >
< previous page
page_136
next page >
Page 136 (7.18) larà 1-ca-rà SET2 she her-go-AN ‘her going’ (Hewitt 1979:112) (7.19) sarà SET2 I me-by your-see-AN ‘my seeing you’ (lit. ‘your seeing by me’) (ibid.: 112) (7.20) SET2-SET2 they them-by my-to you-give-AN ‘their giving me to you’ (lit. ‘my giving to you by them’) (ibid.) Thus, all the three languages presented in this section treat the As in transitive ANCs in the same way as the agents in finite passive clauses (either dynamic or stative). This nominalization pattern applies both to predominantly dependent-marking languages such as Georgian and Russian, and head-marking languages such as Abkhaz, and both to complement-balancing and complement-deranking languages. The three languages also illustrate different sentence types—nominative-accusative, split-ergative (or active), and ergative. 7.1.2. Languages with passive ANCs: Turkish In some languages, action nominals can appear with overt passive markers, the same as those used in finite clauses. In these constructions, the marking of the A and the P can be characterized in the same way as in pure ERG-POSS ANCs, that is, the P genitivizes, while the A is opposed both to the S, P and G. Examples (7.21a–7.21d) from Turkish illustrate the case of a language with passive ANCs. (7.21a) Hasan mektub-u yaz-dı Hasan letter-ACC write-3SG.PAST ‘Hasan wrote the letter.’ (Comrie and Thompson 1985:364, ex. 67)
< previous page
page_136
next page >
< previous page
page_137
next page >
Page 137 (7.21b) Hasan-ın mektub-u yaz-ma-sı Hasan-GEN letter-ACC write-AN-3SG.POSS ‘Hasan’s writing of the letter’ (ibid.: ex. 68) (7.21c) Mektub Hasan-ın tarafından yaz-ıl-dı letter Hasan-GEN by write-PASS-3SG.PAST ‘The letter was written by Hasan.’ (ibid.: 365, ex. 69) (7.21d) mektub-un Hasan-ın tarafından letter-GEN Hasan-GEN by yaz-ıl-ma-sı write-PASS-AN-3SG.POSS ‘the letter’s being written by Hasan’ (ibid.: 365, ex. 70) In (7.21b), the P is put in the accusative and the A appears in the genitive. The whole ANC, thus, follows the POSS-ACC pattern and corresponds to the active finite clause of (7.21a). In (7.21d), the action nominal appears with the passive suffix -ıl, the P is put in the genitive and the A is marked with the postposition tarafından. The same passive suffix and the same postposition are present in the passive finite clause of (7.21c). 7.2. PURE OBL-POSS NOMINALIZATIONS VS. PASSIVES: POSSIBLE CONNECTIONS 7.2.1. Two approaches to OBL-POSS nominalizations Before proceeding to the last group of those mentioned in section 7.0., the formal similarities between pure OBL-POSS nominalizations and passive clauses will be discussed in more detail. The defining characteristic of this nominalization type is the same dependent-marking for the Ss and the Ps as opposed to the As. The Ss and Ps genitivize, while the As take on another marking. The question arises whether the formal identity in the marking of the Ss and the Ps in such ANCs reflects important common features of these arguments as opposed to the As, or whether the only function of this marking is to distinguish between the two possible arguments of transitive action nominals. This question is especially interesting since most of the languages under consideration are nominativeaccusative, which means that finite clauses in them manifest a different formal opposition within the
< previous page
page_137
next page >
< previous page
page_138
next page >
Page 138 set of verbal arguments, namely, between the subject (the S and the A) and the direct object (the P). Traditional grammar distinguishes between the subjective and objective genitives in nominalizations (these terms, originally applying to languages with morphological cases, have now acquired a much wider usage). According to the traditional view, in nominalizations of the OBL-POSS and DBL-POSS (see section 8.1.) types, the formal identity of the S and the G, and of the P and the G merely reflects two different functions and meanings of the genitive, the subjective and the objective genitive respectively. According to the second approach, there is no need to distinguish between the subjective and the objective genitive, since there is only one, subjective genitive. This proposal is based on the observation that for many languages, the A in OBL-POSS ANCs is marked in the same way as in the corresponding finite passive clauses. It has been suggested that all the genitives could be accounted for as having the subjective function and would no longer be ambiguous. Instead, it is the action nominal itself which is ambiguous, or rather, corresponds to two different finite verb forms—an active one and the corresponding passive one. This explanation is especially attractive for those languages in which ANCs can be formed according to both the POSS-ACC and OBL-POSS patterns and where the alternations in the marking of the dependents can be directly correlated to voice oppositions in the corresponding finite clauses. Two such languages, Maori and Hebrew, will be cited in the next section. This approach can be illustrated by Veyrenc’s (1972) analysis of Russian ANCs. Veyrenc tries to show that the impossibility of the ‘subjective’ and the ‘objective’ genitives co-occuring within the same ANC cannot be explained by surface factors. More specifically, in Russian NPs, there is no surface morphological restriction as to the number of genitives following one another, which allows for such combinations as skorost’ russkoj armii—‘velocity of movement:GEN of the Russian Army:GEN’ (note that this example involves stacked genitives and does not say anything about the ability of Russian nouns themselves to have more than one genitive dependent). In addition, there are cases where one and the same action nominal can have two genitive dependents, as in brata ego nasledstva—‘deprivation of the brother:GEN of his heritage:GEN’, where both genitives have the ‘objective’ reading, cf. brata ego nasledstva—‘to deprive:INF the brother:ACC of his heritage:GEN’ (some of Veyrenc’s other examples seem impossible). Note that in this case, and in all the other allowed cases, the verbal noun is derived
< previous page
page_138
next page >
< previous page
page_139
next page >
Page 139 from a verb which itself takes one or more genitive objects. Of the two ‘objective’ genitives in the example, the former corresponds to the direct object, which take the accusative case in active finite clauses, while the latter corresponds to an oblique NP which retains its sentential form in ANCs in accordance with the usual behaviour of obliques in nominalizations. Thus, although ANCs may have two genitive dependents, they are not allowed to combine simultaneously with the ‘subjective’ and the ‘objective’ dependents, where the latter would turn up as the accusative object of a finite active clause. Veyrenc concludes that the restriction on the co-occurrence of the ‘subjective’ and the ‘objective’ genitives within nominalizations can be accounted for by the derivational source of these constructions. More specifically, intransitive ANCs are derived from the corresponding active finite clauses, while transitive ANCs are derived from the corresponding passive finite clauses. Transitive ANCs with only the A present are treated in the same way as intransitive ones. Thus, a transitive action nominal neutralizes the distinction of voice typical for finite verb forms, in the same way as it neutralizes the distinctions of tense and mood. What Veyrenc’s proposal and similar ones do is only to introduce a rule which, in a rather mechanical way, identifies the processes of nominalization (or, rather, the process of deriving transitive action nominals) and passivization on the basis of the formal morphological identity between the A of transitive ANCs and the agent of the corresponding finite passive clauses. Note also that nothing is said about the reasons for deriving these nominalizations via passivization. Though the formal identity of the A in ANCs and the agent in passive clauses does not imply that they are similar in other important respects, the rule itself implies this. That is, to justify the derivation of transitive ANCs via passivization, the following three assumptions should hold: (1) transitive action nominals share some properties with the corresponding finite passive verb forms: among other things, they have similar restrictions on derivation; (2) the Ps and Ss in ANCs and also the subjects of the corresponding finite intransitive and passive clauses share important properties and are thus, in this way, opposed to the likewise mutually similar As in ANCs and agents of the corresponding finite passive clauses; (3) in ANCs, there is no difference between the genitives corresponding to the Ss vs. Ps. For most of the OBL-POSS languages, I lack adequate information which would reveal to what degree these assumptions hold. A few examples will be given below to illustrate the method of comparing OBLPOSS ANCs with the corresponding finite passive clauses in the
< previous page
page_139
next page >
< previous page
page_140
next page >
Page 140 languages with both constructions. The data, insufficient for drawing general conclusions, show, however, that there need not be any direct correspondence between the two. The next subsection will deal with two of the languages with an opposition between the POSS-ACC and OBL-POSS ANCs—Maori and Hebrew—and present some of the language-specific arguments for not correlating this opposition directly with the ‘active/passive’ opposition, as it is manifested in the realm of finite clauses. In the ensuing subsections, we will see some other, more general arguments, showing that the three above mentioned assumptions do not always hold. The arguments are as follows: 1 ANs and passive verbs can have different derivational possibilities, i.e., different restrictions can apply to the derivation of passive verbs and to that of ANs, while there can also be passive verbs which code their arguments in another way than the corresponding ANs; 2 some syntactic features of the As in OBL-POSS nominalizations (their subject properties) bring them together with the Ss and the subjects of the corresponding active clauses, rather than with the agents of the corresponding passive clauses; 3 the Ss and the Ps in OBL-POSS ANCs, in spite of the formal similarity, have different properties, i.e., the subjective and the objective genitives should be distinguished. 7.2.2. Languages with an opposition between pure OBL-POSS and POSS-ACC types: Maori and Modern Israeli Hebrew Maori (Chung 1978), a VSO language, uses exclusive dependent-marking both in clauses and noun phrases (including ANCs). In independent clauses, subjects are not marked, while direct objects of ‘canonical’ transitive clauses are marked with the ‘accusative’ preposition i. (‘Middle’ verbs will not be discussed here.) The agent of passive clauses is marked by e. (7.22) I tae a Hutu ki raro. PAST arrive NOM PROP Hutu PREP underworld ‘Hutu arrived in the underworld.’ (Chung 1973:651)
< previous page
page_140
next page >
< previous page
page_141
next page >
Page 141 (7.23) Ka whakamau a Hata i te taura. UNS fasten NOM PROP Hata ACC ART rope ‘Hata is fastening the rope.’ (ibid.: 652) (7.24) Ka whana te hoīho i a Hōne. UNS kick the horse ACC PROP John ‘The horse kicked John.’ (Chung 1978:66, ex. 35) Passive (7.25) Ka whana-a a Hōne e te hoīho. UNS kick-PASS PROP John AGT the horse ‘John was kicked by the horse.’ (ibid.: 66, ex. 34) Within non-derived noun phrases, the G is marked by one of the two particles a—Dominant possession and o—Subordinate possession. (This Polynesian variant of the distinction between ‘alienable’ and ‘inalienable’ possession will be discussed in section 10.4.) According to Chung (ibid.: 25), ‘as the terms suggest, a is used for possessors who actively control the possessed noun; o is used for other possessor-possessed relationships, including that of the whole to its parts, of animate beings to their body parts, of individuals to immutable elements of culture or environment, and so on’. Non-pronominal Gs follow their Hs, while pronominal Gs either precede or follow them. (7.26) te poti a Paerau ART boat DOM.POSS Paerau ‘the boat of Paerau’ (Chung 1978:108, ex. 11a) (7.27) te poaka o te rangatira ART pig SUB.POSS ART chief ‘the chief’s pig’ (Biggs 1961:24) ANCs in Maori, as in many Polynesian languages, are derived by replacing the tense marker of a sentence with an article (or with a possessive pronoun) and adding some variant of the suffix — (C)(a)nga to the verb. (In Maori, different verbs lexically select among eight alternatives of this suffix.) Maori ANCs with non-pronominal arguments show several patterns. The Ss are always marked with the
< previous page
page_141
next page >
< previous page
page_142
next page >
Page 142 possessive particle o (Subordinate). In transitive ANCs, the first NP following the head normally genitivizes. Thus Maori action nominals without any overt passive markers can combine with their arguments in accordance with the following patterns: (1) POSS-ACC type. Both the Ss and the As genitivize, although in different ways (see section 10.4.)— the Ss are marked with the possessive particle o (Subordinate), while the As are marked with the possessive particle a (Dominant). The Ps retain their sentential marking, being marked with the accusative particle i. Word order is the same as in the corresponding active finite clauses—AN-A-P. (7.28) [te tae-nga o Hutu ki raro], katahi ART arrive-AN SUB.POSS Hutu PREP below then ka haere ia ki te kimi i a Pare. UNS go he PREP ART look.for ACC PROP Pare ‘When Hutu arrived in the underworld he went to look for Pare.’ (Chung 1973:651, ex. 21; cf. 7.22) (7.29) ka kite mātou [i te whakamau-tanga UNS see we ACC ART fasten-AN a Hata i te taura]. DOM.POSS Hata ACC ART rope ‘We saw Hata’s fastening of the rope.’ (ibid: 652, ex. 23; cf. 7.23) (2) OBL-POSS type. The Ss and Ps are marked with the Subordinate particle o, while the As are marked with the agentive particle e. Word order is AN-P-A. (7.30) ka kite mātou [i te whakamau-tanga UNS see we ACC ART fasten-AN o te taura e Hata]. SUB.POSS ART rope AGT Hata ‘We saw the fastening of the rope by Hata.’ (ibid: 652, ex. 24; cf. 7.29) (7.31) te epa-nga o te kupenga e te ART throw-AN SUB.POSS ART net AGT ART tangata man ‘the throwing of the net by the man’ (Chung 1978:302, ex. 79)
< previous page
page_142
next page >
< previous page
page_143
next page >
Page 143 (3) MIX type. The A is marked with the agentive particle e, while the P is marked with the accusative particle i, that is, it retains the sentential marking. Word order is AN-P-A. te epa-nga i te kupenga e te ART throw-AN ACC ART net AGT ART tangata man ‘the throwing of the net by the man’ (ibid.: 301, ex. 78b) Nominalizations with the passive suffix exist as a separate type in Maori. Of special interest is the fact that the arguments in them are marked in the same way as in OBL-POSS nominalizations described above. Does this mean that it would be reasonable to derive both of these nominalization types from a single source? In other words, do both of these types correspond to finite passive clauses? Chung (1978:301–4) does not discuss the transparently passive nominalizations, taking it for granted that they should be derived from the passive verbs. However, for the pure OBL-POSS ANCs, she proposes a special Agent Postposing rule, distinct from Passive. The relevant facts are as follows. First, these rules may have different impacts on the underlying P expressed by means of a prepositional phrase with the preposition i in the corresponding finite active clause. That is, the crucial argument here is constituted by the existence of the MIX type in Maori. While one of the main effects of Passive consists of promoting underlying direct objects to subjects, the former can still be marked with i in the corresponding ANCs, even when the A is marked with e. Secondly, Agent Postposing differs from Passive in not attaching any variant of the passive suffix -Cia to the action nominal. In Maori, the feature [+passive] spreads to certain verbal modifiers, which thus get their own passive suffixes. (Since these suffixes may differ from the passive suffix on the verb, it is the feature and not the suffix itself which is copied.) In OBL-POSS and MIX ANCs, verbal modifiers lack passive suffixes. Compare the examples below where the combinations ‘V/AN+verbal modifier’ are italicized. (Unfortunately, Chung does not give any example of an ANC with an overt A which could illustrate her point. However, it is obvious that the only possible marking for an overt non-pronominal A would be by means of the preposition e.)
< previous page
page_143
next page >
< previous page
page_144
next page >
Page 144 (7.33) Whati-ia poro-tia i waenganui. break-PASS strike-PASS at middle ‘(He) was completely broken off in the middle.’ (Chung 1978; 303, ex. 80a) (7.34) Kāore e makere [i ngā ngutu-nga not NONPAST lost CAUS ART.PL talk-AN kōrero o te hūmārie-tanga]. say SUB.POSS ART beautiful-AN ‘(They) are not forgotten, because (people) still talk of (her) beauty.’ (ibid.: 303, ex. 81) Thus, according to Chung, Maori nominalizations with the e- marker on the A should not be derived from the corresponding passive finite clauses. What is common for both types of constructions is the fact that they contain an A which is demoted from its subject position. Hebrew (Berman 1978) was mentioned in section 6.2. as an example of a language with restricted use of dependent-marking. Within independent clauses, the subjects are in the non-marked form. The distinction between direct and indirect objects seems to be rather vague. The most usual object marker, traditionally classified as ‘a marker of the direct object’, is et, which is used to introduce definite objects and which alternates with before indefinite NPs. The agent in passive clauses has the marker al ydey. Finite verbs agree with their subjects. Basic word order is SVO. (7.35) dan daxa et hahacaa Dan reject:PAST.3SG DEF.OBJ DEF:offer ‘Dan rejected the offer.’ (Berman 1976:71, ex. 24) (7.36) habaxur paras lēxem/et halēxem DEF:boy cut:PAST.3SG bread/DEF.OBJ DEF:bread ‘The boy cut bread/the bread.’ (Berman 1978:123, ex. 9) (7.37) hahacaa al ydey david. DEF:offer reject:PASS.PAST.3SG by David ‘The offer was rejected by David.’ (ibid.: 127, ex. 16) Non-derived possessive NPs show three different patterns mentioned in section 6.2.—the ‘bound’ form, the ‘separate’ form, and the ‘double’ form. The head always precedes its modifiers. The ‘bound’
< previous page
page_144
next page >
< previous page
page_145
next page >
Page 145 form involves only head-marking, where the head-noun appears in a special Construct Form. The ‘separate’ form involves only dependent-marking, where the G is preceded by the preposition . The ‘double’ form uses both head-marking and dependent-marking. (The H is in the Construct Form and takes the pronominal suffix which agrees with the G introduced by .) All three forms are synonymous in most contexts, the choice being influenced by stylistic considerations, although there are some exceptions. Even more interesting is the fact that these forms are mutually exclusive, that is, a noun cannot have more than one G of any kind. In the normative usage (Sven Malkior, personal communication), it is not possible to build chains of more than one smixut -dependent of the type [the teacher’s [picture of the girl]]—or [[the teacher’s picture] of the girl]. This restriction, however, concerns only certain varieties of Modern Israeli Hebrew. Thus, Ritter (n.d.: 21, ex. 32) quotes such examples as normal in certain dialects (tmunat ha-yalda ha-mora —‘the teacher’s picture of the girl’. ‘Bound’ form (xavur) (7.38a) parat ikar cow:CONSTR farmer ‘(a) farmer’s cow’ ‘Separate’ form (parud) (7.38b) para šel ikar cow:ABS of farmer ‘a farmer’s cow’ ‘Double’ form (kaful) (7.38c) parato šel ikar cow:CONSTR:his of farmer ‘a farmer’s cow’ (Berman 1978:232, ex. 1) The majority of Hebrew ANCs belong to two types—to the POSS-ACC and to the OBL-POSS types. 3 In both cases, ANCs exist in three alternative forms and thus show the same range of surface variation as ordinary possessive NPs. Apart from this variation, Hebrew action nominals combine with arguments according to the following patterns: 1 The POSS-ACC type—the Ss and the As genitivize and the Ps are introduced with et, exactly as in the corresponding active finite clauses.
< previous page
page_145
next page >
< previous page
page_146
next page >
Page 146 (7.39a) dxiyat dan et hahacaa reject:AN:CONSTR Dan DEF.OBJ DEF:offer (cf. ex. 7.35) (7.39b) hadxiya šel dan et hahacaa DEF:reject:AN of Dan DEF.OBJ DEF:offer (7.39c) dxiyato šel dan et hahacaa reject:AN:CONSTR:his of Dan DEF.OBJ DEF:offer ‘Dan’s rejection of the offer’ (Berman 1976:71, ex.24) 2 The OBL-POSS type—the Ss and the Ps genitivize, while the As are introduced with the agent marker al ydey, exactly as in the corresponding passive clauses. (7.40a) dxiyat hahacaa al ydey Dan reject:AN:CONSTR DEF:offer by Dan (7.40b) hadxiya šel hahacaa al ydey Dan DEF:reject:AN of DEF:offer by Dan (7.40c) dxiyata šel hahacaa al ydey Dan reject:AN:CONSTR:her of DEF:offer by Dan ‘Dan’s rejection of the offer’ (ibid.; cf. ex. 7.37) The choice between these two types of ANCs seems to be rather free, apart from one restriction: the POSS-ACC type can only be used when the P is definite and, thus, introduced with the marker et. This restriction means that ANCs with indefinite Ps manifest only the OBL-POSS type, which constitutes an argument against relating such nominalizations directly to the corresponding passive clauses. More specifically, passivization promotes the underlying P to the subject position and, thus, typical cases of passivization involve relatively highly referential, definite Ps (since this is one of the most characteristic subject features). Among other things, this implies that passivization is normally avoided when the P is indefinite and that finite verbs with indefinite Ps are normally used in the active form. For the corresponding ANs, there is also only one option, but in this case it is the OBL-POSS type.
< previous page
page_146
next page >
< previous page
page_147
next page >
Page 147 7.2.3. ANCs of the OBL-POSS type vs. passives: restrictions on derivation If ANCs of the OBL-POSS type are to be correlated with the corresponding finite passive clauses, they should presumably have similar restrictions on derivation as the verbs in the latter. Any discrepancies may be used as arguing against common derivation for these two types of constructions. The cases from Dutch and Russian, mentioned in section 1.4.1., involve verbs which can be passivized but whose ANs do not combine with their arguments according to the OBL-POSS pattern. It has been suggested that nominalization is much more sensitive to the thematic roles of arguments (and thus to the semantics of the underlying verb) than passivization, which is mostly concerned with purely syntactic subcategorization. Still more revealing are cases when there are no passive clauses corresponding to OBL-POSS ANCs. For example, for Georgian, Harris (1981:177–8) states that the range of OBL-POSS ANCs is greater than that of finite passive clauses, since (a) ‘agents are generally limited to animate or personified nominals’, while the postposition mier in ANCs may be used with ‘nominals that are neither animate nor personified’; and (b) mier marks the logical subject of ‘affective’ verbs which do not have any passive (like ‘forget’).4 On the basis of this, Harris rejects the idea that nominalization in Georgian should be directly correlated to passivization. In Modern Israeli Hebrew, in contrast to Georgian, passivization in finite clauses is much less restricted than derivation of OBL-POSS ANCs. A few of the verbs which take objects marked with b-, may be passivized and combine with their dependents in the same way as ‘usual’ passive verbs, that is, those whose active counterparts take et- objects. However, action nominals derived from verbs with b -objects are not able to form ANCs of the OBL-POSS type, the only available option being the POSS-ACC pattern. For some verbs, such as yacar —‘to create’ and xanak —‘to strangle’, the passive forms are only used with middle-voice functions in agentless constructions, but the corresponding ANs can combine with their arguments, including the A, according to the OBL-POSS pattern (Sven Malkior and David Gil, personal communication). 7.2.4. ANCs of the OBL-POSS type vs. passives: different syntactic properties of the As and the agents in Russian This section will show some syntactic properties, which are shared by
< previous page
page_147
next page >
< previous page
page_148
next page >
Page 148 the A of OBL-POSS ANCs and the subject of active finite clauses in Russian, but not the agent of passive clauses. The relevant examples concern the behaviour of the different arguments in ANCs and in finite clauses with respect to reflexivization and coordinate reduction, as well as the preferred word order in them. 7.2.4.1. Reflexivity The process of reflexivization involves at least two coreferential noun phrases, one of which is treated as the antecedent of the other. Normally, there exist severe restrictions on the type of possible antecedents of reflexive elements, which, as a rule, can be formulated in terms of syntactic functions of such NPs. Thus, the subject (in those languages for which this function is relevant) tends to be able to control reflexivization (Keenan 1976:315). For the languages with the OBL-POSS type of nominalizations, the following questions are of interest: 1 If reflexivization is possible in OBL-POSS ANCs, which argument can control it? 2 In the same language, which argument within finite clauses can control reflexivization? 3 Is it possible to account for reflexivity within ANCs of the OBL-POSS type on the basis of the corresponding finite clauses? If so, how? That is, what type of finite clause corresponds to these nominalizations? In Russian, reflexivization in OBL-POSS ANCs is controlled by the A, that is, by the NP in the instrumental case. The use of the reflexive pronouns in these cases is not obligatory. In contrast to this, the ability of the agents in passive clauses to serve as antecedents for the reflexive object pronoun sebja is very restricted (for example, the agent cannot control reflexivization of the subject of the same passive clause), while the reflexive possessive pronoun svoj can never be controlled by the agent (Padučeva 1985:197–8). (7.41a) oproverž-enie Gegel-em sam-ogo seb-ja refute-AN Hegel-INSTR self-M.GEN 3.RFL-GEN ‘Hegel’s refutation of himself’ (Padučeva 1985:197–8, ex. 1) (7.41b) *oproverž-enie Gegel-ja sam-im sob-oj *refute-AN Hegel-GEN self-M.INSTR 3.RFL-INSTR ‘the refutation of Hegel by himself’
< previous page
page_148
next page >
< previous page
page_149
next page >
Page 149 (7.41c) Gegel’ oproverg sam-ogo seb-ja. Hegel:NOM refuted self-M.ACC 3.RFL-ACC ‘Hegel refuted himself.’ (7.42a) Presledov-anie Karenin-ym svo-ej/?ego pursue-AN Karenin-INSTR 3.POSS.RFL-GEN/?his žen-y vyzvalo tolki v svet-e. wife-GEN caused rumours in society-LOC ‘Karenin’s pursuing his wife caused rumours in the society.’ (7.42b) Presledov-anie Ann-y *svo-im/ee pursue-AN Anna-GEN *3.POSS.RFL-INSTR/her muž-em vyzvalo tolki v svet-e. husband-INSTR caused rumours in society-LOC Anna’s being pursued by her husband caused rumours in the society.’ (7.42c) Karenin presledoval svo-ju Karenin:NOM pursued 3.POSS.RFL-ACC žen-u. wife-ACC. (7.42d) Ann-a presledovalas’ svo-im Anna-NOM was.pursued 3.POSS.RFL-INSTR muž-em. husband-INSTR ‘Anna was pursued by her husband.’ In general, reflexive relations in nominalized clauses are much more restricted than those within finite clauses. Thus, it is often difficult to find natural examples that could be translated into different languages by means of OBL-POSS nominalizations with reflexive markers. Informants prefer either finite clauses or non-reflexive ANCs. My data on Estonian and Abkhaz show the same tendency as in Russian, since possible relations between antecedents and reflexive elements in OBL-POSS ANCs are similar to those in the corresponding active finite clauses. In Estonian, reflexivization in OBL-POSS ANCs is controlled by the A (the postpositional phrase with the postposition poolt ), although such constructions seem to be rather unnatural. In passive clauses, the agent cannot serve as the antecedent for reflexive pronouns.
< previous page
page_149
next page >
< previous page
page_150
next page >
Page 150 (7.43a) Anna oli jälitatud oma/*tema mehe poolt. Anna was pursued her:RFL/*her husband by ‘Anna was pursued by her husband.’ (7.43b) Anna jälitamine tema/?oma mehe poolt Anna:GEN pursue:AN her/?her:RFL husband by ‘the pursuit of Anna by her husband’ In Abkhaz (Hewitt 1979:77–85), in finite active clauses and in ANCs, reflexivity is expressed in several ways. When expressed by the verbal affix , the antecedent is the A of the clause (as in ex. 7.44a). When expressed by the noun , preceded by the appropriate possessive prefix, or by the reflexive possessive adjective -xa+to, the antecedent may be the A (as in 7.45a), the S (as in 7.46a) or their modifers. To judge from Hewitt’s examples (see 7.44b–7.46b below), the range of reflexive relations in ANCs is the same as in the corresponding active finite clauses. Antecedent=A (7.44a) larà SET2 she her+RFL-she-kill-DYN-FIN ‘She killed herself.’ (Hewitt 1979:77) (7.44b) . the-RFL-kill-AN it-he-want-STAT ‘He wants to kill himself.’ (ibid.: 84) (7.45a) a-sark’ SET2-SET3 the-mirror-in my-head/self it-I-see-FIN. ‘I saw myself in the mirror.’ (ibid.: 78) (7.45b) [ a-ba-rà] . SET2 SET1-SET2 his-head/self its-see-AN it-he-want-STAT ‘He wants to see himself.’ (ibid.: 84)
< previous page
page_150
next page >
< previous page
page_151
next page >
Page 151 Antecedent=S (7.46a) sara s-ò-w+p’ zako’àn-s I I-be-STAT my-head/self law-PRED y-à-mo-w SET2 who-it-have-NONFIN ‘I am a law unto myself.’ (lit. ‘I am the one who is pos sessed to myself as a law.’ (ibid.: 83) (7.46b) [ zako’àn-s your-head/self//your-RFL head/self law-PRED à-ma-zaa-ra] . SET2 SET1 its-have-SUFF-AN it-good-STAT ‘It is good to be a law unto oneself.’ (lit. ‘it is good to have yourself as a law.’ (ibid.: 84) Greek (Joseph and Philippaki-Warburton 1987:82) shows another tendency. Thus, in OBL-POSS ANCs, the reflexive can be the P, but in this case the antecedent is most often not present within the ANC itself, that is, it is ‘either implied or it is present in the clause in which the (action) nominal functions as a constituent’. When both the A and the P should be present in a reflexive ANC, Greek prefers to put the A in the genitive and to introduce the P by an appropriate preposition (such POSS-ADN ANCs in Greek are lexically determined, according to Hans Ruge, personal communication). 7.4.2.2. Coordinate reduction The rule of coordinate reduction involves at least two coordinated coreferential NPs with the same (or similar) syntactic functions, one of which is deleted. Ability to control coreferential deletions and to be deleted has been used as a test for deciding whether it is reasonable to distinguish the subject function in a language (according to Keenan 1976:315, basic subjects are among the possible controllers for coreferential deletions), and for determining the subject of a clause. Similar tests can be applied to coordinated ANCs for checking whether the different arguments differ in their propensity to control
< previous page
page_151
next page >
< previous page
page_152
next page >
Page 152 deletion and to be deleted. Thus, for the purpose of the present discussion, the test might look like the following: two ANCs are coordinated, an intransitive one and a transitive one. They contain three NPs, the S, the P and the A. The S is coreferential to either the P or the A. If it is possible to delete an argument in this situation, which one would be deleted? (a) The A in the first ANC, coreferential to the S in the second one, can be deleted. The most natural interpretation would not mention any A at all, but the resulting construction can be understood as having the deleted A. Although this interpretation is, perhaps, forced by pragmatic factors rather than by purely syntactic ones, there are no similar possibilities for the cases described under (b) and (c) below. (7.47) Axejcy napali na trojancev posle poxiščenija Greeks attacked on Trojans after abduct:AN Eleny i (posledujuščego) vozvraščenija Parisa Helen:GEN and following return:AN Paris:GEN v Troju. in Troy ‘The Greeks attacked the Trojans after the abduction of Helen and Paris’ subsequent return to Troy.’ (b) The P in the first ANC, coreferential to the S in the second, can hardly be deleted. Deletion is made especially difficult due to the problems in marking the A. That is, if the P should be deleted, the A must be put in the genitive, which is normally interpreted as referring to the P. On the other hand, the A cannot be put in the instrumental because of the formal restrictions on its use in ANCs. (The instrumental is used in transitive ANCs only when both of their arguments are present, see section 7.1.) (7.48) Zavoevanie ?axejcev/*axejcami i padenie conquer:AN Greeks:GEN/INSTR and fall:AN Troi byli rezul’tatom Trojanskoj vojny. Troy:GEN were result Trojan war ‘The Greeks’ conquest and the fall of Troy were the result of the Trojan war.’ (c) The S in the first ANC is coreferential to the P in the second. The situation is simular to the one described in the preceding paragraph.
< previous page
page_152
next page >
< previous page
page_153
next page >
Page 153 (7.49) Padenie Troi i zavoevanie fall:AN Troy:GEN and conquer:AN ?axejcev/*axejcami byli rezul’tatom Trojanskoj vojny. Greeks:GEN/INSTR were result Trojan war ‘The fall of Troy and the conquest (of it) by the Greeks were the result of the Trojan war.’ It seems that Russian tries to completely avoid deleting coreferential arguments within ANCs and instead replaces them with suitable personal pronouns. In cases of deletion, though marginal, the As are treated as having a similar (syntactic) function within transitive ANCs as the Ss have within intransitive ones, which is parallel to their behaviour in finite intransitive clauses coordinated with active transitive ones, where they are subjects. 7.4.3.3. Word order Word order in Russian ANCs is fairly free under the condition that the head precedes its dependents, that is, the mutual order of the A and the P can be both A-P and P-A. However, according to Veyrenc (1972:228), it is much more usual for the A to precede the P than otherwise. His list of randomly chosen examples with transitive ANCs includes nineteen phrases, of which only one has the order AN-PA, while eighteen have the order AN-A-P. This means that in the majority of transitive ANCs, the order of the dependents is the same as in the corresponding active finite clauses, where the A (subject) precedes the P (object), and is different from the corresponding passive clauses, in which it is the P which precedes the A. 7.2.5. The subjective and the objective genitives: any differences? As we have seen above (section 7.2.1.), the idea of relating ERG-POSS ANCs to passive clauses is connected to the idea of eliminating the distinction between the subjective and objective genitives in ANCs in favour of the former. However, to what degree are the S and the P in ERG-POSS ANCs morphologically similar? For English, with its ambiguous phrases such as the shooting of the prisoners and the prisoners’ shooting, Vendler (1968) has suggested that the subjective and objective genitives are morphologically different, since they are formally different for two groups of words—pronouns and proper names. Thus, the subjective reading of the above mentioned examples would correspond to such phrases as the
< previous page
page_153
next page >
< previous page
page_154
next page >
Page 154 shooting of his and the shooting of Mary’s, while the objective reading corresponds to the shooting of him and the shooting of Mary . However, if a small group of words within a much greater group makes a formal distinction between two categories, is it reasonable to consider that the whole group distinguishes between them? This is a classical problem for grammatical descriptions, especially for descriptions of case systems. Phenomena similar to those quoted by Vendler are not uncommon for other languages. Thus, in some OBL-POSS languages, ANCs with pronominal or other highly referential dependents follow other patterns and, thus, distinguish between the S and P (see section 10.2.). We may conclude from sections 7.2.2.–7.2.5. that pure OBL-POSS ANCs and passive clauses, though showing a number of similarities, need not be directly correlated to each other. Still, these similarities remain to be accounted for. 7.2.6. OBL-POSS type and Relational Grammar In all the pure OBL-POSS languages mentioned so far, the A in ANCs is marked in the same way as the agent in finite passive clauses. The other common feature of these two constructions is the similar treatment of the S and the P, as opposed to the A. In the preceding sections, the view that the two constructions are directly related to each other was argued against. However, it can be suggested that they are products of certain rules which have essential similarities and, thus, lead to an indirect correlation between these constructions. An example of such an explanation is offered by the Relational Grammar, as applied by Harris (1981) to Georgian OBL-POSS ANCs. Relational Grammar considers grammatical relations in a clause such as the subject-of, direct object-of, indirect object-of, etc., as primitive notions of the syntactic theory. They build the hierarchy S>DO>IO…. Elements of a clause can bear different grammatical relations to each other on different syntactic levels, postulated by the theory. Thus, it is possible to distinguish between initial and final grammatical relations (those that exist on the deepest and on the superficial syntactic levels respectively). Languages possess rules that can change relations between the elements in a clause. One such rule is a Passive rule (an example of so-called advancement rules). In the simplest case it advances the former direct object to the subject position on a more superficial level, which corresponds to the intuitive idea about the nature of passivization. The crucial point concerns the question of what happens with the former subject of the clause. Its
< previous page
page_154
next page >
< previous page
page_155
next page >
Page 155 relation is now taken by the promoted nominal and it becomes ‘unemployed’, or chômeur—neither the subject nor an object. (Harris distinguishes further between chômeurs and other types of retired nominals.) Passivization is not the only rule that results in making some NP ‘unemployed’. Derivation of causative constructions offers another example. Just as languages have specific syntactic means for signalling such relations as the subject and the direct object, they can have special formal means for expressing chômeurs or retired terms of different kinds. For Georgian, Harris quotes several constructions with retired indirect objects and retired subjects in which these NPs are marked in a very consistent way. One of these constructions is an action nominal construction. If we accept this view of the nature of ANCs, it is easy to explain the identical marking of the A in pure OBL-POSS nominalizations and the agent in passive clauses in the languages with both types of constructions. These nominals are initial transitive subjects, both have become retired terms on the most superficial syntactic level, and it is reasonable that they should be marked in the same way. Thus, Relational Grammar offers some kind of unified description and explanation for ANCs of the OBLPOSS type. However, it leaves some questions open. The most important point of criticism concerns the essence of nominalization. Why do the arguments of verbal nouns become retired nominals and how do we know that they do so? While both the subject (transitive and intransitive), direct and indirect objects in nominalizations in Georgian change their marking as compared to the corresponding finite active clauses, in the other languages with the same nominalization type, the indirect object remains intact. Is it a retired nominal or not? Equally questionable is the possibility of applying the same model to the other nominalization patterns. Is it justified to talk about retired subjects in the SENT type, or should we assume that there cannot be any universal definition of nominalization, which in some languages creates retired terms of different kinds, while in others leaves their initial relations intact? Even in languages with personal passives, the agents and the As in OBL-POSS nominalizations can be marked differently. Some examples of such discrepancies are mentioned by Comrie and Thompson (1985:385–7) and include Modern Welsh (the passive agent is marked with the preposition gan—‘by, with’, while the A in ANCs takes the preposition o—‘from’); Italian (the passive agent is usually marked with the preposition da—‘from’, ‘by’, though da parte di—‘from (the) part of’ is also possible, while the A in ANCs takes only the latter); and German (the passive agent takes the
< previous page
page_155
next page >
< previous page
page_156
next page >
Page 156 preposition von—‘by, of’, while the A in ANCs is only introduced with durch—‘through’. Durch marks the passive agents ‘which are not strictly agentive’). Does the existence of such cases imply that the marking of chômeurs is dependent on the exact type of rules that cause their ‘unemployment’? If so, it becomes more difficult to argue that all these different rules have the same syntactic effect on the nominals in question. 7.3. LANGUAGES WITHOUT AGENTIVE PASSIVES The sample contains four languages (Hixkaryana, Samoan, Tongan and Classical Arabic) in which both the Ss and Ps in ANCs genitivize, but the dependent-marking of the As cannot be correlated with that of agents in passive clauses, because the languages do not have agentive passives. Hixkaryana (Derbyshire 1979), an OVS language, does not use any dependent-marking for the S, A or P of finite verbs. Only the indirect object is followed by the ‘relator’ wya . (The same relator follows the causee-NP in causative constructions derived from transitive verbs.) Finite verbs are head-marked for their S, or both the A and the P (by means of portmanteau prefixes). (7.50) namryekno romuru. he:went:hunting my:son ‘My son went hunting.’ (ibid.: 37, ex. 84a) (7.51) yaryako Waraka. 3.POSS.RFL:brother:POSS he:took:him Waraka ‘Waraka took his own brother.’ (ibid.: 63, ex. 145) (7.52) kamara hutxho yonyhoye Waraka karaywa wya jaguar skin:of he:caused:see:it Waraka non.Indian by ‘Waraka let the non-Indian see the jaguar skin.’ (ibid.: 90, ex. 202a) In non-derived NPs, only head-marking is used. The H is marked in two ways: its suffix indicates that it is possessed together with a kind of possession (present or past), while its person-marking prefix agrees with the person of the G. In the case of a third person possessor, an optional G can precede the H. Possessive prefixes, except for the third person, are the same as verb prefixes which mark the direct object,
< previous page
page_156
next page >
< previous page
page_157
next page >
Page 157 occurring when the subject is third person. Among the possessive prefixes, one, has the meaning of ‘third person reflexive’. (7.53a) my-canoe-PRES.POSS ‘my canoe’ (ibid.: 98) (7.53b) ro-kanawa-tho my-canoe-PAST.POSS ‘my former canoe’ (ibid.) (7.54a) his-canoe-PRES.POSS ‘his own canoe’ (ibid.: 45, ex. 99b) (7.54b) Waraka Waraka canoe-PRES.POSS ‘Waraka’s canoe’ (ibid.) In ANCs, both head-marking and dependent-marking are used, albeit in a rather limited way. The S and the P genitivize, i.e., they remain non-marked, while the AN takes nominal person-marking prefixes (as well as tense suffixes which are typical for most types of Hs). The A takes the relator wya —‘to, by’. (7.55) , [wewe ] hona. I:was:afraid, tree 3.POSS:fall:AN:PRES.POSS to ‘I was to be afraid of the trees falling.’ (ibid.: 65, ex. 151) (7.56) [Waraka wya honyko ] Waraka by peccary 3.POSS:shoot:AN:PRES.POSS xe wehxaha. desire I:am ‘I want Waraka to shoot peccary.’ (ibid.: 92, ex. 205a) The account given by Relational Grammar for OBL-POSS nominalizations seems to suit the Hixkaryana facts. More specifically, though Hixkaryana lacks passivization, its causative constructions derived from transitive verbs contain ‘retired transitive subjects’ (their initial relation is now taken by the causer, the subject of the predicate
< previous page
page_157
next page >
< previous page
page_158
next page >
Page 158 ‘to cause’). That is why the A (the initial transitive subject) of ANCs assumes the same marking as the causee in such causative constructions, and the whole ANC follows the OBL-POSS pattern. The other three languages mentioned in the beginning of this section present greater problems, since the A in their ERG-POSS ANCs looks exactly like its sentential counterpart. I have suggested classifying such nominalizations as belonging to the SENT-POSS subtype of the ERG-POSS type. They constitute an exception to one of the generalizations formulated at the end of chapter 3: ‘Of the two arguments in transitive ANCs, the one most likely to lose its sentential dependent-marking and head-marking is the A’ (p. 64). Let us scrutinize these cases. Samoan and Tongan, two West Polynesian languages, have ERG-POSS ANCs similar to those in Maori (see section 7.2.2.), but differ crucially from Maori in being ergative. In other words, while the A in Maori pure OBL-POSS nominalizations takes the same marking as the agent in passive clauses, the same particle e in Samoan and Tongan introduces the A of ergative finite clauses. On the whole, the system of Polynesian languages from the point of view of ergativity -nominativity is rather unstable. This is seen, for example, in Pukapukan, which has nominative-accusative and passive finite clauses of the Maori type, as well as ergative clauses of the Samoan type. Ergative clauses look exactly like passive clauses, but the verbs in them do not have the passive suffix (Chung 1978:323). The instability of the Polynesian system is revealed also by Samoan, which has an ergative case system with basic word order VSO. Its transitive verbs can sometimes take the ‘transitive’ suffix -Cia, in which case the A and the P retain the same marking as in ergative clauses, but ‘the clause exhibits VOS word order at least as often as VSO’ (ibid.: 53). Note that historically it is the same suffix as the passive suffix in Maori. The resemblance between the passive marking in Eastern Polynesian (nominative-accusative) languages and the ergative marking in Tongan and Samoan has given rise to at least two conflicting theories as to which of the systems was the original one. On the basis of the nominalization patterns found in these languages, Chung (1973) argues for the derivation of the ergative system from the original nominativeaccusative one (see section 11.6). In Classical Arabic (Wright 1971:2, 54), SENT-POSS ANCs are relatively common. In these phrases, the verbal noun is in the Construct State, the P immediately follows it and is put in the genitive, while the A in the nominative follows the P. I do not have any explanation for this situation.
