Himalayan Languages and Linguistics
Brill's Tibetan Studies Library Edited by
Henk Blezer Alex McKay Charles Ramble
Languages of the Greater Himalayan Region Edited by
George L. Van Driem
VOLUME 5/12
Himalayan Languages and Linguistics Studies in Phonology, Semantics, Morphology and Syntax
Edited by
Mark Turin and Bettina Zeisler
BRILL
LEIDEN • BOSTON 2011
This book is printed on acid-free paper. Library of Congress Cataloging-in-Publication Data
ISSN 1568-6183 ISBN 978 90 04 19448 9 Copyright 2011 by Koninklijk.e Brill NV, Leiden, The Netherlands. Koninklijk.e Brill NV incorporates the Imprints Brill, Hotei Publishing, IDC Publishers, Martinus Nijhoff Publishers and VSP. All rights reserved. No part of this publication may be reproduced. translated. stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission from the publisher. Authorization to photocopy items for internal or personal use is granted by Koninklijk.e Brill NV provided that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910, Danvers, MA 01923, USA Fees are subject to change.
CONIENTS Notes on contributors .........................................................................vii Introduction .......................................................................................... 1
Mark Turin and Bettina Zeisler PART ONE: The Himalayas in history Lost in the sands oftime somewhere north of the Bay of Bengal.. .... 13
George van Driem PART TWO: Phonology and script A key to four transcription systems ofLepcha ................................... 41
Heleen Plaisier Dialectal particularities of Sogpho Tibetan - an introduction to the ''Twenty-four villages' patois" ..................................................... 55
Hiroyuki Suzuki PART TIIREE: Semantics (words and word classes) On the Old Tibetan Term Khrin in the legal and ritual lexicons ........ 77
Brandon Dotson A functional analysis of adjectives in Newar ..................................... 99
Kazuyuki Kiryu PART FOUR: Morphology and syntax The role of animacy in the verbal morphology of Dongwang Tibetan .............................................................................................. 133
Ellen Bartee The Sampang verbal agreement system ........................................... 183
Rene Huysmans Ergativity in Kundal Shahi, Kashmiri and Hindko ........................... 219
Khawaja A. Rehman Kenhat, the dialects of Upper Ladakh and Zanskar .......................... 235
Bettina Zeisler
Vl
CONTENTS
Index ................................................................................................. 303
NOTES ON CONTRIBUTORS ELLEN BARTEE and her husband, Ken Hugoniot, work with SIL and live with their daughter in Shangrila, Yunnan. They are currently working on a project in cooperation with the Yunnan Minority Affairs Bureau to conduct a linguistic survey of the dialects spoken in Diqing Tibetan Autonomous Prefecture. BRANDON DOTSON is a Visiting Researcher at Ludwig-Maximilians-Universitat in Munich, where he leads the research project ''Kingship and Religion in Tibet." He is the author of The Old Tibetan Annals: An Annotated Translation of Tibet's First History (Vienna: Verlag der Osterreichischen Akademie der Wissenchaften, 2009~ GEORGE VAN DRIEM has directed the Himalayan Languages Project since 1993 and has occupied the chair for Language Documentation at Leiden University since 1999. He was with the Department of Comparative Linguistics from 1983 until its dissolution in 2004. He collaborates with colleagues in the Human Genetics department at Leiden and with foreign partners in the international research programme Languages and Genes of the Greater Himalayan Region.
RENE
HUYSMANS studied Italian linguistics at the University of Amsterdam (UvA), where he graduated in 1993. In 1995 he became a member of the Himalayan Languages Project and since then he has been working on Sampang, a Kiranti language spoken in eastern Nepal.
KAZUYUKI KIRYU trained in linguistics at Kobe University and is now Associate Professor at Mimasaka University, Japan. He works on the N ewar language and in 2002 compiled a reference grammar and glossary for an intensive Newar course at the Research Institute of Language and Culture in Asia and Africa. He also works on Meche, a variety of Bodo (Tibeto-Burman) spoken in south-east Nepal. HELEEN PLAISIER holds a PhD from Leiden University and is an expert on Lepcha language, literature and culture. She has published both a grammar of Lepcha and a catalogue of the world's largest collection ofLepcha manuscripts. She is currently completing an analytical edition of a native 19th-century Lepcha dictionary.