< previous page
page_158
next page >
< previous page
page_159
next page >
Page 159 7.4. HEAD-MARKING IN THE ERG-POSS TYPE The majority of the languages with ERG-POSS ANCs have some form of head-marking on the clause level, except for the Polynesian languages. Head-marking in non-derived NPs is found in eight languages of the twenty-five pure ERG-POSS languages presented in the sample.5 Three of these, Hixkaryana, Abkhaz and Ancient Egyptian, are predominantly head-marking and use almost no dependent-marking at all. Classical Arabic is double-marking, while the other three use dependentmarking within clauses, but not for referring to Gs. Of the eight languages mentioned above, only Kurdish has special suffixes on the H which are not used in any other functions (there are no pronominal possessive elements, but only ‘Ezafe’-suffixes showing that the head of a NP has dependents without giving any indication as to their kind). In Ancient Egyptian, possessive suffixes are similar to verbal subject suffixes. It seems significant that in the remaining six languages, the head-marking in possessive NPs is more similar to the object verbal pronominal elements, than to the subject ones. Thus, in Abkhaz, the H takes pronominal prefixes of Set 2, which are used within verbal complexes for marking the indirect object (see section 7.1.). In Hixkaryana, possessive prefixes are almost identical to verbal prefixes marking the direct object when the subject is third person (cf. section 7.3.). In Persian, verbs can take special pronominal suffixes for referring to objects (direct and indirect), and the same suffixes appear as possessive suffixes on nouns (Amin-Madani and Lutz 1972). In Classical Arabic and Hebrew, verbal object suffixes are the same as possessive suffixes, except for the first person singular. (For Hebrew, this is a diachronic statement, since object suffixes are obsolete and are hardly ever used in the modern language at all.) In Welsh, prenominal possessive pronouns cause mutation of the H. The head-verb in so called ‘periphrastic forms’ is subject to the same type of mutation caused by pronominal object forms (Awberry 1976:16). We will return to these formal features of the head-marking in the languages under consideration in section 11.5. 7.5. FURTHER CHARACTERISTICS OF THE PURE ERG-POSS TYPE There are mainly nominative-accusative languages among the languages included in this group. Panjabi, Kurdish and Georgian manifest split ergativity; Abkhaz, Samoan and Tongan are purely
< previous page
page_159
next page >
< previous page
page_160
next page >
Page 160 ergative languages; Pukapukan has both nominative-accusative and ergative clauses; while for Hixkaryana it is difficult to determine the type because no dependent-marking is used for expressing the major functions within clauses and head-marking on transitive verbs consists of portmanteau forms. Word order in the languages of the group under consideration shows much more variation than in the preceding two groups. In the sample, the ERG-POSS group includes three SOV/GN languages, four SOV/NG languages, seven SVO/NG languages, one SVO/GN language (Estonian), one SVO language with both preposed and postposed Gs (English), eight V-initial/NG languages and one OVS/GN language (Hixkaryana). Of the seven SVO/NG languages, four have special possessive pronouns and/or other possessive forms which precede their heads. The same is true for Dutch and German, SOV languages. Both complement-balancing and complement-deranking languages are included in the group, through the former constitute the absolute majority. Very few verbal features are retained in nominalizations of the ERG-POSS type. Thus, only in Hixkaryana do verbal nouns manifest tense distinctions, which are also typical for all types of head-nominals in NPs. Information about other verbal features is very limited. It seems that most languages either do not allow adverbs within their ERG-POSS nominalizations or, as in the Polynesian languages, do not have any distinction between adverbs and adjectives. The situation in Hebrew was described in section 6.4., while some fluctuations in this respect in Georgian are mentioned in note 2 for this chapter. As regards negation, almost all the languages for which I have information treat their ERG-POSS nominalizations as ordinary noun phrases, the notable exception again being Hebrew (see section 6.4.). Among the languages of the group under consideration, there are six languages which do not have any dependent-marking for the Gs in possessive NPs. Of the remaining nineteen, which thus have some kind of overt genitives in their possessive phrases, these latter can be used with non-adnominal functions (i.e., outside NPs) in nine languages. In most cases these functions include marking objects in finite clauses. In Welsh, constructions with periphrastic verb forms, used in some tenses, look like possessive NPs in several respects (namely mutation of the head, position and form of pronominal objects and possessors). This is another example for the affinities in the expression of Gs and objects in the languages with the ERG-POSS type of
< previous page
page_160
next page >
< previous page
page_161
next page >
Page 161 nominalizations, besides the similarities discussed in the preceding section. For more discussion of this issue see section 11.5. 7.6. INTRA-LEXEMIC SYNTACTIC VARIATION For the pure ERG-POSS type, the following variation is attested: SENT: Tongan and Classical Arabic (languages with SENT-POSS ANCs); POSS-ACC: the Polynesian languages Maori, Pukapukan and Tongan (the latter only in ANCs with pronominal As), rarely Polish, English, Classical Arabic and Modern Israeli Hebrew; NOMN: Greek, English, Estonian, Hebrew (in non-normative usage) and, very rarely, Georgian (DBLPOSS); INC: Estonian, Kurdish, Persian; MIX: Maori and, marginally, Bulgarian and Dutch. ‘Passive’ ANCs occur in languages with POSS-ACC ANCs (Amharic, English, Inuit, Maori, Swahili and Turkish). The main features of the ERG-POSS languages are summarized in Table 7.3 below. 7.7. SUMMARY 0 Definition of the ERG-POSS type. Genitivization of the Ps and the Ss in ANCs and their opposition to the As. Pure ERG-POSS ANCs subdivided into SENT-POSS and OBL-POSS ANCs and opposed to passive ANCs (with overt passive markers). 1 Further syntactic devices of pure ERG-POSS ANCs: a head-marking. More frequent head-marking on the clause level than in the preceding two types. Within NPs less frequent head-marking than in the POSS-ACC type, but more frequent than in the SENT type. ANCs follow the nominal pattern. Frequent affinities in possessive and object affixes; b word order. Where word order of the head and its dependents in NPs is different from in clauses, ERG-POSS nominalizations follow the nominal pattern. 2 Retention of verbal features. No oppositions in tense or aspect (except Hixkaryana). Overt passive markers in some languages (the passive type). Retention of adverbial modifiers uncommon. 3 Nominal features. Possibility of combining verbal nouns with adjectives in most languages and negating them as non-derived nouns (except for Hebrew, which uses a special kind of negation in nominalizations).
< previous page
page_161
next page >
< previous page
page_162
next page >
Page 162 Table 7.3 Languages of the ERG-POSS type (see comments to Table 5.1 in section 5.5., pp. 105–8) Language Word order Case marking Balanced Head-marking Other Other system complements FC NP ANC patterns features Pure: SENT-POSS Arabic, V-1/NG ACC + ACC+ sp ?1, 2 Classical SAPpr Samoan V-1/GprNG ERG + − − − 4Apr Tongan V-1/GprNG ERG/ACCpr + − − − 1, 2pr OBL-POSS Abkhaz SOV/GN − − ERG-SAP+ sp Bulgarian SVO/*NG ACCpr + ACC-SA − − 4A*, ?5 Dutch SOV/*NG ACCpr + ACC-SA − − 4A*, ?5 ADV Egyptian, V-1/NG − + ACC+ sp 2pr, 5pr Ancient SAPpr English SVO/GNG ACCpr + ACC-SA − − 2, 3′, 4A ADJ Estonian SVO/GN ACC + ACC-SA − − 4A, 6 French SVO/GprNGACCpr + ACC-SA − − 4Apr ADJ Georgian SOV/GN SPL + ACC-SAP− − ?4Aa ADJ ADV German SOV/*NG ACC + ACC-SA − − 4A*, 6 Greek, ModernSVO/NG ACC + ACC-SA − − ?4A Hebrew, SVO/NG ACC + ACC-SA +b sp 2, ?4A ADV ADJ Modern NEGAN Hixkaryana OVS/GN − − ACC+ sp TNS SAP c Irish, Old V-1/NG ACC + ACC-SA − − Kurdish SOV/NG SPL + SPL-SAP + sp 6 Maori V-1/GprNG ACC −d − − − 2, 3′, 5 NEGNP Panjabi SOV/GN SPL + ACC-SA − − Persian SOV/NG ACC + ACC+ sp 6 SAPpr Polish SVO/*NG ACC + ACC-SA – − ?2, 4A* ADV ADJ RFL Pukapukan V-1/GprNG A/E + − − − 2 Rumanian SVO/NG ACC + ACC-SA − − Russian SVO/*NG ACC + ACC-SA − − 4A* ADJ Welsh V-1/GprNG ACC + ACC-SA +pr?e
< previous page
page_162
next page >
< previous page
page_163
next page >
Page 163 Passive ANCs Amharic SOV/GN ACC + ACC-SAP +f sp 2 English SVO/GNG ACCpr + ACC − − 2, 3A, 4A ASP NEG ADV Inuit SOV/GN ERG − ERG-SAP + sp 2′ Maori V-1/GprNG ACC −d − − − 2, 3A, 5 NEGNP Swahili SVO/NG − + ACC-SAP − − 2 NEG Turkish SOV/GN ACC (+) ACC-SA + sp 2 TNS Key: ?—rare pattern in the language !—pattern with problematic or indeterminate attribution Xpr—patterns relevant only for pronouns X*—patterns relevant only for highly referential nominals (in addition to pronouns) NG—besides the usual NG-order in NPs, there are possessive NPs with preposed highly referential Gs (including possessive pronouns). TNS, ASP—tense, aspect distinctions CRD—coordination of ANCs with non-derived NPs is allowed RFL—reflexive relations expressed as in finite clauses POL—politeness VLD—validator TOP—topic marker ADV, ADJ—adverbs, adjectives compatiable with ANs NEG—negation expressed as in finite clauses NEGAN—negation restricted to ANCs −—the category absent from ANCs Notes: a A rare pattern in Georgian, see n. 1 to this chapter. b See n. d to Table 6.1. c This is a somewhat vague description of verbal agreement in Hixkaryana, see section 7.3. d See n. e to Table 6.1. e I have no information on ANCs with pronominal arguments in Welsh. f See n. c to Table 6.1.
< previous page
page_163
next page >
< previous page
page_164
next page >
Page 164 4 Other patterns. The SENT, POSS-ACC, DBL-POSS, MIX and INC types. 5 Typological features of the languages. a mostly nominative-accusative languages, three ergative languages, three examples of split ergativity and one example of a non-defined pattern (Hixkaryana); b for languages with pure ERG-POSS ANCs, predominating word order V-1/NG (eight languages) and SVO/NG (seven languages), three examples of SOV/NG, four examples of SOV/GN, one example of SVO/GN and one OVS/GN language. Many examples of preposed possessive pronouns; c mostly predominantly dependent-marking languages with head-marking on the clause level, four examples of dependent-marking languages with no head-marking (Polynesian languages), four examples of split marking (Ancient Egyptian, Hebrew, Persian and Kurdish), two head-marking languages (Abkhaz and Hixkaryana) and one double-marking language (Classical Arabic). 6 Genetic and areal affinity of the languages. Caucasian (Abkhaz-Adygean and Kartvelian), IndoEuropean, Semitic, Austronesian and Carib languages, representing Eurasia, North Africa, Austronesia and Central America.
< previous page
page_164
next page >
< previous page
page_165
next page >
Page 165 8 Type 4: ‘Nominal’ 8.0. INTRODUCTION The Nominal type (NOMN) is characterized by genitivization of both S and A and assimilation of P to some dependents in non-derived NPs. This type exists in two varieties, the Double-Possessive (sub)type and the Possessive-Adnominal (sub)type. The Double-Possessive type (DBL-POSS) is characterized by genitivization of the Ss, Ps and As in ANCs. In most cases, the A and the P have the same marking. However, if a language has two formally distinct genitives, like English, the A and the P are marked differently. In most of the sample languages, ANs, combining with their arguments according to the DBL-POSS pattern, can be the heads of ANCs of other types. The Possessive-Adnominal type (POSS-ADN) is characterized by genitivization of the S and A in ANCs, while the P gets the same dependent-marking as some oblique NPs. 8.1. THE DBL-POSS SUBTYPE 8.1.1. Ideal representatives: English, Finnish and Japanese English distinguishes between two formally different genitives, the prenominal s-genitives and the postnominal of -genitives, which both serve as unmarked dependents of a head-noun with somewhat overlapping, but not exactly identical functions. In constructions with transitive action nominals, these two genitive positions are opposed to each other as subjective genitive and objective genitive respectively, at least when both arguments are present, as in Peter’s shooting of the prisoners . This is by no means the only type of ANCs with -ing-nouns
< previous page
page_165
next page >
< previous page
page_166
next page >
Page 166 in English, and in some cases, in particular when only one argument is present, the two genitives can have the same interpretation within ANCs (more about this is said in section 10.3.). However, the main tendency is clear. As shown in section 3.5.3., the prenominal s-genitive and the postnominal of -genitive, respectively, correspond to the subject and the object of clauses, and therefore they have these functions in ANCs. Discussing the differences between English and Russian ANCs—namely, that where English ANCs can be of several types, Russian ANCs follow only the ERG-POSS pattern—Comrie (1976:183) attributes them to the more restrictive internal structure of the Russian noun phrase overall: there is only one (normally postnominal) genitive position available, it may be filled by either a subjective or an objective genitive, but not both simultaneously. The ungrammaticality of *razrušenie goroda vraga (‘destruction city:GEN enemy:GEN’) is directly paralleled by the ungrammaticality, in English, of *the enemy’s city’s destruction and *the destruction of the city of the enemy in the appropriate sense. Thus, he suggests that the possibility of a language having ANCs of the DBL-POSS type is determined by the number of possible adnominal syntactic positions that are available in the language. Comrie defines the concept of different adnominal syntactic positions in a negative way, by showing the factors to which it cannot be directly related. Firstly, the differences in the linear order between the head of a noun phrase and its dependents do not always determine the number of unmarked adnominal syntactic positions (that is, the number of genitives) in a language. Thus, Comrie mentions that it is possible in Russian to put the genitive before the head, even if the opposite is the preferred order. This fact, however, ‘does not increase the number of possible genitive adjuncts to a given head noun, which remains at one’ (ibid.). Secondly, syntactically different adnominal dependents are not necessarily formally (morphologically) different. Thus, as Comrie writes, in German, prenominal and postnominal genitives represent different positions, though they are morphologically the same. Since they are syntactically different, they ‘both can be filled in an action nominal like: Herrn Dührings Umwälzung der Wissenschaft—‘‘Mr Dühring’s overturning of science”’, (ibid.: 184). Of particular importance here is the fact that in German, a given nominal may occupy only one of the two available positions within a noun phrase, so that DBL-POSS nominalizations in German have a limited range
< previous page
page_166
next page >
< previous page
page_167
next page >
Page 167 of application (a similar situation arises in languages with preposed possessive pronouns and postposed genitives; see section 10.2.). Thirdly, semantically (and morphologically) different genitives are not necessarily different syntactically. Thus, in Maori, the semantic distinction between Dominant and Subordinate possession (signalled by the prepositional particles a and o respectively) is maintained with action nominals, namely, the As take a, while the Ps take o (see section 7.2.2.). According to Comrie and Thompson (1985:383), ‘the a/o opposition in Maori is semantic rather than syntactic. In particular, the a and o genitives do not have distinct syntactic positions, so that we cannot have a single head noun qualified by both an a and an o genitive, i.e. *te patunga te tangata te wheke —“the killing of the man of the whale”.’ The idea that DBL-POSS nominalizations are only possible in languages with more than one unmarked adnominal dependent in non-derived NPs is appealing. However, it is not clear what criteria can, in general, be used to determine the number of such positions in a specific language. Thus, in section 7.2.1., we have seen Russian examples in which an AN combined with two genitive dependents, which, respectively, correspond to the accusative and the genitive objects of the finite verb. These constructions might argue against Comrie’s suggestion to consider only one adnominal genitive position in Russian ANCs. If not, what are the independent arguments for his proposal? There is a danger of getting into a vicious circle, that is, DBL-POSS nominalizations are possible in languages with several syntactically different genitives, while the fact that a language has DBL-POSS nominalizations can show that there are several syntactically distinct genitive positions within NPs of the language. English is a rare example of a language which has two formally distinct genitives occupying two different genitive positions within a noun phrase. A similar situation is found in Gbadi, a Kru (Niger-Congo) language. According to Koopman (1984:107–8), it has two prenominal markers, nI, which appears on constituents governed by [_N’], and na, which appears on constituents governed by [_N]. In ANCs, the P is introduced with nI, while the A and S appear with na (ibid.: 21–2).1 Most languages differ from English and Gbadi in this respect, namely, all unmarked adnominal dependents in them appear in the same form. This is also true for the sample languages with DBL-POSS ANCs, and, consequently, ANCs in most of them have both the A and the P in the same form. In principle, we could assume that this corresponds to one of the following situations: (a) all such languages have only one genitive position within NPs; or (b) all
< previous page
page_167
next page >
< previous page
page_168
next page >
Page 168 these languages have several genitive positions, but the form of all these dependents is identical; or (c) for some of the languages (a) is true, while for the others, (b) is true. Comrie tacitly assumes the second alternative. I do not have sufficient data for making a definite choice among these assumptions, which would be a topic for another typological study. Finnish, discussed below, presents a case of a language in which it is reasonable to distinguish between two structurally different but formally similar genitives. In Japanese, described below, genitives have a very wide range of application. However, I am not aware of any proposals to distinguish among structurally different types of Japanese genitives. In Finnish, an SVO-language, finite verbs agree with their subjects in person and number. Prototypical subjects are always in the nominative, while direct objects and subjects of existential verbs can choose between the partitive, accusative and nominative, the choice depending on certain semantic and syntactic conditions. (8.1a) Silja juo maidon. Silja:NOM drink milk:ACC ‘Silja is drinking the milk.’ (8.1b) Silja juo maitoa. Silja:NOM drink milk:PART ‘Silja is drinking milk.’ (Karlsson 1978:95) (8.1c) Osta kirja! buy:IMP book:NOM ‘Buy a book!’ (ibid.: 114) In non-derived NPs, the G is put in the genitive and precedes the H. Possessive suffixes appear on the H in the absence of an independent non-pronominal G in the same NP, which is, thus, the only case of head-marking in NPs. Though Finnish NPs normally manifest a strict order of constituents, genitive attributes seem to appear at quite different places with respect to the other adnominal dependents. It has been suggested by Jokinen (1991) that a distinction might be made between two different types of genitive attributes, which, though being formally identical, can modify one and the same head, occupy different positions and have different functions. The first genitive serves as a determiner, which specifies the reference of the head. The second, following the first genitive attribute and/or adjectival attributes, has characterizing, descriptive functions. In many cases,
< previous page
page_168
next page >
< previous page
page_169
next page >
Page 169 combinations of the descriptive genitive with the head-noun correspond to compound nouns or combinations of an adjectival dependent with the head-noun in other languages. (8.2) nuo muutamat Liisa-n ruskeat mäyräkoira-n those few Liisa-GEN brown dachshund-GEN pennut puppies ‘those few brown dachshund puppies of Liisa’s’ (Kristiina Jokinen personal communication) In ANCs with minen -nouns as a head, the S, P and A precede the head and are put in the genitive case. When both the A and the P are present, the A precedes the P. (8.3) [Vanhempien taloudellisen tuen antaminen] on parents:GEN economic support:GEN give:AN is riippuvaista tuloista. dependent incomes:PART ‘Parents’ giving of economic support is dependent on their incomes.’ (Hakulinen and Karlsson 1979:395, ex. 198) Constructions such as that quoted in example (8.3) are difficult to interpret, since the first genitive could also be understood as being dependent on the second one. Thus, the ANC Aimo-n kana-n syöminen (Aimo-GEN chicken-GEN eat-AN) is three ways ambiguous, namely: (a) ‘Aimo’s chicken eating’, (b) ‘the eating by Aimo’s chicken’, and (c) ‘the eating of Aimo’s chicken’ (Jokinen 1991:13). As a consequence of this, Finnish tries to avoid transitive ANCs with both arguments present. The two genitively marked arguments of transitive action nominals are similar to the two structurally distinct genitive attributes within non-derived NPs. Thus, the first of these arguments, the subjective genitive, might correspond to the genitive dependent of the determiner type, while the objective genitive might correspond to the characterizing genitive dependent. If this is true, Finnish is similar to English in having two distinct genitive positions available within noun phrases, which, in nominalizations, turn up as the subjective and the objective genitives. Japanese, a consistent SOV language, uses only dependent-marking within clauses and NPs. In finite clauses, formed according to the nominative-accusative pattern, the topic is marked with the postposition wa, which sometimes replaces the postposition which would
< previous page
page_169
next page >
< previous page
page_170
next page >
Page 170 otherwise signal the grammatical relation of the noun phrase in question. For non-topics, these postpositions include ga for subjects, o for direct objects and ni for indirect objects (the description is limited to clauses with intransitive and dynamic verbs, see Hinds 1986:191–5). (8.4) Taroo ga Hanako ni sono hon o yatta. Taroo NOM Hanako DAT that book ACC gave ‘Taro gave that book to Hanako.’ (Shibatani 1987:870, ex. a) In non-derived NPs, all the dependents precede the head noun. The G is marked with the genitive postposition no, which has many other functions and appears in phrases with quantity, material and part-whole relations. As a result, a nominal can have several genitive dependents. (8.5) watakushi no hon I GEN book ‘my book’ (8.6) renga no uchi brick GEN house ‘a house of bricks’ No -phrases also appear outside NPs. Thus, they can function as the subjects in certain types of embedded clauses, for example, in relative clauses (where they alternate with ga-phrases, the usual kind of subjects). (8.7) [NP[S sensei ga/no kat-ta]S hon]NP teacher NOM/GEN buy-PAST book ‘the book the teacher bought’ (Shibatani 1977:802, ex. 42) With respect to ANCs in Japanese, the actual situation is as follows: (1) ‘verbal nouns’ are not derived from verbs, rather, they give rise to predicates by combining with the auxiliary verb suru—‘to do’ (Martin 1975: section 14.3); (2) it is possible to derive abstract nouns from any infinitive by the suffix kata, but the resulting nominals denote the manner of performing the action, and not the action itself (ibid.: section 14.7), that is, they should rather be considered as ‘mode nominals’ (nomina modi, see section 2.2.3. above); (3) many nouns are derived from infinitives (Martin 1975: section 14.5), or, as Hinds describes it, ‘the stem of a verb is a nominal’ (Hinds 1986:364). In many cases such derived nominals mean a process, as tabe—‘eating’,
< previous page
page_170
next page >
< previous page
page_171
next page >
Page 171 or nomi —‘drinking’, and, thus, seem to be the best candidate for the category ‘action nominals’ in Japanese. Unfortunately, the available data are unclear as to what extent the derivation of them is regular (Hinds (ibid.) states only that it is a regular process). It is, however, clear that these nouns cannot be derived from all verbs and that it is not always possible to predict their exact meaning or range of meanings. Thus, it is not obvious whether Japanese has a class of words which satisfies all the criteria for action nominals. However, all three types of words described above behave in the same way with respect to their arguments—both the underlying subject (the S and the A) and the direct object (the P) are marked with the genitive postposition no . All the dependents precede the head of a nominalized construction. Of special interest is the fact that the mutual order of dependents seems to be irrelevant, as can be seen in the translations of the examples below. (All the cases of stacked genitives are omitted in the translations.) Verbal nouns (8.8a) Gakusya ga sizin o kenkyuu suru scholars NOM poets ACC study:VN do:PRES ‘Scholars study poets.’ (8.8b) Sizin ga gakusya o kenkyuu suru. poets NOM scholars ACC study:VN do:PRES ‘Poets study scholars.’ (8.8c) [Gakusya no sizin no kenkyuu] dá. scholars GEN poets GEN study:VN COP (1) ‘It is a study, by scholars, of poets.’ (2) ‘It is a study of scholars by poets.’ (Martin 1975:869) Infinitive-derived nouns (8.9a) Tanaka-shushō ga keisatsu o Tanaka-prime.minister NOM police ACC shiraberu investigate:PRES ‘Prime Minister Tanaka investigates the police.’
< previous page
page_171
next page >
< previous page
page_172
next page >
Page 172 (8.9b) Keisatsu ga tanaka-shushō o police NOM Tanaka-prime.minister ACC shiraberu. investigate:PRES ‘The police investigate Prime Minister Tanaka.’ (8.9c) [Tanaka-shushō no keisatsu no Tanaka-prime.minister GEN police GEN shirabe] dá. investigate:AN COP (1) ‘It is an investigation, by Prime Minister Tanaka, of the police.’ (2) ‘It is an investigation of Prime Minister Tanaka by the police.’ (Yukiyo Wallenström and Bengt Ek personal communication) Mode nominate (8.10) Watasi ga mé o miháttá no wa, I NOM eye ACC open.wide:PAST NMLZ TOP mázu dái-iti ni primarily one-number DAT/ADV [gakki é no káre-ra no nazimi-káta] instruments to GEN they GEN be:familiar-MN de aru. COP ‘What caught my eye was, first of all, the way they were familiar with their instruments.’ (Martin 1975) It can be suggested that the freedom in the mutual order of the arguments within nominalizations may be an argument against postulating several distinct genitive positions in Japanese noun phrases. 8.1.2. Other languages In both Finnish and Japanese, DBL-POSS is the only pattern of ANCs. This seems to be true for Gbadi, Ancient Greek and Latvian as well. In English, Estonian and Latin action nominals which can combine with their dependents according to the DBL-POSS type, can also form constructions of other types. For Latin, Woodcock (1959) also mentions nominalizations with the Ps introduced by one of the prepositions erga, adversus —‘opposite’, in—‘towards’ and the As in the genitive case, which is an example of the POSS-ADN type. In
< previous page
page_172
next page >
< previous page
page_173
next page >
Page 173 Estonian, action nominals can be the head of ERG-POSS, and INC ANCs, while in English they can also choose among the pure ERG-POSS, POSS-ACC and passive types. In the eight languages mentioned so far, DBL-POSS nominalizations are quite common, and, in some cases, they constitute the only pattern of ANCs available in the language. However, in yet other languages this pattern, though possible, is considered as substandard. Thus, for Modern Greek the grammars (Mackridge 1985, Joseph and Philippaki-Warburton 1987) quote ANCs of the ERG-POSS and POSS-ADN types and mention that it is not possible to have two genitives in the same ANC (except for the cases with pronominal As, see section 10.2.). However, translating the questionnaire, my informant consistently gave DBL-POSS nominalizations as one of the alternatives for the transitive ANCs. The same is true for Georgian and Modern Israeli Hebrew (see chapter 7, notes 1, 3). This uncertainty in the degree of acceptability and naturalness of the data makes it impossible to draw any definite conclusions about the general properties of the languages with the DBL-POSS nominalizations. Some more languages have DBL-POSS ANCs with pronominal or other highly referential arguments (see section 10.2). The boundary between such languages and those discussed here is rather vague, as even ‘classical’ DBL-POSS languages, such as English, can have restrictions on the referentiality of the arguments in such nominalizations. Some of the DBL-POSS languages, such as Latin and Modern Greek, tend to differentiate between the pronominal A and P in ANCs—the former appears as a possessive pronoun, while the latter is coded by the genitive of the pronoun. This might be interpreted as an argument for the existence of several genitive positions in all other types of noun phrases. (For the similar discussion in connection with the ERG-POSS type see section 7.2.4.) On the whole, the connection between ANCs of the DBL-POSS type and the number of unmarked adjuncts within non-derived noun phrases remains only a hypothesis, though plausible. All the languages under consideration are complement-balancing. The predominating word order is SVO with either genitives following the head (Modern Greek, Modern Hebrew), or preceding the head (Estonian and Finnish). English has both prenominal and postnominal genitives. Basic word order in Japanese, Latin and Georgian is SOV. Ancient Greek had free order with predominating SOV and SVO orders. Gbadi alternates between SVO and SOV word order
< previous page
page_173
next page >
< previous page
page_174
next page >
Page 174 Table 8.1 Languages of the DBL-POSS type (see comments to Table 5.1 in section 5.5., pp. 105–8) Language Word order Case marking Balanced Head-marking Other FC/NP system complements FC NP ANC patterns English SVO/GNG ACCpr + ACC-SA − − 2, 3B, 3′ Estonian SVO/GN ACC + ACC-SA − − 3B, 6 Finnish SVO/GN ACC + ACC-SA + ?a Gbadi SOV/GN − + − − − ?Georgian SOV/GN SPL + ACC- − − 3B SAP Greek, Ancient SVO/SOV/NG ACC + ACC-SA − − (?b )Greek, SVO/NG ACC + ACC-SA − − 3B Modern ?Hebrew, SVO/NG ACC + ACC-SA +/ ? 2, 3B Modern −c Japanese SOV/GN ACC + − − − Latin SOV/GN/NG ACC + ACC-SA − − Latvian SVO/GN ACC + ACC-SA − − Restricted to highly referential As Bulgarian SVO/*NG ACCpr + ACC-SA − − 3B, ?5 Dutch SOV/*NG ACCpr + ACC-SA − − 3B, ?5 Frenchd SVO/GprNG ACCpr + ACC-SA − − 3B German SOV/*NG ACC + ACC-SA − − 3B Polish SVO/*NG ACC + ACC-SA − − ?2, 3B Russian SVO/*NG ACC + ACC-SA − − 3B Samoand V-1/GprNG ERG + − − − 3A
< previous page
page_174
next page >
< previous page
page_175
next page >
Page 175 Key: ?—Rare pattern in the language Xpr—patterns relevant only for pronouns X*—patterns relevant only for highly referential nominals (in addition to pronouns) *NG—besides the usual NG-order in NPs, there are possessive NPs with preposed highly referential Gs (including possessive pronouns). ADV, ADJ—adverbs, adjectives compatible with ANs Notes: a I have no examples on Finnish ANCs with pronominal arguments. b In standard use, the DBL-POSS type is restricted to ANCs with pronominal As. c In the two examples of DBL-POSS ANCs in Modern Israeli Hebrew, quoted by Berman (1978:131–2) as being accepted by colloquial usage, the H is marked either for the A or the P, but not for both. Presumably, two ‘separate’ forms (see section 7.2.2.) may also be used in such ANCs. d As in DBL-POSS ANCs are restricted to pronouns.
< previous page
page_175
next page >
< previous page
page_176
next page >
Page 176 depending on tense, aspect and mood of the clause, but its basic word order, according to Koopman (1984), is SOV. The main features of the DBL-POSS languages are summarized in Table 8.1. For the explanation of the symbols in the table see the comments to Table 5.1. in section 5.5., pp. 105–8. 8.2. THE POSS-ADN SUBTYPE 8.2.1. Swedish Swedish, an SVO language, uses only dependent-marking in clauses and noun phrases. In finite clauses, only pronouns distinguish between the subject and object forms (the nominative and accusative cases respectively). (8.11) Nazityskland införlivade Danzig. ‘Nazi-Germany incorporated Danzig.’ In non-derived noun phrases, the G in the genitive precedes the H, while all other nominal dependents (prepositional phrases) follow it. The prenominal genitive has a wide range of uses and in many cases may be replaced by some postnominal prepositional dependent. (8.12) flick-or-s/flick-or-na-s böcker girl-PL-GEN/girl-PL-DEF-GEN books ‘(the) books of girls/the girls’ books’ Swedish has at least two types of action nominals—those with the suffix -ande/-ende, which are built regularly from any verb stem and could probably be considered as verb forms, and those with the suffix -(n)ing, which show a number of various idiosyncratic features and, thus, should be classified as derived nominals. As concerns their combination with the S, A and P, they behave identically, for the most part. That is, the S and the A normally turn up as prenominal genitives, while the P is marked with the preposition av and follows the head. Thus, the word order within ANCs is the same as in finite clauses, A-AN-P. (8.13) [Nazityskland-s införliv-ande av Danzig] Nazi.Germany-GEN incorporate-AN of Danzig utlöste Andra Världskriget. triggered Second World. War ‘The incorporation of Danzig by Nazi-Germany triggered the Second World War.’
< previous page
page_176
next page >
< previous page
page_177
next page >
Page 177 The example above seems to be quite similar to English ANCs of the DBL-POSS type. However, even if the Swedish av-phrases are reminiscent of the English of -genitives, it is not justified to talk about postnominal genitives in Swedish. While, in English, of -phrases constitute by far the largest part of postnominal attributes and may replace prenominal s-genitives in the vast majority of cases, this is not the case in Swedish. As shown by Pitkänen (1979), prenominal genitives in Swedish may be replaced by different adnominal prepositional attributes depending on the exact meaning of the genitive ‘hypotagm’. The exact substitution rules are additionally complicated by lexical and syntactic factors. In the majority of cases, av-dependents cannot replace prenominal genitives (and the use of this preposition has decreased in Modern Swedish during the last few centuries), and any other preposition could hardly be considered as introducing unmarked adnominal adjuncts either. 8.2.2. Other languages Another language in the sample which regularly uses the POSS-ADN type of nominalizations is Icelandic. The P in such constructions turns up as a prepositional phrase with the preposition á—‘on, at’ (which governs the dative case). In both Swedish and Icelandic, the above mentioned prepositions are used with action nominals derived from (almost) all transitive verbs, in contrast to other languages, such as Russian and Dutch (see section 1.4.1.), in which the choice of a preposition is determined by the semantics of action nominals. My data suggest two other candidates for languages with POSS-ADN ANCs—Egyptian Arabic and Latin. In Egyptian Arabic (see section 6.2.), the P in transitive ANCs with two arguments present may appear in two forms—either in the non-marked form or preceded by the preposition li —‘to, for’, which may introduce indirect objects in clauses. Thus, in ANCs, the P either retains its sentential form, or turns up in the same form as indirect objects or benefactives in clauses. In other words, Egyptian Arabic alternates between the POSS-ACC and the POSS-ADN types. Different sources present different answers to the question of whether there is any difference between those with li and those without li . Gary and Gamal-Eldin (1982) write nothing about the possibility of having li in such cases and the only available pattern mentioned by them has two non-marked dependents. Mitchell (1962:100) writes that ‘in accordance with the rule that a noun may not be defined more than once, when the verbal noun governs either two nouns or a pronominal
< previous page
page_177
next page >
< previous page
page_178
next page >
Page 178 suffix and a noun then the particle li must be included before the final noun’. Wise (1975:81) and the reference grammar of Egyptian Arabic (Abdel-Massih et al. 1979, vol. 3:313) describe the two constructions as being in free variation with each other. In most of the sentences offered to the informant, ANCs with non-marked Ps and with li -Ps can be interchanged. However, in some situations, one of these possibilities is preferred. Only one dependent in ANCs, the last one, may be introduced with li, and in combination with action nominals derived from ditransitive verbs, there are two dependents in the non-marked form. In such constructions, the last two dependents correspond to the direct and indirect objects in the underlying finite clauses, where the order is either DO IO (the IO being introduced by li ), or IO DO (both noun phrases in the non-marked form). What makes Egyptian Arabic rather special as compared to Swedish and Icelandic, is the fact that the preposition li is commonly used in finite clauses as a marker of indirect object or other types of oblique constituents. In contrast to this, the use of av-PPs in Swedish clauses is much more restricted than in NPs, while á-PPs in Icelandic appear freely both in NPs and clauses, but only with locative, semantic functions. That is, ANCs involving li -Ps in Egyptian Arabic can perhaps be regarded as a counterpart to ANCs of the POSS-ACC type, in the same sense as SENT-POSS ANCs are regarded as a counterpart to OBL-POSS ANCs. In the latter case, the P genitivizes, while the A either retains its sentential dependentmarking (in the SENT-POSS type) or has the same marking as ‘downgraded’ As (in the OBL-POSS type). In the former case, the A genitivizes, while the P either retains its sentential dependent-marking (in the POSS-ACC type), or has the same marking as non-direct objects (in the Egyptian variant of the POSSADN type). However, it is unclear to what extent this analogy is justified. More specifically, though the As in OBL-POSS ANCs have the same marking as the transitive subjects demoted by other ‘advancement’ rules (see sections 7.2.5 and 7.3.), it is not at all certain that Egyptian Arabic has any similar processes by which direct objects are demoted and marked with the preposition li . One of the possible candidates for such a rule is the rule of Dative shift mentioned above, but it leaves the demoted direct object unmarked. For Latin, Woodcock (1959) writes that the P may be coded with one of the prepositions erga, adversus —‘opposite’ or in—‘towards’, but he does not say to what extent such constructions are regular (that is, whether all transitive action nominals can combine with such prepositional phrases as the P), nor which factors determine the choice among these prepositions. Due to this lack of information,
< previous page
page_178
next page >
< previous page
page_179
next page >
Page 179 Table 8.2 Languages of the POSS-ADN type (see comments to Table 5.1 in section 5.5., pp. 105–8) Language Word order Case marking Balanced Head-marking Other Other FC/NP system complements FC NPANC patterns features Arabic, V-1/NG − + ACC+ sa 2 ADV ADJ Egyptian SAPpr NEGAN Icelandic SVO/NG ACC + ACC-SA − − 6 ADJ Swedish SVO/GN ACCpr + − − − 6 ADJ Key: Xpr—patterns relevant only for pronouns ADV, ADJ—adverbs, adjectives compatible with ANs NEGAN—negation expressed in a special way restricted to ANCs
< previous page
page_179
next page >
< previous page
page_180
next page >
Page 180 Latin will not be included among the POSS-ADN languages. Many more languages have nominalizations of the POSS-ADN type restricted to certain types of verbal nouns, where the choice of the preposition to mark the P depends on the semantics of the head-noun. In such languages ANCs manifest lexical splits (see section 10.1.). The data on Swedish, Icelandic, and Egyptian Arabic are summarized in Table 8.2. For the explanation of the symbols in the table see comments to Table 5.1 in section 5.5., pp. 105–8. 8.3. SUMMARY 0 Definition of the NOMN type. In the Double-Possessive type, genitivization of the S, A and P. In the Possessive-Adnominal type, genitivization of the S and A and assimilation of the P to oblique NPs. 1 Further syntactic devices: a head-marking. For the DBL-POSS type, (almost?) complete absence of head-marking in NPs. For the POSS-ADN type, use of nominal head-marking in Egyptian Arabic; b word order. Assimilation to the word order in NPs. 2 Retention of sentential features. Retention of certain adverbs in Hebrew and Egyptian Arabic. 3 Nominal features. Compatibility with adjectives in most languages. 4 Other patterns. The POSS-ACC, ERG-POSS (OBL-POSS), MIX and INC types. 5 Typological features of the languages: a exclusively nominative-accusative languages; b predominating word orders—SVO/NG and SVO/GN, no examples of V-initial languages; c exclusively complement-balancing languages; d preference for dependent-marking: with no head-marking (Japanese, Swedish) and with verbal agreement (English, Estonian, Georgian, Ancient and Modern Greek, Icelandic, Latin and Latvian). Three examples of head-marking in NPs (Modern Israeli Hebrew, Egyptian Arabic, Finnish), one example of a language without marking on clause level (Gbadi). 6 Genetic and areal affinity of the languages. Afro-Asiatic, Altaic, Indo-European, Kartvelian (marginal), Niger-Congo and Uralic language families.
< previous page
page_180
next page >
< previous page
page_181
next page >
Page 181 9 Less frequent types 9.0. INTRODUCTION This chapter describes the four nominalization patterns that are much less frequent among the languages in the sample than the preceding four. The Mixed type is a hybrid between the PossessiveAccusative and Possessive-Ergative types described in chapters 6 and 7. Its existence contradicts the generalization that, in ANCs, either all the arguments retain their sentential dependent-marking, or at least one of them genitivizes. The Incorporating, the Relative and the more or less hypothetical Argument-Reducing types use ‘valency-lowering’ strategies, in the sense that the transitive action nominal has only one external argument, its other argument either building a compound with it, being expressed within a relative clause referring to the action nominal, or being absent altogether. 9.1. TYPE 5: ‘MIXED’ The Mixed type (MIX) is characterized by genitivization of the S, assimilation of the A to some oblique NP (perhaps the agent in passive sentences) and retention of the sentential marking for the P. The sample includes four languages with this type of nominalization, Maori, Bulgarian, Dutch and Thai, where in at least two of them (Bulgarian and Dutch) such ANCs are extremely marginal. The Maori case was described in section 7.2.2. It should be clear from the description that MIX nominalizations are by no means common and should rather be considered as something intermediate between the POSS-ACC and ERG-POSS types, both of which exist in the language. In Dutch, according to Hoekstra (1986), ANCs with ‘bare’ infinitives can follow either the ERG-POSS or the MIX patterns. However, several other Dutch informants consider ANCs of the
< previous page
page_181
next page >
< previous page
page_182
next page >
Page 182 MIX type as highly exceptional, if possible at all. In Bulgarian, MIX nominalizations are very rarely exploited in ‘bureaucratic’ style as an alternative to the ERG-POSS type under certain conditions. More specifically, the preposition na—‘of’, which normally introduces the P to ANs, may be omitted when the P itself has an additional dependent introduced by the same preposition (Ivan Derzhanski personal communication). Thus, in the combination uveličav-ane-to na proizvodstvo-to na fabriki-te —increaseAN-DEF.N of production-DEF.N of factories-DEF.PL (‘the increase of the production of the factories’), the first na may be omitted and the P appears in the same non-marked form as in the corresponding finite clause. The situation in Thai is somewhat different, since it does not build action nominal constructions of the ERG-POSS type. Thai, an SVO language with no inflectional (and almost no derivational) morphology, signals the difference between the subject and the direct object only by means of word order. (9.1) he buy food ‘He buys food.’ (Hudak 1987:768) In non-derived noun phrases, the G either follows immediately after the H, or is separated from the latter by means of the preposition —‘of’. (9.2) dèk book (of) child ‘child’s book’ (ibid.: 770) Action nominals are built by combining one of the compounding head-nouns with the verb to be nominalized. The two most common and productive head-nouns are the noun kaan —‘activity, affair’ and khwaam—‘meaning, message’. These are used with dynamic and stative verbs respectively with the meaning ‘the action of’ and ‘the state of’. Some verbs can build nominalizations of both types with different meanings. Transitive action nominals take Ps in the same non-marked form as in clauses. The A appears less frequently, either with the possessive particle , or with the preposition dooy—‘by’.
< previous page
page_182
next page >
< previous page
page_183
next page >
Page 183 Table 9.1 Languages of the MIX type (see comments to Table 5.1 in section 5.5., pp. 105–8) LanguageWord order Case marking Balanced HeadOther Other FC/NP system complements marking patterns features FC NPANC ? SVO/*NG ACCpr + ACC- − − 3B, 4A* Bulgarian SA ?Dutch SOV/*NG ACCpr + ACC- − − 3B, 4A* ADV SA Maori V-1/GprNG ACC −a − − − 2, 3B, 3′ NEGNP Thai SVO/NG − + − − − 2 Key: ?—Rare pattern in the language Xpr—patterns relevant only for pronouns X*—patterns relevant only for highly referential nominals (in addition to pronouns) *NG—besides the usual NG-order in NPs, there are possessive NPs with preposed highly referential Gs (including possessive pronouns). ADV, ADJ—adverbs, adjectives compatible with ANs NEGNP—negation expressed as in NPs Note: a See n. g to Table 6.1.
< previous page
page_183
next page >
< previous page
page_184
next page >
Page 184 (9.3a) mìi dooy naay-phraan AN-kill bear by hunter ‘the killing of bears by hunters’ (9.3b) mìi naay-phraan AN-kill bear of hunter ‘the killing of bears by hunters’ (Anthony Diller and Preecha Juntunamalaga personal communication) As seen in the description above, Gs in Thai often appear in the same non-marked form as the subjects and direct objects, which makes the exact interpretation of examples (9.3a–9.3b) somewhat difficult. Thus, it would be possible to treat the P in Thai ANCs either as having genitivized (in which case (9.3a) is an example of the ERG-POSS type, while (9.3b) represents the DBL-POSS type) or as retaining its sentential form (in which case (9.3b) should be interpreted as a POSS-ACC nominalization, while (9.3a) manifests the MIX type). However, in contrast to the G in non-derived NPs, the P in ANCs cannot be introduced with the possessive particle. In general, when the P in ANCs appears in the same form as in the corresponding finite clauses (even if this is the non-marked form), it is only natural to see this as a retention of the sentential marking, when there are no clear counter-arguments, such as the form of the head-marking (cf. with Egyptian Arabic in section 6.2.). We may, however, suggest that this common identity in the form of the direct objects and that of the Gs motivates the existence itself of MIX ANCs in Thai. According to Anthony Diller and Preecha Juntunamalaga (personal communication), constructions similar to those in (9.3a–9.3b) are ‘bureaucratic’ or journalistic, not used much in normal colloquial Thai nor in more ‘artistic’ literature. Some authorities would consider such phrases to be ‘translationese’ based on western models. The data on the MIX languages in the sample are summarized in Table 9.1 (for the explanation of the symbols see the comments to Table 5.1., section 5.5, pp. 105–8.). 9.2. TYPE 6: ‘INCORPORATING’ In the Incorporating type (INC), the P forms the first part of the complex action nominal. Nominalizations of this type are ‘valency-lowering’ in the sense that their head nominals, derived from transitive verbs, have only one argument (of the set (A, P)), as compared with the corresponding
< previous page
page_184
next page >
< previous page
page_185
next page >
Page 185 verbs with two. The other argument, the P, is compounded with the action nominal to give rise to a larger action nominal. This process is reminiscent of noun incorporation, in which compounding a nominal stem together with a verbal stem results in a larger verbal stem. According to Mithun (1984:847), all languages with noun incorporation into verbs have syntactic paraphrases for such morphological structures. Applied to ANCs, this is true for the majority of the sample languages, which have nominalizations of other types as well. The only language in the sample which resorts solely to this pattern is Ewe. Hungarian uses exclusively ‘valency-lowering’ strategies, as its ANCs follow either the INC, or the REL patterns. In Ewe (Westermann 1930), a Kwa language, the syntactic relations between a finite verb and its nonpronominal subject and objects are signalled by means of word order alone, SVO. Pronouns, however, have bound forms which distinguish between subjective and objective functions. (9.4a) atí lá 1SG.SBJ:see tree DEF ‘I saw the tree.’ (9.4b) lá dí xewízìwó child DEF look, for birds’.eggs ‘The child was looking for birds’ eggs.’ (Westermann 1930:48) In non-derived NPs, the G precedes the head and is separated from it by means of the particle (which may be omitted in some cases). (9.5) fia chief of house ‘the chief’s house’ (ibid.: 48) In ANCs, the S and the A genitivize, while the P is compounded with the AN, ‘an objective possessive…is generally combined with its verb-noun to form one word’ (ibid.: 124). (9.6a) didí child of seek:AN ‘the seeking of the child’ (that is, the child is seeking)
< previous page
page_185
next page >
< previous page
page_186
next page >
Page 186 (9.6b) child:seek:AN ‘the seeking for a child’ (ibid.: 124) It is worth mentioning that ANs derived from verbs with two objects (both of which follow the verb and are in the non-marked form) build compounds with their ‘near’ object, while the ‘remote’ one follows the compound. (9.7a) tó dziláwo hold ear parents ‘hold ear towards parents=obey parents’ (9.7b) dziláwo he:of ear:hold:AN parents ‘his obedience to the parents’ (ibid.) For Ewe, Westermann explicitly states that the P and the verbal noun in ANCs form a compound. The situation in some other languages of the Kwa group is somewhat less clear. Thus, in ANCs in Ifè, also an SVO-language (sometimes classified as a dialect of Yoruba), the P precedes the verbal noun, although in non-derived NPs the dependents follow the head (Hélène Boëthius personal communication). My source does not explicitly state any process of compounding in this case and I have not found any other indication in favour of this. For example, the Ps and the verbal nouns in ANCs are written separately and the Ps can contain different types of determinants. Accordingly, ANCs in Ifè should rather be considered as belonging to the POSS-ACC type, since the P appears in the same nonmarked form as in the corresponding finite clause. In this case, these ANCs constitute an exception to one of the generalizations mentioned in chapter 3: ‘Head-dependent order is the same in ANCs and nonderived NPs’ (where ‘dependents’ refer to external dependents and not to parts of compounded ANs). Hyman (1975:132–3) quotes nominalizations in some Kwa languages, such as Nupe, as an argument against Givón’s hypothesis that ‘the syntactic order of sentences creeps into the noun phrase via nominalization’ (we will return to this hypothesis in section 12.1.). Nupe has changed from being an SOV language to being an SVO language. It has a productive process of nominalization, its only structural effect being the change of the order between the verb and its object, from VO, like —‘buy yams’ to OV, like in —
< previous page
page_186
next page >
< previous page
page_187
next page >
Page 187 ‘yam-buying’. The order in NPs is NG (like in Ifè). Action nominals form compounds with their Ps, as is suggested by Hyman’s spelling and by the explicit reference to them as compounds in Givón’s later work (1984:217), and, thus, Nupe ANCs do not contradict the generalization stated above. According to Hyman, ‘although structurally ‘‘yam-buying” is mod-N (more specifically, poss-N), when ordinary possessives (genitives) were subjected to the word order change (so that my yams became yams my), nominalizations were exempted’. Hyman writes that this resistance of nominalizations to the change from the order GN to NG is not surprising, since otherwise it would be impossible to distinguish a nominalization from the corresponding verb-object combination. In the case of Ewe and Ifè, however, verbal nouns are formed by reduplication and are quite distinct from the corresponding finite verbs, which reduces the relevance of word order differences between finite clauses and nominalizations. In the above mentioned languages, the A (and the S) in ANCs genitivize. This is by far the most common, though not the only case in nominalizations of the Incorporating type. Another variant of this type is presented by Imbabura and Cuzco Quechua, in which the A retains its sentential dependentmarking (for examples see section 5.5). The two subtypes of the INC type will be referred to as the Possessive-Incorporating (POSS-INC) and the Sentential-Incorporating (SENT-INC) types respectively. Recall that both Imbabura and Cuzco Quechua have also ANCs of the SENT type, where both the S and A are in the nominative case. Thus, complex action nominals behave just like intransitive action nominals in taking an argument with the sentential dependent-marking. The similarity in the treatment of the S and the A of INC ANCs is especially obvious in Cuzco Quechua. In contrast to Imbabura Quechua, this language allows for genitivization of the S in ANCs, which is also allowed for the A of transitive action nominals, but only of the INC type (more specifically, when the P is in the non-marked form). It has been repeatedly noted that noun incorporation in verbs can have different syntactic consequences (see, e.g. Rosen 1989 and the references there). More specifically, while in certain languages an incorporated noun satisfies one of the argument positions of the verb (and, thus, reduces its valency), in others the incorporation of an object into a verb does not affect its transitivity. The first variant is represented by Chukchee, a Paleosiberian language, in which transitive verbs with unincorporated objects pattern ergatively and agree with both the A (in the ergative case) and the P (in the absolutive case), while the same verbs with incorporated objects agree only with
< previous page
page_187
next page >
< previous page
page_188
next page >
Page 188 the A in the absolutive case (see section 9.4.). The second variant is represented by many Native American languages, like Seneca, a Northern Iroquoian language, in which transitive verbs both with unincorporated and with incorporated objects take the same agreement prefixes indicating agreement with the subject and the object (Rosen 1989:302). It may be suggested that ANCs of the INC type manifest a somewhat similar range of variation. Thus, in most of the INC languages, the incorporation of the P reduces the valency of the AN in the sense that the A of a complex action nominal patterns with the S of intransitive action nominals, while the A of transitive action nominals with unincorporated Ps patterns differently. Estonian has another variant of INC ANCs, in which the P forms a compound with the action nominal and the A is marked with the postposition poolt —‘by’, in the same way as the agent in (stative) passive clauses and the A of ERG-POSS ANCs, which also exist in Estonian. Thus, ANs of this type are still treated as transitive, at least as regards their relation to A. The analogy to finite clauses with incorporation in Seneca is only partial, since ANCs of this type, which may be called the Oblique-Incorporating type (OBL-INC), cannot combine with a new P and in this respect manifest a reduced valency. OBL-INC ANCs in Estonian coexist with POSS-INC nominalizations. The following examples show ANCs of the ERG-POSS, POSS-INC and OBL-INC types (9.8a, 9.8b and 9.8c respectively), as well as the corresponding passive finite clause (9.8d). (9.8a) maja-de ehita-mine Peetr-i poolt house-PL.GEN build-AN Peter-GEN by ‘the building of houses by Peter’ (9.8b) maja-ehita-mine Peetr-i poolt house-build-AN Peter-GEN by (9.8c) Peetr-i maja-ehita-mine Peter-GEN house-build-AN ‘Peter’s housebuilding’ (9.8d) Maja on ehitatud Peetri poolt house:NOM is built Peter:GEN by ‘The house is built by Peter.’ (Diana Krull personal communication) In all the sample languages with INC ANCs, verbs do not normally incorporate their Ps and compounding characterizes the process of nominalization itself. There is also a theoretical possibility that a
< previous page
page_188
next page >
< previous page
page_189
next page >
Page 189 language with productive incorporation within the realm of finite verb forms can derive complex action nominals from the corresponding complex verbs. The only incorporating language in the sample which has action nominals is Chukchee.1 However, it remains unclear for me whether action nominals can incorporate their Ps. It is difficult to form a clear picture of the distribution of this pattern. For Quechua, the existence of the pattern is explicitly stated in the grammars (see section 5.5. for examples), but nothing is said about the range of its application. Some of the sentences in my questionnaire have been translated with INC nominalizations into Persian, Kurdish, Hungarian, Estonian, Swedish and Icelandic, but it is difficult to determine to what degree such combinations are lexicalized. The ability of a language to form such constructions seems to be correlated with its general ability to form N+N compounds. Thus, in Russian, where such compounds are extremely rare, action nominals with incorporated Ps represent a very limited class of words (often referring to highly ritualized activities, cf. čaepitie—‘tea-drinking’ and kofepitie —‘coffee-drinking’, but not *sokopitie —‘juice-drinking’) and the process of building such words is not productive at all. In Swedish, N+N compounds are quite common and the process of building INC nominalizations is highly productive. Kurdish is rather special as regards its compounding. In the majority of N+N compounds, it is the second constituent which functions as an attribute, or modifier of the first one, as in –‘cousin’ (= −‘son’+ap −‘uncle’). There exists, however, a class of complex nouns in which it is the first constituent stem which is an attribute to the second one, xane −‘house’, as in kitêbxane−‘library’ (= kitêb−‘book’+ xane −‘house’). According to Bakaev (1973:240), these compounds are untypical for Kurdish, and it would be more ‘correct’ to construct such words with the rearranged constituents. Complex action nominals follow the latter, unusual pattern, as their first nominal part refers to the P of the second part, derived from the verb. The same is true for other types of complex nouns where one of the constituents is a verbal derivative, like soldirû −‘a shoemaker’ (= sol − ‘shoe’+dirû−PAST of the verb dirûn −‘to sew’) (ibid.: 241). For the languages which can choose between the INC type and some other nominalization type(s), it still remains a question exactly when this process can take place, that is, what factors favour the choice of the INC pattern. It is clear that among the determining factors are the referential and formal characteristics of the P and the kind of the activity denoted by the action nominal. Thus, the Ps in such constructions tend to be non-referential and non-individuated, they cannot
< previous page
page_189
next page >
< previous page
page_190
next page >
Page 190 Table 9.2 Languages of the INC type (see comments to Table 5.1 in section 5.5., pp. 105–8) Language Word order Case marking Balanced Head-marking Other Other FC/NP system complements FC NPANC patterns features SENT-INC Quechua, SOV/GN ACC (+)a ACC-SAP+ sa 1, 6B TNS-VLD Cuzco Quechua, SOV/GN ACC −b ACC-SA − − 1 TNS-VLD Imbabura POSS-INC Estonian SVO/GN ACC + ACC-SA − − 3B, 4A, 6C Ewe SVO/GN ACCpr − − − − German SOV/*NG ACC + ACC-SA − − 3B, 4A* Hungarian SVO/GNc ACC + ACC-SA + sa 7 Icelandic SVO/NG ACC + ACC-SA − − 4B Kurdish SOV/NG SPL + SPL d + sa 3B Persian SOV/NG ACC + ACC+ sa 3B SAPpr Quechua, SOV/GN ACC (+)a ACC-SAP+ sa 1, 6A TNS-VLD Cuzco Swedish SVO/GN ACCpr + − − − 4B OBL-INC Estonian SVO/GN ACC + ACC-SA − − 3B, 4A, 6B Key: Xpr—patterns relevant only for pronouns X*—patterns relevant only for highly referential nominals (in addition to pronouns) *NG—besides the usual NG-order in NPs, there are possessive NPs with preposed highly referential Gs (including possessive pronouns) TNS—tense distinctions VLD—validator Notes: a See note l to Table 5.1. b See note m to Table 5.1. c Hungarian has relatively free word order with slightly predominating SVO-order. d In Kurdish, verbs agree with one argument according to the split-ergative pattern.
< previous page
page_190
next page >
< previous page
page_191
next page >
Page 191 usually have dependents of their own, while the whole process of compounding reflects the degree to which the activity is institutionalized and is ‘name-worthy’ in its own right (Mithun 1984:848). In section 8.1., I mentioned the ambivalent character of genitive attributes in Finnish. Those genitives that stand closest to the head nominal have characterizing, descriptive functions, and constructions with such attributes are reminiscent of complex nouns in other languages. This is by no means unique for Finnish. The existence of such cases shows that ANCs of the INC and of the DBL-POSS types may be essentially quite similar. The features of the languages with INC ANCs are summarized in Table 9.2. (As mentioned above, it is difficult to obtain information on which languages have a regular process of forming such nominalizations.) 9.3. TYPE 7: ‘RELATIVE’ In the Relative type (REL), the P of transitive action nominals is marked as its direct dependent, while the A is expressed within the relative clause referring to the action nominal. Only two languages in the sample, Hausa and Hungarian, manifest this pattern of action nominal constructions. This strategy in forming ANCs is considered as ‘valency-lowering’ in the sense that a verbal noun derived from a transitive verb may have only one overt argument directly depending on it, at the same time as the corresponding finite verbs have two. In Hausa, an SVO language, verbal nouns have a somewhat extended range of uses which depends upon the general characteristics of verbs in the language. The verb itself is not inflected for tense, mood or aspect, but receives these characteristics through preverbal pronouns which have special markers attached to them. These pronouns are obligatory in all verbal sentences, even in the presence of an overt lexical subject, which in this case precedes the corresponding pronoun. For non-pronouns, neither the subject, nor the direct object take any marking.
< previous page
page_191
next page >
< previous page
page_192
next page >
Page 192 (9.9a) máháukácìyáa táa káshèe shí dà wúkáa crazy.woman 3F.PRFV kill he.OBJ with knife ‘The crazy woman (she) killed him with a knife.’ (Newman 1987:721) (9.9b) ya 3M.PRFV take money ‘He took money.’ (Ščeglov 1970:174. Unfortunately, this source does not indicate tonal differences in the examples.) Non-derived noun phrases have the order NG and use exclusive head-marking for signalling possessive (and many other types of) relations, namely, the head gets the ‘linker’ -n or attached to it, the choice between the two being determined by the gender of the head noun. (9.10) kàáká-n yáaròo grandfather-of boy ‘the boy’s grandfather’ (Newman 1987:720) Verbal nouns in Hausa replace non-inflected verbs in the continuous tenses. However, they can also be used in all syntactic positions typical for NPs and have all the inflectional features characteristic for nonderived nouns. In ANCs, the dependents follow the head. Combined with either the S, or with one of the arguments A or P, the verbal noun takes the ‘linker’, so that the resulting ANC totally assimilates to non-derived possessive phrases. When both the P and the A are present, the A is expressed within the relative clause with the verb yi—‘to do’, which refers to the action, denoted by the verbal noun, and indicates the time of the action. In other words, English constructions of the form ‘X’s V-ing of Y’ in Hausa correspond to constructions of the form ‘V-ing of Y that X did/does/will do, etc.’. What exactly happens with the P depends on the verb from which the action nominal is derived. The action nominal either gets the linker (that is, the combination AN-P assimilates to possessive NPs), or not; if not, the P follows it either in the non-marked form, or in the form of the indirect object. The same changes accompany derivation of complex verbs with yi from simple ones, with the meaning of ‘generalized action’ (Ščeglov 1970:174, 253, 273).
< previous page
page_192
next page >
< previous page
page_193
next page >
Page 193 Table 9.3 Languages of the REL type (see comments to Table 5.1 in section 5.5., pp. 105–8) Language Word order FC/NP Case marking system Balanced complements Head-marking Other patterns FC NPANC Hausa SVO/NG ACC + ACC-SA + spa Hungarian SVO/GNb ACC + ACC-SA + sp 6B Notes: a On the treatment of the P in Hausa ANCs see above (p. 191). b Hungarian has a relatively free word order, but SVO predominates slightly.
< previous page
page_193
next page >
< previous page
page_194
next page >
Page 194 (9.11 a) zu-wa-n Dauda come-AN-of Daud ‘Daud’s coming’ (ibid.: 76) (9.11b) da ya yi take-AN-of money that 3M.PRFV do ‘his taking of money’ (ibid.: 175) The other language in the sample which uses the same strategy in expressing both the A and the P of ANs is Hungarian (see section 1.4.1 .). 9.4. TYPE 8: ‘ARGUMENT-REDUCING’ (ARG-RED) As mentioned in section 3.3., Chukchee provides an example of a language in which ANs never combine with both the A and the P at the same time. Thus, derivation of ANs from transitive verbs is always associated with reduction of at least one of their arguments, hence the name ‘the Argument-Reducing type’. I will consider the Chukchee case very briefly, as its details are not yet clear. Chukchee, an ergative SOV incorporating language, uses both dependent-marking and head-marking on the clause level. Besides ergative clauses, where the A and the P take on the ergative and the absolutive cases respectively and the verb agrees with both (as in 9.12a), Chukchee has various antipassive constructions in which the A appears in the absolutive case, the verb agrees with the absolutive (in the same way as intransitive verbs do) and the P either stands in some oblique case or is incorporated into the verb (as in 9.12b) (Polinskaja and Nedjalkov 1987). (9.12a) , bear-ERG wolf:ABS 3SG.A-call-AOR-3P reindeer-ABS 3SG.A-eat-AOR-3P ‘The bear called the wolf to this place, wolf] ate up the reindeer’s meat.’
< previous page
[=the bear/the
page_194
next page >
< previous page
page_195
next page >
Page 195 (9.12b) , bear-ABS wolf-DAT 3S-call-AP-AOR-3SG.S reindeer-ABS 3SG.A-eat-AOR-3P ‘The bear called the wolf, [=the bear/the wolf] ate the reindeer’s meat.’ (Polinskaja and Nedjalkov 1987:246) In NPs, the G usually appears in the Possessive form (with the suffix— (n)in and its variants) and either precedes the head-nominal or is incorporated into it. (In 9.13b, the prefix ge- and the suffix -e together refer to the expression of the comitative case.) (9.13a) Rultyn-POSS-PL friend-PL.NOM ‘Rultyn’s friends’ (9.13b) ge-Kelvi-nine-tumg-e with-Kelvi-POSS-comrade-COMIT ‘with Kelvi’s comrade’ (Skorik 1961:228–9) In ANCs, the S is either in the absolutive case and, thus, retains its sentential dependent-marking (as in 9.14b), or in the Possessive form and, thus, assimilates to the G (as in 9.14a). The situation with ANs derived from lexically transitive verbs is much less clear. The examples demonstrate two options. If an AN is directly correlated with a transitive verb, it can take the A in the Possessive form, but seems not to be able to combine with the P (see 9.14c). If an AN contains an antipassive affix and is, thus, correlated with a detransitivized verb, it may take either the A or the P in the Possessive form, although not both (see 9.14d). (9.14a) father-ERG not.want 3SG.S-think-3S.PRES [ ] daughter-POSS marry-AN-ABS ‘Father does not want his daughter to get married.’ (Nedjalkov et al. 1983:227) (9.14b) [ ] 1SG.S-not.want-PRES pour-AN-DAT water:ABS ‘I do not want the water to pour out.’
< previous page
page_195
next page >
< previous page
page_196
next page >
Page 196 (9.14c) read-AN-ABS Pushkin-POSS ‘Pushkin’s reading’ (i.e., ‘Pushkin is reading something’) (9.14d) hunt-PTCP-POSS kill-AP-AN-ABS ‘the hunter’s killing’/‘the killing of the hunter’ (either ‘the hunter is killing’ or ‘someone is killing the hunter’) (Vladimir Nedjalkov personal communication) It is, thus, not clear whether ANs derived from transitive verbs without anti-passive affixes can combine with their Ps at all. It is clear, though, that both the A and the P of anti-transitive ANs can be expressed within ANCs, although not at the same time. At the present moment, I lack information on the distribution of such ANCs and on their actual formal properties (among other things, on the necessity of overt detransitivization in the derivation of their head).
< previous page
page_196
next page >
< previous page
page_197
next page >
Page 197 10 Splits and variation 10.0. INTRODUCTION Chapters 5 to 9 have been devoted to the description of nominalization patterns across languages. Apparently, many languages have several nominalization patterns, and it may be supposed that the choice among such coexisting patterns is not accidental. While, on the whole, I lack information about determining factors, there are a few examples which show that different types of ANCs in a language may have complementary distribution, that is, languages may have ‘split’ nominalization systems. The next sections present some examples of the factors that give rise to splits and, thus, illustrate several kinds of such split systems. The last section of the chapter shows which nominalization patterns have been attested as coexisting within a language, as well as how nominalization patterns may develop. 10.1. LEXICAL, SEMANTIC AND FUNCTIONAL SPLITS As mentioned in section 1.4.1., in some languages it is possible to distinguish between ‘standard’ nominalization patterns, which are manifested by the majority of action nominals (among other things, by those derived from prototypical dynamic transitive verbs), and other patterns, which are restricted to ANs derived from certain types of transitive verbs. Splits of this type could be called ‘lexical splits’. Another type of split, ‘semantic splits’, arises when an action nominal combines with its arguments in different ways depending on its exact meaning. Let us look at a few examples. In Russian, pronominal As in transitive action nominal constructions can appear in two ways: either as prenominal possessive pronouns, or as postnominal personal pronouns in the instrumental.
< previous page
page_197
next page >
< previous page
page_198
next page >
Page 198 This opposition between DBL-POSS and ERG-POSS ANCs correlates with the opposition between the process/event and fact readings of the nominalizations (Padučeva 1974:201), as in examples (10.1a) and (10.1b) respectively: (10.1a) Jego včerašnee ispolnenie Ravelja bylo his yesterday:ADJ perform:AN Ravel:GEN was velikolepno. wonderful ‘Yesterday’s performance of Ravel by him was wonderful.’ (10.1b) Ispolnenie im Ravelja včera bylo perform:AN he:INSTR Ravel:GEN yesterday was neumestno. out.of.place ‘It was out of place for him to perform Ravel yesterday.’ Russian ANCs with the proposition reading are similar to those with the fact reading. Some Daghestanian languages of the Lezghian group alternate between the ergative and genitive marking of the A in transitive action nominal constructions, in other words, between the SENT and POSS-ACC types. The formal opposition between the two types of nominalizations corresponds to the semantic opposition between the fact and the process or manner readings respectively. I suspect that my sources include both the fact and the proposition readings under the label ‘fact’. Agul (10.2a) t hindis q’abul-xindaw c ohuji šer he:DAT not-liked brother:ERG water:NOM wargub carry:AN ‘He did not like the fact that the brother carried water.’ (10.2b) t hindis q’abul-xindaw c ohujin šer he:DAT not-liked brother:GEN water:NOM wargub carry:AN ‘He did not like the way in which the brother was carrying water.’ (Boguslavskaja 1989)
< previous page
page_198
next page >
< previous page
page_199
next page >
Page 199 Thus, in both Russian and Daghestanian languages, nominalizations with the non-fact reading are structurally closer to ordinary noun phrases than those with the fact (and, presumably, proposition) reading. More specifically, the A in non-fact nominalizations genitivizes in contrast to fact nominalizations. Note also that Russian ANCs with the fact reading allow for a very limited number of time adverbials (like včera —‘yesterday’ in example (10.1b)), while in non-fact nominalizations such adverbs have to be substituted for by the corresponding adjectives (like včerašnee in example (10.1a)). English with its ‘verbal’ and ‘nominal’ gerunds, i.e., gerunds and -ing-nouns, is another example of a language with the same tendency. Thus, ‘nominal’ gerunds (a type of Vendler’s ‘perfect nominals’) cannot refer to facts, while the ‘verbal’ gerunds (Vendler’s ‘imperfect nominals’) are much freer in their application (see section 1.5.2). Polynesian languages with the Semantic types of nominalizations (see section 10.4 below) also have a variant of ‘semantic split’, in those cases where the fact/process opposition is interpreted as an opposition in the degree of control exerted by the agent upon the action. In a number of languages, ANCs are characterized by functional splits, when the choice among different nominalization patterns may be determined by the exact function of the ANC. Thus, in Cuzco Quechua, when an ANC has the subject function in a larger clause, its S or A must appear in the genitive case, while in other cases, they may retain their dependent-marking, namely the nominative case (see section 5.5.). A somewhat different example of a functional split is presented by the Siberian Turkic languages, e.g. Tuva. While the different nominalization patterns in Cuzco Quechua appear in different complement positions, Tuva ANCs differentiate between complement and adverbial functions. When an ANC constitutes a complement within a larger clause, its S or A may be either in the nominative or in the genitive (see section 5.5.). When an ANC has an adverbial function, its S or A do not have any choice and must stand in the nominative case, thus retaining their sentential dependent-marking (Ubrjatova and Litvin 1984:175–6). This difference can be seen in the following two examples where the non-past action nominal with the suffix -Vr is the head of a complement to the verb ‘to believe’ (ex. 10.3a–10.3b) and of a temporal adverbial clause (ex. 10.3c–10.3d). The other difference between the two constructions concerns the head-marking of the head. In both uses action nominals take (nominal) possessive suffixes, which for the third person is zero in adverbial clauses, as opposed to an overt expression in complements. Thus, the structure
< previous page
page_199
next page >
< previous page
page_200
next page >
Page 200 of the head-marking paradigm for ANs in adverbial uses is similar to that of finite verbs, which also lack any overt expression for the subject in the third person. (10.3a) [ dirig ün-är-in-ge] he:GEN alive rest-AN.NONPAST-3SG.POSS-DAT büzüräär men! believe 1SG ‘I believe that he will rest alive!’ (Ubrjatoba and Litvin 1986:78) (10.3b) [ šaandakkı ır-lar-nı You:GEN old song-PL-ACC ], sing-AN.NONPAST-2SG.POSS-ACC men men. ‘I have not yet heard you singing old songs.’ (Ubrjatoba and Litvin 1984:73, ex. 59) (10.3c) [Mašina kä-är-ge] car:NONM come-AN.NONPAST-DAT] xooraj baar bis. ‘When the car comes, we will go to the town.’ (Ubrjatoba and Litvin 1986:78) (10.3d) [ava-zı inek-ti saa-p mother-3SG.POSS cow-ACC milk-CONV ka-ar-ga] Kara-kıs šala-nı AUX-AN.NONPAST-DAT Kara-kıs floor-ACC ču-p wash-CONV AUX-FUT-3SG ‘When mother milks the cow, Kara-kys will wash the floor.’ (Bergel’son and Kibrik 1987:21, ex. 9) As suggested in Čeremisina (1981a) and Ubrjatoba and Litvin (1984), this distinction may be interpreted as an indication of an ongoing process by which action nominals are differentiated from converbs (adverbial participles). In particular, for -Vr-gA -forms in adverbial uses, Bergel’son and Kibrik (1987) argue that they no longer belong to the living case paradigm of action nominals, but constitute
< previous page
page_200
next page >
< previous page
page_201
next page >
Page 201 a part of the switch reference system for temporal chaining constructions. More specifically, an -Vr-gAform refers to an event which is connected to the one referred to by the matrix predicate without specifying the exact temporal relationship between the two, and has a different subject from the matrix predicate. A similar process may be attributed to Bulgarian, where ANCs with the P in the sentential form are reported to be typically used in adverbial functions, while in ANCs with other functions, the P tends to genitivize, cf. the contrast between pri podavane molbata—‘by submitting the:application’ vs. srok za podavane na dokumentite—‘a deadline for submitting the:papers’ (Revzin 1973:96).1 A plausible explanation for these Bulgarian data would be that there is a grammaticalization process going on which creates converbs out of a combination of a preposition and an action nominal. The typological parallel between Tuva and Bulgarian is striking: in both languages, ANCs with adverbial functions are structurally closer to finite clauses than ANCs with complement functions. 10.2. REFERENTIALITY/ANIMACY SPLIT Languages which build their possessive NPs in different ways depending on the possessor’s degree of animacy/referentiality were discussed in section 3.5. in connection with the hypothesis that the types of possessive phrases in a language influence or determine its choice of nominalization patterns. As expected, ANCs in these languages manifest different patterns for arguments with a high versus low degree of animacy/referentiality. The relevant facts include the following: (1) As mentioned in 3.5.3., a number of VO languages with predominant NG order, have a class of prenominal possessive forms. In the most typical case, these are restricted to possessive pronouns, as, for example, in French, cf. le livre de la fille, de Marie (lit. ‘the book of the girl, of Marie’) vs. son livre (‘her book’), but, in some languages, this class includes other highly referential words, such as proper names (as in Russian and German), or definite animate nouns (as in Czech). Languages with these prenominal possessive forms have two genitive positions within NPs, even though one of them, the prenominal genitive position, may be occupied by a very limited class of words. Prenominal possessive forms are subject-like in at least two respects. Firstly, as regards their order relative to the other genitive position and/or to the head (similarly to the subject position in VSO and SVO languages respectively), and, secondly, as regards their high
< previous page
page_201
next page >
< previous page
page_202
next page >
Page 202 referentiality (highly referential NPs can always occur as subjects, according to Keenan 1976:319). As expected, these two genitive positions are usually exploited in ANCs in such a way that the prenominal position is occupied by the S and A, while the postnominal position is reserved for the P. In other words, in a number of languages, the DBL-POSS pattern is restricted to ANCs with pronominal As (and some other highly referential As, in the case of Russian, Czech and German), while all other ANCs follow another pattern. Examples (10.4a–10.4b) illustrate this (see also ex. (10.15) and (10.16) from Samoan in section 10.4. below). French (10.4a) son interprétation de Chopin his/her interpret:AN of Chopin ‘his/her interpretation of Chopin’ (10.4b) l’interprétation de Chopin par le musicien the:interpret:AN of Chopin by the musician ‘the interpretation of Chopin by the musician’ The languages which allow DBL-POSS nominalizations with highly referential As include Bulgarian, Russian (and Czech), Dutch, German, French and Samoan. As mentioned in section 8.1.2, English, a prototypical DBL-POSS language, tends to restrict prenominal s-genitives to definite and animate As, which shows that there is hardly any sharp distinction between the two groups of DBL-POSS languages, namely, those discussed in 8.1. and those discussed here. (2) Ancient Egyptian (Gardiner 1957:226), in which ANCs with lexical As follow the ERG-POSS pattern, resorts to ANCs of the POSS-ACC type when the pronominal A appears as possessive suffixes on the verbal noun as in —give:AN-3SG.M.POSS he.OBJ as prince prince (i.e., ‘appointing (lit. giving) him to be prince’).2 Neither the P in finite clauses nor the G are marked, and dependents in NPs follow the head. This might disfavour POSS-ACC nominalizations with overt non-pronominal As and Ps, since, in such ANCs, As and Ps would have just appeared one after the other in the non-marked form. Another peculiarity of the language is the behaviour of pronominal Ps corresponding to third person singular neutral, third person singular female, third person dual and third person plural, which always appear in ANCs in the same form as in finite clauses, ex. m st ntf (through say:AN it:OBJ 3SG.M:AG) —‘through the saying of it on his part in the presence of the (proper) functionary’ (ibid.: 225). This means
< previous page
page_202
next page >
< previous page
page_203
next page >
Page 203 that ANCs with such Ps and non-pronominal As follow the MIX pattern. (3) In two languages, Modern Israeli Hebrew and Macedonian, the choice of the POSS-ACC type of nominalization depends on the referentiality of the P. Thus, in Hebrew, POSS-ACC ANCs involve definite Ps, that is, those with the marker et, while indefinite Ps are allowed only in ERG-POSS ANCs (see section 7.2.2.). Macedonian presents the opposite case. The object of finite clauses is not marked in any way, either by head-marking or dependent-marking, and the same non-marked form may appear in ANCs, but only when the P is unspecified and the whole construction does not have the definite article, as in čitańe kniga —‘reading a book’ (Topolińska 1981:149). (4) The choice of INC nominalizations in those languages which have some other types of ANCs is also connected to the degree of referentiality of the P (see the discussion in section 9.2.). 10.3. AMBIGUITIES Up to now, we have mostly considered transitive ANCs in which both the A and the P are present simultaneously. In the majority of cases, these arguments are discriminated by means of dependentmarking and/or head-marking. However, what happens when only one of these two arguments is overt? The structure of such ANCs is governed by two opposite tendencies—the tendency to retain the distinction between the A and the P, and the tendency to assimilate ANCs with a single argument to non-derived noun phrases. Which tendency has the strongest impact on the output of nominalization depends on the particular language. In those cases where the expression for the A or the P in an ANC depends on the presence vs. absence of the other argument, we may talk about ‘anti-X systems’, as suggested in Verhaar (1990:155) in connection with English ANCs. Verhaar’s use of the term ‘an anti-X system’, defined as a system ‘with a ‘‘primary default” for a complete Argument load, and a “secondary default” for a reduced Argument load’, is inspired by Comrie’s (1975) analysis of anti-ergative systems found for Welsh and Finnish clauses. Let us now look at transitive ANCs with a reduced argument load for each of the nominalization types. The available information, although less systematic than that on transitive ANCs with a full argument load, supports the general impression that the distinction between the A and the P is neutralized much more often in ANCs than in clauses. (1) Among the languges of the SENT type, the Daghestanian
< previous page
page_203
next page >
< previous page
page_204
next page >
Page 204 languages tend to retain the distinction between the A and the P. While the A may be put in the genitive in almost all the members of the Lezghian group (except Budux), the use of the objective genitive in them is extremely restricted (Boguslavskaja 1989, Klimov and Alekseev 1980:234). As described in chapter 5, the exact type of ANCs in Burushaski (the choice being between the SENT and the POSS-ACC) cannot be determined due to the lack of sufficient information in my source (Lorimer 1935). Lorimer notes variation between the nominative and genitive marking of the P (in the absence of the A) and of the S, as in —me:GEN kill:AN:on (‘by the slaying of me’) vs. —me:NONM kill:AN:DAT (‘to kill me’) (Lorimer 1935:69), and in — me:GEN/me:NONM laugh:AN (‘my laughing’) (ibid.: 351). Due to the great similarity between the ergative and the genitive cases in Burushaski, the possibility of putting the P in the genitive in transitive ANCs without an overt A, leads to ambiguities in the interpretation of this sole dependent. For the other languages with the SENT type, I lack any explicit information on transitive ANCs with only one argument present. More specifically, I have no information about whether such constructions are possible and, if they are, whether they may be ambiguous as regards the interpretation of the dependent NP in them. (2) Among the languages with the POSS-ACC type, ambiguities between the A and the P interpretation of the only argument in a transitive ANC arise in Classical Arabic, Modern Israeli Hebrew, Maori, Tahitian, Polish (in which verbal nouns can combine with their arguments both according to the POSSACC and ERG-POSS patterns) and in Egyptian Arabic (which has both POSS-ACC and POSS-ADN nominalizations). In Evenki (a Tungus language), which has only POSS-ACC ANCs, the phrase Puškin taŋ-in-in—Pushkin read-AN-POSS.3 SG has two readings—‘Pushkin’s reading (of something)’ and ‘someone’s reading of Pushkin’ (Vladimir Nedjalkov personal communication). For the other languages in the group, such ambiguities have not been registered in my sources. (3) Among the languages with the ERG-POSS type of nominalization, transitive ANCs with only one argument present tend to be ambiguous. Georgian, described in section 7.1., shows that this need not be true in all the languages of this group. (4) Among the languages with the NOMN type, some DBL-POSS languages do not distinguish between the A and the P of transitive ANCs even when both are present (see section 8.1.1.). The others, in their ANCs, have either two distinct genitive positions for the A and
< previous page
page_204
next page >
< previous page
page_205
next page >
Page 205 the P (in the case of the other DBL-POSS languages, including those with referentiality/animacy splits), or put only the A in the genitive, while the P appears as an oblique NP (in the case of POSS-ADN languages). This distinction between the A and the P tends to be retained even in those cases when only one of the arguments is present. Thus, Czech, a language with a referentiality split, in which DBLPOSS ANCs involve singular definite As, shows an opposition between matčina ztráta—mother:POSS lose:AN (‘mother’s loss (of something)’) vs. ztráta matky—lose:AN mother:GEN (‘(someone’s) loss of (his) mother’) (Comrie and Thompson 1985:377). This is quite similar to the situation in English, in which there are strong tendencies to interpret the prenominal genitive as coding the A and the S, and the postnominal of -genitive as coding the P. However, even if examples of this kind are numerous, there are exceptions to the above mentioned tendencies. Thus, Corbett (1987:329–30) observes that prenominal possessive adjectives in all the Slavonic languages may be used for marking the P of action nominals, although the range and frequency of this use is normally much more restricted than for marking the A. According to Padučeva (1984:57–8), in Russian, animate and inanimate Ps differ as to their ability to appear as preposed possessive pronouns. Inanimate Ps are quite freely allowed to undergo such preposing. Animate Ps, on the other hand, only turn up as possessive pronouns with a very restricted class of action nominals. Compare the examples below: (10.5a) Knigi ja poslal; izvestite menja ob ix books:ACC I:NOM sent; inform me about their polučenii. receive:AN ‘I have sent the books; inform me about the receipt of them.’ (Padučeva 1984:58, ex. 5) (10.5b) *On raskaivaetsja v tvoem oskorblenii. he:NOM regrets in your insult:AN:LOC ‘He regrets having insulted you.’ (ibid: 57, ex. 1)
< previous page
page_205
next page >
< previous page
page_206
next page >
Page 206 (10.5c) On sčitaet svoj arest he:NOM considers his:RFL arrest:AN protivozakonnym. illegal ‘He considers his own arrest to be illegal (that he has been arrested illegally)’ (ibid: 58, ex. 4) Padučeva writes that she has not found any semantic difference between the action nominals that allow or do not allow their animate Ps to appear as preposed possessive pronouns. My impression is that action nominals of the former type correlate semantically with the perfective aspect of the corresponding verb in its resultative meaning (the arrest of X correlates with (the fact) that X has been arrested), but this does not, however, explain the impossibility of (10.5b) above. The use of possessive pronouns for inanimate Ps (which do not normally include either the speaker or the hearer and are therefore typically referred to by third person pronouns) is, perhaps, facilitated by the fact that the possessive and genitive forms of the third person pronouns differ only in their position relative to the head. In addition, inanimates are rather poor candidates for being As and, thus, no confusion arises when they are referred to by the possessive pronouns. Germanic and Romance languages with NOMN ANCs also have a class of action nominals which admit their Ps to appear as preposed possessive nominals or pronouns, the form otherwise typical for the As. According to the explanation given in Comrie and Thompson (1985:371–2), in the English phrase Bill’s execution, the preposed genitive must be interpreted as corresponding to the P, because the underlying verb requires a P. When the verb corresponding to an action nominal does not require, but only allows for a P, there is a tendency to interpret the preposed genitive as referring to the A, cf. John’s reading and ?the book’s reading . This explanation, however, does not cover all such cases even in English, not to mention the other languages with similar properties. Another explanation, advocated by Anderson (1979, quoted in Giorgi and Longobardi 1988) attributes the difference between action nominals which allow or do not allow the P (‘object’) in the prenominal genitive form to the difference between affected and unaffected objects respectively. Anderson proposes the generalization that, if the action expressed by the head noun does not affect (modify) the state of the object, the latter cannot appear as the prenominal genitive. Giorgi and Longobardi (1988:
< previous page
page_206
next page >
< previous page
page_207
next page >
Page 207 146–50) offer a number of examples which show that this generalization cannot, per se, explain the differences between English and the Romance languages. The class of action nominals with Ps in the prenominal possessive form is much more restricted in English, as can be seen in the contrast between (10.6a) from English, on one hand, and (10.6b) from Italian and (10.6c) from French, on the other: (10.6a) *The novel’s understanding (10.6b) A proposito di quel romanzo, la sua comprensione richiede notevoli capacità ermeneutiche. ‘Concerning that novel, its understanding requires remarkable hermeneutic skills.’ (10.6c) (Ce problème est très difficile.) Sa compréhension éxige beaucoup de travail. ‘(This problem is very difficult.) Its understanding needs a lot of work.’ (Giorgi and Longobardi 1988:147–8) Giorgi and Longobardi hypothesize that the contrasts illustrated in (10.6) may be connected with the general typological differences between English (and Germanic in general) and Romance with respect to the parameters ‘Head-complement’ and ‘Head-subject’, which are set in the same direction in Romance, but in different directions in Germanic. German, however, is less restricting than the other Germanic languages in allowing ‘unaffected’ Ps to appear as prenominal genitives or possessive pronouns: (10.7) Was die Algebra betrifft, ist ihre Kenntnis sehr wichtig. ‘As far as algebra is concerned, its knowledge is very important.’ (ibid.: 153) Among the POSS-ADN languages, Icelandic marks both the A and the P, when only one is present, with the genitive. Swedish presents an example of a language in which transitive ANCs with both the A and the P are potentially ambiguous. Its nominalizations have been classified as following the POSSADN pattern, since the A is put in the genitive and the P turns up as a prepositional phrase with the preposition av. The latter has several meanings, among others, referring to ‘part-whole’ and ‘reason’ relations. It is also used to introduce the agent phrase in passive clauses. Thus, transitive ANCs in Swedish involve one dependent in the genitive and the other in the same form as the agent in passive clauses, which makes them look very similar to the ERG-POSS nominalizations of other languages. The older
< previous page
page_207
next page >
< previous page
page_208
next page >
Page 208 sources on Swedish (Noreen 1904, Svenska akademiens ordbok 1898) report that such ANCs in Swedish are completely ambiguous and should be avoided in the absence of clear indications as to which of the readings should be preferred. For example, the ANC Konstantinopel-s eröfr-ing af turkarna — Konstantinopolis-GEN conquer-AN of the Turks (Svenska akademiens ordbok 1898, vol. 1:88) is unambiguously interpreted as an ERG-POSS ANC (‘the conquest of Konstantinopolis by the Turks’), while the ANC Handelsbanken-s övertag-ande av Mälarbanken ‘Handelsbank-GEN take.over-AN of Mälarbank’, is described by Wellander (1973) as completely ambiguous, between ‘the take-over of Handelsbanken by Mälarbanken’ and ‘the take-over of Mälarbanken by Handelsbanken’ and, thus, unacceptable. For Modern Swedish, all my informants either reject the ERG-POSS interpretation as absolutely impossible, or accept such constructions as extremely marginal, un-Swedish, or influenced by German. In fact, the connection between preposed genitives and the S or the A of ANCs is so strong, that the distinction between the A and the P is normally maintained even when only one of them is present. Some exceptions, however, are possible. Thus, long and formally complex and/or indefinite As appear as oblique av- phrases, while singular definite Ps may turn up as preposed genitives. The exact state of affairs has still to be investigated. (5) In both languages with the REL type of nominalizations, Hausa and Hungarian, the only overt argument of transitive ANCs genitivizes, which thus makes such constructions completely ambiguous. The exposition above shows that ERG-POSS and REL transitive ANCs with only one argument present tend to be ambiguous more frequently than similar ANCs of the SENT, POSS-ACC and NOMN types (with the exception of certain DBL-POSS languages). Note that in the former case, ambiguities arise because the A, in the absence of any P, genitivizes, but when both arguments are present, it is the P which genitivizes. In the latter case (with the exception of the SENT type), however, it is the A which normally genitivizes and ambiguities would arise if the P were to genitivize in the absence of an A. In other words, in ERG-POSS and REL ANCs, it is often the presence of the P which prevents the A from being genitivized, while in the SENT, POSS-ACC and NOMN types the presence of the A is not the only reason for non-genitivization of the P. This finding suggests that the A is easier to genitivize than the P, in accordance with Keenan’s (1974) idea that possessors and subjects are semantically
< previous page
page_208
next page >
< previous page
page_209
next page >
Page 209 similar (see section 3.4.). Genitivization of the P, on the other hand, occurs under certain additional conditions, one being when possessive phrases in the particular language show affinities with verbobject combinations (see sections 3.5. and 11.4.). Up to now, this section has concentrated on potential ambiguities in ANCs of the types that normally discriminate between the A and the P. My data offer two examples of the opposite tendency; that is to discriminate between the different arguments of action nominals in those types that normally do not do this. Basque normally resorts to SENT ANCs built on a binary opposition between the A in the ergative case and both the S and the P in the absolutive (see ex. 10.8a–10.8b). However, in the northern dialects (N’Diaye 1970:141, Saltarelli 1988:155), the P may be put in the genitive when no other argument in the ergative or dative is present (see ex. 10.8c). Since this option is neither available for the S (which normally has the same marking as the P), nor for the A, the P in ANCs with one argument present is opposed to both the A and the S: (10.8a) [Mikel-ek ama-ri diru-a Mikel-ERG mother-SG.DAT money-SG.ABS ema-te-a] oso ondo irudi-tzen give-AN-SG.ABS very well appear-HAB za-it. (3ABS-PRES)-AUX1-1SG.DAT ‘Mikel’s giving the money to mother appears very good to me.’ (10.8b) Mikel gaizto-a iza-te-a Mikel.-ABS bad-SG.ABS be-AN-SG.ABS ‘Mikel’s being bad…’ (10.8c) [gizon-aren ikus-te-ak] mesede egi-n man-SG.GEN see-AN-SG.ERG help do-PRF z-io-n. 3SG.ERG-(PAST-3ABS-AUX2)-3SG.DAT-PAST ‘The sight (seeing) of the man did him/her good.’ (Saltarelli 1988:155) Tahitian (a Polynesian language) has nominalizations of the POSS-ACC type, in which the S is introduced by the possessive particle o (having assimilated to the G of non-derived NPs) and, thus, appears in the same form as the A. This type is, on the whole, based on a binary opposition among the S, A and P. However, sometimes the S of intransitive ANCs can appear not only as the G, but in the non-marked form, or in other words, in the same form as the subject of
< previous page
page_209
next page >
< previous page
page_210
next page >
Page 210 the corresponding finite clause: (10.9) i te pee-raa ’tu te hue PREP ART approach-NOM away NOM ART gourd ‘As the gourd approached him’ (Chung 1973:657, ex. 36) Under such conditions, the S is opposed to both the A and the P of ANCs. Another peculiarity of these constructions in Basque and Tahitian is their deviation from the general cross-linguistic tendency of intransitive ANCs to be the easiest to assimilate to non-derived NPs. 10.4. ALIENABLE VS. INALIENABLE POSSESSION AND ANCs Seiler (1983, for discussion see section 3.4. above) shows that, in a number of languages, there is an exact correspondence between inalienable possessives and object markers. Affinities of this kind are connected to inherent POSSESSION, while affinities between possessives and subject markers have establishing POSSESSION as their basis. If this is correct, we would expect that at least some of the languages which consistently distinguish between inalienable and alienable possession may also use this distinction within nominalizations. Polynesian languages offer us interesting examples of this. Chung (1973) examines fifteen Polynesian languages and shows that the system of dominant vs. subordinate possession, which works almost identically in non-derived NPs, is exploited differently by different languages in nominalizations. (For the description of the Polynesian opposition in possession see section 7.2.2.) These languages manifest three major types of nominalizations with regard to the basis for assigning possessives within them—the Syntactic type, the Semantic type and the Neutralized type. These types with their variants are illustrated below. Note that Chung’s classification is based on parameters other than mine, and that the two intersect each other. More specifically, Chung proceeds from ANCs in which at least one of the arguments genitivizes, and studies the principles for choosing a possessive marker for this argument. The structure of ANCs on the whole is not relevant for the classification. In the Syntactic type, ‘possessives are assigned by syntactic category to nominalizations’ (Chung 1973:651), that is, it is the syntactic category of the dependent within an ANC which determines whether
< previous page
page_210
next page >
< previous page
page_211
next page >
Page 211 it will be marked with o (subordinate possession) or with a (dominant possession). Two variants of this type are attested, depending on which syntactic opposition is valid for nominalizations—either that between the A versus both the S and the P, or that between the A and S versus the P. The first variant of the syntactic type is illustrated by Maori, in which the S and the P are treated alike in contrast to the A, that is, they assimilate to subordinate or dominant possessives respectively. (For examples see section 7.2.2.) Tongan manifests the second variant of the syntactic type, since its S and A are now treated alike and assimilate to dominant possessives, in contrast to the P which assimilates to subordinate possessives. Thus, ANCs in Maori, a nominative-accusative language, exploit the subordinate/dominant opposition according to the ergative-pattern, while ANCs in Tongan, an ergative language, exploit the same opposition according to the nominative-accusative pattern. (For discussion, see section 11.6.) Tongan (10.10) ‘oku ou hela’ia [he fa’a kata mai PRES I be.tired ART often laugh:AN here ’a kiate au] DOM.POSS Alan PREP me I’m tired of Alan’s always laughing at me.’ (Chung 1973:654, ex. 29) (10.11) ’oku ou ‘ohovale [‘i he’ene kai PRES I be.surprised PREP his:DOM.POSS eat:AN ’a e me’a kai] ABS ART thing eat ‘I am surprised at his eating the food.’ (ibid.: 655, ex. 30) (10.12) ’oku ’ikai tonu [’a e taki ’o PRES NEG good ABS ART lead:AN SUB.POSS e fonua ’e he tu’i] ART land ERG ART king ‘The leading of the country by the king is not good.’ (ibid.: 656, ex. 33) In the Semantic type, possessives are assigned on a semantic basis to nominalizations, that is, dependents in nominalizations take different possessive markers depending on their semantic category. Here again two variants of the major type can be distinguished. Pukapukan illustrates the first variant, combining syntactic and semantic features and characteristic for the whole group of the
< previous page
page_211
next page >
< previous page
page_212
next page >
Page 212 Samoic-Outliers. The syntactic opposition between the A and P is reflected as the opposition between dominant and subordinate possessives, exactly as in the Syntactic type. The Ss, however, may select either a or o, and the determining factor in the choice between them is the degree of agency or control: ‘ a focuses on the subject’s handling of the action, o makes no particular statement about it’ (Chung 1973:660). This contrast is manifested in different ways. Thus, a may emphasize the way of performing the action or imply that the action has been successfully completed. In any case, although these implementations of the contrast may seem rather disparate, they are alike in using a to draw attention to the competence of the subject… In contrast, o is used to state the action without focusing on any NP in particular. The semantic constrast which is produced is essentially one of dominance (a) versus non-dominance (o) of the possessor. As such, it is virtually the same as the contrast of a and o which occurs in ordinary possessive constructions. (ibid.: 661) Pukapukan (10.13) welele loa [i tā lātou run they PREP ART:DOM.POSS they welele-ngaia], oko ki Yato. run-ANDEM reach DIR Yato ‘They ran towards Yato.’ (lit. ‘They ran in their running, reached Yato’.) (ibid.: 661, ex. 46) (10.14) e mea kino [te yau-nga ō-na ki PRES thing bad ART come-AN SUB.POSS-he PREP ]. Auckland ‘His coming to Auckland is a bad thing.’ (ibid: 660, ex. 45) Samoan has the second variant of the Semantic type, since both the S and A may be assigned different possessive markers depending on the degree of control exerted by the referent of the nominal under consideration. The Ps are always marked as subordinate possessives.