V111
NOTES ON CONTRIBUTORS
KHAWAJA A REHMAN teaches at the Post Graduate College Muzaffarabad and is a PhD candidate at Quaid-i-Azam University, Islamabad. He is currently researching Kundal Shahi, an endangered language spoken in the Neelam Valley of Pakistan-administered Kashmir, and is compiling a Kundal Shahi-Urdu-English dictionary. lllROYUKI SUZUKI holds a D.Litt from Kyoto University and is a postdoctoral research fellow at the Japan Society for the Promotion of Science and National Museum of Ethnology. His primary research interests are descriptive and historical linguistics of the languages spoken in the eastern Tibetan cultural area. MARK TURIN is a linguistic anthropologist and holds a PhD from Leiden University where he was a member of the Himalayan Languages Project. He has worked on Thangmi, a Tibeto-Burman language spoken in Nepal, and has supervised the Linguistic Survey of Sikkim. He is presently based at the University of Cambridge where he continues to direct the Digital Himalaya Project BETTINA ZEISLER holds a Dr.Phil. degree from the Free University Berlin. After a seven-year postdoctoral research project in the Collaborative Research Centre 441 'Linguistic data structures' at the University of Tiibingen she is now directing a new project on verb semantics and valency: the Valency Dictionary ofLadakhi Verbs. Her research interests focus on the grammar and history of Tibetan languages, from the th century Old Tibetan to contemporary Ladakhi.
INTRODUCTION MARK TuRIN AND BETIINA ZEISLER
The greater Himalayan region is the principal meeting point for the two largest language families of the world, Indo-European and TibetoBunnan. The same massifs have also been home to two smaller language families (Austroasiatic and Dravidian), and to two language isolates (Burushaski and Kusunda). Despite their physical prominence, the Himalayas constitute not so much an insurmountable barrier but rather a region of interaction between these various language families. Indo-Iranian languages, and perhaps also Burushaski, were in all probability spoken along the northern slopes of the western Himalayas before the spread of Tibetan to this area in the second half of the frrst millennium CE. Much earlier still, Tibeto-Bunnan languages established themselves on the southern slopes, while the speakers of IndoAryan and Iranian languages entered these mountainous regions along the river valleys from the south and, then again, from the northwest Whether such incursions were motivated by conquest or were rather of a more peaceful nature, the resulting linguistic situation was usually characterised by coexistence, with specific niches supporting the survival of specific languages. In those instances where speech forms were abandoned, and where a shift to a dominant language occurred, as in the case of Balti and Shamskat Ladakhi (from Eastern Iranian and/or Dardic languages to Tibetan), the transition took much longer than allowed for by the standard three-generation model for languages under extreme pressure. Multilingualism was surely the norm rather than the exception in many parts of the Himalayas, and this linguistic pluralism left traces in the languages involved. For example, a significant proportion of the lexicon of Newar, a Tibeto-Bunnan language spoken in Nepal, derives from Indo-Aryan; while Ladakhi, spoken in Jammu & Kashmir (India), and Balti spoken in Pakistan, have not only acquired and retained certain traits from Dardic languages, but have even influenced the latter. The geopolitical shifts brought about by the British colonisation of the Indian subcontinent, and the subsequent introduction of Englishmedium education together with the spread of modern mass communi-
2
MARK TURIN & BETIINA ZEISLER
cation, helped Indo-European (whether in the form of Hindi, Urdu or English) to gain ground, furthering the decline of many regional and local languages. While the diversity of the sub-groupings within Tibeto-Burman is on a par with that of the Germanic or Romance languages of Europe, (even if their speaker numbers are relatively low), this richness in variation and the opportunity for original, fimdamental research has not yet translated into a profusion of linguists working in Himalayan areas. Furthermore, while the Himalayan region is the nucleus for languages of the Tibeto-Burman family, and also the area of their greatest diversity, this mountainous territory remains somewhat peripheral for scholars working on other language families. This is reflected in the interests of the participants of the Himalayan Languages Symposia held to date, with the majority of researchers focussing on Tibeto-Burman languages. This collection, too, originating at the 11th Himalayan Languages Symposium, held at Chulalongkom University in Bangkok, Thailand, in December, 2005, brings together a number of papers on the sub-groupings of Tibeto-Burman, such as Kiranti, Lepcha, Newar and Tibetan, one paper on Indo-Aryan, and another which addresses data from Austroasiatic. The present volume contains a selection of six of the best papers frrst presented at the Bangkok meeting, albeit fimdamentally revised and significantly updated with more recent field data, and all situated in a wider conceptual and comparative context. Three additional papers were invited on the basis of topical interest and new research fmdings coming to light (van Driem, Huysmans and Plaisier). Each of these nine contributions highlight the results of primary research, whether from the field or based on comparative material. THE HIMALAYAS IN HISTORY
Focussing on the eastern Himalayas, George van Driem provides a helpful introduction to the background context and macro-history of two Himalayan language families, Austroasiatic and Tibeto-Burman. In so doing, he draws on archaeology, population genetics, and linguistic reconstruction, and what may be termed linguistic paleaontology (the attempt to understand the material culture of a language family on the basis of its inherited lexicon). According to van Driem, these two language families hold the key to understanding the population prehistory of northeastern India and the Indo-Burmese borderlands.
INTRODUCTION
3
The original Tibeto-Bunnan homeland nominates itself on accmmt of the density of its speakers in the northeastern segment of the Indian subcontinent. The location of the ancestral Austroasiatic homeland, however, lying somewhere between South Asia and Southeast Asia, in the area around the northern coast of the Bay of Bengal, may be posited on the basis of linguistic palaeontology, the distribution of modem Austroasiatic language communities and the deep phylogenetic divisions in the family tree. Linguistic palaeontology further identifies the ancient speakers of Austroasiatic as likely being the earliest cultivators of rice, since a rich repertoire of reconstructible roots representing ancient rice agriculture is robustly reflected in all branches of this language family. PHONOLOGY AND SCRIPT
Closer to the ground, the second section opens with a discussion of how spoken and written languages are best transcribed and transliterated. Due to India's colonial heritage and the wide reach of English medium education, many speakers of languages indigenous to the subcontinent rely on conventions of English orthography when giving their tongues written form in Roman script. On occasion, this can result in arbitrary transcriptions signalling contrasts where there are none. A standardised system of transcription, then, based on linguistic criteria and understanding, would be beneficial for both native speakers and concerned fieldworkers. Heleen Plaisier discusses these issues in detail for the case of Lepcha (spoken in Sikkim and K.alimpong), carefully evaluating three established orthographic systems (two introduced by Western authors, the other by a native scholar), each of which have different shortcomings, and contrasts them with her own more systematic approach. Tibetan languages and dialects vary considerably in the degree to which they preserve the original phonological complexity documented in Old or Classical Tibetan orthography. Initial clusters, for example, are best preserved in West Tibetan (Balti and Ladakhi), followed by the Eastern Tibetan Amdo dialects spoken in nomadic regions. The non-nomadic Amdo dialects, as well as the Khams Tibetan dialects, typically occupy an intermediate position between the phonologically conservative dialects and the phonologically most innovative Central Tibetan dialects, where practically all clusters are reduced to a single
4
MARK TURIN & BETIINA ZEISLER
consonant in word-initial position. Such rough classifications will need further qualification and nuancing, however, as more hitherto undescribed Tibetan dialects are documented and analysed. While Sogpho Tibetan, as described by Hiroyuki Suzuki in this volume, is classified as a Khams Tibetan dialect by virtue of its location (it is spoken in Danba County in western Sichuan), it nevertheless shows several conservative features which are otherwise only recorded among the phonologically conservative Amdo dialects. Examples include preaspiration resulting from former clusters with dental and velar preinitials, the retention of clusters with labial stop initials as prelabialised consonants and the partial retention of clusters with a labial initial followed by a palatal glide or alveolar trill. In most Tibetan dialects, these clusters have developed into palatal or alveopalatal affricates and retroflex stops or affricates. Sogpho and the nomadic Amdo dialects are examples of an intermediate stage: while the initial remains unaffected or simply weakened, the post-initial glide or trill undergoes the major change, and the fmal stage is reached only by dropping the initial labial. Such fmdings are extremely important for our understanding of the mechanics and history of sound changes in Tibetan, and this paper is therefore an original and important contribution to this end. SEMANTICS (WORDS AND WORD CLASSES)