< previous page
page_212
next page >
< previous page
page_213
next page >
Page 213 Samoan (10.15) sā ’ou ofo [i lana PAST I be.surprised PREP his:DOM.POSS kukaina o le i’a] cook:AN SUB.POSS ART fish ‘I was surprised at the way he cooked the fish.’ (Chung 1973:662, ex. 49) (10.16) sā ’ou ofo [i lona PAST I be.surprised PREP his:SUB.POSS tīa’i o le i’a] ’ina ’ua ’ama ’ona kuka discard:AN SUB.POSS ART fish ‘I was surprised that he threw out the fish after it was cooked.’ It is worth mentioning that inanimate Ss and As are almost always marked with o (subordinate possession), which corresponds exactly to the distribution of inanimate possessors within ordinary possessive constructions. The Neutralized type existed in nineteenth-century Tahitian. At that time, Tahitian still had an opposition between subordinate and dominant possession in ordinary NPs, but not in nominalizations. In nominalizations, the first non-pronominal NP after the verbal noun or the pronoun preceding it is marked with the subordinate form of the possessive, o (in those nominalizations where it is marked with the possessive at all). Tahitian (10.17) e tei ni’a i te a’au te tahi tapua’e avae and is.on top on the.SG reef the.Sg other print foot [i to’na here ra’a ’tu i te rā]. PREP his:SUB.POSS noose AN away ACC ART sun ‘And upon the reef is the other footprint, which he made as he noosed the sun.’ (Chung 1973:657, ex. 34) (10.18) ua motu te ’ofa’i [i te taviri-ra’a TMA break the.Sg stone PREP ART twist-AN o taura mau taura a Mā-ū-i ra]. SUB.POSS that PL rope DOM.POSS Maui DEM ‘Around it are spiral indentations, said to be from the twisting of Maui’s rope (by someone).’ (ibid.: 657, ex. 37)
< previous page
page_213
next page >
< previous page
page_214
next page >
Page 214 What the first four of the above mentioned types have in common is the assimilation of the Ps (objects) to the possessor-nominals in NPs with subordinate (inalienable, inherent) possession, and the assimilation of the most prototypical As (those which exert the greatest control upon the action) to the possessor-nominals in NPs with dominant (alienable, establishing) possession, fitting Seiler’s theory perfectly. Due to their conceptually intermediate status, the other dependents of action nominals which may appear with possessive particles, more specifically, the Ss and the less prototypical As (those which exert a low degree of control upon the action), are combined with one of the former categories in different ways. ANCs in the Polynesian languages give rise to two other conclusions. Firstly, to repeat what was said in section 8.1.1.: semantically different possessor-nominals do not necessarily occupy syntactically distinct positions within NPs, which is demonstrated by the fact that transitive ANCs in Polynesian languages cannot have two possessively marked postposed dependents at the same time. Secondly, the existence of an overt distinction in the type of possession does not imply that it should necessarily be exploited in nominalizations, which is shown by the existence of the Neutralized type. 10.5. PATTERN SHIFTS AND INTRA-LEXEMIC VARIATION The sample includes numerous languages in which a transitive action nominal may combine with its arguments according to different patterns, the phenomenon which is here called ‘intra-lexemic variation’ of ANs. Besides that, in several cases the data show how languages can change nominalization types during their development. This section will be devoted to such diachronic pattern shifts and synchronic intra-lexemic variation of ANs, showing that these phenomena are subject to certain limitations. 10.5.1. Diachronic pattern shifts My sources note the following cases: SENT → POSS-ACC: Classical Mongolian vs. Modern Mongolian and Buryat. Lewis (1967:169) mentions that in older Turkish, the subjects of ANCs could be in the non-marked form, while Modern Turkish has only POSS-ACC nominalizations (see, however, section 6.5.).
< previous page
page_214
next page >
< previous page
page_215
Page 215 Table 10.1 Intra-lexemic variation of ANs REL INC SENTPOSSINC INC SENT Quechua, Quechua, ImbaburaCuzco Quechua, Cuzco POSSACC
ERGPOSS
NOMN
SENTPOSS OBLPOSS
DBLPOSS POSSADN
MIX
next page >
NOMN DBL-POSS POSSADN
ERG-POSS POSSSENTOBL-POSS ACC POSS Tongan Agul Arabic, Arabic, Classical Classical ?Tuva Tonganpr Maori English Arabic, Arabic, Hebrew, Thai ?Hebrew, Egyptian Classical Modern (Egyptian, Modern (Tongan)1English Ancientpr)1Polishpr Maori Polish Pukapukan Samoanpr
Estonian ?Bulgarian Estonian Kurdish ?Dutch English Persian Maori ?Georgian Egyptian, Greek, Ancientpr Modern Bulgarianpr Dutch* Frenchpr German* Polish* Russian* Estonian Icelandic Swedish
INC Hungarian Note: 1 POSS-ACC ANCs in Ancient Egyptian and Tongan are restricted to combinations with pronominal As.
< previous page
page_215
next page >
< previous page
page_216
next page >
Page 216 DBL-POSS → ERG-POSS: Latin vs. Modern Romance languages (French, Rumanian, Italian), Ancient Greek vs. Modern Greek. ERG-POSS → MIX: this development is suggested by Revzin (1973:95) for Bulgarian and Macedonian, which combine the common Slavonic ERG-POSS system with the ‘Turkic’ POSS-ACC system. However, Old Church Slavonic texts contain some rare examples of POSS-ACC nominalizations (Večerka 1963, quoted in Comrie 1976:193). 10.5.2. Synchronic intra-lexemic variation In discussing possible intra-lexemic variation of ANs, it may be appropriate to distinguish between the subtypes of the ERG-POSS, the NOMN and the INC types. Thus, for example, the shift between the SENT and the ERG-POSS type concerns only the two languages with the SENT-POSS subtype of the latter. Table 10.1. summarizes the attested cases. Putting aside ANCs restricted to pronominal arguments, the data in Table 10.1. can be generalized in still another way, presented in Figure 10.1. This figure covers both the diachronic pattern shifts and cases of synchronie variation attested in the sample, but excludes the INC and the REL types. The latter, valency reducing types, are somewhat outside of this general scheme. The ARG-RED type is not, of course, compatible with any other ANC type. What Figure 10.1 essentially shows is that the different patterns
Figure 10.1 Diachronic pattern shifts and synchronie intra-lexemic variation of ANs
< previous page
page_216
next page >
< previous page
page_217
next page >
Page 217 according to which action nominals combine with their arguments in a single language either at the same time or at different stages of its development, tend to be relatively similar to each other. In most cases, the differences between them concern the marking of only one argument—either the A (as in the shifts between the SENT and the POSS-ACC types, between the POSS-ACC and the MIX types, and between the OBL-POSS subtype of the ERG-POSS type and the DBL-POSS subtype of the NOMN type), or the P (as in the shifts between the SENT type and the SENT-POSS subtype of the ERG-POSS type, between the OBL-POSS subtype of the ERG-POSS type and the MIX type, and between the POSS-ACC type and the POSS-ADN subtype of the NOMN type). The only exception to this generalization is constituted by the transition between the POSS-ACC and the ERG-POSS types, which concerns both the A and the P.
< previous page
page_217
next page >
< previous page
page_218
next page >
page_218
next page >
Page 218 This page intentionally left blank.
< previous page
< previous page
page_219
next page >
page_219
next page >
Page 219 Part 3 Discussion
< previous page
< previous page
page_220
next page >
page_220
next page >
Page 220 This page intentionally left blank.
< previous page
< previous page
page_221
next page >
Page 221 11 Explanation of type-membership 11.0. INTRODUCTION The preceding six chapters present a detailed description of the different nominalization patterns represented by the languages in the sample. It is now reasonable to ask for some explanation for this typology. As Stassen (1985:6) points out, in connection with typologies we can put forth at least two kinds of explanatory questions, the question which concerns ‘the occurrence of attested and nonattested categories’, and the question which concerns ‘ the attested distribution of languages over the types in the typology, that is, it is a question about the explanation of type-membership’. An answer to the first question of why the typology contains just these attested types, instead of other, non-attested but also imaginable alternatives, may lead to a more profound understanding of the limitations imposed on human languages. The second question ‘assumes that the grouping of languages in the typology reflects a division into natural classes, and attempts to formulate a basis for the explanation of this naturalness’. In order to do this we can try to identify some ‘determining ‘‘outside” factor’, that is, set up a new typology which will be based on this new parameter, the ultimate aim consisting of formulating implicational universals of the form ‘If a language belongs to the category X in typology A, it belongs to the category Y in typology B’. Chapters 11 and 12 are devoted to both explanatory questions mentioned by Stassen, starting with the second one. The discussion primarily concerns pure nominalization types. These are particularly interesting since the resulting transitive ANCs are directly correlated to transitive finite clauses and are, thus, products of the nominalization process itself. In the case of derived nominalization types (the passive variant of the ERG-POSS and the anti-passive variant of the POSS-ACC types), there are only intransitive finite clauses which
< previous page
page_221
next page >
< previous page
page_222
next page >
Page 222 correspond to ANCs, more specifically, those in which the head verb is passivized or anti-passivized. Thus, the only effect of nominalization on such structures is assimilation of the ‘subject’ to the G of nonderived NPs. I have not succeeded in finding the determining factor for the typology of ANCs, and this may not even be possible. The following discussion is intended to show in what ways the attested types are correlated to some other features of the languages in the sample. It should be emphasized here that the sample is too limited to give secure empirical justification to some of these correlations, which makes them relatively provisional and speculative. Quite a number of them should rather be considered as hypotheses calling for future investigation (see, e.g., sections 11.4. and 11.5. on head-marking and on the structure of non-derived NPs). 11.1. GENETIC AND AREAL DISTRIBUTION OF THE NOMINALIZATION TYPES A natural question in connection with typologies is: to what extent are the types spread across languages of the world, and, in particular, to what extent is the membership in a type predetermined by or correlated to the genetic and areal affinity of the language? Table 11.1 shows the distribution of the nominalization patterns among the language families in the sample. For each family, it quotes the number of the languages in the sample and the number of the languages with the pattern under consideration. As seen from Table 11.1, the number of the languages in the sample with each nominalization pattern does not offer an adequate measure for the actual spreading of each pattern across languages. Thus, the most numerous group in the sample, that of the ERG-POSS languages, is grossly over-represented due to the disproportionately large share of the Indo-European languages in the data. This case, however, is interesting from another point of view, as it shows a pervasiveness of a nominalization pattern in genetically related languages which have otherwise developed a considerable number of typologically distinct features, such as different word order, nominativity vs. split ergativity, different grades of analyticity vs. syntheticity. (Of the nineteen Indo-European languages from seven genetic groups in the sample, fourteen languages from six groups have ERG-POSS nominalizations.) No investigation has been made of the extent to which such ANCs are remnants of the Proto-IndoEuropean linguistic system, are products of independent internal development, or are
< previous page
page_222
next page >
< previous page
page_223
next page >
Page 223 results of language contacts, but the following facts are relevant. Firstly, in the case of Slavonic languages, ERG-POSS ANCs can be attributed to their common ancestor, as the same construction is found already in Old Church Slavonic (Nilsson 1972:42). Concerning Old Irish, Disterheft (1980:142) concludes that its verbal nouns are closest to ‘infinitives’ in Proto-Indo-European.1 For the latter category, Disterheft reconstructs the following features: it is included within the nominal paradigmatic system; has a genitive object which is ‘almost always mandatory’; and ‘equi deletion often does not apply (in fact does not ever apply when the verbal noun is passive)’ (Disterheft 1980:191). Disterheft’s usage of the term ‘passive’ is rather unfortunate, for ‘passive’ verbal nouns and infinitives in Old Irish, Rigveda and Avesta are not morphologically distinct from the corresponding ‘active’ ones. The relevance of the above mentioned reconstruction for this study may be limited, since Disterheft uses quite different terminology and is mainly concerned with the development of infinitives in Indo-European. A possible interpretation of this reconstruction (reformulated in the terms adopted here) would imply that Proto-Indo-European had action nominal constructions of the ERG-POSS type. In this case it would mean that both Old Irish and Slavonic languages inherited these constructions from the ancestor language. It is noteworthy that my sources on Latin (Woodcock 1959) and Ancient Greek (Smyth 1956, Schwyzer 1950) do not mention such nominalizations, while in the daughter languages (at least, those included in the sample) these are quite common. Table 11.1 gives no information on areal affinity of languages with similar patterns and should be supplemented by the following facts. The majority of the SENT languages are found in the ‘Indo-Asian Language Area’, as defined by Masica (1976). Semitic and Niger-Congo languages have ANCs of the POSS-ACC type. Indo-European languages with NOMN ANCs include English, Icelandic, Latvian and Swedish, of which the two latter languages are spoken in the vicinity of the two Fennic NOMN languages, Finnish and Estonian. Ps in ANCs in Bulgarian (and Macedonian) may appear in the sentential form, which is typical for the Turkic family, but quite uncommon among Indo-European languages (see, however, section 12.3.). 11.2. COMPLEMENT SYSTEMS In section 2.2. the notions of ‘complement-balancing’ and ‘complement-deranking’ languages were introduced. It turns out that the property of being ‘complement-balancing’ or ‘complement-
< previous page
page_223
next page >
< previous page
page_224
next page >
Page 224 Table 11.1 Genetic distribution of the nominalization patterns in the sample SENT P-ACC ERGNOMN MIX INCREL Total number of languages in the POSS sample 1 2 2′3A 3B 3′ 4A 4B5 6 7 0 Afro-Asiatic 1a 4+1pr 1 2 1 1b 1 1pr 1 6 Altaic 3c 5 1 1 8 Andean 3 2 3 Austronesian 1 4+1pr 2d 2 1 1pr 1 6 Basque 1 1 Burushaski 1 1 Carib 1 1 Caucasiane 4 1 2 1 6 Dravidian 1 1 Eskimo-Aleut 1 1 1 Gum 1 1 Indo2f 14 1 6+*4+1pr 2 2 6 19 European Kam-Tai 1 1 1 E New 1 1 Guinean Nivkh 1 1 Niger-Congo 4g 1 1 1 6 Paleosiberian 11 Uralic 1 1 2 2 1 4 Yokuts 1 1 Yukagir 1 1 Total 17 25+2pr1 3 22 6 12+*4+2pr3 4+1pr11 2 1
< previous page
page_224
next page >
< previous page
page_225
next page >
Page 225 Key: 1=SENT, 2=pure POSS-ACC, 2′=anti-passive, 3A=SENT-POSS, 3B=OBL-POSS, 3′=passive, 4A=DBLPOSS, 4B=POSS-ADN, 5=MIX, 6=INC, 7=REL, 0=7ARG-RED; *X—patterns restricted to highly referential arguments, Xpr—patterns restricted to pronominal arguments. Notes: a A rare pattern in Classical Arabic. b A rare pattern in Modern Israeli Hebrew. c Altaic languages with SENT ANCs include both Classical Mongolian, Tuva and Korean, whose genetic relationship is rather questionable. d On the similarities between the SENT-POSS ANCs in Tongan and Samoan vs. the OBL-POSS ANCs in Maori and Pukapukan, see section 7.3. e The genetic relationship between the three groups of Caucasian languages (Northeast Caucasian, Northwest Caucasian and South Caucasian) is unclear (Voegelin and Voegelin 1977:96). SENT occurs in Northeast Caucasian, while OBL-POSS occurs both in Northwest and South Caucasian. f A rare pattern in Polish. g On the difficulties in classifying ANCs in Ifè see section 9.2.
< previous page
page_225
next page >
< previous page
page_226
next page >
Page 226 deranking’ determines, to a large extent, the possible choice of a nominalization pattern in a language. Here only the four major nominalization patterns will be considered. The instances of the MIX and REL are too few to allow any conclusions to be drawn. As regards the INC, it would appear that the general property of noun-noun compounding in a language is not connected to its complement system. Of the four major nominalization patterns, the SENT and the NOMN turn out to be opposite poles with regard to the property of being complement-deranking vs. complement-balancing within the languages with such ANCs. More specifically, fourteen languages of the seventeen in the SENT type are complement-deranking, while all the languages with the NOMN types (and also the REL and ARG-RED types) are complement-balancing. It is thus clear that complement-balancing languages favour assimilation of ANCs to non-derived NPs, while complementderanking languages favour retention of sentential dependent-marking in ANCs. While in the case of the NOMN there is an implication between the type-membership of a language and its being complement-balancing, in the case of the SENT languages, the correlation is distorted by the existence of three complement-balancing languages with that pattern. In one of them, Classical Arabic, SENT ANCs are extremely rare. Unfortunately, I lack information about the spreading of this pattern in Tongan, the second complement-balancing SENT language, which also has nominalizations of other types. For the other two major nominalization patterns, the proportion of complement-deranking languages is twelve languages of the twenty-five in the POSS-ACC type (48 per cent) and three languages of the twenty-five in the ERG-POSS type (12 per cent). However, the relation between the POSS-ACC and the ERG-POSS types is not as straightforward as these figures suggest due to the genetic biases of the sample. For example, all of the fourteen Indo-European languages with ERG-POSS ANCs are complement-balancing. Let us now look at the genetic groupings of the languages with POSS-ACC and ERG-POSS ANCs. Of the eleven genetic groupings in the POSS-ACC group, five are represented in the sample only by complement-deranking languages (Altaic, NE Caucasian, Uralic, Yokuts and Yukagir), four are represented only by complement-balancing languages (Afro-Asiatic, Gum, Indo-European and Kam-Tai) and two by both complement-deranking and complement-balancing languages (Austronesian and Niger-Congo). Of the seven genetic groupings with
< previous page
page_226
next page >
< previous page
page_227
next page >
Page 227 the ERG-POSS type, two are represented by only complement-deranking languages (Carib and NW Caucasian), four by only complement-balancing languages (Afro-Asiatic, South Caucasian, IndoEuropean and Uralic) and one by both complement-deranking and complement-balancing languages (Austronesian languages). This information, supplemented by the corresponding data for the SENT and the NOMN, is summarized in Table 11.2. On the basis of the data in Table 11.2 we can use a simple method for characterizing the major nominalization types with respect to the opposition ‘complement-deranking vs. complement-balancing’ languages. More specifically, for each nominalization type a unit corresponds to a collection of languages which both are genetically related and belong to the same group in the classification into ‘complement-balancing’ and ‘complement-deranking’ languages. A group consisting only of complementbalancing languages gets the value 1, while a complement-deranking group gets the value 0. However, if a certain nominalization type exists in both complement-balancing and complement-deranking languages of the same family, Table 11.2 Complement-balancing vs. complement-deranking genetic groupings with the four major nominalization patterns Complement-deranking Complement-balancing Heterogeneous SENT 7- Altaic, Andean, Burushaski, Nivkh, NE 3- Afro-Asiatic, Austronesian, Basque Caucasian, Dravidian, E New Guinean POSS- 5- Altaic, Yukagir, NE Caucasian, Penutian, 4- Afro-Asiatic, Indo-European, Gum, 2- Austronesian, ACC Uralic Kam-Tai Niger-Congo, ERG- 2- Carib, NW Caucasian 4- Afro-Asiatic, S Caucasian, Indo1- Austronesian POSS European, Uralic NOMN 6- Afro-Asiatic, Altaic, S Caucasian, Indo-European, Niger-Congo, Uralic
< previous page
page_227
next page >
< previous page
page_228
next page >
Page 228 I consider this family to be represented by two units, one with the value 1 and one with the value 0. The resulting value for each type is the total of the values of its representatives divided by the number of these representatives. This method is of course far from being satisfactory. Its drawbacks include indifference to the frequency of a certain type among the languages of a certain family, as well as to the portion of complement-balancing vs. complement-deranking languages in different families. Thus, Afro-Asiatic languages have the value 1 in both the SENT type and the POSS-ACC type, though, in the first case, the pattern appears only in one language and is used very rarely. Both Niger-Congo and Austronesian languages in the POSS-ACC type have two representatives, contrary to a general impression that the former family has a greater portion of complement-deranking languages than the latter. The resulting values for the SENT, POSS-ACC, ERG-POSS and NOMN types are 30 per cent, 46 per cent, 62.5 per cent and 100 per cent. Figure 11.1 visualizes the resulting correlations. Each bar represents a nominalization type, with a maximum height corresponding to the value 1, while the shaded portion of the height corresponds to the value of the type. The width of each bar reflects the number of the language families in it.
Figure 11.1 The distribution of complement-balancing vs. complement-deranking languages with the four major nominalization patterns
< previous page
page_228
next page >
< previous page
page_229
next page >
Page 229 11.3. WORD ORDER Table 11.3 shows the distribution of word order types among the nominalization patterns. In Table 11.3, we observe the following relationships between word order types and nominalization patterns: 1 The absolute majority of languages with the SENT type of ANCs are SOV/GN languages, except for the two V-initial languages, Classical Arabic and Tongan. 2 There are no V-initial languages with INC in the sample. 3 Languages with ERG-POSS nominalizations show the greatest range of variation with regard to basic word order, favouring especially V-initial word order and disfavouring GN word order. Each of these relationships will be discussed separately below. 11.3.1. The absolute majority of languages with the SENT type of ANCs are SOV/GN languages, except for the two V-initial languages, Classical Arabic and Tongan In general it should be stressed that there is a considerable correlation between SOV word order and the property of being complement-deranking. Thus, in the sample, there are twenty-four complementderanking languages with the word order SOV/GN, two V-initial languages (Maori and Tahitian), two SVO languages (Ewe and Ifè) and one OVS language (Hixkaryana). This correlation is not yet fully understood. For example, Dik (1983), discussing the properties of ‘Prefield’ and ‘Postfield’ subordinated verbs (i.e., those that precede or follow the matrix verb respectively) writes: The fact that Prefield subordinated verbs tend to be non-finite is itself in need of explanation. Speculating about this, one might assume it to be due to a principle which says that the first finite Verb to occur in a complex sentence is preferably the main Verb. In Postfield patterns, this provides no problems, since the main Verb will occur in initial or second position in the clause; in Prefield patterns, in which the main Verb will occur in sentence-final position, or at least rather late in the complex sentence, the principle would still be fulfilled by giving a non-finite form to all the subordinate Verbs which precede the main Verb. Further research will be required to see to what extent such a principle can be maintained. (Dik 1983:285)
< previous page
page_229
next page >
< previous page
Page 230 Table 11.3 Word order and nominalization patterns SOV/GN SOV/NG V-1/NG SVO/GN SENT 14+?1 1+?1 POSS-ACC 12 6 ERG-POSS 3 4 8 1 NOMN 2+?1 1 4 MIX ?1 1 INC 2 3 4 REL 1 ARG-RED 1 Total 31 4 11 6 Notes: a English, an SVO language with two genitive positions (GNG) b Hixkaryana, an OVS/GN language c Ancient Greek, a language with a free word order d Latin, an SOV language with a free word order in NPs (GN/NG)
< previous page
next page >
page_230
page_230
SVO/NG
Other
5+?1 7 1+?2 1+?1 1 1
1a 2a,b 3a,c,d
14
4
Total 15+?2 24+?1 25 11+?3 2+?2 10 2 1
next page >
< previous page
page_231
next page >
Page 231 Dik’s tentative principle does not give any satisfactory account for the problem under consideration and is itself in need of explanation. Since many complements in SOV languages occur clause-medially, it is possible to assume that the great frequency of non-finite subordinate clauses in SOV languages is somehow connected to the perceptual difficulties in processing centre embedding. At least two slightly different hypotheses can be suggested, as follows. Grosu and Thompson (1977:145) observed that there are differences in the acceptability of heavy constituents and complements with initial subordinators and that the acceptability of sentences with non-final heavy constituents other than NP clauses improves substantially when the weight of the following clause-mate is increased; i.e., what seems to matter in the first place is the relative, rather than the absolute, weight of such phrases. In contrast, the acceptability of sentences with non-final NP clauses does not improve nearly to the same extent under comparable circumstances. To explain these differences, Grosu and Thompson hypothesize that ‘the hearer’s decision to accept or reject a sentence is made (i) for sentences with long constituents of any type, after the entire sentence has been received, and (ii) for sentences with NP-clausal arguments, at the point at which the noun clause in question begins to be received’ (ibid.: 147). This hypothesis (the Decision Point Hypothesis) predicts that the presence vs. absence of an initial marker of subordination in a complement is highly relevant for its acceptability. Thus, clause-medial complements with initial subordination markers are likely to be rejected, ‘because at the point where the decision is made, the complementizer is signalling the exocentricity (and potential disruptiveness) of the clause’ (ibid.: 148). On the other hand, in SOV languages without subordination markers (such as Mojave and Wappo, North-American Indian languages) or with final subordination markers (as in Japanese), the hearers miss the point at which they can judge the acceptability of clause-medial complements and, thus, do not reject them as centre-embedded constructions. If the hypothesis of Grosu and Thompson is correct, it may be suggested that non-finite complements arise as a ‘reward’ for the hearer’s expectations (or, rather, for the absence of any expectation), that is, he does not expect to hear a centre-embedded clause, and so he does not hear one either. In complement-deranking SOV languages with POSS-ACC ANCs, perception of such intra-clausal constructions is still more simplified. More specifically, the S and the A of ANCs (the left-most
< previous page
page_231
next page >
< previous page
page_232
next page >
Page 232 constituent in a considerable number of cases) look like the Gs of non-derived NPs. Upon receiving such a ‘genitive’, the hearer expects to receive the head-nominal and is, thus, not exposed to indeterminacy as to the status of the received constituent. The existence of SENT ANCs and of the centre-embedded finite clauses mentioned above argue for a slightly different view on the perception of SOV clauses, as suggested by Jan Anward (1985). This proposal is based on the assumption that it is much more difficult to process several clauses at the same time than to store long sequences of symbols without attributing any structure to them. In this view, SOV languages represent an extreme case of ‘look-ahead’ languages, in which ‘the speaker is allowed to hear symbols subsequent to a given terminal before making up his mind as to the appropriate or most probable tree structure’ (Kimball 1973:21). The quoted passage refers to English, an SVO language, for which Kimball assumes that ‘it may require less computation to hold a symbol in storage without tree attachment while one, but probably no more than two, subsequent symbols are scanned, than to build a tree only to have to return to alter it’ (ibid.). In the case of SOV languages, as suggested by Jan Anward, the hearer stores much longer passages, more specifically, long sequences of NPs, without making any decision about their relations to each other and to the clause. Only upon encountering a verb, does the hearer ‘go backwards’ and combine it with the appropriate NPs. By applying such strategies, the hearer avoids processing several clauses at the same time. In this connection, the distinction between finite and non-finite verb forms serves to direct ‘going backwards’. More specifically, only upon encountering a finite verb form, is the hearer ‘ordered’ to empty the stack of non-attached NPs by combining them with the verb form. After that, he is ready to start storing new NPs and to process new sentences. To conclude, the problem of the high frequency of non-finite subordinate clauses in SOV languages and of the high correlation between the property of being complement-deranking and having SOV/GN word order still needs much more study. As a closing remark, I will mention the case of Amharic, a Semitic language which has acquired SOV/GN word order and other features of SOV languages (influenced by the surrounding Cushitic languages). In a number of cases, complement-deranking SOV languages acquire some types of balanced complements, most often under the influence of other, complementbalancing languages (these complements seem to appear after finite verbs, in contrast to ordinary NPs; Jan Anward and Anders Holmberg, personal communication). However, for Amharic,
< previous page
page_232
next page >
< previous page
page_233
next page >
Page 233 a complement-balanced language, the opposite tendency has been observed, namely, the decay of finite subordinate clauses. A study of different Amharic texts based on one hundred sentences randomly picked out from each text, shows a dramatic increase in the use of ‘infinitives’ (verbal nouns) within a time space of nearly a hundred years, from two instances in the earliest case to forty-eight instances in the latest case (Kapeliuk 1980). This change has occurred at the expense of finite subordinate clauses, which are now used less and less often. Concerning this, Kapeliuk talks about a gradual process of ‘nominalization’ of the whole verbal system in Amharic (ibid.: 100). To use the terms adopted here, Amharic is gradually developing to a complement-deranking language. If this development continues, Amharic will become a very peculiar Semitic language, different from (most of) the others in two respects: (a) having an SOV word order; and (b) being complement-deranking. Turning to the two exceptional V-initial languages in the SENT group, it is worth pointing out that Classical Arabic and Tongan are also included among the languages which constitute another exception in the group as being complement-balancing (see section 11.2.). As suggested by Jeffers (1976), there might be a high correlation between VSO word order and a high frequency of subordination through nominalization. In this respect, at least, certain VSO languages approach complement-deranking languages in their frequent use of deranked complements. 11.3.2. There are no V-initial languages with INC in the sample This observation is in accordance with Solveig Pflueger’s hypothesis: ‘compounds with a relationship between nominal and verbal element are possible only in a language with a predicate node’ (quoted in Lehmann 1975:153). However, it may be argued that in at least some V-initial languages, ‘the verb and its complements do form a unit at some level of structure’ (Anderson and Chung 1977:24), but compounding seems to be particularly sensitive to the difference between continuous and discontinuous verb phrases. As Lehmann comments, ‘such compounds are productive only when O’s and V’s are contiguous in the underlying syntactic strings’ (ibid.: 153). He mentions the case of Hittite, in which the mutual position of a verb and its object is not definite and which has virtually no process of compounding. Insular Celtic, a VSO language constituting Lehmann’s other
< previous page
page_233
next page >
< previous page
page_234
next page >
Page 234 example, has abandoned most types of compounds typical for Proto-Indo-European, an SOV language. 11.3.3. Languages with ERG-POSS nominalizations show the greatest range of variation with regard to basic word order, favouring especially V-initial word order and disfavouring GN word order ERG-POSS languages exhibit all the word order types present in the sample including OVS. The only OVS language, Hixkaryana, has ANCs only of the ERG-POSS type. The ERG-POSS type favours verb-initial word order. There are only two verb-initial languages in the sample, Egyptian Arabic and Tagalog, which do not have ERG-POSS ANCs. One of them, Tagalog, prefers the word order VOS (in contrast to the other ten VSO languages in the sample) and, in general, utilizes a different basis for building finite clauses than most of the other languages in the sample (the ‘Philippine’ system). All the other nine languages, representing the Afro-Asiatic, Austronesian and IndoEuropean families, are found in the group under discussion. The ERG-POSS type disfavours GN word order. It is found only in three SOV/GN languages, the largest word order group in the sample. None of these three SOV/GN languages is nominative-accusative, being either ergative (Abkhaz) or manifesting split ergativity (Georgian and Panjabi). Note that SOV/GN languages are in the absolute majority among the SENT and POSS-ACC languages, which means that SOV/GN word order favours retention of sentential dependent-marking for the P. 11.4. THE STRUCTURE OF NON-DERIVED NPs Concerning the ways in which the structure of non-derived NPs in a language may influence its choice of nominalization patterns, we can distinguish between the following aspects: 1 Are more than one genitive positions available in NPs of the language (for the term ‘genitive’ as it is understood here, see section 3.5.2.)? 2 What types of affinities between the non-derived phrases and ‘verb-P’ or ‘verb-A’ structures exist (and predominate) in the language? A positive answer to the first question implies that the language can have DBL-POSS ANCs (either with lexical or pronominal arguments,
< previous page
page_234
next page >
< previous page
page_235
next page >
Page 235 see sections 8.1.1. and 10.2.) and includes two possibilities: different genitive positions are occupied by either formally identical or formally distinct dependents. However, as mentioned in connection with the DBL-POSS type, the opposite implication seems not to be true (or, at least, no evidence has been found in its favour). It may also be suggested that in some languages with DBL-POSS ANCs, NPs do not have multiple genitive positions, but the genitive itself has an extremely wide range of uses, and NPs with genitives are similar to both ‘verb-A’ and ‘verb-P’ structures. An answer to the second question may be relevant for the choice between those nominalization patterns which involve assimilation of dependents to the G (‘genitivization’), as it influences which of the dependents in transitive ANCs can become genitivized—the A or the P. Sections 3.4. and 3.5. examine some of the respects in which possessive phrases can be similar to either ‘verb-P’ or ‘verb-A’ structures. These include head-marking, dependent-marking, word order, as well as referential characteristics of the whole structure and of its head. It is also shown that possessive phrases need not be ‘harmonic’ in the sense that they are similar to one and the same structure (either ‘verb-P’ or ‘verb-A’) in all the above mentioned respects. Thus, the choice of nominalization patterns is influenced not only by the inherent similarities referred to above, but also by their relative force with respect to each other. As a hypothesis, I suggest this type of explanation for a possible choice between the POSS-ACC vs. ERG-POSS types, and between the ERG-POSS vs. POSS-ADN types. In ERG-POSS languages, possessive NPs are similar to ‘verb-P’ structures in the following cases: dependent-marking—in most of the Indo-European SVO languages (Russian, Polish, French, German) and in Estonian, genitives may be used in objective functions in finite clauses, this marking being determined either by subcategorization of the head verb, or by certain syntactic conditions (such as negation); head-marking—in five of the seven ERG-POSS languages with head-marking within NPs, this headmarking is closely related or almost identical to verbal object affixes (for Modern Israeli Hebrew this is true only diachronically, see section 7.4.); in Welsh constructions with ‘periphrastic’ verb forms, used in certain tenses and aspects, pronominal Ps cause the same mutation in the verb as pronominal Gs cause in the H;
< previous page
page_235
next page >
< previous page
page_236
next page >
Page 236 word order and referentiality—in ten ERG-POSS languages (six Indo-European and four Polynesian), there is an opposition between possessive NPs with (highly) referential and/or animate Gs which precede their head, and other Gs, which follow their head. Only the latter, more similar to ‘verb-P’ structures, give rise to ERG-POSS nominalizations (see sections 3.5.3. and 10.2.). In POSS-ACC and POSS-ADN languages, possessive NPs are similar to ‘verb-A’ structures in the following ways: dependent-marking—here such similarities are the least pronounced. The relevant examples include Icelandic, in which finite clauses can have subjects in the genitive, and the ‘negative’ case, that is, the absence of dependent-marking for both the As (or subjects) and the Gs, as in Egyptian Arabic, Modern Israeli Hebrew and Evenki; head-marking—in some languages, possessive nominal affixes are similar to verbal subject affixes (as in Buryat, Evenki, Selkup, Tuva and Turkish); word order and referentiality—in some cases, the Gs get a special dependent-marking only when referential and/or definite (as in Selkup, Modern Mongolian, Tuva and Turkish); in other cases, the presence of a G in a NP limits the range of possible referential characteristics of the head, in other words, the Gs function as determiners (as in Semitic languages, English, Icelandic and Swedish) (see section 3.5.1). Due to our limited knowledge about the structure of NPs across languages, the correlations mentioned in this section are particularly hypothetical. Note also that some of these facts in isolation would lead to false predictions about the occurrence of POSS-ACC, ERG-POSS and POSS-ADN ANCs. Thus, Amharic, Egyptian Arabic and Tagalog manifest some of the similarities between possessive NPs and ‘verb-P structures’ mentioned above, but they have no ERG-POSS ANCs. Modern Israeli Hebrew and Icelandic resemble each other in a number of respects (being SVO/NG languages with Gs of the determining type). Still only the first language allows ERG-POSS nominalizations (see section 3.5.3.). 11.5. HEAD-MARKING In the preceding section, some cases were mentioned in which the similarities in head-marking used within NPs and within finite clauses
< previous page
page_236
next page >
< previous page
page_237
next page >
Page 237 may influence the choice of nominalization patterns. Another question is whether the morphological marking type of a language per se has an impact on its nominalization possibilities. The following facts should be mentioned. Both the SENT and the DBL-POSS types disfavour head-marking within non-derived NPs. The only exceptions in the former case are Classical Arabic, Cuzco and Huallaga Quechua, and, possibly, Tuva; in the latter case, Modern Israeli Hebrew and, to some extent, Finnish. A possible explanation for the disfavoured character of nominal head-marking in SENT languages, proposed in section 6.3., proceeds from the observation that action nominals tend to acquire the nominal ability to be head-marked (among other nominal inflectional features), which results in a potential conflict in the status of the A in SENT ANCs. More specifically the A, at the same time, is qualified as a verbal dependent by its dependent-marking and as a nominal dependent by the head-marking of the verbal noun. Head-marking in DBL-POSS ANCs would lead to the presence of two markers of the same type in an action nominal. Modern Israeli Hebrew, which has each of head-marked, double-marked and dependent-marked possessive phrases, in its extremely rare DBL-POSS ANCs resorts to dependentmarking for at least one of the dependents. In Finnish, head-marking in the form of possessive suffixes is used only in the absence of an independent non-pronominal G in the same phrase. Thus, the head of a DBL-POSS ANC would have two possessive suffixes only in the case of pronominal A and P. This seems to be a highly hypothetical situation. Transitive ANCs with both arguments present are, in general, very rare in Finnish. For ANCs with pronominal As and Ps, my informants have not been able to give a single example and do not have any clear intuition about the structure of such phrases. In addition, there are the following connections between the morphological marking types of languages and their choice of nominalization patterns: (1) In the four (predominately) head-marking languages (with no dependent-marking on As and Ps of finite clauses and Gs of possessive NPs) in the sample, nominalization leads to reduction of headmarking on transitive ANs as compared to the corresponding finite clauses, so that of the two arguments, the A and the P, only one is marked on the head. These languages are Hixkaryana, Ancient Egyptian and Egyptian Arabic, all of which use dependent-marking for indirect objects, and Abkhaz, the only consistent head-marking language in the sample. Let us consider the relevant facts.