Following a more traditional philological approach, Brandon Dotson offers a precise study of an item of the Old Tibetan legal and ritual lexicon: khrin. Philology and linguistics diverged from one another long ago, and while pertinent linguistic issues are almost entirely neglected in Tibetan studies, most modern linguists lack the experience and skills to make sense of Tibetan texts, which are often opaque even to the specialist The rare vocabulary and lack of lexical resources in Old Tibetan add to the difficulties of decoding the syntax of a language that is as yet imperfectly understood. Dotson's contribution is an important step in the re-integration of linguistics with philology, in that he demonstrates the context-sensitivity of lexical items, particularly when used in a technical context As a legal term, khrin should be translated as 'judicial punishment', even though the most basic meaning of the word khrin is 'lead' or 'tether', an instrument for leading (sacrificial) animals along as well as the recipients of legal punishment The noun khrin 'tether' can be
INTRODUCTION
5
analysed as a derivation of the verb 'khrid in Dotson's transliteration 'lead along', with an archaic, 1mproductive derivational sufftx (-d/ -n) added to the root *khri. Beyond the ideological reshaping of concepts and meanings, Dotson's paper also offers insights into the daily practice of ancient Tibetan bureaucrats, who, 1mder certain circumstances at least, decided legal cases by rolling dice. The definition of parts of speech has long been the awkward stepchild of Tibeto-Burman studies. Word classes are typically established rather loosely according to the intuition of mostly non-native scholars, following in broad strokes the classifications fo1md in dominant European languages of which they are native speakers. Adjectivals, namely words describing qualia, constitute a particularly intricate class, oscillating between their more state-like properties, which are linked with nominals, and their more dynamic properties, which are linked with verbs. In some languages, such as Tibetan, adjectivals are split between basic verbal forms used as predicates and derived nominal forms used attributively. Newar, spoken in the Kathmandu valley of Nepal, exhibits an even more complex picture. While some researchers of Newar have denied the existence of a separate class of adjectives altogether, Kazuyuki Kiryu argues for the existence of a distinct word class and, based on morphosyntactic criteria, proposes that Newar adjectives be divided into three subclasses. Discussing thirteen conceptual domains usually associated with adjectivals, Kiryu identifies nine concepts that are realised as adjectives. He further distinguishes between verb-like adjectives, non-verb-like adjectives, and non-predicative adjectives. MORPHOLOGY AND SYNTAX
In many Tibeto-Burman languages, and among the Tibetan languages in particular, the distinction between conjunct information related to an (intentionally acting or controlling) 'informant' (that is, speaker in assertions, addressee in questions) and disjunct information not related to such an informant has become grammaticalised along with the concept of evidentiality, namely the distinction between varying sources of knowledge. Ellen Bartee reports on the Southern Khams Tibetan dialect of Dongwang, spoken in Shangri-la Co1mty in north-west Yunnan, which, like most other modem Tibetan varieties, shows a conjWlct/disjWlct system besides evidential marking. Unlike all other Tibetan dia-
6
MARK TURIN & BETIINA ZEISLER
lects, however, Dongwang, and possibly other Southern Khams dialects also, displays an animacy distinction in existential linking verbs and auxiliaries based on these very verbs. This distinction operates on S and 0 (more precisely possessed) arguments and interacts with the conjunct/disjunct system. Having discussed all possible combinations of conjunct/disjunct and animacy marking and their uses in various non-existential clauses, Bartee investigates the possible origins of the animacy distinction. A similar animacy distinction is found in Lolo-Burmese languages, such as Naxi, and since the speakers of Dongwang Tibetan have been in contact with the Naxi people for several centuries, the innovation in Dongwang most likely results from intensive linguistic contact Bartee briefly outlines the history of interaction between the two communities, and also provides a short description of the animacy distinction in Naxi. According to her analysis, the combination of the Dongwang conjunct/disjunct opposition with the animacy distinction has resulted in something akin to a cross-referencing agreement system, thus distinguishing itself from other evidential systems in modern Tibetan. Nepal's Kiranti languages are known for their intricate verbal agreement systems, perhaps the most complex in Tibeto-Burman. In his paper, Rene Huysmans provides the frrst published analysis of the verbal agreement system of Sampang, a Kiranti language spoken in eastern Nepal. While Sampang intransitive verbs agree with their only argument, transitive verbs show agreement both with the agent, i.e. transitive subject, and the non-agent, i.e. the undergoer, recipient, patient, or beneficiary. Singular, dual, and plural number is marked on all actants, and non-singular frrst person actants are further distinguished for inclusion or non-inclusion of the addressee. The Sampang verbal string is home to eleven afftxal slots (one preftxal, the rest sufftxal) for cross-referencing, tense marking, negation, and for additional copy morphemes. Many of the morphemes in question have several allomorphs, including zero-allomorphs, and, to complicate the picture further, several morphemes have fused into portmanteau morphemes. As a side effect, there is even more complex marking for combinations of frrst person patient with third person agents in past tense, reflecting a higher level of semantic transitivity. After laying bare the elaborate verbal morphology of Sampang, and having discussed several analytical alternatives, Huysmans demonstrates that a large portion of its morphemic inventory can be re-
INTRODUCTION
7
lated in a straightforward manner to the Proto-Kiranti verbal agreement model. Sampang verbal motphology is particularly reminiscent of the verbal agreement system of Kulung, but also shows similarities to the verbal morphology ofLimbu and Lohorung. In short, Kiranti morphological change can be seen to be primarily semantically driven: while the formal systems are indices of meaning, their re-analysis and re-modelling are propelled by ever re-interpreted shades of meaning. While the agreement patterns of Indo-Aryan languages are certainly less complicated than those of their Tibeto-Burman counterparts, there are variations in whether a verb agrees only with the least agent-like argument (always in the absolutive/nominative) or also with further arguments. In his contribution, Khawaja A Rehman discusses three different languages spoken in the Neelam Valley in Pakistanadministered Kashmir: his own native language Kashmir~ the dominant language Hindko and the highly endangered Kundal Shahi. All three languages show full agreement with the least agent-like argument, while Kashmiri, as spoken in the Neelam valley, displays additional agreement in person and number with third person plural and second person agent-like arguments. In contrast to other Indo-Aryan languages, where intransitive verbs may display alternations between absolutive and ergative matking, Neelam Kashmiri further shows obligatory ergative marking for a set of more agentive or volitional one-place verbs, including reflexives such as 'bathe', sound production verbs such as 'laugh', and motion verbs such as 'dance'. Interestingly, most of these verbs do not display gender-based agreement, that is, they only exhibit masculine agreement markers. Dialect classifications are often based on the most evident differences in phonology and lexicon. With respect to the Tibetan varieties, Bettina Zeisler shows that such an approach may be misleading. In the case of the two dialect groups spoken in Upper and Lower Ladakh (Jammu & Kashmir, India), Kenhat and Shamskat, the essential difference is not between the dialects that preserve a conservative phonology with initial clusters, close to 'Jh century Old Tibetan, and those that have lost all clusters, but rather between the dialects that distinguish morphologically between agents and possessors, and those that do not. Other differences at the level of syntax include various ways of marking tense and evidentiality. In terms of phonology, the Kenhat dialects are particularly intriguing as their minor differences provide a model for the progressive stages
8
MARK TURIN & BETIINA ZEISLER