< previous page
page_237
next page >
< previous page
page_238
next page >
Page 238 Each of Abkhaz, Ancient Egyptian and Hixkaryana have ERG-POSS nominalizations in which the head is marked for the P by means of nominal possessive affixes, while the A takes dependent-marking. Thus, transitive ANs in these languages are head-marked only for the P, while the corresponding transitive finite verbs are marked both for the A and the P simultaneously. In contrast to that, transitive ANs in Egyptian Arabic cannot be head-marked for their P, while this is possible for the corresponding finite verbs. More specifically, in finite clauses, pronominal objects (either the direct object or the indirect object, though not both at the same time) can be marked on the verb, in the absence of an overt independent NP. To judge from the available descriptions and from informant work, such pronominal object clitics do not appear on action nominals. In other words, while transitive ANCs with non-pronominal Ps may follow either the POSS-ACC or the POSS-ADN pattern, those with pronominal Ps exhibit only the POSS-ADN pattern. (2) Among double-marking languages (i.e., those which in several constructions use both headmarking and dependent-marking, see Nichols 1986:72), it is reasonable to distinguish between those in which transitive finite verbs are regularly marked for more than one dependent, and those which have verbal agreement with only one argument. In the former case, non-derived nouns and finite verbs are opposed to each other both in the number of their head-marked dependents (one for nouns vs. more than one for transitive verbs), and in the form of the markers. (a) Languages in which finite verbs may be marked for more than one argument. Classical Arabic, Cuzco Quechua and Huallaga Quechua have transitive ANCs in which the head-nominal is marked both for the A and the P at the same time. Dependent-marking in Quechua ANCs follows the SENT pattern, while head-marking in such constructions conforms rather to the POSS-ACC type, comprising nominal possessive affixes for the A and verbal object affixes for the P. In Classical Arabic, double-marking within finite clauses concerns only the subject, since pronominal object clitics on finite verbs appear only in the absence of independent objects (cf. with the situation in Egyptian Arabic described above). Classical Arabic has SENT, POSS-ACC and ERG-POSS ANCs, the SENT being very rare. As mentioned in section 5.3., SENT ANCs do not employ head-marking for referring to the A, since this marking is intimately connected to the presence of a genitive dependent in the same noun phrase. In contrast to Egyptian Arabic, Classical Arabic allows for transitive POSS-ACC ANCs with
< previous page
page_238
next page >
< previous page
page_239
next page >
Page 239 pronominal arguments. In these, the verbal noun may have a (nominal) possessive suffix referring to the A, and a (verbal) object suffix referring to the P. The fact that the only transitive action nominals in the sample which are simultaneously head-marked for two dependents occur in the double-marking, rather than in the head-marking languages, appears to be significant and should be explained. A reasonable hypothesis is that head-marking has a different nature in head-marking and double-marking languages. In dependent-marking and double-marking languages, nominals are marked as different types of dependents and are, thus, opposed to (finite) verbs as being dependent-marked. A category shift from a verb to a nominal involves acquisition of the nominal inflectional properties, including, primarily, the ability to be dependent-marked in a number of ways and, in double-marking languages, the ability to be head-marked for the G. As long as verbal agreement is not in conflict with nominal head-marking, it can be retained by verbal nouns. In headmarking languages, however, the situation is different. Nouns are typically not at all or very sparsely marked as different types of dependents, but they are marked as being the head of a NP and have normally no more than one dependent. This is in contrast to (finite) transitive verbs, which are marked for being the head of a clause and being able to have more than one dependent. A category shift from a verb to a noun involves acquisition of the nominal marking potential, which means that derived nominals in head-marking languages should be marked for only one dependent and bear nominal affixes. Another consistent double-marking language, Inuit, has no transitive ANCs in the strict sense. In its anti-passive nominalizations, the head is marked only for the A. (b) Languages in which finite verbs agree with only one argument. In Turkish and Buryat, POSS-ACC ANCs are consistent in their head-marking and dependent-marking, in the sense that action nominals bear nominal possessive suffixes referring to that argument (the S or the A) which has been genitivized. In Tuva, the head AN bears possessive suffixes independently of the exact expression of the A, which alternates between the non-marked form and the genitive case (the SENT and the POSS-ACC type respectively). In Kurdish, the head of ERG-POSS ANCs takes nominal suffixes indicating that the head-nominal has dependents, without, however, specifying their kind. (3) Languages with split-marking (i.e., those which ‘have some
< previous page
page_239
next page >
< previous page
page_240
next page >
Page 240 head-marked and some dependent-marked patterns’, Nichols 1986:72) are very heterogeneous. The following cases may be distinguished. a Swahili uses head-marking within clauses and dependent-marking within NPs. Its POSS-ACC ANCs retain verbal object agreement. b Basque and Georgian use double-marking in clauses and dependent-marking within NPs. Both languages have no head-marking within ANCs, though in Basque, in SENT ANCs, all the arguments retain their dependent-marking. Close to them is Amharic, which exhibits two possibilities for pronominal Gs in NPs—either to appear as independent NPs with a special prefix, or to appear as possessive suffixes on the head-nominal. Amharic ANCs follow the POSS-ACC pattern and exploit head-marking to the same extent as non-derived NPs. c Evenki and Persian use double-marking in clauses (verbal agreement with the subject and, in the case of Persian, pronominal object clitics in the absence of independent objects) and head-marking in NPs (see, however, note 5 to chapter 7). ANCs in Evenki follow the POSS-ACC pattern, while those in Persian follow the ERG-POSS pattern. Modern Israeli Hebrew is close to Evenki, but has several options for NPs (dependent-marking, head-marking, double-marking). Its ANCs follow the POSSACC, ERG-POSS and, very rarely, DBL-POSS patterns. Selkup resorts mainly to dependent-marking, its only head-marking features being verbal agreement with the subject and possessive suffixes (used in the absence of independent Gs). Its ANCs follow the POSS-ACC pattern. In all the four languages, ANCs exploit head-marking to the same extent as non-derived NPs. (4) Predominantly dependent-marking languages constitute the majority of the sample languages and are represented in all the nominalization types, except the REL. In Daghestanian languages, ANCs (of the SENT type) retain verbal class agreement with the P, in contrast to all the other predominantly dependent-marking languages. In conclusion, morphological marking type may be included among the possible determining factors for the choice of nominalization pattern. Head-marking in NPs disfavours the choice of the SENT and the DBL-POSS types. Predominantly head-marking languages tend to choose nominalization patterns which ‘reduce’ the number of head-marked arguments as compared to the corresponding finite clause. The other facts mentioned above do not invite straightforward generalizations about possible connections between marking types and
< previous page
page_240
next page >
< previous page
page_241
next page >
Page 241 nominalization patterns. This may depend partly on the limited data, and partly on real differences between languages, for example, due to different origin of head-marking, as shown by Nichols (1986, cf. with Steele’s (1978) distinction between ‘copy agreement’, ‘semi-copy agreement’ and ‘inflectional agreement’). 11.6. NOMINATIVITY VS. ERGATIVITY Nominative-accusative languages are in the absolute majority in the sample and are frequent in all the nominalization types. Ergative languages and languages with split ergativity exhibit the SENT pattern (Basque, Burushaski, Daghestanian languages, Ingush and Tongan) and the ERG-POSS pattern (the SENT-POSS subtype in Samoan and Tongan and the OBL-POSS subtype in Abkhaz, Georgian, Kurdish, Panjabi). Agul, an ergative Daghestanian language, alternates between SENT and POSS-ACC ANCs, though the former pattern is much more common. Kurdish, a language with split ergativity, can build POSS-INC ANCs (besides ERG-POSS ones). Finally, Inuit resorts to antipassive and passive ANCs, while Chukchee avoids transitive ANCs with both the A and the P present (the ARG-RED type). The high percentage of ergative languages among the SENT and SENT-POSS languages as compared to their almost total absence among the POSS-ACC (the exception being Agul) and the NOMN languages seems significant. In other words, ergativity favours retention of sentential marking by the A and disfavours its genitivization. One possible explanation for this fact follows from the assumption that ergativity is, on the whole, motivated by semantic factors and that morphological marking in ergative languages is oriented towards differentiation of semantic roles, rather than towards differentiation of surface syntactic relations (see, e.g., Kibrik 1987:135). If this assumption is true, it follows that the ergative dependent-marking of the A in a clause need not be connected to its finiteness or independence and, thus, need not be a sentential feature. Since the A in finite ergative clauses and in ANCs has the same semantic role, it may also have the same kind of marking in both types of constructions and will hardly ever genitivize. According to Klimov (1973:189–90), the existence of a special genitive case and, especially, its functioning as a subjective or an objective genitive is very rare in ergative languages. Klimov states that the presence or absence of the genitive in a language can be used for dividing all the ergative languages into two groups—the ‘western’ group (which includes, among others, Daghestanian,
< previous page
page_241
next page >
< previous page
page_242
next page >
Page 242 Vejnax, Kartvelian, Tibeto-Burman languages and Basque) and the ‘eastern’ group (which includes Eskimo-Aleut, Australian, Paleosiberian, Abkhaz-Adygean and Polynesian languages and Burushaski) respectively. The rise of a subjective (and, rarely, objective) genitive has occurred only in those ‘western’ ergative languages that are going through a ‘relatively late period of ergativity’. In my sample, the ‘western’ group is represented by Daghestanian languages, Ingush (a Vejnax language) and Basque, which have ANCs of the SENT type, and by Georgian (a Kartvelian language), whose ANCs are of the ERG-POSS type. As for the ‘eastern’ group, it is true that many languages within it do not have a special genitive case for nominal dependents, opposed to all the other forms, but this, in different cases, has different reasons and has different consequences for nominalizations. Thus, Eskimo-Aleut languages, which have the same case for marking the As of finite ergative clauses and the Gs of non-derived NPs, lack ANCs of the SENT type (see section 6.1.). In these languages, the A in ANCs, although having the same marking as the A of finite ergative clauses, cannot be correlated to the latter, but rather to the A in the absolute case of anti-passive and passive clauses. In Burushaski, the ergative case is almost identical to the genitive case, but they do differ for a number of words. Unfortunately, the exact type of its nominalizations is not clear from my source of information (see section 5.2.). Abkhaz (a Northeast Caucasian language) is typically head-marking and does not have a special genitive form in the same way as it does not have distinct cases for the A and the P of finite clauses. Its ANCs manifest the ERGPOSS pattern. Polynesian languages have possessive particles used in nominalizations as well (more specifically, in ANCs of the POSS-ACC and ERG-POSS types), thus constituting a clear counter-example to Klimov’s statement. An especially interesting question in this connection is whether there is any dependence between ‘ergative-looking’ or ‘accusative-looking’ nominalizations (the terms used in Harrison 1981:181) and the syntactic typology of clauses. ‘Ergative-looking’ and ‘accusative-looking’ nominalization patterns include, firstly, the ERG-POSS vs. the POSS-ACC types respectively and, secondly, the Maori vs. the Tongan uses of the dominant/subordinate possessive particles in nominalizations respectively (the two versions of Chung’s ‘Syntactic type’; see section 10.4.). The problem of accusativity/ergativity in nominalizations vs. finite clauses has at least three sides: firstly, whether accusativity/ergativity of clauses can rule out or motivate changes in nominalization pat-
< previous page
page_242
next page >
< previous page
page_243
next page >
Page 243 terns; secondly, whether the existence of ‘accusative-looking’ or ‘ergative-looking’ nominalizations in a language may be viewed as an indicator of the accusativity vs. the ergativity respectively of its clauses; and thirdly, ‘whether an ‘‘ergative-looking” or “accusative-looking” possessor-marking pattern in nominalization can arise independent of the syntactic typology of clauses’ (Harrison 1981:181). The first question has been discussed in connection with the Polynesian languages, among which Maori, a nominative-accusative language, uses the dominant/subordinate opposition according to the ergative pattern, while Tongan, an ergative language, uses this opposition according to the accusative pattern.2 To reconstruct the Proto-Polynesian nominalization pattern, Chung (1973:669) assumes that grammatical changes ‘will only be historically plausible if they can be shown to be motivated by other structural facts of the languages involved’. According to Chung, a change from ergative to accusative possessive marking in Tongan nominalizations is historically reasonable, as it identifies the S and the A, producing a category of subject, the relevance of which in Tongan is demonstrated independently by processes of equi-deletion and Raising. Maori, however, does not seem to possess an independent category of ‘absolutive’ and, thus, a change from accusative to ergative possessive marking in Maori nominalizations, identifying the S and the P and producing a category of ‘absolutive’, is implausible. On the basis of these considerations, Chung concludes that the Maori ‘ergative-looking’ version of the Syntactic type should be viewed as a retention of the Proto-Polynesian syntax. Applied to the three problems formulated above, Chung’s conclusion amounts to saying that only a change of nominalization patterns in a certain direction, but not a nominalization pattern per se, is an indicator of clausal accusativity/ergativity. Clark does not agree with Chung, at least in the first half of her conclusion, because she does not take into consideration cases of spontaneous innovation. He writes, ‘if languages without absolutive/ ergative rules in their grammar can ever acquire them, then clearly there must always be a first such rule in the language. No reasons are given as to why this first rule should not be a possessor marking rule’ (Clark 1981, quoted in Harrison 1981:180). In connection with the second question stated above (whether the existence of ‘accusative-looking’ or ‘ergative-looking’ nominalizations in a language may be viewed as an indicator of the accusativity vs. the ergativity of its clauses), it is interesting that several nominative-accusative languages have ERG-POSS ANCs, though the
< previous page
page_243
next page >
< previous page
page_244
next page >
Page 244 reverse situation, ergative languages with POSS-ACC ANCs, is almost non-existent in the sample (Agul being an exception). This issue is further developed in chapter 12. Even if this asymmetry may be caused by the limited sample, it still seems to argue against any direct connection between ergativity of nominalizations and clauses. Such a conclusion may, however, be too hasty and straightforward. It is generally recognized that ‘a language is never wholly and exclusively either ergative or active or accusative, in all its grammatical patterns’ (Lehmann 1982:245). Some patterns, relating to the semantic role of agent, favour the active or the accusative system, while others, relating to topic, favour the ergative system (Schachter 1976, Comrie 1981:110–16). It would be interesting to study what kinds of implications there are among different active/ergative/accusative patterns. In other words, it may turn out that at least some of the predominantly nominative-accusative languages with ERG-POSS ANCs are less consistent than those with POSS-ACC ANCs. To quote one example, it has been demonstrated by Kozinskij (1983) that certain grammatical processes in Russian, predominantly nominative-accusative, make use not only of subject, but also of absolutive and patient (which combines direct objects and subjects of passive sentences). In connection with the third question (namely, whether the origin of ‘ergative-looking’ or ‘accusativelooking’ nominalizations should be sought in the syntactic typology of clauses), Harrison (1981:181) suggests that nominalization patterns may arise independently of the clausal syntax. This idea has bearing on the first two questions. More specifically, neither a change of nominalization patterns, nor a nominalization pattern per se need be an indicator of clausal accusativity/ergativity, since nominalization patterns may emerge as a result of some independent processes. Thus, for example, the morphological and syntactic affinities between possessive NPs and ‘verb-P’ complexes (see sections 3.4. and 11.4.) can give rise to ERG-POSS ANCs even in the absence of clausal ergativity. Nominalizations have sometimes been suggested as a possible source for ergativity. To use the terms adopted here, ergativity could have arisen in some erstwhile nominative-accusative languages due to the existence of ERG-POSS ANCs in these languages. In such languages, ‘ergativity would be a logically possible development if finite verbal forms were to be replaced by nominalizations’ (Comrie 1978:374). In his discussion of this hypothesis, Comrie states that he does not know any examples of such a development. In all cases known to him ‘where ergativity is alleged to derive via nominalization, it seems a priori plausible that a development similar to that in Iranian
< previous page
page_244
next page >
< previous page
page_245
next page >
Page 245 occurred, with nominalizations originally a device for forming passive constructions, and subsequent reinterpretation of these passives as ergatives’ (Comrie 1978:377). On the other hand, Comrie quotes Chol and Jacaltec, two Mayan languages, as examples of erstwhile ergative languages, which have become nominative due to the similar treatment of the A and the S in nominalizations. Though I do not know whether these languages have clausal or lexical nominalizations (see section 2.4. on the distinction between lexical and clausal nominalizations), I can still envisage the following scenario for the rise of nominativity in them. Both Chol and Jacaltec are head-marking, with verbal ergative affixes identical to nominal possessive affixes, as opposed to verbal absolute affixes. I suggest that it is the character of head-marking which has been the prime factor in introducing nominativity. More specifically, as has been mentioned in connection with the SENT type, intransitive ANCs are the first to assimilate to nonderived NPs. Transitive ANCs, on the other hand, are less prone to such assimilation. We might suppose that clausal nominalizations are likely to exhibit similar tendencies. Thus, a plausible hypothesis is that intransitive nominalizations in Chol and Jacaltec assimilated to NPs, that is, verbal absolutive affixes were replaced by nominal possessive affixes, while transitive nominalizations retained their sentential head-marking. As a result of this differential treatment of intransitive and transitive nominalizations, the S in intransitive nominalizations came to be marked in the same way as the A in transitive nominalizations, which, in its turn, was reanalysed as subject marking. Anderson (1977:330) mentions the fact that, among asymmetries in mixed ergative and accusative languages, ‘languages may have ergative marking in main clauses and accusative marking in subordinate clauses, but apparently not vice versa’ with the possible exception of Ngarlumain Australia. By ‘subordinate clauses’ he, presumably, means finite subordinate clauses. Extended to complement-deranking languages, Anderson’s generalization does not seem to be correct. Thus, ANCs in Maori and Hixkaryana, which are the only noun clauses in these complement-deranking languages, may follow the ERG-POSS type, i.e., subordinate clauses are morphologically ergative, while matrix clauses are either nominative-accusative or, in the case of Hixkaryana, indeterminate. 11.7. MORPHOLOGICAL TYPE Jaxontov (1981) offers some generalizations on dependences between the morphological type of a language and its complement system.
< previous page
page_245
next page >
< previous page
page_246
next page >
Page 246 According to him, in fusional languages, complements of verbs of cognition, verbs of speech, verbs of perception and verbs of emotions are expressed by finite subordinate clauses, while complements of the other complement-taking verbs are expressed by infinitives. In agglutinating languages, complements correspond typically to various non-finite clauses, that is clauses with the main predicate expressed by participles, converbs and verbal nouns. Here it is possible to distinguish between two groups of languages. In the first (Uzbek, Buryat, Evenki), complements of verbs of cognition, speech, emotions, and perception are expressed by participles or action nominals, while complements of the other complement-taking verbs correspond either to the same forms or to converbs. In the second group, complements are treated in almost the same way as in fusional languages, the only difference lying in their relation to the infinitive—this form either does not exist at all, or is the same as the form of the converb, or there are several infinitives in the language. Expressed in the terms adopted here, Jaxontov’s generalizations amount to saying that agglutinating languages can be both complementbalancing and complement-deranking, while fusional languages are complement-balancing. Since the property of being complement-balancing or complement-deranking has direct bearing on the choice of nominalization patterns in a language, as stated in section 11.2., the morphological type of the language might also be a determining factor in this choice. It should be mentioned here that Jaxontov’s article is based on a very limited number of languages. Furthermore, complement systems are not always treated consistently presumably due to inconsistencies in the descriptions of the languages themselves. Thus, Jaxontov does not mention action nominal constructions in either English or Russian, though, as we have seen, they play an important part in both languages. The distinction between agglutinating and fusional languages cannot be applied in a straightforward manner to a great number of languages in my sample, since many of them (such as Basque, Daghestanian languages and Finnish) use a combination of agglutination and fusion. As stated by Lehmann (1973) and confirmed by Bybee (1985:46), agglutinating languages tend to exhibit SOV word order. Since there is a high correlation between complement-deranking and SOV word order (see section 11.3.), it is not clear how all the three features, SOV word order, agglutination and complementderanking, are related to each other and to the choice of nominalization patterns.
< previous page
page_246
next page >
< previous page
page_247
next page >
Page 247 12 Explanation of the attested types 12.0. INTRODUCTION Let us return now to the first explanatory question stated in section 11.0., namely: why the typology of the ANCs contains just the types described in chapters 5 to 9. Among other things, we should try to explain the generalizations formulated in section 3.4. and repeated here: 1 in all languages, ANCs involve marking (both dependent-marking and head-marking) used in either finite clauses or in possessive NPs; 2 of the two arguments in transitive ANCs, the one most likely to lose its sentential dependent-marking and head-marking is the A (except for the marginal SENT-POSS type); 3 either all the arguments in a transitive ANC retain their sentential dependent-marking or at least one of them genitivizes (except for the marginal MIX type); 4 head-dependent word order is the same in ANCs and non-derived NPs (with the possible exception of some Kwa languages, see section 9.2.); 5 if an ANC retains its sentential head-marking for some argument, it also retains its sentential dependent-marking for that argument (except for one type of intransitive ANCs in Tabassaran). On the basis of these generalizations, we can draw two conclusions. First, ANCs across languages are always sentence-like and/or NP-like with respect to their internal syntax, the extent of these similarities varying across languages. Second, there exist cross-linguistic limitations on the possible combinations of such sentence-like and NP-like properties of ANCs.
< previous page
page_247
next page >
< previous page
page_248
next page >
Page 248 12.1. FINITE CLAUSES VS. NPs VS. ANCs: DIRECT ANALOGY OR MODELLING? How can we explain the fact that ANCs across languages are sentence-like and/or NP-like? What are the possible links, synchronic or diachronic, between the three constructions? Let us start by looking at one attempt to relate clauses, NPs and nominalizations to each other, namely, Givón’s (1975) ‘nominalization model’. Givón challenges the general view that the parallelism between OV word order in clauses and GN word order in NPs is due to direct analogy between different types of ‘operator-operand’ (Vennemann 1973) or ‘head-modifier’ (Greenberg 1963) structures, and claims instead that NP syntax is dependent on that of VP and is based on an unidirectional analogy (that is, from VP syntax to NP syntax). To quote Givón, in essence, the claim of ‘direct analogy’ is really a claim that speakers, and in particular children during first-language acquisition, are able to analogize along Chomsky’s X-bar convention. That is, they are able to see the modifier standing in the same relation to the head noun as complement or object to the main verb. I think this is most unlikely. It would further require that the acquisition of noun modification coincide with the acquisition of verb complementation, again a proposition that is not at all supported by current studies on the acquisition of language. (Givón 1975:107) To avoid the two above mentioned problems, Givón proposes a ‘nominalization model’ according to which VP syntax is transferred into NPs via nominalizations. Let us have a look at one of the instances of this model which accounts for the prenominal position of ‘genitival’ or ‘genitive-derived’ nominals in SOV languages. Unidirectional analogy from clauses to NPs is accomplished in two steps. First, nominalization rules ‘create OBJECT-VERB nominal compounds (eventually re-analyzed, in many languages, as NOUN-NOUN com-
Figure 12.1
< previous page
page_248
next page >
< previous page
page_249
next page >
Page 249 pounds)’, so that ‘the OV word order in the verb phrase is transferred to a GEN-N order in the noun phrase’ as in Figure 12.1. In other words, Givón means that there is an obvious analogy between VPs and nominalizations thereof, which provides for their identical word order at the first step from VP syntax to NP syntax. The second step is provided by an analogy between nominalizations and other types of NPs. To quote Givón, if a language has ‘real possessive’ nominals involving the same genitive case marker as the one used in the nominalizations described above (as is indeed the case in Indo-European, Semitic, Bantu as well as many and maybe most other languages), then the analogy from the genitive derived by nominalization on to the ‘true possessive’ genitive is a more viable hypothesis. (Givón 1975:90) Thus, the ‘nominalization model’ accounts for the dependent status of NP syntax as compared with VP syntax and makes use of the concrete and obvious analogies ‘VP—nominalized VP’ and ‘nominalized VP —NP’ instead of much more abstract principles based on the ‘head—modifier’ opposition and the like. In connection with Givón’s suggestion, it is reasonable to distinguish between two different questions, namely: (1) does the ‘nominalization model’ adequately explain the parallelism in the NP and VP word order in those cases where the same ‘genitive case marker’ is used with ‘real possessive’ nominals and with dependents in nominalizations; and (2) is the ‘nominalization model’ plausible as a description of the general language mechanisms responsible for the structure of VPs, NPs and nominalizations? The two questions will be considered separately. To start with the first question, I do not exactly understand the role of compounding in the model. Contrary to what Givón states in the quoted article, compounds and nominalized OV compounds belong to patterns with conservative word order, a fact mentioned by Givón himself in a later work (1984:216). This point was discussed in section 11.3. in connection with word order in the INC type. In general, I do not see much point in attaching so much weight to compounding in the ‘nominalization model’, since what is important is the existence of nominalizations, in which the P is treated in the same way as the G in non-derived NPs. Putting compounding aside, there are still problems with languages in which it is the A of nominalizations which genitivizes. For example, in Bantu languages the basic word order is SVO/NG. At least some of them, such as Swahili and Luganda, build
< previous page
page_249
next page >
< previous page
page_250
next page >
Page 250 ANCs of the POSS-ACC type, in which the transitive AN precedes both its dependents. If the syntax of possessive NPs in these languages is dependent on that of VPs or clauses, what role can be attributed to nominalizations? More specifically, the order NG is parallel to the order VO, but the Ps in nominalizations retain their sentential marking, thus, the analogy between them and Gs is somewhat farfetched. On the other hand, the As of ANCs are marked in exactly the same way as the Gs in nonderived NPs and would be viable candidates for analogy, but the order between the A and the head in ANCs and in finite clauses is different. Thus, it seems that the ‘nominalization model’ gives wrong predictions for some of the cases in which it is applicable. Turning to the second question, we can first investigate the extent to which the ‘nominalization model’ can account for the parallelism between word order in clauses and in NPs, more specifically, between OV and GN across languages. Thus, the model is not applicable at all to the languages with the SENT type of nominalizations, most of which are SOV/GN languages, but in which neither dependent in nominalizations is marked in the same way as the Gs. In this case, there is no formal similarity between the P of nominalizations and the G, but they still occupy the same position with respect to their head. Consequently, the ‘nominalization model’ cannot account for all cases of word order similarities between NPs and VPs, and we are still left with some variant of ‘Head-Dependent’ explanation. Is it plausible to suggest that NP syntax is dependent on VP syntax and that nominalizations are a type of a bridge between the two constructions? That this is not always true is demonstrated by Persian and Kurdish, SOV/NG languages. Vennemann (1973:352–4, quoted in Mallinson and Blake 1981:403) mentions Persian as an example of a language in which ‘minor’ constructions can change their order while the clauses retain their old word order. In this case, the order within NPs is innovative (due to Arabic influence, according to Vennemann), while that in clauses is conservative. Of particular relevance for the present discussion is the fact that both Persian and Kurdish resort to ERG-POSS ANCs in which the P follows the head nominal in contrast to the corresponding finite clause. In other words, the word order in nominalizations is derived from that of non-derived NPs, and not the other way round. Thus, Givón’s ‘nominalization model’ can hardly account for the parallelism in word order of OV and GN sequences in the general case. Still, it may be true that nominalizations with the Ps that assimilate to the Gs, can exert pressure on non-derived NPs when the basic word order in independent clauses has changed.
< previous page
page_250
next page >
< previous page
page_251
next page >
Page 251 Let us return to the problems formulated at the beginning of this section: (1) how can we explain the fact that ANCs across languages are always sentence-like and/or NP-like? and (2) what are the possible links between finite clauses, NPs and ANCs? The generalization underlying the first question can be expressed in another way: no language has syntactic means which are exclusively used in ANCs. This is a highly significant fact which shows a fundamental difference between ANCs and both finite clauses and NPs. Thus, many languages have syntactic means, restricted to only one of the latter two constructions, such as the genitive case in Quechua or Korean, which is only possible in NPs, or the accusative case in Russian, which is not allowed in either non-derived NPs or in ANCs. ANCs are, in a sense, dependent on finite clauses, NPs or both. This fact, together with the facts mentioned in connection with Givón’s model, provides evidence for the following view on the nature of ANCs as being modelled on finite clauses and NPs.1 It has been common in Generative Grammar to describe the process of language learning as if it were instantaneous. For example, Chomsky (1975:121–2) concludes that this ‘obviously false’ simplifying assumption is ‘nevertheless a legitimate one and provides a proper basis for pursuing a serious inquiry into human cognitive capacity’. Chomsky’s focus is on how learners construct ‘linguistic hypotheses’ in the light of the ‘experience’ they get from the environment. His argument showing that we do not find ‘substantial differences in the result of language learning depending on such factors as order of presentation of data, time of presentation, and so on’ (ibid.: 121) does not take into account that there is a certain order in which new items are learned, independently of the input. It is here that the assumption of instantaneous learning leads to a theory that misses interesting generalizations. An important optimization aspect of a developing grammar, besides construction of ‘linguistic hypotheses’, is how items and constructions that are learned at a given point are fitted into a system consisting of items and constructions learned earlier. A minimal requirement is that the changes that are made when a new item or construction is added to the system are such that the system still functions afterwards. Beyond this requirement, different strategies are possible. The view that will be taken here is that when a new construction is added to a grammar, it is, as far as possible, modelled on already existing constructions. Action nominal constructions and other types of subordinated clauses are fairly complex constructions corresponding to quite advanced communication needs, appearing relatively late in language
< previous page
page_251
next page >
< previous page
page_252
next page >
Page 252 acquisition and, probably, in language development. As such, they must fit the already existing language system in such a way that both production and perception of the new system will not be inhibited. Note that the ‘modelling hypothesis’ is in a way reminiscent of Givón’s ‘unidirectional analogy’, that is, ANCs appearing later in language acquisition have a syntax dependent on that of constructions acquired earlier. However, which constructions should be taken as the basis for modelling? The present study shows that, for ANCs, the choice is severely restricted to finite clauses and non-derived NPs. More specifically, since action nominals both morphologically and semantically occupy an intermediate position between prototypical verbs and prototypical nouns, it is justified for them to have syntactic properties of either one or both of these word classes. Prototypical constructions for verbs and nouns are finite clauses and NPs respectively, in which these words are lexical heads. Thus, ANCs with ANs as lexical heads are modelled on the basis of either one or both of these two constructions. An interesting point in this connection is that of all NPs, it is possessive NPs which serve as a model for ANCs. A similar explanation has been proposed for comparative constructions by Stassen (1985), who claims that such constructions are modelled on constructions with temporal chains. Let us now compare ‘modelling’ as applied to comparative constructions and ANCs. The notion ‘comparative construction’ is defined semantically, and a typology of comparative constructions comprises cross-linguistic types of codification of the concept ‘comparison’. It turns out that the choice between comparative types in a language is determined by the possible consecutive chains, which are also defined semantically as encoding the concept ‘temporal chaining’. In this sense, comparative constructions are not ‘elementary’, ‘basic’ or ‘autonomous’, but ‘parasitic’ upon consecutive chains. To quote Stassen, in order to express this relation of dependency between the linguistic codifications of comparison and temporal chaining, we will say that, in natural languages, the linguistic codification of comparison is claimed to be modelled upon or borrowed from the ways in which the concept of temporal chaining can be encoded in natural language systems. (Stassen 1985:105). According to Stassen, this is plausible because the perception of temporal ordering between events is a very elementary psychological process and is much more basic than the perception of comparison (ibid.: 255). On the basis of these correlations, it is stated that
< previous page
page_252
next page >
< previous page
page_253
next page >
Page 253 ‘the underlying structures of comparative constructions in natural languages have the form of temporal chains’ and the surface structures of both types of constructions are derived by means of the same syntactic procedures from the underlying structures (ibid.: 334). To summarize, the mental operation by which two objects are compared may not be an independent, ‘primitive’ operation; at least as far as its codification into language is concerned, the mental act of comparison must be seen as a conceptual extension of the mental operation by which two events are ordered with respect to their occurrence in time. (ibid.: 105) Thus, codification of comparison in a language is predetermined by codification of the more basic concept, ‘temporal chaining’, and, in this sense, comparative constructions are modelled on consecutive chains, even if the two lack overt surface parallelism. In the present study, however, ‘modelling’ applies to ANCs which manifest overt surface parallelism with finite clauses and/or NPs, even if the choice of an ANC type is not predetermined by the structure of these latter constructions. This is the crucial difference between Stassen’s and my use of the term ‘modelling’, which can partly be attributed to the fundamental differences between comparative constructions defined semantically in language-independent terms and ANCs, which for each language are determined in a specific way. What Stassen claims is that expressions for less basic concepts may be predetermined by expressions for basic ones. What I claim is that linguistic constructions per se may be more or less basic. In other words, even in the absence of obvious conceptual reasons, the increased communication needs give rise to complex linguistic constructions which may be modelled on more basic, ‘primitive’, primary constructions. This follows from the fact that ANCs are absent from many languages, since there is no universal concept which calls for their existence, while all languages have comparative constructions, necessary for encoding the concept of comparison. Another, not uncontroversial question is whether constructions which primarily meet communicative, rather than conceptual needs can still be conceptually different from all the other constructions in the language. Thus, for example, Langacker (1987) claims that nouns and verbs can be attributed universally valid notional characterizations and that ‘nominalizing a verb necessarily endows it with the conceptual properties characteristic of nouns’ (ibid.: 90).
< previous page
page_253
next page >
< previous page
page_254
next page >
Page 254 12.2. SENTENCE-LIKE VS. NP-LIKE PROPERTIES OF ANCs In the present study, ANCs are classified with respect to dependent-marking for the arguments within them. The attested types may be ordered in accordance with the degree to which they are sentencelike, as shown in Figure 12.2. If we order the nominalization types according to their degree of NP-likeness (Figure 12.3), we obtain a scale which is close to being the opposite of the scale presented in Figure 12.2. One of the exceptions concerns the MIX type, which is less sentence-like than the SENT, because the marking of the A is not the same as in the unmarked, active finite clauses. It is, however, more sentence-like than both the POSS-ACC and the SENT-POSS, because both the A and the P in the MIX ANCs are marked as in some types of finite clauses, while in POSS-ACC and SENT-POSS ANCs, one of the arguments (the A or the P respectively) genitivizes and, thus, the relation between the head nominal and this argument is signalled in a way alien to finite clauses. However, neither the SENT nor the MIX are NP-like and, therefore, occupy the same position in Figure 12.3. In the general case, there is no difference between the SENTPOSS and SENT-INC vs. OBL-POSS and OBL-INC in their degree of NP-likeness, which is why they also occupy the same position in Figure 12.3. If we consider only the major nominalization patterns, the scales in Figures 12.2 and 12.3 will be reduced to that shown in Figure 12.4. Retention of sentential marking for arguments in ANCs is correlated
Figure 12.2 Scale of sentence-like dependent-marking in nominalization types
Figure 12.3 Scale of NP-like dependent-marking nominalization types
< previous page
page_254
next page >
< previous page
page_255
next page >
Page 255
Figure 12.4 Scale of sentence-like and NP-like dependent-marking in the major nominalization types with retention of other sentential properties in ANCs, such as the ability to combine with adverbs, to be negated in the sentential way and to exhibit tense oppositions. As mentioned previously, my data of this kind are very limited. The only satisfactory and relevant information concerns tense distinctions within ANCs. Thus, among the seventeen SENT languages, eight have distinct tenses/aspects in ANCs, five languages among the twenty-five POSS-ACC languages and one language (Hixkaryana2) of the twentyfive ERG-POSS languages in the sample. ANCs of the other types do not exhibit distinctions in tense. With respect to this feature, the major nominalization patterns form a hierarchy which is the same as that of Figure 12.4. Though ANCs across languages vary considerably in their degree of similarity to finite clauses and possessive NPs, this variation is severely restricted, in at least two respects. Firstly, as generalizations (4) and (5) above demonstrate, the three syntactic means, word order, head-marking and dependentmarking, form a hierarchy of their nominal and sentential properties. Secondly, as generalization (2) shows, different arguments vary as to their proneness to retain sentential marking. Let us look at these points separately. Note that ‘sentential’ here refers to all properties that distinguish finite clauses from nonderived NPs. However, as follows from the above mentioned generalizations, in some cases it is appropriate to oppose ‘sentential’ properties, which are restricted to finite clauses, to ‘verbal’ properties. In the realm of word order, there are no distinctions between more sentence-like and more NP-like ANCs, as ANCs across languages have NP-like word order. More specifically, arguments in ANCs precede or follow the head in the same way as Gs precede or follow the H in non-derived NPs. (In the case of Swedish POSS-ADN ANCs, the P follows the AN just like prepositional dependents in non-derived NPs, while the A precedes the AN, just like the G in non-derived NPs.) In other words, the very fact that a word has nominal inflectional characteristics seems to imply a certain relative order between the word and its dependents, even though the form of the dependents can vary. Head-marking in ANCs tends to be more NP-like than clause-like. However, there are some instances in which ANCs retain sentential
< previous page
page_255
next page >
< previous page
page_256
next page >
Page 256 head-marking, namely, SENT ANCs in Cuzco and Huallaga Quechua, POSS-ACC ANCs in Swahili and Classical Arabic which retain verbal agreement with the P, and SENT ANCs in Archi, Tabassaran and other Daghestanian languages, which retain agreement with the S and the P. These cases give rise to generalization (5), because in all of them, retention of sentential head-marking for a certain argument in an ANC is accompanied by retention of sentential dependent-marking for the same argument. Note that head-marking for the A is never retained. Archi and Tabassaran retain verbal class agreement with the S, which is avoided by other languages. This agreement, at least in Tabassaran, is peculiar in still another respect, as it is retained even when the S is genitivized, which, thus, constitutes a counterexample to generalization (5). Putting aside the Daghestanian case, my data show that agreement with the S and the A, as it is manifested in finite clauses, is a stronger indicator of being a sentence than head-marking of the P and, thus, the data confirm the general impression that ‘subject agreement’ is intimately connected to finiteness of clauses and is a typical sentential property. Head-marking of the P, on the other hand, is rather a verbal property, and may be present or absent in ANCs just like other verbal properties. However, as generalization (5) shows, it is a ‘more verbal’ property than dependentmarking of arguments. Dependent-marking in ANCs (the basis for the typology of ANCs) shows the greatest range of variation. Still, there are some limitations to this variation, connected with the relative proneness of the different arguments to retain sentential marking. Thus, it is relatively unusual for languages to combine sentential dependent-marking for the S with non-sentential marking for the A or the P, the only exception being Tahitian and Basque as mentioned in section 10.3. The reverse situation, genitivization of the S combined with sentential marking for the A and the P, is more common. This is easy to understand, in the general case, as intransitive ANs are similar to non-derived NPs in having only one unmarked dependent. In general, languages in the sample do not normally use distinct dependent-marking in ANCs for each of the three arguments, the S, the A and the P. Possible exceptions to this tendency are constituted by MIX languages, which oppose genitivized Ss to both the As and Ps in ANCs. However, in all these languages MIX ANCs alternate with either POSS-ACC or ERGPOSS ones, in which one of the dependents is marked in the same way as the S in intransitive ANCs. Another interesting point in connection with dependent-marking concerns the A, more specifically, the fact that the A is less likely to
< previous page
page_256
next page >
< previous page
page_257
next page >
Page 257
Figure 12.5 Hierarchy of sentential vs. NP-like syntactic means in ANCs retain its dependent-marking than the P. This is probably another piece of evidence for the special character of this argument as compared to the P. There are several possible explanations for this phenomenon, but these will not be discussed here. Pronominal Ss and As are more likely to genitivize than non-pronominal ones. To conclude, syntactic means exploited by ANCs form a hierarchy (Figure 12.5) which predicts which combinations of nominal and sentential properties are possible. This hierarchy orders syntactic means according to their relative proneness to assimilate to the patterns found in NPs. The hierarchy of nominal and sentential syntactic means in ANCs may be useful in studies of major word classes, as it suggests which features tend to be clustered together and, thus, are implied by a certain categorization of a word. 12.3. ANCs AS PRODUCTS OF COMPETING MOTIVATIONS How can we explain that the typology of ANCs presented in the preceding chapters contains just the types it contains? In other words, even if we assume that ANCs are modelled on the basis of either finite clauses or possessive NPs and if the hierarchy presented in Figure 12.5 is true, what factors cause the attested variation in nominalization patterns across languages? This typology is probably a result of several competing principles, or ‘competing motivations’ (on this notion see, for example, DuBois 1985). Some tentative principles which motivate the existence of different nominalization types are proposed below. Principle 1 The more uses ANCs have where they are semantically independent, the more they will be similar to independent clauses. This principle accounts for the fact that ANCs in many languages are more or less sentence-like, more specifically, that they exhibit the SENT, the POSS-ACC, the ERG-POSS and the MIX patterns. Ideas
< previous page
page_257
next page >
< previous page
page_258
next page >
Page 258 close to this principle have been formulated earlier by different linguists, as, for example, Givón’s (1980) ‘binding hierarchy’ and the Interclausal Relations Hierarchy of Foley and Van Valin Jr (1984). Givón suggests that the more the agent of the complement clause is influenced by the agent of the matrix clause, the less the complement would tend to be coded as an independent clause. Foley and Van Valin Jr state that ‘the more tightly linked the dependent clause is to the adjacent clause, the more the casemarking possibilities in it are reduced’ (1984:278). These correlations can be considered as one of the instances of Haiman’s ‘motivation hypothesis’: ‘Given two minimally contrasting forms with closely related meanings, the difference in their meaning will correspond to the difference in their form’ (Haiman 1985:20), or applied here, ‘the greater the formal distance between X and Y, the greater the conceptual distance between the notions they represent’ (ibid.: 106). What facts confirm Principle 1 in the data? As mentioned previously, the actual range of uses typical for ANCs in the sample languages is left outside the present study. However, it would appear that the distinction between complement-deranking and complement-balancing languages bears witness to Principle 1. More specifically, in complement-deranking languages, ANCs constitute the only type of noun clauses and, as such, are used even for coding relatively independent complements, such as complements to utterance predicates. These complements have independent time reference and, in many cases, independent truth-value, to use Noonan’s (1985) terms,3 and the control exerted by the agent of the main-clause verb over the agent of the complement is minimal, to use Givón’s (1980) characteristics. Thus, according to Principle 1, ANCs in such languages will tend to be relatively similar to independent clauses, which is corroborated by the strong correlations between the properties of being a complement-deranking vs. complement-balancing language and of having SENT vs. NOMN ANCs. The distinction between complement-deranking and complement-balancing languages gives a very rough approximation to the actual uses of ANCs across languages. Thus, for example, complementderanking languages may have other types of deranked complements which compete with ANCs and are even preferred in many situations. For example, ANCs in Abkhaz, a complement-deranking language, have relatively restricted uses, since most subordination involves deranked s-like clauses (see section 2.2.2. for discussion of Abkhaz). On the other hand, even complement-balancing languages may vary as to their frequency of use of ANCs for coding more or less independent
< previous page
page_258
next page >
< previous page
page_259
next page >
Page 259 complements. It is reasonable to suppose that cross-linguistic differences of this kind may be correlated to the differences in the choice of nominalization patterns. Thus, discussing ANCs with the sententially marked Ps which are found in Bulgarian, Macedonian and sporadically in Polish, Comrie (1976:193) writes that, in these languages, ‘the action nominal is used much more frequently, and in a greater range of productive constructions, than in other Slavonic languages; this may well be a factor in promoting the closer tie between action nominal and verbal syntax in the former group of languages’. Also, as mentioned in section 11.3., a number of VSO languages, though not being complementderanking, frequently resort to subordination through nominalization (Jeffers 1976). I feel that it is not only (or not so much) the frequency of ANCs, but their semantic range which favours retention of sentential syntax, as suggested by Principle 1. This is supported by the few cases of functional splits, arising when an action nominal combines with its arguments in different ways depending on its exact meaning (see section 12.1.). This problem needs further investigation. Principle 1 predicts that ANCs across languages may be more or less sentence-like, in other words, that independent clauses constitute a pole to which ANCs are attracted. Thus, if only this principle were at work, there would not necessarily be any system in ‘deviations’ from sentence structure exhibited by ANCs. However, as is clear from Figures 12.2–12.4 this is not the real state of affairs. The important point is that ANCs, with marginal exceptions, ordered according to the extent of their sentence-likeness, are in almost reverse order according to the extent of their NP-likeness. In other words, NPs constitute the other pole to which ANCs are attracted. This may be accounted for by Principle 2 below. Principle 2 Words that share inflectional features of nouns, tend to assume nominal syntactic features.4 Note that this principle is particularly corroborated by the behaviour of intransitive ANCs, which, in the absolute majority of the sample languages, assimilate to non-derived NPs. This suggests that languages try to minimize conflicts between nominal and verbal properties of one and the same word, as far as this is possible. It is reasonable to imagine that Principle 2 and the like are responsible for categorization of words into word classes, which presupposes that certain morphosyntactic properties tend to be clustered together. However, I will not dwell on this matter here.
< previous page
page_259
next page >
< previous page
page_260
next page >
Page 260 The two principles, 1 and 2, account together for the fact that ANCs across languages are stretched between the two poles, independent clauses and NPs. However, what motivates the existence of both the POSS-ACC and ERG-POSS types? As mentioned in section 11.6., the present study shows an asymmetry in the distribution of the two types, namely, while both ergative, split-ergative and nominative-accusative languages have ANCs of the ERG-POSS type, ANCs of the POSS-ACC type are extremely rare in non-nominative-accusative languages. Let us first discuss what can favour the existence of ERG-POSS ANCs in nominative-accusative languages. It turns out that in many languages, transitive ANCs with two overt arguments are very rare (which, among other things, increased my difficulties in obtaining data for the present study). For example, it has been shown by Hopper and Thompson (1980:285) that out of 100 cases with transitive ANCs in English, only 5 had two arguments. In other words, in most cases, transitive ANs are similar to intransitive ANs by having only one argument. Typically, this overt argument is the P. To use DuBois’ (1985:349) terminology, for English and, at least, some other languages, S/P constraint is the ‘preferred argument structure’ in ANCs, ‘that grammatical structuring of arguments which is statistically ‘‘preferred” in clause tokens in discourse’. This suggests two hypotheses: in many languages, ‘AN-P’ complexes are formally identical to intransitive ANCs either directly, due to essential similarities between the P and the S in them, or indirectly, due to their assimilation to non-derived NPs, in each case occurring independently. In other words, according to the former hypothesis, the ‘preferred argument structure’ in ANCs is founded on essential similarities between the S and the P in them, for example, on their similar, possibly discourse, functions. According to the latter hypothesis, the P and the S need not themselves be similar, but it is the ‘AN-P’ complex, much more frequent than the ‘AN-A’, which assimilates to nonderived NPs. The former hypothesis is reflected by Principle 3a, while the latter corresponds to some variant of economy principles, such as Principle 3b: Principle 3a Dependents with similar (discourse) functions get the same marking. Principle 3b More frequent complexes tend to be syntactically simpler. Note that these principles are potentially valid for the ERG-POSS, POSS-REL and, partly, DBL-POSS types.