of the loss of initial consonant clusters, in the vocalisation of the fmal sibilant and in the trans-phonologisation from voice to register tone distinctions. More importantly, even though their phonology is less conservative than that of the Shamskat dialects, Kenhat dialects show more lexical and grammatical archaisms, allowing us to trace the morphological developments that both dialect groups, along with other Tibetan varieties, must have undergone. This is an important lesson in itself, since historical reconstruction in general, and that of the Tibeto-Burman languages in particular, tends to be based on somewhat superficial aspects of phonology and neglects the development of grammatical systems. The phonologically innovative Central Tibetan varieties, to date largely disregarded in attempts at reconstruction, may be as rich in lexical and syntactic archaisms as the Kenhat dialects of Upper Ladakh, while the phonologically most conservative dialects might not necessarily represent the oldest layers of the language. Where van Driem focuses on the most southeasterly section of the Himalayas and their prolongation into the Bay of Bengal, Rehman and Zeisler focus on the extreme north-west and its extension into the Hindu Kush. Towards the center of these geographical extremes, Suzuki and Bartee lead us to the northern and southern regions of Tibet, Plaisier to Sikkim, and Huysmans and Kiryu to eastern and central Nepal respectively. Between the two poles of macro-history, as represented by van Driem's overview of tens of thousands ofyears of language development and population migration, and micro-history, as embodied in Dotson's careful reconstruction of the semantics of a specific lexical item used in the bureaucratic jargon of the Tibetan Empire around one thousand years ago, we fmd Bartee's description of the last one thousand years of interaction between two communities speaking TibetoBurman languages, and Zeisler's analysis of the Upper Ladakhi varieties, offering us a glimpse into the prehistory of Tibetan speech forms. This volume has been a long time in the making. Five years have elapsed since the initial conference that brought many of the contributors together and the fmal publication of this edited book. While the delays are regrettable, and remain our responsibility alone, we believe that the collection has matured in the oak and is the better for the wait. In our capacity as editors, we wish to thank Dr Krisadawan Hongladarom, a linguist at Chulalongkorn University and founder of the impressive Thousand Stars Foundation that promotes understanding of
INTRODUCTION
9
Tibetan religion and culture, both for convening the conference and for her support in the early stages of this volume's production. In addition, we are very grateful to the students and staff of Chulalongkom University for helping to make the conference a success; to the former Collaborative Research Centre 441 at the University of Tiibingen for organisational and fmancial assistance; to the anonymous reviewer of the manuscript for such careful reading and helpful comments; to Patricia Radder and the team at Brill for their responsiveness and interest; to Eleanor Wilkinson for her assistance with preparing the index and to all of our contributors for their considerable patience and good cheer.
PART ONE: THE HIMALAYAS IN HISTORY
LOST IN THE SANDS OF TIME SOMEWHERE NOR1H OF THE BAY OF BENGAL GEORGE VAN DRIEM
In the northeastern portion of the Indian subcontinent, the presence of two language families stretches back into prehistory. What light can historical linguistics, linguistic palaeontology, archaeology, palaeoethnobotany and human population genetics shed on the ancient origins of Austroasiatic and Tibeto-Bunnan? Which new questions arise from these interdisciplinary insights?
1. OLD AND NEW LINGUISTIC PHYLA IN THE EASTERN HIMALAYAN REGION
The world's two most populous families of languages meet in the Himalayas. These are the Tibeto-Bunnan phylum, which includes Cantonese and Mandarin Chinese, and Indo-European, to which languages such as English and Bengali belong. In addition to these two great linguistic phyla, Kra-Dai alias Daic, Austroasiatic and Dravidian language communities skirt the eastern Himalayan region. For example, the Austroasiatic language Khasi is spoken in the Meghalaya. Ahom, a now extinct Kra-Dai tongue, was once a prominent cultural language in northeastern India, where scattered Daic language communities are still settled today. The Dravidian tongues Dhangar and Jhangar, dialects of Kurukh or Uraon, are spoken in Nepal's eastern Terai, and Kurukh is also spoken in scattered communities throughout northeastern India by people displaced by British colonial policies in India involving resettlement, not unlike the policy of transmigrasi in the former Dutch East Indies. Yet the Himalayas would appear to be peripheral to our understanding of the prehistory of Dravidian. The crux to the ethnolinguistic prehistory of the eastern Himalayan region are the language families Tibeto-Burman, Austroasiatic, Indo-European and, to a lesser extent, Kra-Dai.