< previous page
page_260
next page >
< previous page
page_261
next page >
Page 261 A relevant question in this connection is why transitive ANCs with two arguments are so rare, or why nominalization ‘is typically associated with a reduction in actual valency [the number of arguments that receive overt expression]’ (Mackenzie 1985:32). To use the terminology and the explanation offered by Hopper and Thompson (1980:283), nominalizations are low in Transitivity (where Transitivity is understood as a global property of clauses) because their main function, at least in narratives, is to be a constituent part in the backgrounded portions of discourses, ‘that part of a discourse which does not immediately and crucially contribute to the speaker’s goal, but which merely assists, amplifies, or comments on it’ (ibid.: 280). According to Hopper and Thompson, the two most important characteristics of foregrounded clauses as opposed to backgrounded ones consist of the fact that (a) ‘the foregrounded portions together comprise the backbone or skeleton of the text, forming its basic structure; the backgrounded clauses put flesh on the skeleton, but are extraneous to its structural coherence’; (b) ‘the foregrounded clauses are ordered in a temporal sequence; a change in the order of any two of them signals a change in the order of real-world events. Backgrounded clauses, however, are not ordered with respect to each other, and may even be movable with respect to the foregrounded portions’ (ibid.: 281). Hopper and Thompson claim that foregrounding is correlated with high Transitivity, while backgrounding is correlated with low Transitivity. Nominalizations, thus, provide a support for this claim, being backgrounded and characterized by low Transitivity (they not only tend to have one argument, but this argument is typically non-referential, ibid.: 283). What motivation can be proposed for the existence of the POSS-ACC type and its rarity among ergative languages? A possible explanation is that POSS-ACC ANCs in nominative-accusative languages are relatively sentence-like not only in their retention of sentential marking for the P, but also in their retention of the sentential oppositions among the arguments (the similar treatment for the S and the A as opposed to the P). Thus, in nominative-accusative languages, POSS-ACC ANCs are motivated by Principle 1, while in ergative languages, there is no evident motivation for POSS-ACC ANCs. On the other hand, ERG-POSS ANCs in nominative-accusative languages are motivated by Principle 3, while those in ergative languages are motivated both by Principle 3 (a or b), and by Principle 1. If this explanation is correct, the scales of sentence-likeness for ANCs in nominative-accusative and ergative languages will be distinct. Note that the same tendency for ANCs to retain the distinction
< previous page
page_261
next page >
< previous page
page_262
next page >
Page 262 between the S and the A, as opposed to the P, is at work in the POSS-ADN type. However, in at least one ergative language in the sample, Agul, POSS-ACC ANCs are possible as an alternative to SENT ANCs. Thus, there is probably yet another principle motivating the existence of the POSS-ACC type, besides the one referred to in the preceding paragraph. A plausible hypothesis is that marking of arguments tends to reflect the degree of independence characterizing their referents. Thus, marking of As may be more sensitive to the character of clauses than marking of Ps, because the agents of subordinate clauses, and in particular, of ANCs, are always less independent than the As of independent clauses, while for the P this is hardly true. Whatever the explanation may be, the sample provides evidence for the special ‘sentential’ status of the A in contrast to the P (formulated in generalization (2), see section 12.0.). It is highly probable that As in nominative-accusative and ergative languages differ in this respect. The existence of the SENT and the SENT-POSS types (i.e., those which retain sentential marking for the A) poses a challenging problem: what distinguishes the A of independent clauses in SENT and SENTPOSS languages from those in all the other languages? Interestingly, two of the three SENT-POSS languages, i.e., those in which the P genitivizes while the A retains its sentential marking, are ergative (Tongan and Samoan). Tongan is also one of the two exceptional, V-initial complement-balancing languages with SENT ANCs. The other one, Classical Arabic, also has SENT-POSS ANCs. In other words, ANCs in Tongan and Classical Arabic follow both patterns in which the A retains its sentential marking, in both cases contradicting the general tendencies of the sample languages. I believe that there may be some common source in each of the two languages for both deviations. None of the principles referred to above offers explanation for the existence of the MIX type, which is special in two respects. Firstly, though not absolutely sentence-like, MIX ANCs are not at all NP-like. Secondly, they distinguish between all the three arguments, the S, the A and the P, though finite clauses distinguish only between subjects and objects. (For Thai this is true only with respect to word order.) The very provisional principles proposed above motivate the existence of different nominalization patterns across languages, that is, they predict that languages may resort to different solutions, as regards marking of the dependents in ANCs. However, what is even more interesting than formulation of the competing motivations
< previous page
page_262
next page >
< previous page
page_263
next page >
Page 263 themselves, is developing ‘a theoretical framework for describing and analyzing their interaction within specified contexts, and ultimately for predicting the resolution of their competition’ (DuBois 1985:344). I feel that there are at least two different aspects in this problem. Firstly, the motivations themselves may have different relative force. Thus, for example, Principle 1 seems to be superordinate to Principle 2, since ANCs in many complement-deranking languages are not at all NP-like (with respect to their internal structure). In other words, different choices of nominalization patterns across languages can be partly accounted for by the relations among the motivating principles themselves, whereas the latter sometimes reinforce each other, or exhibit significant differences in relative force. This is the internal source for resolution of competing motivations. However, there may be an external source for such resolution, namely, certain formal features of languages that opt out or reinforce motivations. These features are likely to include those discussed in chapter 11—word order, morphological marking type and structure of non-derived NPs. For example, the different structure of NPs in Icelandic, English and Russian restricts the choice of nominalization patterns by the languages in different ways (the POSSADN type is chosen by Icelandic, the ERG-POSS by Russian, the DBL-POSS and the ERG-POSS by English, putting aside more verbal, POSS-ACC gerunds in the latter). The head-marking character of Abkhaz may be responsible for its choice of the ERG-POSS and not of the SENT type, while ‘nonharmonic’ word order SOV/NG excludes POSS-ACC ANCs in Persian. The limited size of the sample makes it impossible to attribute a high degree of certainty to the principles and their interaction as they are described in this section. Still they are worth considering as a plausible framework for the discussion of ANCs, other types of complements and derived nouns. 12.4. INFLECTION VS. DERIVATION AND ANCs This study presents a typology of ANCs as if these were a homogeneous class of phenomena, constructions with nominals derived from verbs. Is it true that all ANs in the sample constitute a natural class or do they vary in their categorial status, from being verbal derivatives, nouns, to being inflectional verb forms? In this connection, two separate problems should be discussed: (1) are ANs across languages both derivational and inflectional verb forms? (2) if so,
< previous page
page_263
next page >
< previous page
page_264
next page >
Page 264 what is the connection between the categorial status of an AN and its nominalization pattern? Let us start with the first question and look at whether ANs can be inflectional verb forms. Anderson (1985:163) writes, ‘any process which involves a shift in word class between the basic and the derived forms (as for instance nominalization) could probably be called derivational’, even if such ‘derivational formations may have virtually complete productivity (as for instance the class of -ing nominalizations in English)’. As shown in the present study, ANs, though having acquired nominal inflectional features, may retain some verbal properties, and, thus, it is not quite obvious what is meant by ‘shifts in word class’. If the acquisition of nominal inflectional properties implies a shift in word class, ANs, according to Anderson, are always derivational. Bybee (1985:83) mentions two types of differences between derivational and inflectional categories— relevance (the extent to which ‘the meaning of the category directly affects the lexical content of the… stem’, (ibid.: 15)) and lexical generality. Thus, derivational processes normally produce larger meaning changes and are applicable in a more restricted semantic, syntactic and phonological domain than inflectional categories. Let us consider how the two parameters relate to ANs. Applying the relevance criterion to derivation, Bybee correlates derivations that change the syntactic category of a word with the amount of semantic change they effect. She claims that a category shift always implies meaning changes, since ‘the syntactic category of a word is an inherent part of its semantic representation’ (ibid.). Still, in some cases, such processes, for example, derivation of English -ing-gerunds, that describe the same situation as the ‘underlying’ verb, ‘add little further meaning, and thus border on inflection’ (ibid.). In connection with the generality principle, Bybee allows for the possibility ‘that the principle that morphology that changes the syntactic category of a word is always derivational is a false principle, and that -ly, gerundial -ing and comparable morphemes in other languages should be considered inflectional’ (ibid.: 85), because these morphemes, though leading to category changes, have full lexical generality. As Bybee shows, the relevance and generality parameters do not lead to any discrete division between derivation and inflection. Still, I do not understand why English gerundial -ing morphemes are viewed as derivational and what properties corresponding morphemes in other languages need in order to be considered inflectional. While Anderson’s and Bybee’s criteria hardly provide an answer for
< previous page
page_264
next page >
< previous page
page_265
next page >
Page 265 the question whether ANs exist as inflectional verb forms, Chomsky’s (1970) argumentation suggests that this is, in fact, so. Derived nominals, such as conquest, refusal, arrival, are lexically restricted and may have unpredictable semantic and syntactic properties as compared to the verbs they are derived from, not to mention idiosyncrasies in formation. On the other hand, they form NPs in exactly the same way as prototypical, non-derived nouns, which means that their syntactic properties need not be deduced by rules from verbs, but should follow from the representation of these words in the lexicon. By contrast, -ing-gerunds need not appear in the lexicon as separate lexical entries, but should be derived by transformations from the corresponding finite clauses. Though Chomsky does not use the word ‘inflectional’, this would be the natural term for those items which are derived by transformations. In this case, ANs in some languages are an inflectional category. For instance, to judge from the available sources, ANs in Tamil and Quechua satisfy all the criteria for transformationally derived nominals: they have full lexical generality, they do not show idiosyncrasies in meaning or syntax as compared to the verbs they are derived from, and they build ANCs according to the SENT pattern, which, thus, differ considerably from non-derived NPs. Let us now consider the second question, whether there are any connections between the categorial status of an AN and its nominalization pattern. Thus, for instance, it could very well be that the nominalization patterns were neatly divided between derivational and inflectional forms, which would invalidate the whole typology of ANCs as it is presented here. To use a metaphor, it may turn out that I am comparing apples and pears. However, it appears that Chomsky’s criteria may come into conflict with each other. Thus, in Russian and Hebrew, derivation of ANs is, in many cases, connected to unpredictable syntactic and semantic changes, but they have syntactic properties different from nonderived NPs. For instance, Russian ERG-POSS ANCs involve dependents in the instrumental, while Hebrew POSS-ACC ANCs involve dependents with the object marker et, both excluded from non-derived NPs. It is somewhat more difficult to find ANs derived from verbs in a regular way (with full lexical generality and without any semantic and syntactic idiosyncrasies), but still showing nominal syntactic properties. Swedish verbal nouns with the suffix -ande and its variants, and English nominal gerunds are plausible candidates. If we return to the scales of sentence-like vs. NP-like ANCs (Figures 12.2–12.4), their extreme points show a high degree of correlation between Chomsky’s criteria, namely, SENT ANCs
< previous page
page_265
next page >
< previous page
page_266
next page >
Page 266 (sentence-like) typically involve regular verb formations (though Classical Arabic, at least, is an exception), while NOMN ANCs (NP-like) typically involve clearly derived nominals. Still, the other types involve both regular and irregular forms. The above mentioned facts confirm Bybee’s (1985:87) claim that ‘the distinction between derivational and inflectional morphology is not discrete, but rather a gradient phenomenon’. ANCs are heterogeneous with respect to their categorial status, but the same typology is valid for all of them. In other words, apples sometimes grow on pear trees, while pears sometimes grow on apple trees. This result suggests that it is possible to develop a common formalism which would offer a unified description of ANCs across languages. 12.5. WHY AN ACTION NOMINAL CONSTRUCTION? In chapter 1, discussing semantics of ANCs in different languages, the emphasis was primarily on those cases where such constructions are the only available means of expression. Something should be said here about situations in which ANCs are chosen among several competing expressions. ANCs may be used for at least two opposing purposes—to compress texts and to elaborate them. These cases will be considered separately below. Let us start with text compressing uses of ANCs. As early as 1924, Jespersen suggested that ‘abstract nouns’ (including action nominals) help avoid many clumsy expressions, where finite subordinate clauses would otherwise be necessary, as in His display of anger was equivalent to an admission of belief in the other’s boasted power of divination (Jespersen 1924:136). This is especially important when verbal nouns have attributes, as seen in the following comparison: The Doctor’s extremely quick arrival and uncommonly careful examination of the patient brought about her very speedy recovery vs. The Doctor arrived extremely quickly and examined the patient uncommonly carefully; she recovered very speedily (ibid.: 137). Note that the extent to which similarly clumsy constructions may arise is highly correlated to specific communication needs, which ultimately leads to uneven distribution of nominalizations among different styles of speech. Thus, the complexity of scientific discourse, reflecting the complexity of interrelated factors in scientific thought, favours nominalizations and other means of maximizing the amount of information in texts of comparable length. However, in other cases, this
< previous page
page_266
next page >
< previous page
page_267
next page >
Page 267 strategy is far from being satisfactory. To cite Jespersen, when we express by means of nouns what is generally expressed by finite verbs, our language becomes not only more abstract, but more abstruse, owing among other things to the fact that in the verbal substantive some of the life-giving elements of the verb (time, mood, person) disappear. While the nominal style may therefore serve the purposes of philosophy, where, however, it now and then does nothing but disguise simple thought in the garb of profound wisdom, it does not lend itself so well to the purposes of everyday life. (Jespersen 1924:139) Which constructions appear as relatively more awkward, finite subordinate clauses or ANCs, is to a certain extent a language specific question and depends primarily on the morphological complexity of their heads. For Amharic, for example, Kapeliuk (1980:100) reports on studies of schoolchildren’s talk, showing that ‘infinitives’ (action nominals) belong to the simplest verb forms. This is invoked by her as one of the reasons for the gradual replacement of finite subordinate clauses by ANCs (cf. section 11.2.), along with other factors, such as the ease with which ANCs may be embedded into matrix clauses. The choice between finite subordinate clauses and ANCs is largely determined by requirements on their discourse function in a particular context. Thus, as discussed by Noonan (1985) and Hopper and Thompson (1980; see also section 12.3. in this book), nominalizations are, by their nature, discourse dependent and backgrounded. However, this feature is, to some extent, language specific, as has been shown by Nichols (1988) on the basis of Russian scientific texts and their translation equivalents in English. She states that Russian verbal nouns, exactly like English verbal nouns, cannot assert, but, in contrast to the latter, they are often used in contexts where English would prefer assertion and/or a finite verb. To illustrate the point, let me give one of the Russian examples quoted by Nichols with its literal translation and a free translation.
< previous page
page_267
next page >
< previous page
page_268
next page >
Page 268 (12.1) Ob”jasnenie fonemnyx sootvetstvij meždu explanation of.phonemic correspondences between jazykami obščnost’ju ix proisxoždenija iz languages by.commonality of.their descent from opredelennoj isxodnoj jazykovoj sistemy predpolagaet particular source linguistic system assumes neobxodimost’ rekonstrukcii ee s cel’ju necessity of.reconstruction of.it with goal izučenija vozniknovenija i putej preobrazovanija of.study of.rise and paths of.transformation istoričeski zasvidetel’stvovannyx rodstvennyx of.historically attested related jazykovyx sistem. linguistic systems Literally: ‘(An) explanation of phoneme correspondences between languages by means of (the) commonality of their descent from an original linguistic system assumes (the) necessity of (a) reconstruction of it (the original system) with the goal of (a) study of the rise and paths of transformation of historically attested cognate linguistic systems.’ Free translation: ‘If phonemic correspondences between lan guages are to be explained as due to a common descent from an original system, then that system must be reconstructed in order that the origin of the attested cognate languages and the changes undergone by them may be studied.’ (Nichols 1988:403) As Nichols points out, besides stylistic faults, the literal translation contains one major error of content: it fails to get across the main point of the sentence, which is that reconstructions must be made. Even if we streamline the original sentence, no English version which contains nominalized necessity or a synonym of it communicates that point unambiguously. (ibid.) Analysis of cases such as example (12.1.) has led Nichols to conclude that ‘the choice of assertion vs. non-assertion, and the choice of finite vs. nominalized verb form, is at least partly based on languagespecific inventories of grammatical categories’ (ibid.: 400), and partly on differences in the relations among reader, writer, text and knowledge, underlying texts in different languages (and styles). Nichols’ main claim is that differences in the use of verbal nouns in
< previous page
page_268
next page >
< previous page
page_269
next page >
Page 269 English and Russian are due to the English-Russian differences in definiteness, new participants, and information increment. Thus, English has a grammatical category of definiteness, which is lacking in Russian. English resorts typically to special presentative constructions, with motion and stance verbs, for introducing new participants, while indefinite articles cannot ‘gracefully’ introduce new participants. In Russian, new participants may be introduced into the discourse as arguments in an ordinary clause. Although English and Russian prefer to introduce no more than one chunk of new information per clause, only in Russian can a clause introduce a new issue and make a new point about it at the same time. These differences have an important bearing on the use of verbal nouns in both languages. To quote Nichols, in Russian, the absence of English new-participant constraint means that a verbal noun carries no secondary presupposition of prior mention in the text. The absence of definiteness in Russian means that a nominalization carries no presupposition about the hearer’s knowledge or familiarity with the text. It can be used to name events, etc. previously asserted, and it can be used with equal ease to introduce newly mentioned topics. Importantly, it can be used to make statements which are the writer’s original claims, presented for the first time. Since claims articulated in this fashion are not asserted, the effect of using a verbal noun for new claims is to blur the distinction between writer’s opinion, analysis, findings, etc. and general, received knowledge of truth. (Nichols 1988:405) On the basis of such considerations, Nichols concludes that Russian scientific texts are gnomic rather than personal, they state the ‘knowledge available to the scientific tradition’ and characterized by other depersonalizing properties. Let us turn now to text elaborating uses of ANCs. Many languages have a special type of construction which consists of a semantically empty or half-empty verb and a verbal noun, the whole construction meaning almost exactly the same as the verb which underlies the verbal noun, like to inflict a defeat (cf. to defeat), to do damage (cf. to damage) etc. (For the syntactically different types of such combinations see Mel’čuk and Žolkovskij 1970.) These constructions seem to be quite superfluous and have often been criticized by stylists. However, they have positive features and may even be preferred in some situations. As Platzack (1977:14) has shown in experiments with messages in Swedish, sentences with such ‘clumsy’ and ‘superfluous’
< previous page
page_269
next page >
< previous page
page_270
next page >
Page 270 analytical constructions are much easier to remember than the corresponding non-elaborated sentences. This is due to the fact that the former are much more redundant, which makes them much easier to remember and makes it possible for the hearer to predict the structure and the meaning of the sentence at an early stage of processing. Gak (1976) has demonstrated some advantages of similar analytical constructions in Russian. Thus, they cover many more aspect oppositions, due to the combination of the aspect oppositions signalled by the finite verb and to the number distinctions signalled by the verbal noun. Of further importance is the greater possibility of argument reduction in constructions with ANs (cf. section 1.4.2.) as compared to non-elaborated sentences. Thus, analytical constructions allow subjects and objects to be deleted in situations where these are generic or indefinite, while at the same time retaining the typical bipartitional sentence structure. ANs may have still other language specific functions. In Semitic languages, for example, a combination of a finite verb together with an action nominal derived from the same verb is used with emphatic functions. In Tamil, Modern Mongolian and Selkup, negative sentences involve ANs. In Turkish and Quechua, relativization is based on ANCs (see section 2.2.5.). In Welsh and Hausa, ANs are exploited to express certain tenses and aspects, while in Kwa languages, they appear in sentences corresponding to English cleft-sentences. In Amharic, ANs combined with copula verbs are gradually replacing finite verbs. Finally, let us consider the interesting asymmetry between nominalization and verbalization noted by Hopper and Thompson (1984:745): ‘if a language has category-deriving morphology at all, what we find is that it has noun-deriving, but not verb-deriving processes’. The explanation, according to Hopper and Thompson, lies in the fact that a nominalization qualifies as a metaphor: The metaphorical process takes something abstract and treats it as if it were concrete precisely because human cognition can deal with concrete entities more easily than with abstraction; this process is thus unidirectional. That is why languages have nominalization processes, where an abstract event is treated as a concrete entity; this results in a form which is both an event and an entity, and tends to be marked with signals of this fact. But there is no analogous ‘verbalization’ process, since we do not metaphorically treat concrete things as events. (ibid.: 746)
< previous page
page_270
next page >
< previous page
page_271
next page >
Page 271 Summary The purpose of this study has been to arrive at a typology of action nominal constructions which would answer the question: what are the means of signalling the relations between an action nominal and its dependents (subjects and objects) as compared to the relationships in corresponding finite clauses and in non-derived NPs? The resulting classification of ANCs in seventy languages, based on dependent marking as the main parameter, consists of four major and three, or possibly four, minor (less frequent) types. These types may be ordered in accordance with the degree to which they are sentence-like and/or NP-like. The fact that no language has syntactic means which are exclusively used in ANCs is an interesting finding and gives support to the view that constructions which appear later in language acquisition and language development are modelled on already existing constructions as far as possible. Thus, ANCs are syntactically modelled on finite clauses and/or NPs, those constructions with which they also share semantic and morphological properties. I have suggested several competing principles motivating the existence of different types of ANCs. Further research is necessary on the modelling hypothesis in order to elucidate the question of what constructions may ‘legitimately’ be modelled as well as what type of constructions may be subject to modelling. The great variation in the internal syntax of ANCs is severely restricted in at least two respects. First, word order, head-marking and dependent-marking differ in their range of allowed variation. Second, different arguments differ as to their proneness to retain sentential marking. Together these restrictions may be generalized as a hierarchy predicting which combinations of nominal and sentential properties are possible. The choice of nominalization patterns in a language is hardly accidental. The following factors may be suggested as determining or
< previous page
page_271
next page >
< previous page
page_272
next page >
Page 272 restricting possible choice of nominalization patterns in a language: basic word order, the property of being complement-balancing vs. complement-deranking, the structure of non-derived NPs, accusativity/ergativity and morphological marking type. Finally, the facts discussed in the present study invite further research on at least the following issues: 1 word classes across languages—which syntactic properties tend to be clustered and are, thus, implied by the morphological category of a word? When is it possible to talk of category shift? 2 strategies of balancing and deranking—where are they manifested in languages? 3 affinities between non-derived NPs and clauses—semantic, syntactic and morphological; 4 complement systems across languages—which formal complement types tend to coexist vs. exclude each other? What are the connections between the semantics and syntax of the different complement types?
< previous page
page_272
next page >
< previous page
page_273
next page >
Page 273 Appendix A Alphabetical language index The numbers in the parentheses after nominalization types refer to the sections which are devoted to the problems with type determination or to deviations from the standard patterns. For the examples see Appendix B. (For any abbreviations or symbols not explained in Abbreviations see key to Table 5.1 on p. 106.)
< previous page
page_273
next page >
< previous page Page 274 Language ABKHAZ
ARABIC, CLASSICAL
Affiliation Caucasian: Northwest Caucasian: Northeast: Daghestanian Papuan: Gum Afro-Asiatic: Semitic Afro-Asiatic: Semitic
ARABIC, EGYPTIAN
Afro-Asiatic: Semitic
ARCHI
Caucasian: Northeast: Daghestanian Isolate
AGUL AMELE AMHARIC
BASQUE
page_274 ANC Type(s) ERG-POSS: OBLPOSS SENT, POSS-ACC POSS-ACC POSS-ACC
next page >
Source Hewitt 1979, 1987 Boguslavskaja 1989 Roberts 1987 Amsalu Aklilu personal communication, Cowley et al. 1976, Gasser 1983, Kapeliuk 1980 Wright 1971
SENT, POSS-ACC, ERG-POSS: SENTPOSS POSS-ACC (6.2.), Ahmed Elgendi personal communication, Abdel-Massih NOMN: POSS-ADN et al. 1979, Gary and Gamal-Eldin 1982, Mitchell 1962, Wise 1975 SENT Kibrik 1977
Oskar Aranzabal personal communication, N’Diaye 1970, Saltarelli 1988 BULGARIAN Indo-European: ERG-POSS: OBLMila Stoyčeva personal communication, Ivan Slavonic: South POSS, MIX, NOMN: Derzhanski personal communication, Revzin 1973, *DBL-POSS Stojanov 1964, Topolińska 1981 BURUSHASKIIsolate SENT (7.2., 10.3.) Lorimer 1935 BURYAT Altaic: Mongol POSS-ACC (6.2.) Skribnik 1983, 1988, Ubrjatova and Litvin (eds) 1984, 1986 CHUKCHEE Paleosiberian: ARG-RED (9.4.) Vladimir Nedjalkov personal communication, Nedjalkov (CHUKCHI) Chukotkoet al. 1983, Skorik 1961 Kamchatkan CZECH (see Indo-European: Comrie 1976, Comrie and Thompson 1985 Russian) Slavonic: West
< previous page
SENT (10.3.)
page_274
next page >
< previous page
page_275
Page 275 DUTCH IndoERG-POSS: OBL-POSS, ?MIX, European: *NOMN: DBL-POSS Germanic: West EGYPTIAN,Afro-Asiatic: ERG-POSS: OBL-POSS, POSSANCIENT Egyptian ACCpr, MIXpr ENGLISH IndoPOSS-ACC, ERG-POSS: OBLEuropean: POSS, NOMN: DBL-POSS Germanic: West ESTONIAN Uralic: Finno- ERG-POSS: OBL-POSS, Ugric: Baltic- NOMN: DBL-POSS, INC: Finnic POSS-INC, INC: OBL-INC EVENKI Altaic: POSS-ACC (6.2.) Tungus: Northwestern EWE Niger-Congo: POSS-INC FINNISH Kwa NOMN: DBL-POSS Uralic: FinnoUgric: BalticFinnic FRENCH IndoERG-POSS: OBL-POSS, European: NOMN: DBL-POSSpr (10.3.) Italic FULA Niger-Congo: POSS-ACC West-Atlantic GBADI Niger-Congo: NOMN: DBL-POSS Kwa: Kru GEORGIANCaucasian: ERG-POSS: OBL-POSS, ? Kartvelian NOMN: DBL-POSS GERMAN IndoERG-POSS: OBL-POSS, European: *NOMN: DBL-POSS (10.3.) Germanic: West
< previous page
next page >
Kees Hengeveld personal communication, Dik 1985, Hoekstra 1986 Gardiner 1957 Chomsky 1970, Lees 1960, 1966, Vendler 1967, 1968, 1970, Verhaar 1990
Diana Krull personal communication, Haman 1962 Vladimir Nedjalkov personal communication, I.V. Nedjalkov 1983, 1985 Westermann 1930 Kristiina Jokinen personal communication, Dubrovina 1972, Hakulinen and Karlsson 1978, Jokinen 1991, Karlsson 1979 Danielle Cyr personal communication, Bartning 1986, Giorgi and Longobardi 1988, Tutescu 1972 Arnott 1970, Labatut 1982 Koopman 1984 Harris 1981, 1984, Hewitt 1983, Vamling 1989 Esau 1973, Giorgi and Longobardi 1988, Helbig and Buscha 1981
page_275
next page >
< previous page Page 276 Language GREEK ANCIENT GREEK, MODERN HAUSA HEBREW, MODERN
Affiliation Indo-European: Greek Indo-European: Greek Afro-Asiatic: Chadic Afro-Asiatic: Semitic
page_276 ANC Type(s) NOMN: DBL-POSS
next page >
Source Schwyzer 1950, Smyth 1956
ERG-POSS: OBL-POSS, ? Irene Fotopoulou personal communication, NOMN: DBL-POSS Joseph and Philippaki-Warburton 1987, Mackridge 1985 REL Newman 1987, Ščeglov 1970
Sven Malkior personal communication, David Gil personal communication, Berman 1976, 1978, Ritter n.d. HIXKARYANACarib Derbyshire 1979 HUNGARIAN Uralic: FinnoFerenc Kiefer personal communication, Ugric: Ugric Majtinskaja 1960 ICELANDIC Indo-European: NOMN: POSS-ADN, INC: Kjartan Ottósson personal communication Germanic: North POSS-INC IFÈ Niger-Congo: POSS-ACC Hélène Boëthius personal communication Kwa INGUSH Caucasian: SENT Nichols forthcoming Northeast: Vejnax INUKTITUT Eskimo-Aleut: Juanasie Koperkualuk and Danielle Cyr personal (see West Inuit communication Greenlandic) IRISH, OLD Indo-European: ERG-POSS: OBL-POSS Disterheft 1980, Thurneysen 1946, Jeffers 1976 ITALIAN Celtic (10.3.) Comrie and Thompson 1985, Giorgi and (see Indo-European: Longobardi 1988 Rumanian) Italic JAPANESE Altaic NOMN: DBL-POSS Yukiyo Wallenström and Bengt Ek personal communication, Hinds 1986, Martin 1975 KOBON Papuan: East SENT Davies 1981 New Guinea Highlands
< previous page
POSS-ACC, ERG-POSS: OBL-POSS, ?NOMN: DBL-POSS ERG-POSS: OBL-POSS INC: POSS-INC, REL
page_276
next page >
< previous page Page 277 KOREAN
Altaic (isolated)
KURDISH
page_277 SENT
Kim 1987, Ramstedt 1968, Martin and Lee 1969, Vasil’ev 1961, 1963, Hang-Seok Choi personal communication, Lev Koncevič personal communication ERG-POSS: OBL- Farhad Shakely personal communication, Bakaev 1973, POSS, INC:POSS- Kurdoev 1978 INC NOMN: DBLWoodcock 1959, Groot 1957 POSS, (?NOMN: POSS-ADN 8.2.2.) NOMN: DBLLaine Lasmanis personal communication, Lasmane 1981 POSS
IndoEuropean: Iranian LATIN IndoEuropean: Italic LATVIAN IndoEuropean: Baltic LUGANDA Niger-Congo: (see Swahili) Bantu MACEDONIAN Indo(see Bulgarian) European: Slavonic: South MAORI Austronesian:POSS-ACC, PAS, Polynesian: ERG-POSS: OBLEast POSS, MIX MONGOLIAN, Altaic: SENT (5.5.) CLASSICAL Mongol MONGOLIAN, Altaic: POSS-ACC (6.5.) MODERN Mongol (=KHALKHA) NIVKH=GILYAKIsolate SENT (5.2.) PANJABI PERSIAN
IndoEuropean: Indic IndoEuropean: Iranian
< previous page
next page >
Debora Mulindwa personal communication, Ashton et al. 1954 Topolińska 1981
Chung 1973, 1978 Grønbech and Krueger 1955 Binnick 1979, Beffa and Hamayon 1975, Jan-Olof Svantesson personal communication
Vladimir Nedjalkov and Otaina personal communication, Krejnovič 1979a ERG-POSS: OBL- Ravinder Singh personal communication, Gill and POSS Gleason 1963 ERG-POSS: OBL- Hassan Djamshidpey personal communication, AminPOSS, INC:POSS- Madani and Lutz 1972 INC
page_277
next page >
< previous page
page_278
next page >
Page 278 Language POLISH
Affiliation ANC Type(s) Source IndoERG-POSS: OBL-POSS, ? Comrie and Thompson 1985, Topolińska 1981 European: POSS-ACC, *NOMN: DBLSlavonic: POSS West PUKAPUKAN Austronesian: POSS-ACC, ERG-POSS: Chung 1973, 1978 Polynesian: OBL-POSS West QUECHUA, Andean SENT, INC: SENT-INC, Lefebvre and Muysken 1988, Muysken 1989 CUZCO INC: POSS-INC (5.5.) QUECHUA, Andean SENT Weber 1983 HUALLAGA (HUÁNUCO) QUECHUA, Andean SENT, INC: POSS-INC Cole 1982 IMBABURA RUMANIAN IndoERG-POSS: OBL-POSS Mallinson 1986 European: Italic RUSSIAN IndoERG-POSS: OBL-POSS, Padučeva 1974, 1982, 1984, Revzin 1973, Veyrenc European: *NOMN: DBL-POSS 1972, Gak 1976 Slavonic: East SAMOAN Austronesian: ERG-POSS: SENT-POSS, Chung 1973, 1978, Vonen 1988 Polynesian: NOMN: DBL-POSSpr West SELKUP Uralic: POSS-ACC Kuznecova et al. 1980, Tereščenko 1973 Samoyedic SWAHILI Niger-Congo: POSS-ACC Abdulaziz Y.Lodhi personal communication, Sune Bantu Magnusson personal communication SWEDISH IndoNOMN: POSS-ADN, INC: Martin Tamm personal communication, Noreen European: POSS-INC, (?ERG-POSS: 1904, Pitkänen 1979, Wellander 1973, Svenska Germanic: OBL-POSS, 10.3.) akademiens ordbok 1898, Platzack 1977 North TABASSARANCaucasian: SENT (5.3.) Bergel’son et al. 1982, Bogatyrev and Northeast: Boguslavskaja 1982, Kibrik 1985 Daghestanian
< previous page
page_278
next page >
< previous page Page 279 TAGALOG TAHITIAN TAMIL THAI TOKELAUAN (see SAMOAN)
Austronesian: Northwest Austronesian: Polynesian: East Dravidian Kam-Tai
TONGAN
Austronesian: Polynesian: West Austronesian: Polynesian: West
TURKISH
Altaic: Turkic
TUVA
Altaic: Turkic
VATA
page_279 POSS-ACC (6.2.) POSS-ACC (10.2.)
Schachter and Otanes 1972 Chung 1973, 1978
SENT Asher 1982 POSS-ACC, MIX Anthony Diller and Preecha Juntunamalaga personal communication, Hudak 1987 Vonen 1988 SENT, POSS- Chung 1973, 1978 ACCpr, ERGPOSS: SENT-POSSpr POSS-ACC, PAS Lewis 1967, Nilsson 1985, Underhill 1976, Kari Fraurud personal communication, Comrie and Thompson 1985 !SENT(5.2., Čeremisina 1981a, Ubrjatova and Litvin 1984, 1986, 10.1.), POSS- Bergel’son and Kibrik 1987 ACC POSS-ACC Koopman 1984
Niger-Congo: Kwa: Kru WELSH Indo-European: ERG-POSS: Celtic OBL-POSS W.GREENLANDICEskimo-Aleut: ANTI-PAS, PAS Inuit WIKCHAMNI Penutian: POSS-ACC Yokuts YUKAGIR Isolate POSS-ACC
< previous page
next page >
Awberry 1976, Comrie and Thompson 1985 Fortescue 1984 Gamble 1978 Elena Maslova personal communication, Maslova 1989
page_279
next page >
< previous page
page_280
next page >
Page 280 Appendix B ANC types with examples For the definition of the nominalization types see section 3.3. 1. SENTENTIAL (SENT) AGUL: example (10.2a) in section 10.1. ARABIC, CLASSICAL: presumably only for transitive ANs. (1) min [ zayd-un wonder:PRF.1SG of beat:AN-GEN Zayd-NOM ‘amr-an] ‘Amr-ACC or: min [ ‘amr-an wonder:PRF.1SG of beat:AN-GEN ‘Amr-ACC zayd-un] Zayd-NOM ‘I wonder at Zèid’s beating ‘Amr.’ (Wright 1971: part 3:59) ARCHI: examples (5.9) and (5.10) in section 5.1. BASQUE: examples (10.8) in section 10.3. BURUSHASKI: for intransitive ANCs see examples in section 10.3, p. 204. (2) [ hir ] Kiser-ERG/GEN man:ABS bring:AN:ABS seeing ‘seeing Kiser bringing a man’ (Lorimer 1935:351)
< previous page
page_280
next page >
< previous page
page_281
next page >
Page 281 INGUSH (3a) a:z diexar deadar cung, [yz xa:nnah I:ERG request made he:ALL he:ABS on.time ha-va:r] here-come:AN ‘I asked him to come on time.’ (3b) cunna dieza [cuo ši: nanna he:DAT likes/wants she:ERG own:GEN mother:DAT novq’uostal dar] help:ABS do:AN.ABS ‘He wants her to help her/his mother.’ (Nichols forthcoming) KOBON (4a) [Yad wög g-ep] iru g-öp. 1SUBJ work quickly do-AN much do-PERF.3SG ‘I am tired of working quickly.’ (4b) [Yad wög kub g-ep] iru g-öp. 1SUBJ garden big do-AN much do-PERF.3SG ‘I am not interested in making a big garden.’ (Davies 1981) KOREAN: examples (5.5–5.6) in section 5.1. MONGOLIAN, CLASSICAL (5a) [ta ayu-qu] kereg ügei. you:NOM fear-AN matter without ‘Your fearing (is) without reason.’ (Grønbech and Krueger 1955:32) (5b) ünege ber keme-küi-dür fox:NOM TOP what be:PRES say-AN-LOC ‘When the fox said ‘‘What is that?”’ (ibid.: 33) (5c) bükün-i bi küliye-kü bolbacu every thing-ACC I: NOM endure-AN in spite ‘although I am one who has endured everything’ (‘in spite of me enduring everything’) (ibid.: 38) NIVKH: examples (5.15–5.16) in section 5.2.
< previous page
page_281
next page >
< previous page
page_282
next page >
Page 282 QUECHUA, CUZCO (6a) [ hamu-na-n-ta] yacha-ni. Juan-NOM come-AN.UNREAL-3-ACC know-1SG ‘I know that Juan is to come.’ (Lefebvre and Muysken 1988:17, ex. 7) (6b) [ papa-ta mikhu-sqa-n-ta] yacha-ni. Juan-NOM potato-ACC eat-AN.REAL-3-ACC know-1SG ‘I know that Juan eats potatoes.’ (ibid.: 21, ex. 20a) QUECHUA, HUALLAGA (HUÁNUCO) (7a) Mana musyashkaachu [ not know:PAST:NEG:1SG many-NOM chayamu-sha-n-ta] arrive-AN.PAST-3-ACC ‘I did not know that so many arrived.’ (7b) Chawraqa aymismo rigikushkaa noqa then right.there believe:PAST:1SG I [ willa-ma-sha-n-ta] brothers-NOM tell-1OB-AN.PAST-3-ACC ‘Then right there I believed what the brothers told me.’ (7c) [Juan-ta maqa-sha-n-ta] musya Juan-ACC hit-AN.PAST-3-ACC know:1SG ‘I know that he hit John.’ (Weber 1983) QUECHUA, IMBABURA: example (5.18a) in section 5.5. (8) ñuka-ka [Juan kay-pi ka-shka-ta] ya-ni I-TOP Juan this-in be-AN.PAST-ACC think-1SG ‘I think that Juan was here.’ (Cole 1982:33, ex. 103a) TABASSARAN: example (5.17) in section 5.3. (9) izu-s [bali č hi ] I-DAT boy:ERG sister:NOM wash…CLI…wash-AN.ABS lāzim vu -is need-1SG.IO ‘I need the boy to wash his sister.’ (Bergel’son et al. 1982:58)
< previous page
page_282
next page >
< previous page
page_283
next page >
Page 283 TAMIL (10a) siita kaalayile va-nt-atu Sita morning:LOC come-PAST-AN ‘Sita’s coming in the morning.’ (Asher 1982:25, ex. 107c) (10b) [coomu tampikki ] Somu younger.brother:DAT money give:PAST-AN aaccariyamaa iruntatu surprising be:PAST:3SG.N ‘It was surprising that Somu gave younger brother the money’. (ibid.: 110, ex. 516b) TONGAN: only for transitive ANs with non-pronominal As. (11) ’oku ’ikai tonu [’a e taki ’e he tu’i ’a PRES NEG good ABS ART lead:AN ERG ART king ABS e fonua] land ERG ART land ‘The king’s leading of the country is not good.’ (Chung 1973:655) ?TUVA: examples (10.3c–10.3d) in section 10.1. 2. POSSESSIVE-ACCUSATIVE (POSS-ACC) Pure AGUL: example (10.2b) in section 10.1. AMELE (12) [Uqa na cabi od-oc] eu fil. 3SG of work do-AN that different ‘His way of working [lit. ‘doing work’] is different.’ (Roberts 1987:47, ex. 171) AMHARIC (13a) [yä-pitär zäfän azzäfafän] GEN-Peter love-DEF-ACC song sing:AN(MN) asdännäqat mother-POSS.3SG-ACC astonish:PERF:3SG.M:3SG.F ‘Peter’s singing of [i.e., ‘his way of singing’] the love song pleased his mother.’
< previous page
page_283
next page >
< previous page
page_284
next page >
Page 284 (13b) ya-pitär bet-u-n mä-srat GEN-Peter house-DEF-ACC AN-build ‘Peter’s building the house’ (Amsalu Aklilu personal communication) ARABIC, CLASSICAL (a) with lexical dependents (14a) kāna [qatl-u l-xalīfat-i ] was kill:AN-NOM.CONSTR the-chalif-GEN fi s-sanat-i in this the-year-GEN ‘In this year the chalif put to death.’ (lit. ‘the chalif’s killing was in this year’ (b) with pronominal dependents (14b) ‘allama-nī love:AN-1SG.POSS-OB.3SG teach:PRF:3SG.M-OB.1SG t-tanassuk-a the-religiousness-ACC ‘My love of him taught me to be religious.’ (Wright 1971: part 3:58) ARABIC, EGYPTIAN (15a) sahar-u ta‘ab-u. stay.up.late:AN-3SG.POSS tired:3SG.N-3SG.OBJ ‘His staying up late tired him.’ (Gary and Gamal-Eldin 1982:17, ex. 103) (15b) [ ] fikra refuse:AN-3SG.POSS the-job NEG:was:NEG idea:F kuwajjisa good:F ‘His refusing the job was not a good idea.’ (ibid.: 63, ex. 442) (15c) [‘amal muna fustaan li nafs-aha] make-AN Mona the dress for self-her pleased:3SG.M mother-3SG.POSS ‘Mona’s making herself a dress pleased her mother.’ (ibid.: 44, ex. 310)
< previous page
page_284
next page >
< previous page
page_285
next page >
Page 285 BURYAT (16a) Bišixan Darima [minii little Darima I:GEN ere-xe-güj-höö-mni] ajna. come-AN.FUT-NEG-ABL-1SG.POSS is.afraid ‘Little Darima is afraid that I will not come.’ (Skribnik 1988:16) (16b) [Baatar-aj tex-ee Baatar-GEN goat-RFL.POSS xütel-xe-de-ni], tere-ni lead-FUT.AN-LOC-3SG.POSS that-3SG.POSS durata duragüj-göör xojnohoo-ni xülöö tatana willing non-willing after-3SG.POSS drag:PRES ‘When Baatar leads his goat, that one drags along after him willy-nilly.’ (ibid.: 22) EGYPTIAN, ANCIENT: with pronominal Ps, see the examples in section 10.2, p. 202. ENGLISH: see example (1.3) on p. 5. EVENKI (17a) Bi ba-de-m [mu ] I want-NEG-1SG water run.out-AN-ACC-3SG.POSS ‘I do not want water to run out.’ (17b) Nuŋartyn [ŋinakin koška-va-s they dog:NONM cat-2SG.POSS-ACC asa-djari-va-n] chase-AN-ACC-3SG.POSS look-IMPRF-3PL ‘They looked at the dog’s chasing our cat.’ (Vladimir Nedjalkov personal communication) FULA: no examples with both the A and the P present. (18a) wart-u-ki Bello/maako return-ACT-AN Bello/his ‘Bello’s/his return’ (cf. Bello wartii/‘o-wartii—‘Bello/he returned.’) (18b) nodd-u-ki Bello/mo call-ACT-AN Bello/him ‘to call Bello/him’ (cf. ’o-noddii Bello/’o-noddii-mo—‘He called Bello/him.’) (Arnott 1970:378)
< previous page
page_285
next page >
< previous page
page_286
next page >
Page 286 HEBREW, MODERN: examples (7.39a–7.39c) in section 7.2.2. IFÈ: no examples with both the A and the P present. (19a) ìlúu town:of Atakpame create:AN ‘the creation of the town of Atakpame’ (19b) Yèésù ti gba [ ŋa] Jesus already take beat:AN-your PL ‘Jesus has already been beaten by you.’ (Hélène Boëthius personal communication) MAORI: examples (7.28–7.29) in section 7.2.2. MONGOLIAN, MODERN: no examples with both the A and the P present (note the accusative case on the S in ex. 20b). (20a) [Bat-yn türgeb sajn bol-x-yg] bid Bata-GEN soon well become-AN-ACC we think-IMPF baj-na be-PRES ‘We think Bata will be well soon.’ (Binnick 1979:87, ex. 1) (20b) [Aav-yg xödöö jav-sn-yg] bi med-sen father-ACC country go-AN.PRF-ACC I know-AN.PRF ‘I knew father went to the country.’ (ibid.) (20c) Bagš bidn-ees [xičeel-ee sajn teacher we-ABL lesson-RFL.POSS well davt-san] uu? asuu-v prepare-AN.PRF Q COMPL ask-PAST ‘Teacher asked us if we had done our lesson well.’ (ibid.: 88) (The A of the ANC is deleted under coreferentiality with the indirect object of the matrix clause.) ?POLISH: no examples with both the A and the P present. (21) Proszę o [zwoln-ienie z request:1SG about dismiss-AN.LOC from pracy dwie pracownice]. work:GEN two:ACC woman.worker:ACC.PL ‘I request the dismissal of the two women workers.’ (Comrie 1976:191, ex. 46)
< previous page
page_286
next page >
< previous page
page_287
next page >
Page 287 PUKAPUKAN: for intransitive ANs see examples (10.13–10.14) in section 10.4. (22) na takayala [t-a-na waiwai-nga PAST be.mistaken DEF-DOM.POSS-3SG break-AN i na mea o te motu] ACC PL thing SUB.POSS ART motu ‘Mistaken were his breakings of the tapus of the motu.’ (Chung 1973:659, ex. 40) SELKUP: examples (6.9–6.10) in section 6.1. SWAHILI (23a) kw-enda kw-a Peter AN.CL15-go CL15-of Peter ‘Peter’s going’ (23b) ku-soma kw-a Peter kitabu AN.CL15-read CL15-of Peter CL7.book ch-a-ke mwenyewe CL7-of-he own ‘Peter’s reading of his own book’ (23c) [Ku-u-imba kw-a Peter u-le w-imbo AN.CL15-CL7-sing CL15-of Peter CL7-that CL7-song w-a mapenzi] ku-li-mshangaza mama CL7-of love CL15-PAST-astonish mother.CL1 y-a-ke. CL1-of-he ‘Peter’s singing of the love song astonished his mother.’ (Abdulaziz Y.Lodhi personal communication) TAGALOG (24a) Masyado-ng mabilis [ang pag-tugtog niya excessive-LNK fast TOP AN-play GEN/OB:he ng piesa]. OB piece ‘His playing of the piece is too fast.’ (Schachter and Otanes 1972:163)
< previous page
page_287
next page >
< previous page
page_288
next page >
Page 288 (24b) Na-tuwa ang lahat [sa pag-ka-kasagot TNS-be.amused TOP all DIR AN-PRF-answer ni Bob sa titser]. GEN/OB Bob DIR teacher ‘Everyone was amused at the way Bob had answered the teacher.’ (Schachter and Otanes 1972:164) TAHITIAN: example (10.17) in section 10.4. THAI: example (9.3b) in section 9.1. TONGAN: only for pronominal Ss and As, example (10.11) in section 10.4. TURKISH: example (7.21b) in section 7.1.2. TUVA: examples (10.3a–10.3b) in section 10.1. VATA (25) kòfí sàká pī-lì Kofı GEN rice prepare-AN ‘Kofı’s preparing rice’ (Koopman 1984:21, ex. 9) WIKCHAMNI (26a) t ha-ŋ [čanin ] tell-AOR I:NOM that-ACC John:GEN swim:AN-DAT ‘I told her about John’s swimming.’ (Gamble 1978:125, ex. 17) (26b) [nim t ha-ŋ hatat] want I:NOM I:GEN that-ACC lift:AN:ACC ‘I want to lift it’ (lit. ‘I want my lifting it’) (ibid.: 69, ex. 5) YUKAGIR: example (6.19) in section 6.2. 2′. ANTI-PASSIVE INUIT (WEST GREENLANDIC and INUKTITUT): examples (6.16b, 6.17) in section 6.1.
< previous page
page_288
next page >
< previous page
page_289
next page >
Page 289 3. ERGATIVE-POSSESSIVE (ERG-POSS) 3a. Sentential-Possessive (SENT-POSS) ARABIC, CLASSICAL (27) safat-hum [ sweep:PRF.3SG:OB.3PL wind-NOM.CONSTR safy-a r-rimāl-i] the-annihilation:GEN sweep:AN:ACC the-sands-GEN yad-u d-dabūr-i hand-NOM.CONSTR the-west.wind:GEN ‘The wind of annihilation swept them away, as the hand of the westwind sweeps away the sands.’ (Wright 1971: part 3, 58) SAMOAN (28) e lelei [le kukaina e Sali o UNSP be.good ART cook:AN ERG Sally SUB.POSS le i’a] ART fish ‘Sally’s cooking of the fish is good.’ (Chung 1973:665) TONGAN (29) ’oku ’ikai tonu [’a e taki ’o e PRES NEG good ABS ART lead:AN SUB.POSS ART fonua ’e he tu’i] land ERG ART king ‘The leading of the country by the king is not good.’ (Chung 1973:656) 3b. Oblique-Possessive ABKHAZ: examples (7.18–7.20) in section 7.1.1. BULGARIAN (30a) pee-ne-t-o na Petyr sing-AN-DEF-N of Peter ‘Peter’s singing’
< previous page
page_289
next page >
< previous page
page_290
next page >
Page 290 (30b) ubij-stvo-t-o na kuče-t-o ot negovija stopanin kill-AN-DEF-N of dog-DEF-N by his owner ‘the killing of the dog by his owner’ (Mila Stojčeva personal communication) DUTCH (31a) Jan-s lopen John-GEN walk:AN ‘John’s walking’ (Dik 1985:24, ex. 79) (31b) het lopen van Jan the walk:AN of John ‘the walking of John’ (ibid.: 24, ex. 80) (31c) het slaan van de hond door Jan the beat:AN of the dog by John ‘the beating of the dog by John’ (ibid.: 24, ex. 84) EGYPTIAN, ANCIENT (32a) m prt s(t)m at go.forth:AN setem.priest ‘at the going forth of the setem-priest’ (32b) m prt-f tpt at go.forth:AN-3SG.M.POSS first ‘at his first going forth’ (Gardiner 1957:226) (32c) šdt recite:AN glorification by lector-priests ordinary ‘reciting of glorifications on the part of (by) ordinary lector-priests’ (ibid.: 225) ENGLISH (33a) the secretary’s arrival/the arrival of the secretary (33b) the secretary’s dismissal by the chairman/the dismissal of the secretary by the chairman (Verhaar 1990:154, ex. 235) ESTONIAN: example (9.8a) in section 9.2
< previous page
page_290
next page >
< previous page
page_291
next page >
Page 291 FRENCH (34a) la venue de Pierre the:come:AN of Pierre ‘Pierre’s arrival’ (34b) la destruction de Carthage par les Romains the destroy:AN of Carthage by the Romans ‘the destruction of Carthage by the Romans’ GEORGIAN: examples (7.11–7.13) in section 7.1.1. GERMAN (35a) [Das Kommen der Tante] führt immer zu Zwistigkeiten. ‘Aunt’s coming always brings about quarrels.’ (Esau 1973:184, ex. 51) (35b) die Verhaftung eines Verbrechers durch die Polizei ‘the arrest of a criminal by the police’ (ibid.: 128, ex. 64) GREEK, MODERN (36a) o erxomós tu jáni the:M.NOM come:AN.NOM the:M.GEN John:GEN ‘the coming of John; John’s coming’ (Joseph and Philippaki-Warburton 1987:130, ex. 30) (36b) to plísimo ton the:N.NOM wash:AN.NOM the:GEN.PL rúxon apó ton jáni clothes:GEN.PL by the:M.ACC John:ACC ‘the washing of the clothes by John’ (ibid: 129, ex. 28) HEBREW, MODERN: example (7.40) in section 7.2.2. HIXKARYANA: examples (7.55–7.56) in section 7.3. IRISH, OLD (37a) [tīchtu Phatraicc] nī ceiltis go:AN:ACC Patrick:GEN NEG they:hid (Disterheft 1980:136, ex. 113a)
< previous page
page_291
next page >
< previous page
page_292
next page >
Page 292 (37b) it ilsailm hitadbadar there.are many:psalms:NOM in.which.it.is.shown [foisitiu a-pecthae do-duaid] confess:AN:NOM his-sins:GEN to-David:DAT ‘There are many psalms in which the confession of his sins by David is shown.’ (ibid: 137, ex. 114a) KURDISH (38a) xeber-a [xat-in-a min] word-EZ arrive-AN-EZ me:OBL ‘news about my arrival’ (Kurdoev 1978:70) (38b) hirçê ji aliy-ê rawkeran kill-AN-EZ bear from side-EZ hunters ‘the killing of the bear by the hunters’ (Farhad Shakely personal communication) MAORI: examples (7.30–7.31) in section 7.2.2. PANJABI (39) amrika d-a kolambas dvara labh-na America:OBL of-M Columbus by discover-AN ‘the discovery of America by Columbus’ (Ravinder Singh personal communication) PERSIAN (40a) raft-an-e man go.away-AN-EZ I ‘my going away’ (Amin-Madani and Lutz 1972:490) (40b) āmad-an-am come-AN-1SG.POSS ‘my coming’ (ibid: 488) (40c) košt-an-e xarse bevasileje šekarčejan kill-AN-EZ bear by hunters ‘the killing of the bear by hunters’ (Hassan Djamshidpey personal communication)
< previous page
page_292
next page >
< previous page
page_293
next page >
Page 293 POLISH (41) kupowa-nie mięs-a przez Hankę buy-AN meat-GEN by Hanka ‘the buying of meat by Hanka’ (Topolińska 1981:148) RUSSIAN: examples (1.7, 1.9) in section 1.3. PUKAPUKAN: for an intransitive AN see examples (10.13–10.14) in section 10.4. (42) e lelei [te oko-nga o te moa PRES good ART buy-AN DOM.POSS ART chicken e nā wāwine] AGT PL women ‘The buying of the chicken by the women is good.’ (Chung 1973:659, ex. 41) RUMANIAN (43a) pleca-re-a lui Ion depart-AN-DEF he:GEN Ion ‘Ion’s departure’ (Mallinson 1986:213, ex. 1059) (43b) limbilor de learn-AN-DEF language:DEF:GEN.PL foreign:F.PL by englezi towards English:M.PL ‘the studying of foreign languages by the English’ (ibid: 214, ex. 1061) WELSH (44) gwrthodiad y cynnig o’r gweinidog rejection the offer from:the minister ‘the rejection of the offer by the minister’ (Comrie and Thompson 1985:386, ex. 188)
< previous page
page_293
next page >
< previous page
page_294
next page >
Page 294 4. NOMINAL (NOMN) 4a. Double-Possessive (DBL-POSS) BULGARIAN: with highly referential As (pronouns and personal names). (45a) negov-o-t-o pee-ne na ljubovn-a-t-a pesen his-N-DEF-N sing-AN of love:ADJ-F-DEF-F song ‘his singing the love song’ (45b) pee-ne-t-o mu na ljubovn-a-t-a pesen sing-AN-DEF-N his (ENCL) of love:ADJ-F-DEF-F song ‘his singing the love song’ (Mila Stojčeva personal communication) DUTCH: with highly referential As (pronouns and personal names). (46) Jan-s slaa-n van de hond John:GEN beat-AN of the dog ‘John’s beating of the dog’ (Dik 1985:24, ex. 83) ENGLISH: example (1.4) in section 1.2.1. ESTONIAN (47a) Peetr-i laul-mine Peter-GEN sing-AN ‘Peter’s singing’ (47b) Peetr-i maja-de ehita-mine Peter-GEN house-GEN.PL build-AN ‘Peter’s building (of) houses’ (Diana Krull personal communication) FINNISH: example (8.3) in section 8.1.1. FRENCH: only with pronominal As, example (10.4a) in section 10.2. GBADI: no examples with intransitive ANs and transitive ANs with the A present. (48) I wà [bànyÒ na kú ] I like Bagno GEN VPTL look-AN ‘I like to observe Bagno.’ (Koopman 1984:22, ex. 12)
< previous page
page_294
next page >
< previous page
page_295
next page >
Page 295 GERMAN: with highly referential As (pronouns and person names). (49a) Peters Schnarchen ‘Peter’s snoring’ (49b) Peters Singen des Liedes ‘Peter’s singing of the song’ GREEK, ANCIENT (50a) ART.F.NOM ART.M.GEN Laches:GEN ART.GEN.PL ship:GEN.PL command ‘Laches’ command of the fleet’ (50b) ART.F.NOM Phaeacians:GEN occupation ART.F.GEN Corcyra ‘the former occupation of Corcyra by the Phaeacians’ (Smyth 1956) GREEK, MODERN With pronominal As (51a) i tu tis the:F.NOM know:AN his(ENCL) the:F.GEN ksénis foreign:F.GEN language:F.GEN ‘his knowledge of the foreign language’ (Mackridge 1985) With non-pronominal As: sub-standard (51b) tú niku oi the:M.GEN Nick:GEN the:PL.NOM know:AN tis ksénis the:F.GEN foreign:F.GEN language:F.GEN (Irene Fotopoulou personal communication. This example is qualified as unacceptable in Mackridge 1985) HEBREW, MODERN: sub-standard, examples (5,6) in Chapter 7, note 3. JAPANESE: examples (8.8–8.10) in section 8.1.1.
< previous page
page_295
next page >
< previous page
page_296
next page >
Page 296 LATIN (52) Helveti-orum iniuri-is popul-i Helvetian-GEN.PL injury-ABL.PL people-GEN.SG Roman-i Roman-GEN.SG ‘the injuries of the Roman people inflicted by the Helvetians’ (Groot 1957:31) LATVIAN (53) Aleksandr-a iekaro-šana sagrāva Alexander-GEN Egypt-GEN conquer-AN destroyed zemi. country:ACC ‘Alexander’s conquest of Egypt destroyed the country.’ (Laine Lasmanis personal communication) POLISH: only with highly referential As. (54) moj-e pisa-nie list-u my-N write-AN.N letter-GEN ‘my writing of the letter’ (Topolińska 1981:148) RUSSIAN: only with highly referential As (mainly pronouns and personal names), example (10.1a) in section 10.1. SAMOAN: only for pronominal As, examples (10.15, 10.16) in section 10.4. 4b. Possessive-Adnominal (POSS-ADN) ARABIC, EGYPTIAN (55) [akl samiir l-iggibna] teach:AN Samiir to-cheese ma-za’’al-ni-iš NEG-upset:PERF:3SG-1SG.OBJ-NEG ‘Samir’s eating the cheese did not upset me.’ (Abdel-Massih et al. 1979)
< previous page
page_296
next page >
< previous page
page_297
next page >
Page 297 ICELANDIC (56a) söngur Pétur-s/han-s sing:AN Peter-GEN/he-GEN ‘Peter’s/his singing’ (56b) dráp-ið veiðimannanna á birninum kill:AN-DEF hunter:DEF:GEN.PL on bear:DEF:DAT.SG ‘the killing of the bear by the hunters’ (Kjartan Ottósson personal communication) SWEDISH: example (8.13) in section 8.2.1. 5. MIXED (MIX) BULGARIAN: in bureaucratic style. (57) pokriv-ane(t-o) razxod-i-t-e na cover-AN(DEF-N) expense-PL-DEF-PL of komandirovanija ot fondacija-t-a business.traveller by fund-DEF-F ‘the fund’s covering the expenses of the one who makes a business trip’ (Ivan Derzhanski personal communication) ?DUTCH: marginal. (58) het zijn broertje vermoorden door Jan the his brothe kill:AN by John ‘John’s killing of his brother’ (Hoekstra 1986:566) EGYPTIAN, ANCIENT: restricted to certain pronominal Ps in the third person, see the examples in section 10.2, pp. 202. MAORI: example (7.32) in 7.2.2. THAI: example (9.3a) in section 9.1. 6. INCORPORATING (INC) 6a. Sentential-Incorporating (SENT-INC) ?QUECHUA, CUZCO: example (5.19) in section 5.5. QUECHUA, IMBABURA: example (5.18) in section 5.5.
< previous page
page_297
next page >
< previous page
page_298
next page >
Page 298 6b. Possessive-Incorporating (POSS-INC) ESTONIAN: example (9.8c) in section 9.2. EWE: example (9.6b, 9.7b) in section 9.2. HUNGARIAN (59) Péter ujság-olvas-ás-a Peter newspaper-read-AN-3SG.POSS ‘Peter’s reading of newspapers’ (Ferenc Kiefer personal communication) ICELANDIC (60a) bjarna(r)-dràp veiðimanna bear-kill:AN hunter:GEN ‘the hunter’s killing of a bear/bears’ (60b) hús-bygg-ing Pétur-s house-build-AN Peter-GEN ‘Peter’s building a house/houses’ (Kjartan Ottósson personal communication) KURDISH (61a) sitran-got-în-a Peter song-say-AN-EZ Peter ‘Peter’s singing’ (61b) xanî-çêkir-în-a Peter house-build-AN-EZ Peter ‘Peter’s house-building’ (Farhad Shakely personal communication) PERSIAN (62) ketab-xand-än-e Peter book-read-AN-EZ Peter ‘Peter’s reading of a book’ (lit. ‘book-reading’) (Hassan Djamshidpey personal communication) QUECHUA, CUZCO (63) [Xwan-pa hamu-sqa-n-qa] Juan-GEN come-AN.REAL-3-TOP manchari-chi-wa-n-mi. afraid-CAUS-1OB-3-AFFIRM ‘That John has come frightens me.’ (Lefebvre and Muysken 1988:21, ex. 22)
< previous page
page_298
next page >
< previous page
page_299
next page >
Page 299 SWEDISH (64) Peter-s piano-spel-ande Peter-GEN piano-play-AN ‘Peter’s playing the piano’ (Martin Tamm personal communication) 6c. Oblique-Incorporating (OBL-INC) ESTONIAN: example (9.8b) in section 9.2. 7. RELATIVE (REL) HAUSA: example (9.11b) in section 9.3. HUNGARIAN: example (1.15) in section 1.4.2. ?8. ARGUMENT-REDUCING (ARG-RED) CHUKCHEE: example (9.14) in section 9.4.
< previous page
page_299
next page >
< previous page
page_300
next page >
Page 300 NOTES 1 PRELIMINARIES: THE AIM OF THIS BOOK 1 According to Hopper and Thompson (1984:703), ‘the grammars of languages tend to label the categories N and V with morpho-syntactic markers which are iconically characteristic of these categories to the degree that a given instance of N or V approaches its prototypical function. In other words, the closer a form is to signaling this prime function, the more the language tends to recognize its function through morphemes typical of the category—e.g. deictic markers for N, tense markers for V.’ 2 In fact, as shown by Verhaar (1990), even ANs derived from intransitive verbs may show different nominalization patterns depending on their degree of transitivity. Future studies may show whether the differentiation between Ss which are high and low in transitivity has impact on the overall typology of ANCs. 3 Vladimir Plungjan and Ekaterina Raxilina (personal communication) have suggested that the rules which govern reduction of coreferential NPs, or interpretation of empty arguments, in Russian ANCs are not of syntactic but rather of semantic nature. There must exist a ‘semantic agreement’ between the thematic role of a certain argument of the matrix verb and the thematic role of the deleted coreferential argument of the subordinate action nominal. Thus, in the phrase žaždat’ mesti—‘to thirst for revenge’, the subject of the action nominal is coreferential with the subject of the matrix verb, while in the phrase žaždat’ proščenija—‘to thirst for forgiveness’, the object of the action nominal is coreferential with the matrix verb, the difference being attributed to the different thematic structure of the two action nominals. 4 Ransom (1986) offers an excellent semantic classification of English complement clauses, which is, however, too detailed for the present discussion. 2 ANCs AS NON-FINITE NOUN CLAUSES 1 It may be argued that for a number of languages with no inflectional morphology, like Chinese, the notion of finiteness does not make sense. From this point of view, s-like complements in Chinese could hardly be characterized as finite subordinate clauses. For the sake of simplicity, I will leave these cases aside.
< previous page
page_300
next page >
< previous page
page_301
next page >
Page 301 2 The other difference between independent and dependent clauses in Wappo, which Noonan does not mention, consists of the presence vs. absence of a glottal stop at the end of the head verb (Thompson and Longacre 1985:172). 3 In Burjat, according to Skribnik (1983:87), only subjects of intransitive verbs may be put in the accusative in indirect quotes. 4 Balanced subordinated clauses (with subordinators) may also be used independently, like in the following German example, which is introduced by the subordinator dass: Dass du bloß nicht wieder zu spät kommst!—‘Don’t dare to be late again!’ (Martin Haspelmath personal communication). 5 Though Noonan does not clarify the term ‘associative relation’, it is clear from the use of it that it refers to relations between a head nominal in an NP and its unmarked dependents, including possessors. 6 The exact status of action nominals in Modern Mongolian is also complicated by the fact that they can be used as matrix predicates, tending to replace ‘finite’ verbs in all the tenses (Binnick 1979:85, JanOlof Svantesson personal communication). 7 As suggested to me by Martin Haspelmath, the ‘substantivized’ infinitive in German, like das Nehmen ‘the taking’ may continue the old Proto-Indo-European nominative case, while the normal infinitive nehmen goes back to the homophonous directional accusative. If this is true, the former may never have completely lost its nominal properties. 3 NOUN PHRASES VS. FINITE CLAUSES IN DIFFERENT LANGUAGES 1 An interesting question is that of why languages, which may easily use verbs in nominal functions, still have nominalizations with an overt nominalization marker. Clark (1981:79) proposes that ‘unsuffixed nominalizations denote activities or processes, which can be qualified as to manner, or described as beginning or ceasing, whereas suffixed nominalizations denote events, which can be enumerated and located in time’. In Samoan and Tongan, action nominalization normally does not attach an overt nominalization marker to the verb, which, according to Chung (1973:646), is an innovation in these languages. Tokelauan, closely related to Samoan, uses overt nominalizers much more regularly than Samoan (Vonen 1988:86–98). 2 In this study, the discussion is limited to those cases where the arguments of ANs are nouns or NPs, i.e., have the same category status as in the corresponding finite cases. Thus, we will not consider the use of adjectives in such ANCs as the Italian invasion of Albania . 3 Note that these two explanations can lead to different results at some points. Thus, Keenan talks specifically about inalienable possession as the case where the referent of the possessor determines the reference of the whole function. On the other hand, according to Seiler, in situation of inherent, inalienable possession, the possessor appears as being qualified by the possessum. 4 There is, however, a class of constructions in Swedish, such as en folkets teater —‘a theatre for the people’ (lit., ‘a the-people’s theatre’), in which
< previous page
page_301
next page >
< previous page
page_302
next page >
Page 302 a noun is preceded by both an article and a genitive. In these relatively marginal constructions, genitives express beneficiaries and not possessors. 5 Here, however, I lack data which could argue for or against the existence of any connection between verb-subject or verb-object structures and alienable and inalienable possession respectively. Perhaps, one of the reasons for this is the fact that inalienable/alienable opposition is almost never expressed only by means of dependent-marking, as shown by Nichols (1986). 6 The class of possessive forms is subject to great variation across Slavonic languages. Thus, in Czech, the class of preposed possessive forms is much greater than in Russian and involves possessive adjectives formed for singular definite NPs, without any dependents of their own. Nouns of certain morphological classes do not form possessive adjectives (Corbett 1987). 4 THE DATA BASE 1 It may be suggested that the existence in languages of complex constructions, including ANCs, is somewhat correlated with the cultural level and the level of communication needs in the societies which use these languages. 5 TYPE 1: ‘SENTENTIAL’ 1 Krejnovič (1979a:311) states that object markers disappear only when the verb has incorporated its object, otherwise (that is, even when there is an overt external object) they appear on the verb. The question of whether there is a regular process of incorporation in Nivkh is complicated (Vladimir Nedjalkov personal communication). 2 The minor head-marking patterns in Ingush concern both personal and class agreement (see Table 5.1). It would be interesting to obtain data on head-marking in ANCs and compare them with the hypothesis launched in section 5.3., more specifically, that ANCs are more likely to retain class agreement than personal agreement. 3 Nouns in the Construct State differ from those in the Absolute State in lacking the preposed article and the final -n. In fact, it is not obvious whether these phenomena may be at all regarded as an instance of head-marking. 6 TYPE 2: ‘POSSESSIVE-ACCUSATIVE’ 1 According to my source of information about Selkup (Kuznecova et al. 1980:384), in all the cases where there is no possibility of confusing the subject and the direct object of a sentence, the latter can be put in the nominative. The accusative, however, is preferred. Thus, among the 413 cases analysed by Kuznecova et al., only 53 direct objects (13 per cent) are in the nominative. No semantic difference has been registered between the nominative and the accusative direct objects. 2 The nominalization pattern described in this section is not the only one possible in Cairene Egyptian Colloquial Arabic. For more information see section 8.2.2.
< previous page
page_302
next page >
< previous page
page_303
next page >
Page 303 3 Describing the structure of ANCs in Tagalog, Schachter and Otanes (1972:163) write: ‘The performer of the action expressed by a gerund may or may not be designated. If designated, the performer is expressed by a possessive modifier…specifically, either a ng phrase (ng plus unmarked noun, (ni/na) plus personal noun, or the ng form of a deictic or personal pronoun) immediately following the gerund, or the sa form of the personal pronoun plus the linker -ng preceding the gerund.’ Thus, according to this formulation, only personal pronouns in their sa-form can express the S or the A and only ng -phrases follow the head of the construction. Neither of these restrictions are present in the description of possessive phrases. Thus, the question arises whether there are real restrictions which are only peculiar for action nominal constructions (in which case we can no longer say that the As and the Ss in them totally assimilate to Gs), or whether the authors are not consistent in their description of Gs. I presume that the latter is the case. 4 In this book, the terms ‘Khalkha Mongolian’ and ‘Modern Mongolian’ are used interchangeably. ‘The new literary language is close to the spoken vernacular, and no great harm is done pretending that the Khalkha dialect is equivalent to this language’ (Binnick 1979:xxiv). 5 In the list of the conjunctions used for phrasal conjunction in Modern Mongolian, Binnick mentions one, xijgeed (literally ‘having done’) which seems to be used only with verbal nominals (Binnick 1979:50). 7 TYPE 3: ‘ERGATIVE-POSSESSIVE’ 1 Harris describes only one pattern available in Georgian action nominal constructions. However, according to Hewitt (1983:260–2), the situation is much more complicated. Hewitt mentions the following cases of the ‘fluctuating treatment of certain arguments’: (a) the As, which are normally governed by the postposition mier (or by -gan from certain verbs), can be marked by the simple genitive. While this seems to be extremely rare in the presence of the P, as in (1) below (1) giv-is tamaši pexburt-isa Givi-GEN playing football-GEN ‘Givi’s playing (of) football’ such a marking is a commonplace in the absence of any P, as in (2): (2) e. honigman-is es šenišvna E. honigman-GEN observation (cf. es šenišna—‘he observed this’) ‘This observation by E.Honigmann’ As Hewitt puts it, ‘surely, the simple Gen is possible because the nominal force of the form is here somehow stronger than the verbal, which would require the presence of one of the agentive postpositions’. (b) Although Ss are normally treated as Ps, i.e., by being put in the
< previous page
page_303
next page >
< previous page
page_304
next page >
Page 304 genitive, they can sometimes appear with agentive postpositions, as in (3). (3) gadasvla bavšvis mier. many:worded sentences:to transition child:GEN by ‘The transition to many-worded sentences on the part of the child’. (c) IO can sometimes be marked with the genitive in combination with masdars. Thus, action nominals derived from certain ditransitive verbs can manifest ambiguity as regards the interpretation of their genitive modifiers. For example, the phrase in (4) may be a nominalization of either (4′) or of (4″). (4) čemi daxmareba my help(ing) (4′) me vexmarebi. I:NOM someone:DAT I:help:him ‘I help someone.’ (4″) me mexmareba. someone:NOM I:DAT he:helps:me ‘Someone helps me.’ 2 ‘The statement (Harris 1981:153) that masdars may take adjectival qualifiers presents only part of the picture. Note (8) (p. 298) more accurately observes that one such form (namely “answer(ing)”) behaves not only like a masdar but also like an ordinary noun. However, Harris proceeds to state that, when performing the role of masdar, it will be qualified by an adverb (= “correctly answering’), whilst its nominal character requires the presence of an adjective (= “correct answer’’). This wavering between verbal and nominal characteristics on the part of masdars is an interesting feature of G(eorgian) and thus a clear candidate for future investigation’ (Hewitt 1983:260). 3 The two types of action nominal constructions described in section 7.2 are characteristic of normative usage. According to Berman (1978:131–2), many people also allow nominalizations of the DBL-POSS type. Thus, as regards the following example (5) dxiyat hahacaa šel Dan reject:AN.CONSTR DEF:offer of Dan grammars ascribe it only one interpretation, ‘the rejection of Dan’s offer’, while native Hebrewspeaking college students interpret such expressions as ambiguous, where in the second reading ‘Dan’ is taken as the A in relation to the verbal noun. What is even more remarkable is the fact that the order between the modifiers in such constructions is rather free. Thus, colloquial usage accepts also the following example, which is not possible at all according to the purists. (6) dxiyat Dan šel hahacaa reject:AN.CONSTR Dan of DEF:offer ‘Dan’s rejection of the offer’ The normative restrictions on ANCs seem to be related to the general
< previous page
page_304
next page >
< previous page
page_305
next page >
Page 305 restrictions on possessive NPs in Hebrew, mentioned in section 7.2.2., that is, the impossibility of having more than one ‘Possessor’ within one and the same NP. Those dialects that allow such chains could perhaps use different modifiers to distinguish between the P and the A within action nominal constructions. 4 Hewitt (1983:263), in his review of Harris (1981), remarks that there are some difficulties with her statement about the behaviour of masdars derived from ‘affective’ verbs. Thus, Harris translates as ‘hatred’. However, according to Hewitt, this form corresponds to the dynamic verb ‘to conceive hatred’ and thus behaves exactly as a normal dynamic masdar, while the ‘affective’ verb ‘to hate’ has a stative masdar —‘hatred’, which marks its ‘logical Subject’ (experiencer) with the genitive. In some cases, however, one and the same form can be both a stative and a dynamic masdar, as it is the case with , which corresponds to both the ‘affective’ verb ‘to forget’ and the dynamic verb ‘to put something out of one’s mind’. In both cases it takes a mier -argument, which thus corresponds either to the experiencer or to the agent respectively, while the finite ‘affective’ verb cannot be used in a passive form and thus cannot combine with a mier -argument. 5 John Payne (personal communication) suggests that the Ezafe-suffixes in Persian possessive NPs should be considered as markers of the whole construction rather than head-markers. 8 TYPE 4: ‘NOMINAL’ 1 Koopman explicitly states that in Gbadi and Vata, the Ps in ANCs differ, but does not say whether there are any differences in the expression of the As and offers analysis only for Vata. 9 LESS FREQUENT TYPES 1 The Eskimo languages, which are often considered as incorporating languages, have hardly any noun incorporation. What has previously been described as a process of incorporation should rather be considered as a process of verbal derivation by means of verbal affixes. 10 SPLITS AND VARIATIONS 1 As mentioned in section 9.1., ANCs with the P in the non-marked (sentential) form in Bulgarian are characteristic for the bureaucratic style and are normally used when the P itself has dependents introduced with the preposition na—‘of’. 2 On the basis of (rare) examples in which the pronominal object appears in the same form as in finite clause, if it does not immediately follow the AN, Gardiner (1957:226) suggests that ‘whenever the object of an infinitive is separated from it by an extraneous element, as in —“in giving tohim white.bread”, this object is to be regarded, not as a direct genitive, but as an accusative’. In the terms adopted in this book, such constructions would be classified as following the POSS-ACC pattern.
< previous page
page_305
next page >
< previous page
page_306
next page >
Page 306 11 EXPLANATION OF TYPE-MEMBERSHIP 1 As argued by Jeffers in an extremely interesting paper (1976), the morphological category verbal noun in Old Irish and syntactic constructions involving it are probably an innovative morphosyntactic system which developed in the Celtic branch of the Indo-European family. 2 Note again that Chung’s classification of nominalization types (presented in section 10.4.) differs from the one adopted here. Thus, accusativity/ergativity in the dominant/subordinate opposition cuts across the opposition between the ‘accusative-looking’ POSS-ACC type and the ‘ergative-looking’ ERG-POSS type. 12 EXPLANATION OF THE ATTESTED TYPES 1 I have benefited in my thinking on these issues from discussions with Jan Anward and Östen Dahl. 2 Non-derived NPs in Hixkaryana show an opposition between present and past possession; see section 7.3. 3 According to Noonan (1985:92), ‘time reference dependency’ means that the time reference of a complement is a necessary consequence of the meaning of the complement-taking predicate, while ‘truth-value (epistemic) dependency’ means that the complement construction containing a complement involves an explicit qualification of commitment to the truth of the proposition embodied in the complement. 4 This principle shows certain resemblance to Dik’s (1985:3) ‘Principle of formal adjustment’: ‘Derived constructions of type X are under pressure to adjust their formal expression to the prototypical expression model provided by non-derived constructions of type X’.
< previous page
page_306
next page >
< previous page
page_307
next page >
Page 307 Bibliography Abdel-Massih, E.T., Abdel-Malek, Z.N., and Badawi, E.-S.M. in association with McCarus, E.N. 1979. A Comprehensive Study of Egyptian Arabic. Center for Near Eastern and North African Studies, University of Michigan, Ann Arbor. Abney, S. 1986. Functional Elements and Licensing. Unpublished MS. MIT, Cambridge, Mass. Allen, S. 1964. Transitivity and Possession. In Language, vol. 40. 3: 337–43. Amin-Madani, S. and Lutz, D. 1972. Persische Grammatik. Julius Groos Verlag, Heidelberg. Anderson, M. 1979. Noun Phrase Structure. Unpublished Ph.D. Thesis, University of Connecticut. Anderson, S.R. and Chung, S. 1977. On Grammatical Relations and Clause Structure in Verb-Initial Languages. In Cole, P. and Sadock, J.M. (eds), Syntax and Semantics, vol. 8:1–26. Academic Press, New York, San Francisco, London. Anderson, S.R. 1977. On Mechanisms by Which Languages Become Ergative. In Li (ed.) 1977: pp. 317– 64. Anderson, S.R. 1985. Inflectional Morphology. In Shopen, T. (ed.) 1985, vol. 3:150–201. Andrews, A. 1985. The major functions of the noun phrase. In Shopen (ed.) 1985, vol. 1:62–124. Anward, J. 1985. Att förklara språkliga fenomen. Unpulished MS. University of Stockholm. Armbruster, Ch.H. 1960. Dongolese Nubian. A Grammar. Cambridge University Press, Cambridge. Arnott, D.W. 1970. The Nominal and Verbal Systems of Fula. Oxford University Press, Oxford. Asher, R.E. 1982. Tamil. Lingua Descriptive Studies, vol. 7. North-Holland Publishing Company, Amsterdam.
< previous page
page_307
next page >
< previous page
page_308
next page >
Page 308 Ashton, E.O., Mulira, E.M.K., Ndawula, E.G.M., and Tucker, A.N. 1954. A Luganda Grammar. Longmans, Green and Co., London, New York, Toronto. Awberry, G.M. 1976. The Syntax of Welsh. Cambridge University Press, Cambridge, London, New York, Melbourne. Bakaev, C.X. 1973. Jazyk kurdov SSSR. Nauka, Moscow. Bakker, D., Hengeveld, K., Kahrel, P. and Rijkhoff, J. Forthcoming. A Method of Language Sampling. To appear in Studies in Language. Bartning, I. 1986. Aspectes des syntagmes binominaux en de en français. In Travaux de linguistique et de littérature, xxiv, 1:347–71. Beffa, M.-L. and Hamayon, R. 1975. Eléments de grammaire mongole. Pierre-et-Marie Curie University, Centre of Quantitative Linguistics, Dunod, Paris. Bell, A. 1978. Language Samples. In Greenberg, J.H. et al. (eds), vol. 1:123–56. Bergel’son, M.M. and Kibrik, A.A. 1987. Sistema pereključenija referencii v tuvinskom jazyke. In Sovetskaja tjurkologija, vol. 2: 16–32, and vol. 4:30–44. Bergel’son, M.B., Zaliznjak, A.A., and Kibrik, A.E. 1982. Konstrukcii s sentencial’nym aktantom v tabasaranskom jazyke. In Tabasaranskie étjudy. Moscow University Press, Moscow. 44–65. Berman, R.A. 1976. On Derived and Deriving Nominals in Modern Hebrew. In Cole, P. (ed.), Studies in Modern Hebrew Syntax and Semantics: 57–98. North-Holland Publishing Company, Amsterdam. Berman, R.A. 1978. Modern Hebrew Structure. University Publishing Projects, Ltd, Tel-Aviv. Biggs, B. 1961. The Structure of New Zealand Maori. In Anthropological Linguistics, vol. 3, 3:1–54. Binnick, R.I. 1979. Modern Mongolian. A Transformational Syntax. University of Toronto Press, Toronto, Buffalo, London. Bogatyrev, K. and Boguslavskaja, O. 1982. Opredelitel’nye konstrukcii v dvux govorax tabasaranskogo jazyka. In Tabasaranskie étjudy. Moscow University Press, Moscow. 74–95. Boguslavskaja, O. 1989. Atributivnye konstrukcii v dagestanskix jazykax. Unpublished dissertation, Moscow State University. Bokarev, E.A. 1979. Dagestanskie jazyki. In Jazyki Azii i Afriki, 3: 161–172. Bolkenstein, A.M., de Groot, C., and Mackenzie, J.L. (eds) 1985. Predicates and Terms in Functional Grammar. Foris Publications, Dordrecht, Holland; Cinnaminson, USA.
< previous page
page_308
next page >
< previous page
page_309
next page >
Page 309 Bybee, J.L. 1985. Morphology. John Benjamins Publishing Company, Amsterdam, Philadelphia. Čeremisina, M. 1981a. Predikativnoe sklonenie kak baza zavisimoj predikacii v altajskix jazykax. In Čeremisina, M. (ed.), Padeži i ix ékvivalenty v stroe složnogo predloženija v jazykax narodov Sibiri. AN SSSR, Novosibirsk. Čeremisina, M. 1981b. K voprosu o sistemnyx otnošenijax meždu finitnymi, pričastnymi i deepričastnymi formami glagol’nogo slova (na materiale altajskix jazykov). In Ubrjatova, M. and Kim Čer Len (eds), Teoretičeskie voprosy fonetiki i grammatiki jazykov narodov SSSR: 3–18. Novosibirsk University Press, Novosibirsk. Chomsky, N. 1957. Syntactic Structures. Mouton, The Hague. Chomsky, N. 1970. Remarks on Nominalizations. In Jacobs, R.A. and Rosenbaum, P.S. (eds), Readings in English Transformational Grammar. Ginn and Company, Waltham, Mass., Toronto, London. Chomsky, N. 1975. Reflections on Language. Fontana/Collins, Glasgow. Chung, S. 1973. The Syntax of Nominalizations in Polynesian. In Oceanic Linguistics, 12:641–86. Chung, S. 1978. Case Marking and Grammatical Relations in Polynesian. University of Texas Press, Austin and London. Clark, R. 1981. Inside and Outside Polynesian Nominalizations. In Hollyman, J. and Pawley, A. (eds), Studies in Pacific Languages and Cultures in Honour of Bruce Biggs. Linguistic Society of New Zealand, Auckland. Cole, P. 1982. Imbabura Quechua. Lingua Descriptive Studies, vol. 5. North-Holland Publishing Company, Amsterdam. Comrie, B. 1975. The Antiergative: Finland’s Answer to Basque. In Chicago Linguistic Society, 11:112– 21. Comrie, B. 1976. The Syntax of Action Nominals: A Cross-linguistic Study. In Lingua, 40:177–201. Comrie, B. 1978. Ergativity. In Lehmann (ed.) 1978:329–94. Comrie, B. 1981. Language Universals and Linguistic Typology. Basil Blackwell, Oxford. Comrie, B. (ed.) 1987. The World’s Major Languages. Croom Helm, London, Sydney. Comrie, B. and Thompson, S.A. 1985. Lexical Nominalization. In Shopen (ed.), 1985:349–98. Cook, E.-D. 1984. A Sarcee Grammar. University of British Columbia Press, Vancouver.
< previous page
page_309
next page >
< previous page
page_310
next page >
Page 310 Corbett, G.G. 1987. The Morphology/Syntax Interface: Evidence from Possessive Adjectives in Slavonic. In Language, vol. 63, 2: 299–345. Cowley, R., Bender, M.L., Ferguson, Ch.A., Fulass, H., and Haile, G. 1976. The Amharic Language. In Bender, M.L., Bowen, J.D., Cooper, R.L., Ferguson Ch.A. (eds), Language in Ethiopia: 77–89. Oxford University Press, London. Dahl, Ö. 1980. Russian Direct Objects and Functional Case Marking Principles. In G.Jacobsson (ed.), Slavica Gothoburgensia, vol. 7: 17–27. Gothenburg University, Gothenburg. Dahl, Ö. 1985. Tense and Aspect Systems. Basil Blackwell, Oxford. Davies, J. 1981. Kobon. Lingua Descriptive Studies, vol. 3. North-Holland Publishing Company, Amsterdam. Dayley, J.P. 1985. Tzutujil Grammar. University of California Publications in Linguistics, vol. 107. University of California Press, Berkeley, Los Angeles, London. Dench, A. and Evans, N. 1988. Multiple Case-marking in Australian Languages. In Australian Journal of Linguistics, 8:1–47. Derbyshire, D.C. 1979. Hixkaryana. Lingua Descriptive Studies, vol. 1. North-Holland Publishing Company, Amsterdam. Dik, S.C. 1983. Two Constraints on Relators and What They Can Do for Us. In Dik, S.C. (ed.), Advances in Functional Grammar: 267–99. Foris Publications, Dordrecht, Holland; Cinnaminson, USA. Dik, S.C. 1985. Formal and Semantic Adjustment of Derived Constructions. In Bolkenstein, A.M., et al. (eds) 1985:1–28. Disterheft, D. 1980. The Syntactic Development of the Infinitive in Indo-European. Slavica Publishers, Inc., Columbus, Ohio. Dixon, R.M.W. 1972. The Dyirbal Language of North Queensland. Cambridge Studies in Linguistics, 9. Cambridge University Press, Cambridge. Dixon, R.M.W. (ed.) 1976. Grammatical Categories in Australian Languages. Australian Institute of Aboriginal Studies, Canberra; Humanities Press, New Jersey. DuBois, J.W. 1981. The Sacapultec Language. Unpublished Ph.D. dissertation, University of California, Berkeley. DuBois, J.W. 1985. Competing Motivations. In Haiman (ed.) 1985: 343–66. Dubrovina, Z.M. 1972. Infinitivy v finskom jazyke. Leningrad University Press, Leningrad. Ellis, Ch.D. 1971. Cree Verb Paradigms. In International Journal of American Linguistics, vol. 37, 2:76– 95.
< previous page
page_310
next page >
< previous page
page_311
next page >
Page 311 Esau, H. 1973. Nominalization and Complementation in Modern German. North-Holland Publishing Company. Amsterdam, London; American Elsevier Publishing Company, Inc., New York. Foley, W.A. and Van Valin, R.D., Jr 1984. Functional Syntax and Universal Grammar. Cambridge University Press, Cambridge, London, New York, New Rochelle, Melbourne, Sydney. Fortescue, M. 1984. West Greenlandic. Croom Helm Descriptive Grammars. Croom Helm, London, Sydney, Dover, New Hampshire. Gak, V.G. 1976. Nominalizacija skazuemogo i ustranenie ob”ekta. In Zolotova, G.A. (ed.), Sintaksis i stilistika: 85–102. Nauka, Moscow. Gamble, G. 1978. Wikchamni Grammar. University of California Publications in Linguistics, 89. University of California Press, Berkeley, Los Angeles, London. Gardiner, A. 1957. Egyptian Grammar. Oxford University Press, Oxford. Gary, J.O. and Gamal-Eldin, S. 1982. Cairene Egyptian Colloquial Arabic. Lingua Descriptive Studies, vol. 6. North-Holland Publishing Company, Amsterdam. Gasser, M. 1983. Topic Continuity in Written Amharic Narrative. In Givón, T. (ed.) Topic Continuity in Discourse: 95–140. John Benjamins Publishing Company, Amsterdam, Philadelphia. Gill, H.S. and Gleason, H.A., Jr 1963. A Reference Grammar of Panjabi. Hartford Studies in Linguistics, The Hartford Seminary Foundation, Hartford, Connectitut. Giorgi, A. and Longobardi, G. 1988. Typology and Noun Phrases. In Rivista di Linguistica, 1:115–59. Givón, T. 1975. Serial Verbs and Syntactic Change: Niger-Kongo. In Li, (ed.) 1975:47–112. Givón, T. 1980. The Binding Hierarchy and the Typology of Complements. In Studies in Language, vol. 4, 3:333–77. Givón, T. 1984. Syntax: a Functional-Typological Introduction, vol. 1. John Benjamins Publishing Company, Amsterdam, Philadelphia. Greenberg, J.H. 1963. Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements. In Greenberg, J. H. (ed.), 1963: Universals of Language: 73–113. MIT Press, Cambridge, Mass. Greenberg, J.H., Ferguson, Ch.A., and Moravcsik (eds) 1978. Universals of Human Language. Stanford University Press, Stanford.
< previous page
page_311
next page >
< previous page
page_312
next page >
Page 312 Grišina, N.M. 1979. Padežnye pokazateli i služebnye slova v strukture složnogo predloženija ketskogo jazyka. Unpublished dissertation, USSR Academy of Sciences, Institute of Linguistics, Leningrad. Groot, A.W.de 1957. Classification of the Uses of a Case Illustrated on the Genitive in Latin. In Lingua, vol. VI:8–66. Grosu, A. and Thompson, S.A. 1977. Constraints on the Distribution of NP Clauses. In Language, vol. 53, 1:104–51. Grønbech, K. and Krueger, J.R. 1955. An Introduction to Classical (Literary) Mongolian. Otto Harrassowitz, Wiesbaden. Guma, S.M. 1971. An Outline Structure of Southern Sotho. Shuter and Shooter (Ptd) Ltd, Pietermaritzburg. Hagman, R.S. 1977. Nama Hottentot Grammar. Indiana University Publications, Bloomington. Haiman, J. 1980. Hua: A Papuan Language of the Eastern Highlands of New Guinea. John Benjamins Publishing company, Amsterdam. Haiman, J. 1985. Natural Syntax. Cambridge University Press, Cambridge, London, New York, New Rochelle, Melbourne, Sydney. Haiman, J. (ed.) 1985. Iconicity in Syntax. John Benjamins Publishing Co., Amsterdam, Philadelphia. Haiman, J. and Thompson, S.A. (eds), 1988. Clause Combining in Grammar and Discourse. John Benjamins Publishing Co., Amsterdam, Philadelphia. Hakulinen, A. and Karlsson, F. 1979. Nykysuomen lauseoppia. Suomalaisen Kirjallisuuden Seura, Juväskylä. Haman, A. 1962. Lärobok i estniska. Bokförlaget Medborgarskolan, Uppsala. Harris, A.C. 1981. Georgian Syntax. Cambridge University Press, Cambridge, London, New York, New Rochelle, Melbourne, Sydney. Harris, A.C. 1984. Inversion as a Rule of Universal Grammar. In Perlmutter, D.M. and Rosen, C.G. (eds), Studies in Relational Grammar, vol. 2:259–91. The University of Chicago Press, Chicago and London. Harrison, S.P. 1976. Mokilese Reference Grammar. The University Press of Hawaii, Honolulu. Harrison, S.P. 1981. Recent Directions in Oceanic Linguistics: A Review of the Contributions to Studies in Pacific Languages and Cultures. In Oceanic Linguistics, vol. 20, 2:151–233.
< previous page
page_312
next page >
< previous page
page_313
next page >
Page 313 Haspelmath, M. 1989. From Purposive to Infinitive. A Universal Path of Grammaticization. In Folia Linguistica Historica, vol. 10, 1/2:287–310. Haspelmath, M. 1990. Some Typological Evidence Against Determiners as Noun Phrase Heads. Unpublished MS, Free University of Berlin. Hawkins, J.A. 1983. Word Order Universals. Academic Press, New York, London, Paris, San Diego, San Francisco, Sao Paulo, Sydney, Tokyo, Toronto. Healey, P. 1964. Teleéfoól Quotative Clauses. In Linguistic Circle of Canberra Occasional Papers, vol. 3:25–34. (Reprinted in Householder, F.W. (ed.), 1972. Syntactic Theory, 1. Structuralist. Penguin Books Ltd, Harmondsworth). Helbig, G. and Buscha, J. 1981. Deutsche Grammatik. VEB Verlag Enzyklopädie, Leipzig. Hewitt, B.G. 1979. Abkhaz. Lingua Descriptive Studies, vol. 2. North-Holland Publishing Company, Amsterdam. Hewitt, B.G. 1983. Review of Alice C.Harris, Georgian Syntax: A Study in Relational Grammar. In Lingua, 59:247–74. Hewitt, B.G. 1987. The Typology of Subordination in Georgian and Abkhaz. Mouton de Gruyter, Berlin, New York, Amsterdam. Hinds, J. 1986. Japanese. Croom Helm Descriptive Grammars: Croom Helm Ltd, London, Sydney, Dover, New Hampshire. Hoekstra, T. 1986. Deverbalization and Inheritance. In Linguistics, vol. 24, 3:549–84. Hopper, P.J. and Thompson, S.A. 1980. Transitivity in Grammar and Discourse. In Language, vol. 56:251–99. Hopper, P.J. and Thompson, S.A. 1984. The Discourse Basis for Lexical Categories in Universal Grammar. In Language, vol. 60, 4: 703–72. Hudak, T.J. 1987. Thai. In: Comrie (ed.) 1987:757–75. Hyman, L.M. 1975. On the Change from SOV to SVO: Evidence from Niger-Kongo. In Li, (ed.) 1975:113–47. Jackendoff, R. 1977. X-bar Syntax: A Study of Phrase Structure. Linguistic Inquiry Mongraph Series No. 2. MIT Press, Cambridge, Mass. Jaxontov, S.E. 1981. Svjaz’ meždu semantikoj predikatnoaktantnogo glagola i oformleniem ego aktanta v jazykax različnyx tipov. In Semantika i sintaksis konstrukcij s predikatnymi aktantami, USSR Academy of Sciences, Leningrad: 41–56. Jazyki Azii i Afriki, vol. 3, (Sanzheev, G.D., Kononov, A.N. and Vardul, I.F. (eds)) 1979. Nauka, Moscow.
< previous page
page_313
next page >
< previous page
page_314
next page >
Page 314 Jeffers, R.J. 1976. Typological Shift and Change in Complex Sentence Structure. In Steever, S.B. et al. (eds) 1976:136–49. Jespersen, O. 1924. The Philosophy of Grammar. George Allen & Unwin Ltd, London. Jokinen, K. 1991. On the Two Genitives in Finnish. In EuroTyp Working Papers (Programme in Language Typology, European Science Foundation), Series 7, no. 14. European Science Foundation, Strasbourg. Joseph, B.D. 1983. The Synchrony and Diachrony of the Balkan Infinitive. Cambridge University Press, Cambridge, New York, New Rochelle, Melbourne, Sydney. Joseph, B.D. and Philippaki-Warburton, I. 1987. Modern Greek. Croom Helm Descriptive Grammars. Croom Helm Ltd., London, Sydney, Dover, New Hampshire. Kalmár, I. 1979. The Antipassive and Grammatical Relations in Eskimo. In Plank, F. (ed.), Ergativity: Towards a Theory of Grammatical Relations. Academic Press, London: 117–43. Kapeliuk, O. 1980. Evolution de la phrase amharique: la nominalization du verbe. In Tubiana, J. (ed.), Modern Ethiopia from the Accession of Menilek II to the Present: 97–106. A.A.Balkema, Rotterdam. Karlsson, F. 1978. Finsk grammatik. Helsingin Liikekirjapaino Oy, Helsinki. Keenan, E.L. 1974. Generalizing the Notion of ‘Subject of’. In Chicago Linguistic Society, 10:298–309. Keenan, E.L. 1976. Towards a Universal Definition of Subject. In Li (ed.) 1976:303–33. Kibrik, A.E. 1977. Opyt strukturnogo opisanija arčinskogo jazyka. Moscow University Press, Moscow. Kibrik, A.E. 1985. Towards a Typology of Ergativity. In Nichols and Woodbury (eds) 1985:268–323. Kibrik, A.E. 1987. Constructions with Clause Actants in Daghestanian Languages. In Lingua, 71:133–78. Kim, Nam-Kil 1987. Korean. In Comrie (ed.) 1987:881–98. Kimball, J. 1973. Seven Principles of Surface Structure Parsing in Natural Language. In Cognition, 2.1:15–48. Kiparsky, P. and Kiparsky, C. 1970. Fact. In Bierwisch, M. and Heidolph, K.E. (eds), Progress in Linguistics: 147–73. Mouton, The Hague. Klimov, G.A. 1973. Očerk obščej teorii érgativnosti. Nauka, Moscow.
< previous page
page_314
next page >
< previous page
page_315
next page >
Page 315 Klimov, G.A. and Alekseev, M.E. 1980. Tipologija kavkazskix jazykov. Nauka, Moscow. Kölver, U. 1977. Nominalization and Lexicalization in Modern Newari. Arbeiten des Kölner UniversalienProjekts, 30. University of Cologne, Cologne. Koopman, H. 1984. The Syntax of Verbs. From Verb Movement Rules in the Kru Languages to Universal Grammar. Foris Publications, Dordrecht, Holland; Cinnaminson, U.S.A. Koptjevskaja, M. 1984. Sketch of the Tense-Mood-Aspect System of Tahitian. In Dahl, Ö. and KósDienes, D. (eds) 1984, Selected Working Papers from the Tense-Mood-Aspect Project: 22–8. University of Stockholm, Stockholm. Koptjevskaja-Tamm, M. forthcoming. Finiteness. In Asher, R.E. and Simpson, J.M.Y. (eds), Encyclopedia of Language and Linguistics. Pergamon Press, Oxford, and Aberdeen University Press , Aberdeen. Kozinskij, I. 1983. O kategorii ‘‘podležaščee” v russkom jazyke. In Predvaritel’nye publikacii problemnoj gruppy po éksperimental’noj i prikladnoj lingvistike, 156. Institut russkogo jazyka, Moscow. Krejnovič, E.A. 1979a. Nivxskij jazyk. In Jazyki Azii i Afriki, vol. 3:295–329. Krejnovič, E.A. 1979b. Ketskij jazyk. In Jazyki Azii i Afriki, vol. 3: 330–347. Kuipers, A.H. 1967. The Squamish Language. Mouton and Co., The Hague, Paris. Kurdoev, K.K. 1978. Grammatika kurdskogo jazyka. Nauka, Moscow. Kuznecova, A.I., Xelimskij, E.A., Gruškina, E.V. 1980. Očerki po sel’kupskomu jazyku. Moscow University Press, Moscow. Labatut, R. 1982. La phrase peule et ses transformations. Atelier national de reproduction des thèses, Université de Lille III. Langacker, R.W. 1987. Nouns and Verbs. In: Language, vol. 63, 1: 53–94. Langdon, M. 1970. A Grammar of Diegueño. The Mesa Grande Dialect. University of California Publications in Linguistics, 66. University of California Press, Berkeley, Los Angeles, London. Lasmane, V. 1981. Översikt över lettisk grammatik. Lettiska centralrådet i Sverige, Utbildningutskottet, Stockholm. Lazard, G. 1984. La distinction entre nom et verbe en morphologie et en syntaxe. In Modèles linguistiques, vol. 6:29–39. University of Lille Press, Lille.