14
GEORGE VAN DRIEM
Tamangic
~o ~
Sinitic
GG 8
0 O Q
Qiangic
sa. ""Raute~epangic D•ganm8~EV~ ~~
Magaric
Rl!i•-
Midzhuim
Dura
Nwgim
~ ~
E:2G~8~
~ ~ ~ e~l';:'\
~e~eeE;8 Figure 1. The fallen leaves diagram for Tibeto-Bmman represents a relatively agnostic view of the internal phylogeny of the linguistic phylum. Some subgroups are wellestablished, whilst others are less so. Brahmaputran may include Kachinic and Dhimalish. For the sake of argwnent, this diagram breaks up the administrative catch-all 'Qiangic' into the Ersfi cluster and a tnmcated 'Qiangic', not to posit a robust phylogenetic hypothesis but with the intent of presenting a challenge, by emphasising the crucial work in this area that has still been left undone. The precise phylogenetic relationships between Erg5ng, Qiang, Mi-fiag (Milyii), Tangut, the diverse rGyal-rong languages, Ersu, Lfisu, Tosu (Duoxu), Namuyi, ShOOng, Guiqi6ng, Choyo (Quey\1), Zhabit and Prinmi (Pfurn") have yet to be demonstrated What is sometimes called 'Northern Qiangic' is supposed to include the rGyal-rongic group recognised by Jackson Sun (Sfin Tiaroon) and Huang BUfan Hopefully scholars working on the TibetoBurman languages of Sichuan and YUnnan will in the coming years shed light on the structure of this portion of the Tibeto-Burman family tree (van Driem 2001, 2006).
The advent of Indo-European languages and Kra-Dai languages to the northeastern portion of the Indian subcontinent is relatively recent and to some extent historically attested. Yet whilst the impact of Kra-Dai today is marginal, the intrusion of Indo-European into the region has been more robust and is represented by expansive languages such as Assamese, Hindi, Bengali and English. At the same time, historical,
LOST IN 1HE SANDS OF liME
15
ethnographic, archaeological, anthropological and linguiltic data all point to an Austroasiatic and Tibeto-Bmman presence in northeastern India which must date back to some hoary period of ethnolinguistic prehistory. Both families, Amtroasiatic and Tibeto-Bmman, therefore hold the key to understanding the population prehistory of northeastern India and the Indo-Bunnese borderlands.
Figure 2. Geographical distribution of the majoc branches of the Tibeto-Bunnan language family. Each damond rep-esents not a language, but a ml!ior subgroup.
The deeper phylogeny of the Indo-Emopean language family was once conceived by August Schleicher as a branching oak tree, but the phylmn has increasingly assmned a rake-like appeamnce in more recent literatme. The cmrently best-informed family tree structure of the Tibeto-Bmman phylmn is likewise essentially rake-like, a situation for which I introduced the metaphor of fallen leaves (Figme 1). The geographical distribution of the major Tibeto-Bmman subgroups saliently shows the densest concentration in the northeastern portion of the Indian subcontinent, with most major subgroups being exclusive to this area (Figme 2). The most authoritative phylogenetic tree for Austroasiatic at present is the model developed by Gerard Diffi.oth (Figme 3). Some novel insights into the phylogeny of Kra-Dai have been put forward by Edmondson and Solnit (1988, 1997) and Ostapirat (2005).