< previous page
page_315
next page >
< previous page
page_316
next page >
Page 316 Lees, R.B. 1960. The Grammar of English Nominalizations. Mouton, The Hague. Lefebvre, C. and Muysken, P. 1988. Mixed Categories. Nominalizations in Quechua. Kluwer Academic Publishers, Dordrecht, Boston, London. Lehmann, Ch. 1982. Thoughts on Grammaticalization. Arbeiten des Kölner Universalien-Projekts, 48. University of Cologne, Cologne. Lehmann, Ch. 1984. Der Relativsats. Gunter Narr Verlag, Tübingen. Lehmann, Ch. 1985. Ergative and Active Traits in Latin. In Plank, F. (ed.), Relational Typology: 243–56. Mouton Publishers, Berlin, New York, Amsterdam. Lehmann, Ch. 1988. Towards a Typology of Clause Linkage. In Haiman and Thompson (eds) 1988:181– 225. Lehmann, W.P. 1973. A Structural Principle of Language and Its Implications. In Language, vol. 49, 1:47–66. Lehmann, W.P. 1975. A Discussion of Compound and Word Order. In Li, (ed.) 1975:149–62. Lehmann, W.P. (ed.) 1978. Syntactic Typology. The Harvester Press, Sussex. Lewis, G.L. 1967. Turkish Grammar. Oxford University Press, Oxford. Li, Ch.N. (ed.) 1975. Word Order and Word Order Change. University of Texas Press, Austin, London. Li, Ch.N. (ed.) 1976. Subject and Topic. Academic Press, Inc., New York, San Fransisco, London. Li, Ch.N. (ed.) 1977, Mechanisms of Syntactic Change. University of Texas Press, Austin, London. Li, Ch.N. and Thompson, S.A. 1976. Subject and Topic: A New Typology of Language. In Li (ed.) 1976:457–90. Lindenfeld, J. 1973. Yaqui Syntax. University of California Publications, Linguistics, 76. University of California Press, Berkeley, Los Angeles, London. Lord, C. 1976. Evidence for Syntactic Reanalysis: from Verb to Complementizer in Kwa. In: Steever et al. (eds) 1976:179–91. Lorimer, D.L.R. 1935. The Burushaski Language. Institute for Comparative Cultural Studies, Oslo. Lyons, Ch. 1985. A Possessive Parameter. In Sheffield Working Papers in Language and Linguistics, 2:98–104. Lyons, Ch. 1986. The Syntax of English Genitive Constructions. In Journal of Linguistics, 22:123–43.
< previous page
page_316
next page >
< previous page
page_317
next page >
Page 317 McCawley, J. 1968. The Role of Semantics in a Grammar. In Bach, E. and Harms, R. (eds), Universals in Linguistic Theory: 10–47. Holt, Rinehart and Winston, New York. Mackenzie, J.L. 1985. Nominalization and Valency Reduction. In Bolkenstein et al. (eds) 1985:29–47. Mackridge, P. 1985. The Modern Greek Language. Oxford University Press, Oxford. Majtinskaja, K.E. 1960. Vengerskij jazyk. USSR Academy of Sciences, Moscow. Mallinson, G. 1986. Rumanian. Croom Helm Descriptive Grammars. Croom Helm, London, Sydney, Dover, New Hampshire. Mallinson, G. and Blake, B.J. 1981. Language Typology. North Holland Publishing Company, Amsterdam, New York, Oxford. Martin, S.E. 1975. A Reference Grammar of Japanese. Yale University Press, New Haven, London. Martin, S.E. and Lee, Young-Sook C. 1969. Beginning Korean. Yale University Press, New Haven and London. Masica, C.P. 1976. Defining a Linguistic Area: South Asia. University of Chicago Press, Chicago. Maslova, E. 1989. Sootnošenie kommunikativnoj i sintaksičeskoj struktur v prostom predloženii jukagirskogo jazyka. Unpublished dissertation, USSR Academy of Sciences, Institute of Linguistics, Leningrad. Matisoff, J.A. 1972. Lahu Nominalization, Relativization and Genitivization. In Kimball, J.P. (ed.), Syntax and Semantics, vol. 1:237–58. Seminar Press Inc., New York. Mel’čuk, I.A. and Žolkovskij, A.K. 1970. Towards a Functioning ‘Meaning-Text’ Model of Language. In Linguistics, 57:10–47. Merlan, F. 1982. Mangarayi. Lingua Descriptive Studies, vol. 4. North-Holland Publishing Company, Amsterdam. Mitchell, T.F. 1962. Colloquial Arabic. The Living Language of Egypt. The English Universities Press, London. Mithun, M. 1984. The Evolution of Noun Incorporation. In Language, vol. 60, 4:847–94. Moyse-Faurie, C. 1984. L’Opposition verbo-nominale dans les langues d’Océanie. In: Modèles linguistiques, vol. 6, 1:107–24. University of Lille Press, Lille. Munro, P. 1982. On the Transitivity of ‘Say’ Verbs. In Hopper,P. J. and Thompson, S.A. (eds), Studies in Transitivity: 301–18. Academic Press, New York.
< previous page
page_317
next page >
< previous page
page_318
next page >
Page 318 Muysken, P. 1989. Predication Chains: Case and Argument Status in Quechua and Turkish. In Linguistic Inquiry, vol. 20, 4:627–45. N’Diaye, G. 1970. Structure du dialecte basque de Maya. Mouton, The Hague, Paris. Nedjalkov, I.V. 1983. Konstrukcii s predikatnymi aktantami v évenkijskom jazyke. In Xrakovskij (ed.) 1983:198–211. Nedjalkov, V.P. 1990. Osnovnye tipy deepričastij. In Birjulin, L.A., Vardul’, I.F., and V.S.Xrakovskij (eds), Tipologija i grammatika: 36–59. Nauka, Moscow. Nedjalkov, V.P., Inenlikej, P.I. and Raxtilin, V.G. 1983. Cukotskie konstrukcii s sub”ektnym infinitivom. In Xrakovskij (ed.) 1983: 221–34. Newman, P. 1987. Hausa and the Chadic Languages. In Comrie, (ed.) 1987:705–23. Nichols, J. 1986. Head-marking and Dependent-marking Grammar. In Language, vol. 62, 1:59–119. Nichols, J. 1988. Nominalization and Assertion in Scientific Russian Prose. In Haiman and Thompson (eds) 1988:399–428. Nichols, J. forthcoming. Ingush. A Grammatical Sketch. In Job, M. and Smeets, R. (eds), Indigenous Languages of the Caucasus. Nakho-Daghestanian Languages. Caravan Books, Delmar, N.Y. Nichols, J. and Woodbury, C. (eds) 1985. Grammar Inside and Outside the Clause. Cambridge University Press, Cambridge, London, New York, New Rochelle, Melbourne, Sydney. Nilsson, Barbro 1972. Old Russian Derived Nominals in -nie, -tie. Almqvist and Wiksell, Stockholm. Nilsson, Birgit 1985. Case Marking Semantics in Turkish. University of Stockholm, Stockholm. Noonan, M. 1985. Complementation. In Shopen (ed.) 1985, vol. 2: 42–140. Noreen, A. 1904. Vårt språk, 7 vols, vol. 5. C.W. K. Gleerups Publishing Company, Lund. Obolensky, S., Zelelie, D. and Andualem, M. 1964. Amharic. Foreign Service Institute, Washington, D.C. Padučeva, E.V. 1974. O semantike sintaksisa. Nauka, Moscow. Padučeva, E.V. 1977. O proizvodnyx diatezax otpredikatnyz imen v russkom jazyke. In Xrakovskij, V.S. (ed.), Problemy lingvističeskoj tipologii i struktury jazyka: 84–107. Nauka, Moskva. Padučeva, E.V. 1984. Pritjažatel’noe mestoimenie i problema zaloga otglagol’nogo imeni. In Civigor’ev, V.P. (ed), Problemy strukturnoj lingvistiki 1982: 50–66. Nauka, Moscow.
< previous page
page_318
next page >
< previous page
page_319
next page >
Page 319 Padučeva, E.V. 1985. Vyskazyvanie i ego sootnesennost’s dejstvitel’nost’ju. Nauka, Moscow. Payne, J. 1991. The Headedness of Noun-phrases: Slaying the Nominal Hydra. Paper presented at the Talking Heads round table, 21–2 March 1991. University of Surrey. Perkins, R.D. 1988. The Covariation of Culture and Grammar. In Hammond, M.T., Moravcsik, E.A. and Wirth, J.R. (eds), Studies in Syntactic Typology: 359–78. John Benjamins, Amsterdam. Pitkänen, A. 1979. Binominala genitiviska hypotagmer i yngre svenska. Skrifter utgivna av svenska litteratursällskapet i Finland, no 484, Doctoral dissertation, Helsinki. Pjurbeev, G.C. 1979. Tipy složnyx predloženij v mongol’skix jazykax. Nauka, Moscow. Platero, P.R. 1974. The Navajo Relative Clause. In International Journal of American Linguistics, vol. 40, 3:202–46. Platzack, Ch. 1977. Substantivsjuka—ett friskhetstecken? Lund Studies in Scandinavian Linguistics, Serie D, 12, Lund. Pokrovskaja, L.A. 1978. Sintaksis gagauzskogo jazyka v sravnitel’nom osveščenii. Nauka, Moscow. Polinskaja, M.S. and Nedjalkov, V.P. 1987. Contrasting the Absolutive in Chukchee. In Lingua, vol. 71:239–69. Press, M. 1980. Chemehuevi. A Grammar and Lexicon. University of California Publications in Linguistics, 92. University of California Press, Berkeley, Los Angeles, London. Ramstedt, G.J. 1968. A Korean Grammar. Anthropological Publications, Oosterhout N.B. The Netherlands. Ransom E.N. 1986. Complementation: Its Meanings and Forms. Typological Studies In Language, vol. 10. John Benjamins Publishing Company, Amsterdam, Philadelphia. Revzin, I.I. 1973. Transformacionnoe issledovanie konstrukcij s sub”ektnym i ob”ektnym priimennym dopolneniem (genitivus subjectivus i genitivus objectivus). In Zaliznjak, A.A. (ed.), Problemy grammatičeskogo modelirovanija: 88–96. Nauka, Moscow. Ritter, E.n.d. A Head Movement Approach to Construct State Noun Phrases. Unpublished MS. Massachusetts Institute of Technology. Roberts, J.R. 1987. Amele. Croom Helm Descriptive Grammars Series. Croom Helm, London, New York, Sydney. Rosen, S.T. 1989. Two Types of Noun Incorporation: A Lexical Analysis. In Language, vol. 65, 2:294– 317.
< previous page
page_319
next page >
< previous page
page_320
next page >
Page 320 Rothstein, R.A. 1974. Relevancy Marking in Polish Complements. In Brecht, R.D. and Chvany, C.V. (eds) Slavic Transformational Syntax: 53–66. The University of Michigan, Ann Arbor. Saltarelli, M. 1988. Basque. Croom Helm Descriptive Grammars Series. Croom Helm, London, New York, Sydney. Sanzheev, G.D., Kononov, A.N. and Vardul, I.F. (eds) 1979. Jazyki Azii i Afriki, vol. 3. Ščeglov, J.K. 1970. Očerk grammatiki jazyka xausa. Nauka, Moscow. Schachter, P. 1976. The Subject in Philippine Languages: Topic, Actor, Topic-actor, or None of the Above. In Li (ed.) 1976: 491–518. Schachter, P. 1985. Parts-of-speech Systems. In: Shopen (ed.) 1985, vol. 1:3–61. Schachter, P. and Otanes, F.T. 1972. Tagalog Reference Grammar. University of California Press, Berkeley. Schwartz, A. 1970. On Interpreting Nominalizations. In Bierwiesch, M. and Heidolph, K.E. (eds), Progress in Linguistics: 295–305. Mouton, The Hague. Schwyzer, E. 1950. Griechische Grammatik. C.H. Beck’sche Verlagsbuchhandlung, Munich. Seiler, H. 1983. Possessivity, Subject and Object. In Studies in Language, vol. 7, 1:90–116. Semantika i sintaksis konstrukcii s predikatnymi aktantami. 1981. USSR Academy of Sciences, Leningrad. Shibatani, M. 1977. Grammatical Relations and Surface Cases. In Language, vol. 53, 4:789–810. Shibatani, M. 1987. Japanese. In Comrie, B. (ed.) 1987:855–80. Shopen, T. (ed.) 1985. Language Typology and Syntactic Description. Center for Applied Linguistics. Cambridge University Press, Cambridge, New York, New Rochelle, Melbourne, Sydney. Silverstein, M. 1976. Hierarchy of Features and Ergativity. In Dixon (ed.) 1976:112–171. Simpson, J.H. 1983. Aspects of Warlpiri Morphology and Syntax. Dissertation. MIT, Cambridge, Mass. Skorik, P. 1961–77. Grammatika čukotskogo jazyka, 2 vols. Izdatel’stvo AN SSSR, Moscow (vol. 1, 1961), Nauka, Leningrad (vol. 2, 1977). Skribnik, E.K. 1983. Burjatskie konstrukcii s predikatnymi aktantami. In Xrakovskij (ed.) 1983:77–87. Skribnik, E.K. 1988. Polipredikativnye sintetičeskie predloženija v burjatskom jazyke. Nauka, Novosibirsk.
< previous page
page_320
next page >
< previous page
page_321
next page >
Page 321 Smyth, H.W. 1956. Greek Grammar. Harvard University Press, Cambridge, Mass. Stassen. L. 1985. Comparison and Universal Grammar. Basil Blackwell , Oxford. Steele, S. 1978. Word Order Variation: A Typological Study. In Greenberg et al. (eds), vol. 4:585–623. Steever, S.B. et al. (eds) 1976. Papers from the Parassession on Diachronic Syntax, Chicago Linguistic Society, Chicago. Stojanov, S. 1964. Gramatika na bylgarskija knižoven ezik. Nauka i izkustvo, Sofia. Svenska akademiens ordbok (Ordbok öfver svenska språket utgifven av Svenska Akademien. 1-a bandet) 1898. C.W.K. Gleerups Publishing Company. Lund. Tabasaranskie étjudy. 1982. Moscow University Press, Moscow. Tauli, V. 1966. Structural Tendencies in Uralic Languages. Indiana University Publications, Uralic & Altaic Series, vol. 17. Mouton and Co., London, The Hague, Paris. Taylor, Ch. 1985. Nkore-Kiga. Croom Helm Descriptive Grammars. Croom Helm, London, Sydney, Dover, New Hampshire. Tereščenko, N.M. 1973. Sintaksis samodijskix jazykov. Nauka, Leningrad. Thompson, S.A. and Longacre, R.E. 1985. Adverbial Clauses. In Shopen (ed.) 1985, vol. 2:171–233. Thurneysen, R. 1946. A Grammar of Old Irish. The Dublin Institute for Advanced Studies, Dublin. Topolińska, Z. 1981. Remarks on the Slavic Noun Phrase. Polish Academy of Sciences, Institute of the Polish Language, Warsaw. Tutescu, M. 1972. Le Groupe nominal et la nominalization en français moderne. Bucarest. Ubrjatova, E.I. and Litvin, F.A. (eds) 1984. Predikativnoe sklonenie pričastij v altajskix jazykax. Nauka, Novosibirsk. Ubrjatova, E.I. and Litvin, F.A. (eds) 1986. Strukturnye tipy sintetičeskix polipredikativnyx konstrukcij v jazykax raznyx semej. Nauka, Novosibirsk. Ultan, R. 1978. Towards a Typology of Substantival Possession. In Greenberg (ed.) 1978:11–49. Underbill, R. 1976. Turkish Grammar. MIT Press, Cambridge, Mass., London. Uspenskij, B.A. 1968. O sisteme ketskogo glagola. In Ivanov, V. Vs., Toporov, V.N. and Uspenskij, B.A. (eds), Ketskij sbornik: 196–228. Nauka, Moscow.
< previous page
page_321
next page >
< previous page
page_322
next page >
Page 322 Vamling, K. 1989. Complementation in Georgian. Doctoral dissertation. Lund University Press, Lund. Van Valin, R.D., Jr 1985. Case Marking and the Structure of the Lakhota Clause. In Nichols and Woodbury (eds) 1985:363–413. Vasil’ev, A.G. 1961. Infinitiv na -ki v sovremennom korejskom jazyke. In: Učenye zapiski LGU, 294. Leningrad Univeristy Press, Leningrad. Vasil’ev, A.G. 1963. Substantivirovannyj infinitiv na -m v korejskom jazyke. In: Filologija stran vostoka. Leningrad University Press, Leningrad. Večerka, R. 1963. Sintaksis bezpredložnogo roditel’nogo padeža v staroslavjanskom jazyke. In Kurz, J. (ed.), Issledovanijapo sintaksisu staroslavjanskogo jazyka: 183–223. Czechoslovak Academy of Sciences, Prague. Vendler, Z. 1967. Facts and Events. In Linguistics in Philosophy: 12–146. Ithaca, N.Y. Vendler, Z. 1968. Adjectives and Nominalizations. In Papers in Formal Linguistics. The Hague, Paris. Vendler, Z. 1970. Say What You Think. In: Cowan, J.L. (ed.), Studies in Thought and Language: 79–91. The University of Arizona Press, Tucson, Arizona. Vennemann, T. 1973. Explanation in Syntax. In Kimball, J.P. (ed.): Syntax and Semantics, vol. 2: 1–50. Seminar Press, New York. Verhaar, J.W.M. 1990. How Transitive is Intransitive? In Studies in Language, 14, 1:93–168. Veyrenc, J. 1972. Existe-t-il un génitiv de l’objet? In Bulletin de la Société de Linguistique de Paris, vol. 67, 1:215–38. Voegelin, C.F. and Voegelin, F.M. 1977. Classification and Index of the World’s Languages. Elsevier, New York, Oxford. Vonen, A.M. 1988. The Noun Phrase in Samoan and Tokelauan. Oslo University, Oslo. Walter, H. 1981. Studien zur Nomen-Verb-Distinktion aus Typologischer Sicht. Wilhelm Fink Verlag, Munich. Watahomigie, L.J., Bender, J. and Yatamoto, A.Y. 1982. Hualapai Reference Grammar. American Indian Studies Center, UCLA, Los Angeles. Weber, D.J. 1983. Relativization and Nominalized Clauses in Huallaga (Huánuco) Quechua. University of California Press, Berkeley, LA, London. Wellander, E. 1973. Riktig svenska. En handledning i svenska språkets vård. 4th edn. Scandinavian University Books, Stockholm.
< previous page
page_322
next page >
< previous page
page_323
next page >
Page 323 Westermann, D. 1930. A Study of the Ewe Language. Oxford University Press, London, New York, Toronto. Wise, H. 1975. A Transformational Grammar of Spoken Egyptian Arabic. Basil Black well, Oxford. Woodbury, A.C. 1985. Noun Phrase, Nominal Sentence, and Clause in Central Alaskan Yupik Eskimo. In Nichols and Woodbury (eds) 1985:61–88. Woodcock, E.C. 1959. A New Latin Syntax. Methuen and Co. Ltd, London. Wright, W. 1971. A Grammar of the Arabic Language. Cambridge University Press, Cambridge. Xrakovskij, V.S. (ed.) 1983. Kategorii glagola i struktura predloženija. Nauka, Leningrad. Živova, G.T. 1983. Konstrukcii s predikatnymi aktantami v ketskom jazyke. In Xrakovskij (ed.) 1983:129–37.
< previous page
page_323
next page >
< previous page
page_324
next page >
Page 324 Author index Abdel-Massih, E.T. 178 Abney, S. 65 Allen, S. 72 Amin-Madani, S. 159 Anderson, M. 206 Anderson, S.R. 233, 245, 264 Alekseev, M.E. 105, 204 Andrews, A. 89 Anward, J. 232, 304 Armbruster, Ch. H. 80 Asher, R.E. 28, 107 Awberry, G.M. 159 Bakaev, C.X. 189 Bakker, D. 83 Bell, A. 79, 82 Bergel’son, M.B. 107, 200 Berman, R. 4, 18–21, 126, 144–6, 304 Biggs, B. 141 Binnick, R. 4, 37, 127, 301 Blake, B.J. 250 Boëthius, H. 186 Bogatyrev, K.K. 100 Boguslavskaja, O. Ju. 100, 198 Bokarev, E.A. 107 Bybee, J. 246, 264, 266 Čeremisina, M.A. 29, 98 Chomsky, N. 3, 5, 65, 248, 251, 265 Chung, S. 125, 140–4, 157, 210–13, 233, 242, 243, 301, 304 Clark, R. 243, 301 Cole, P. 28, 50, 107 Comrie, B. 4–6, 8, 49, 50, 60, 136, 155, 166–8, 203, 205, 206, 216, 244, 245, 259 Corbett, G.G. 302 Dahl, Ö. 73, 79, 304 Davies, J. 52 Dayley, J.P. 40, 41, 85 Derbyshire, D.C. 156, 157 Derzhanski, I. 182 Dik, S.C. 229, 306 Diller, A. 184 Disterheft, D. 40–2, 223 Dixon, R.M. W. 11, 85 DuBois, J.W. 85, 257, 263 Dubrovina, Z.M. 35 Ellis, Ch. D. 31 Esau, H. 3 Evans, N. 43, 44 Foley, W.A. 22, 103, 258 Fortescue, M. 30, 52, 113–16 Fraurud, K. 123, 126 Gak, V.G. 270 Gamal-Eldin, S. 177 Gardiner, A. 202, 305 Gary, J.O. 177 Gil, D. 147 Giorgi, A. 65, 206–7 Givón, T. 186, 187, 248–52, 258 Greenberg, J. 248 Grišina, N.M. 51, 85 Grosu, A. 231
< previous page
page_324
next page >
< previous page
page_325
Page 325 Grønbech, K. 37, 103 Guma, S.M. 30 Hagman, R.S. 50 Haiman, J. 27, 28, 85, 258 Harris, A.C. 4, 130–2, 147, 154–5, 303–5 Harrison, S.P. 19, 27, 56, 242–4 Haspelmath, M. 42, 65, 301 Hawkins, J. 59, 88 Healey, P. 28 Hewitt, B.G. 28, 30, 31, 134–6, 150, 151, 303–5 Hoekstra, T. 4, 10, 11, 180 Hinds, J. 170, 171 Holmberg, A. 232 Hopper, P.J. 6, 12, 55, 73, 260, 261, 267, 270, 300 Hudak, T.J. 182 Hyman, L.M. 186, 187 Jackendoff, R. 65, 73 Jaxontov, S.E. 107, 245, 246 Jeffers, R.J. 233, 259, 306 Jespersen, O. 266, 267 Jokinen, K. 168, 169 Joseph, B.D. 29, 32, 151, 175 Juntunamalaga, P. 184 Kalmár, I. 58 Kapeliuk, O. 233, 267 Karlsson, F. 35, 43, 168 Keenan, E.L. 66, 67, 69–71, 74, 148, 208, 303 Kibrik, A.A. 200 Kibrik, A.E. 92–4, 99, 241 Kiefer, F. 13 Kim, N.-K. 90 Kimball, J. 232 Kiparsky, C. 17 Kiparsky, P. 17 Klimov, G.A. 105, 241–2 Koncevič, L.F. 37 Koopman, H. 167, 176, 305 Koptjevskaja (-Tamm), M. 29, 125 Kozinskij, I. 244 Kölver, U. 51 Krejnovič, E.A. 96, 97, 302 Krueger, J.R. 37 Kuipers, A.H. 85 Kuznecova, A.I. 34, 70, 110–13, 125, 302 Langacker, R.W. 253 Langdon, M. 51, 85 Lazard, G. 54–5 Lee, Y.-S. C. 90, 92 Lees, R. 3, 17 Lefebvre, C. 4, 104, 105 Lehmann, Ch. 15, 28, 103 Lehmann, W.P. 233, 246 Lewis, G.L. 46–8, 51, 125, 214 Li, Ch. N. 90 Lindenfeld, J. 85 Litvin, F.A. 38, 43–4, 98, 199, 200 Longacre, R.E. 301 Longobardi, G. 65, 206–7 Lord, C. 28 Lorimer, D.L.R. 95, 107, 204 Lutz, D. 159 Lyons, Ch. 71 McCawley, J. 27
next page >
Mackenzie, J.L. 15 Mackridge, P. 175 Magnusson, S. 31 Malkior, S. 127, 145, 147 Mallinson, G. 250 Martin, S.E. 90, 92, 170–2 Masica, C.P. 109, 223 Maslova, E. 89, 118, 119 Matisoff, J.A. 51 Mel’čuk, I.A. 269 Merlan, F. 80 Mitchell, T.F. 177 Moyse-Faurie, C. 54 Munro, P. 27 Muysken, P. 4, 104, 105, 107 N’Diaye, G. 105, 209 Nedjalkov, V.P. 14, 44, 82, 184–6, 204, 302 Nedjalkov, I.V. 125 Newman, P. 191, 192 Nichols, J. 7, 28, 88, 98, 100, 238, 240, 241, 267–9, 302 Nilsson, Barbro 223 Nilsson, Birgitt 37, 70, 126
< previous page
page_325
next page >
< previous page
page_326
next page >
Page 326 Noonan, M. 25–27, 30, 31, 33, 39, 41, 42, 267, 301, 304 Noreen, A. 208 Obolensky, S. 19 Otanes, F.T. 55, 119, 303 Padučeva, E.V. 14, 148, 198, 205, 206 Payne, J. 65, 305 Perkins, R.D. 83 Pflueger, S. 233 Philippaki-Warburton, I. 151, 175 Pjurbeev, G.C. 29 Platero, P.R. 85 Platzack, Ch. 269 Plungjan, V. 300 Pokrovskaja, L.A. 38, 39 Polinskaja, M.S. 194, 195 Press, M. 80 Ramstedt, G.J. 37, 91, 107 Raxilina, E. 300 Revzin, I.I. 201, 216 Ritter, E. 145 Roberts, J.R. 127 Rosen, S.T. 187–8 Rothstein, R.A. 14 Ruge, H. 50 Saltarelli, M. 105, 209 Schachter, P. 55, 119, 244, 303 Schwartz, A. 20 Schwyzer, E. 223 Seiler, H. 67–9, 214, 219 Simpson, J.H. 39, 40 Shibatani, M. 170 Skorik, P. 195 Skribnik, E.K. 29, 32, 43, 122, 301 Smyth, H.W. 223 Stassen, L. 5, 23, 24, 26, 51, 221, 252, 253 Steele, S. 241 Svantesson, J.-O. 37 Ščeglov, J.K. 14, 191, 193 Tauli, V. 39 Taylor, Ch. 123 Tereščenko, N.M. 123 Thompson, S.A. 4, 6, 8, 12, 49, 50, 55, 73, 90, 136, 155, 167, 205, 206, 260, 261, 270, 300–1 Topolińska, Z. 203 Ubrjatova, E.I. 38, 43, 44, 98, 199, 200 Ultan, R. 74 Underhill, R. 46–8 Uspenskij, B.A. 88 Van Valin, R.D., Jr. 22, 85, 103, 258 Vasil’ev, A.G. 37, 93, 94 Večerka, R. 216 Vendler, Z. 5, 17, 18, 153, 154, 199 Vennemann, T. 248, 250 Verhaar, J.W.M. 203, 300 Veyrenc, J. 138, 139, 153 Vonen, A.M. 301 Walter, H. 58 Watahomigie, L.J. 85 Wellander, E. 208 Westermann, D. 185, 186 Wise, H. 126, 178
Woodbury, A.C. 57, 58 Woodcock, E.C. 172, 178, 223 Wright, W. 42, 157 Živova, G.T. 88 Žolkovskij, A.K. 269
< previous page
page_326
next page >
< previous page
page_327
next page >
Page 327 Language index Pages illustrating ANCs are italicized. Abkhaz 28, 30, 32, 42, 60, 86, 88–9, 134–6, 150–1, 159, 161, 164, 234, 238, 241–2, 258, 263, 274, 289 Afro-Asiatic 87, 109, 121, 128, 179, 224, 227–8, 234 Agul 86, 103, 106, 122, 124, 126, 198, 216, 241, 244, 262, 274, 280, 283 Algonquian 31, 82 Altaic 36–7, 43, 46, 70, 86, 109, 121, 128, 180, 224, 226–7 Altay 44 Amele 124, 127, 128, 274, 283 Amharic 19, 81, 123, 127–8, 162–3, 232–3, 236, 240, 267, 270, 274, 283–4 Arabic, Classical 42, 84, 87, 100–2, 104, 106, 108–9, 124, 126, 128, 156–7, 159, 161, 163–4, 204, 215, 226, 229, 233, 238, 262, 266, 274, 280, 284, 289 Arabic, Egyptian 60, 81, 84, 87, 89, 101, 117, 124, 126–8, 177–80, 204, 215, 234, 236–8, 274, 284, 296 Archi 83, 92–4, 102, 106, 256, 274 Australian 39, 43, 80, 85 Austronesian 55, 87, 109, 128, 164, 224, 227, 228, 234 Aztec-Tanoan 89 Basque 81, 99, 102, 106, 209–10, 224, 227, 240, 242, 246 Balkan 38–9 Baltic Finnic 39, 86 Bantu 30, 68, 83, 121, 123, 126, 249 Bulgarian 32, 38, 81, 83, 126, 161, 182–3, 201–2, 214, 216, 259, 274, 289–90, 294, 297 Burmese 51 Burushaski 95–6, 99, 102, 106, 204, 241–2, 274 Buryat 29, 32, 84, 86, 122–4, 214, 236, 239, 246, 301 Celtic 233, 306 Chemehuevi 80 Chinese 24, 26, 32, 54, 300 Chol 245 Chukchee 13–4, 62, 81–2, 86, 187, 189, 194–6, 274 Cree 31–2 Czech 83, 205, 302 Diegueño 51, 85 Dutch 4, 11, 160–1, 173, 180, 183, 202, 215, 275, 290, 294, 297 Egyptian, Ancient 87, 89, 124, 159, 161, 202, 215, 238, 275, 290 English 11, 17–18, 20–1, 28, 45, 65–6, 71–3, 124, 126, 153–4, 161–2, 165–6, 173, 175, 206, 207, 215, 260, 264, 267–9, 275, 290 Eskimo 32, 57–8, 87, 224, 242
< previous page
page_327
next page >
< previous page
page_328
next page >
Page 328 Estonian 73, 81, 149, 150, 161, 173, 188–90, 215, 235, 275 Evenki 38, 81, 117–18, 124, 204, 206, 240, 246, 275, 285 Ewe 84, 185–7, 190, 275 Finnish 34–6, 39, 73, 168, 169, 173, 191, 237, 246, 275 French 6, 31, 71, 161, 173, 202, 207, 215, 235, 275, 291 Fula 124, 275, 285 Gagauz 38–9 Gbadi 84, 167, 172–3, 175–6, 275, 294 Georgian 4, 130–4, 147, 155, 160–1, 173, 175, 215, 240, 242 German 3, 27–8, 42, 71–3, 83, 154–5, 160–1, 166, 173, 201–2, 207, 215, 235, 275, 291, 295, 301 Germanic 83, 206–7 Gilyak see Nivkh Greek, Ancient 50, 172–3, 216, 223, 276, 295 Greek, Modern 38, 50, 81, 151, 161, 173, 175, 215–16, 276, 291, 295 Hausa 13, 14, 191–4, 208, 276 Haya 68 Hebrew 4, 18–21, 76, 116, 117, 124, 126, 144–6, 147, 161, 173, 175, 203, 216, 236, 237, 265, 276 Hixkaryana 156–8, 161, 238, 245, 255, 276 Hungarian 13, 14, 39, 81, 185, 189, 190, 193, 208, 215, 276, 298 Icelandic 73, 177, 179, 188, 190, 207, 215, 236, 276, 297, 298 Ifè 84, 120, 124, 186, 187, 276, 286 Ingush 28, 99, 100, 106, 242, 276, 281 Inuit 83, 122–3, 125, 162 Inuktitut 81, 83, 110, 276 Irish, Modern 39, 41 Irish, Old 161, 223, 276, 291–2, 304 Italian 83, 155, 207, 216, 276 Jacaltec 245 Japanese 169–73, 231, 276 Jinghpaw 51 Kabardian 72 Karaim 39 Kayardild 43, 44 Kobon 102, 103, 106, 276, 281 Korean 36–8, 51, 70, 90, 91, 93, 94, 102, 103, 106, 277 Kurdish 81, 159, 161, 189, 190, 215, 250, 277, 292, 298 Kwa 64, 84, 120, 185–7 Lahu 51 Lakhota 50 Lakk 72 Lapp 39 Latin 172, 173, 174, 178, 180, 223, 277, 296 Latvian 172, 173, 223, 277, 296 Luganda 83, 123, 249, 277 Macedonian 83, 203, 216, 259, 277 Malagasy 66 Mangarayi 80 Maori 124, 140–4, 157, 161, 167, 181, 183, 204, 211, 215, 243, 245, 277 Mayan 40, 85, 245 Micronesian 19, 27, 56 Mojave 231 Mokilese 19, 27, 56 Moldavian 38 Mongol 36 Mongol(ian) 29, 36, 83 Mongolian, Classical 27, 37, 38, 83, 99, 100, 103, 106, 214, 277, 281 Mongolian, Modern (Khalkha) 4, 27, 37, 38, 83, 103, 121, 123, 124, 127, 214, 236, 270, 277, 280 Montagnais 82 Nama Hottentot 50, 85
Navajo 85 Newari 51 Ngarluma245 Nivkh 96–7, 106, 277 Nubian 80
< previous page
page_328
next page >
< previous page
page_329
next page >
Page 329 Nupe 186, 187 Panjabi 159, 161, 234, 241, 277, 292 Persian 42, 81, 159, 161, 189–90, 216, 240, 250, 263, 277, 292, 298 Polish 14, 123–4, 161, 204, 216, 235, 259, 276, 286, 293, 296 Pukapukan 123–4, 157, 161, 211–12, 215, 278 Quechua 4, 14, 43, 83, 102–3, 189, 265 Quechua, Cuzco 83, 100–6, 121, 187, 190, 199, 215, 256, 278, 282, 298 Quechua, Huallaga (Huánuco) 83, 100–1, 106, 121, 256, 278, 282 Quechua, Imbabura 28, 31–2, 50, 83, 99, 101–4, 106, 187, 190, 215, 278, 282 Romance 206–7 Rumanian 38–9, 161, 278, 293 Russian 8–10, 13–14, 32–4, 36, 38–9, 41–2, 45–6, 72–5, 133–4, 138–9, 148–9, 152–3, 161, 166, 173, 197–9, 201–2, 205–6, 215, 244, 267–70 Samoan 157, 161, 173, 202, 212–13, 215, 278, 301 Selkup 34, 36, 110–13, 124, 278 Slavonic, Old Church 216, 223 Southern Sotho 30–2 Swahili 31–2, 81, 83, 123–4, 162–3, 240, 256, 278, 287 Swedish 71, 83, 176–7, 179, 189–90, 207–8, 215, 223, 236, 255, 265, 278, 299 Tabassaran 99–100, 105–6, 247, 256, 278, 282 Tagalog 55, 119–20, 123–4, 279, 287–8, 334 Tahitian 54, 124, 209–10, 213, 229, 279 Tamil 28, 31–2, 102–3, 106, 278, 283 Teleéfoól 28 Thai 124, 182–4, 215, 279 Tibetan 51 Tokelauan 279, 301 Tongan 99, 102, 106, 123, 125, 157, 160–1, 211, 215, 226, 229, 233, 242–3, 262, 279, 301 Tungus 36 Turkic 36, 38, 39 Turkish 6, 14–15, 37–8, 124, 126, 136–7, 279 Tuva 97–8, 100–2, 106, 122–4, 199–201, 215, 234, 239, 279 Uralic 39 Vata 84, 124, 279, 288 Wappo 27, 231, 301 Warlpiri 39, 40 Welsh 155, 159, 161, 163, 279, 293 West Greenlandic 29, 30, 31, 83, 113–16, 125, 279 Wikchamni 124, 279, 288 Yukagir 89, 118, 119, 124, 227, 279 Yuki 27 Yupik, Central Alaskan 87, 88 Zuni 60
< previous page
page_329
next page >
< previous page
page_330
next page >
Page 330 Subject index Numbers in bold refer to pages where terms are defined. act nominal 20 action nominal (AN) 5, 24, 33–8, 41, 42 et passim as opposed to clausal nominalization 49–52 as opposed to converbs 44–5 as opposed to infinitives 33–42 as opposed to converbs 44–5 intermediate nature of 6–7, 252 intransitive 11 problems in defining 5, 6, 49 transitive 11 action nominal construction (ANC) 3, 5 et passim uses of 257–8, 266–70; as complements 43, 44–6, 199, 246; discourse functions of ANCs 260–1, 267–9; in adverbial functions 43, 199–201; in relative clauses 43; in other functions 266, 269–70 internal syntax of 3–4, 8–9 et passim meaning of 15 (see also interpretation) type of 60–4 (see also nominalization pattern) typology of 4, 60–4 ANCs with highly referential/animate/pronominal dependents 97, 116, 123, 205–6 (see also split, referentiality/animacy) adjectives vs. adverbs in nominalizations 18, 33, 35–6, 94, 102, 106, 116, 123–4, 160–1, 180, 199 anti-passive ANCs 62, 63, 110–16, 129 Argument-Reducing type (ARG-RED) 62–3, 194–6 articles in nominalizations 18, 50, 203 aspect in nominalizations 33, 49, 93–4, 106, 123, 162 balancing 23, 24, 49 Binding Hierarchy 258 clause adverbial 23, 24, 43–4 (see also action nominal constructions, uses of) noun 24 (see also complement) ; balanced 24, 26, 31, 32; deranked 24, 26, 30–2; finite 26, 29, 38, 39; indicative 25, 27; non-finite 26, 29; paratactic 25, 26; s-like (sentence-like) 25–32; semantic type of 16; subjunctive 25, 27, 30, 31, 38, 39; type of 24 independent 23, 27, 29 et passim relative 23, 43, (see also Relative type)
< previous page
page_330
next page >
< previous page
page_331
next page >
Page 331 subordinate 23, et passim; functional type of 23 complement 24, 43–4, et passim (see also action nominal constructions, uses of ; (noun) clause) type of 25–6 complement-balancing (see languages, complement-balancing) complement-deranking (see languages, complement-deranking) complement-taking predicate 16, 17, 45–8 compounding 104, 185–9, 191, 248–9 converb 32, 35, 44, 45, 93, 199–201 coordination as opposed to subordination 22, 24, 26 of ANCs 103, 106, 124, 126, 151–3 dependent-marking 7, 59–60, 256–7, et passim deranking 23, 24, 49, 51 Double-Possessive subtype (DBL-POSS) 61, 63, 127, 165–76, 180, 198, 202, 204–5, 208, 215–17, 224, 234–5, et passim (see also Nominal type) embedding 23, 24, 51 Ergative-Possessive type (ERG-POSS) 61–3, 104, 126, 129–164, 198, 202–4, 207–8, 215–17, 222–9, 234–6, et passim (see also Sentential-Possessive and Oblique-Possessive subtypes) finiteness 29, 32, 256 finitivization 31–2 generalizations across ANC types 64, 247 genitive adjectival-genitive vs. determinergenitive 71, 73 and referentiality 70, 74, 76, 235–6 subjective vs. objective 138–40, 153–4, 165–6 genitivize 61, et passim gerund 24 grammaticalization of indirect quotations 28, 29 of infinitive 42 of converbs 199–201 grammatic relation 9, 11, 12 head-marking 7, 59, 98–102, 259, 236–40, 255–6, et passim incorporating type (INC) 61–3, 104–5, 164, 184–91, 203, 215–17, 224, 229, 233, et passim incorporation 104, 185, 187–9 (see also compounding) infinitive 6, 19, 25, 31, 45–6, et passim active vs. passive 40 as opposed to ANs 33–42 origin of 42 inflection vs. derivation 263–6 Interclausal Relations Hierarchy 258 interpretation (of noun clauses/nominalizations) abstract 21 act 20 action 5, 19, 20, 45 ambiguities in 203–9 concrete 21, 93, 105 event 5, 17–20, 45, 47, 198–9 fact 5, 17–20, 45–7, 198–9 manner 5, 18–20, 45, 48 object 20 process 5, 19, 20, 45, 198–9 proposition 5, 17, 18, 21, 28, 45–7, 199 result 5, 20
intra-lexemic syntactic variation (in nominalization patterns) 103–5, 126–7, 163, 214–17 (see also split)
< previous page
page_331
next page >
< previous page
page_332
next page >
Page 332 languages active 89 agglutinating vs. fusional 245–6 analytical 88 complement-balancing vs. complement-deranking 26, 29–32, 38, 45, 102, 123, 223, 226–8, 229–33, 258, et passim dependent-marking 240–1, et passim double-marking 238–9 head-marking 88, 237–8, et passim incorporating 188–9, 194 nominative-accusative vs. ergative 241–5, et passim split-marking 239–40 masdar 30, 36, 42 meaning (of noun clauses/nominalizations) 5, 16–21, 45–9 (see also interpretation) Mixed type (MIX) 61, 63, 181–4, 203, 215–17, 224, et passim mode nominal 19, 48 modelling 251–3 mood dependent 24, 29, 30 indicative 30 subjunctive 30 subordinate 30 superordinate 30 in ANCs 103, 115 motivations for ANC types 257–63 negation in nominalizations 51, 93, 103, 106, 116, 124, 126, 161–2 nominal inflection in nominalizations 5–6, 34–45, 50, 255, 259 Nominal type (NOMN) 61, 63, 165–80, 204, 206, 208, 215–17, 223–8, et passim (see also Double-Possessive and Possessive-Adnominal subtypes) nominalization 25, 33, et passim (see also action nominal construction) clausal 5, 49 from stative verbs 19 lexical 5, 49–52 ‘nominalization model’ 186–7, 248–50 nominalization pattern=action nominal construction, type of standard 11, 12 nouns vs. verbs 6, 54–8, 253 noun phrases vs. finite clauses 1, 53, 55–60, 65–76, 248–53 (see also syntactic means, sentential vs. nominal) Oblique-Incorporating subtype (OBL-INC) 63, 188, 190 (see also Incorporating type) participle 24, 25, 31–2, 42 as opposed to action nominals 14–15, 42–4 uses/meanings of 46–9 passive agentive 130–47 agentless 156–8 passive vs. ERG-POSS 138–56 passive ANCs 62, 63, 127, 129, 136–7, 143, 164 possession inalienable/inherent vs. alienable/establishing 68–70, 210, 214 Subordinate vs. Dominant 141, 167, 210–14, Possessive-Accusative type (POSS-ACC) 61–3, 103, 110–13, 116–29, 140–2, 145, 146, 164, 202–4, 208–9, 214–17, 223–8, 231, 235–6, et passim Possessive-Adnominal subtype (POSS-ADN) 61, 63, 127, 176–80, 207–8, 215–17, 224, 235–6, et passim (see also Nominal type) Possessive-Incorporating subtype (POSS-INC) 63, 184–8, 190, 215 (see also Incorporating type)
< previous page
page_332
next page >
< previous page
page_333
Page 333 possessor vs. subject and/or object 66–77, 235–6 quotation/quote direct 27, 28, 29, 127 indirect 27, 28, 29, 127 reduction of arguments=also valency reduction reflexivization and ANCs 102, 106, 124, 148–52 inter-clausal 28 Relational Grammar 154–8 Relative type (REL) 14, 61–3, 191–4, 208, 215, 224 sample 82–4, 89 Sentential type (SENT) 61, 63, 90–109, 126, 163, 198, 203–4, 208–9, 214–17, 223–33, 252–7, 262, et passim Sentential-Incorporating subtype (SENT-INC) 63, 187, 190, 215 (see also Incorporating type) Sentential-Possessive subtype (SENT-POSS) 61, 63, 157, 215–17 (see also ERG-POSS) split functional 199–201 lexical 197 referentiality/animacy 201–3 semantic 187–99 subordination 22–4, 26 marker of 23 syntactic means 7, 58–60, et passim (see also dependent-marking, head-marking, word order) sentential vs. nominal/NP-like 60–4, 254–7 hierarchy of 257 tense in nominalizations 18, 33, 49, 51, 93–4, 115, 123–4, 160–2, 255 thematic role 9–12 valency reduction=reduction of arguments 12–15, 62, 260–1 verb form dependent vs. non-dependent 23–4, 26–7, 29, 31 finite vs. non-finite 24, 26, 30–2, 36, 38, 43–4, et passim inflectional vs. derived 22, 23 (see also inflection vs. derivation) verbal noun 24, 29, 30, 33–40, et passim voice (see also passive, anti-passive) in ANCs 33, 103, 110, 123, 129 word class 6, 257 (see also nouns vs. verbs) word order 7, 59, 102, 153, 160, 175, 229–34, 248–50, et passim
< previous page
page_333