16
GEORGE VAN DRIEM JOOOAD
OAD
JOOOBC
I
I
I
Korlru Kherw arian Kharia-Juang Koraput Khasian Pakan ic Eastern Palaungic Western Palaungic Khmu ic Vietic Eastern Katuic Western Katuic Western Bahnari Northwestern Bahnaric Northern Bahnaric Central Bahnaric Southern Bahnaric Khmeric Peari.c Monic Nerthern Asli Senoic Southern Asli Nicob arese
ZOOOBC
3000BC
I
4000BC
I
I
5000BC
I
Monda
I
I Khasi-Khmuic
I Vieto-Katuic I I
r- -
Khmero-Vietic
Mon-Khmer
KhmeroBahnaric
-
--
Asli-Monic
J-r----1
Nico-Monic
Figure 3. Austroasiatic with Gerard Diflloth's tentative calibration of time depths for the various branches of the language family (modified from Diflloth 2001, 2005). The precise phylogenetic propinquity of Pearic, after Khmeric loan layers have been stripped off, remains Wicertain except that Diflloth observes that Pearic is MonKbmer and not 'Wle espece de vieux khmer', as some scholars once maintained. This diagram arranges in a tree-shaped phylogeny the fourteen recognised branches of Austroasiatic, i.e. North MWida, South MWida, Khasian, Pakanic, PalalDlgic, Khmuic, Vietic, Katuic, Bahnaric, Khmeric, Pearic, Monic, Astian and Nicobarese.
The location of the Austroasiatic ancestral homeland can be argued from a purely linguistic point of view principally on the basis of linguistic palaeontology and on the basis of the geographical centre of gravity of the family based on the distribution of modem Austroasiatic language communities and the deep phylogenetic divisions in the family tree. The distribution of the modem language communities and the geography of the deepest historical divisions in the family's linguistic phylogeny would put the geographical centre of the family somewhere between South Asia and Southeast Asia, in the area around the north-
LOST IN THE SANDS OF TIME
17
em coast of the Bay of Bengal. Whether we asswne that the deepest division in the family lies between Munda1 and the rest, as an older generation of scholars used to suspect, or asswne the veracity of Diffloth's new tripartite division, then the geography of deep historical divisions in linguistic phylogeny would compel us to look for a homeland on either side of the Ganges and Brahmaputra delta, although we would be unable to say whether this homeland would have to have lain to the east or to the west of the delta. When linguists look beyond what linguistic phylogeny can tell them, they must ask which archaeological transition or modem genetic gradient can be related with confidence to an ancient linguistic intrusion or to the prehistorical spread of a language family. When linguists resort to linguistic palaeontology, they must have recourse to the fmdings of palaeoclimatologists and, more particularly, palaeobotanists. Most conceivable theories about the homelands of TibetoBurman and Austroasiatic have already been put forward. The idea of a Tibeto-Burman homeland situated in or nearby present-day Sichuan has been entertained since the 19th century, especially by British scholars in India. Sinocentrists favour a northern Tibeto-Burman homeland in the lower Yellow River basin on the plains of northern China, whereas some have proposed a provenance within the Himalayan region itself. Scholars have sought to situate the Austroasiatic Urheimat as far west as the Indus valley and as far east as the Yangtze delta or insular Southeast Asia. However, the main contenders today for the Austroasiatic homeland are the Indian subcontinent, mainland Southeast Asia and the middle Yangtze. 2. ARCHAEOLOGY, PALAEOETHNOBOTANY AND LINGUISTIC PALAEONTOLOGY
The fimdamental epistemological question will continue to haunt us whether the spread of a recognisable Neolithic and Bronze Age assemblage can actually ever be taken with certainty to reflect the spread of a language and so of a language family. Archaeology reflects what 1 The presence of many speakers of Munda languages in northeastern India is a legacy of resettlement to Assam by the East India Company, an economic policy mentioned above in connexion with speakers of N orthem Dravidian languages who were also thus affected
18
GEORGE VAN DRIEM
we have been able to glean about the material culture of past communities. In fact, how often can we be certain which language was spoken by ancient stone knappers or by the potters behind a particular ceramic culture in some archaeologically attested pre-literate society? Indeed, we must ask whether the modem geographical distribution of the Tibeto-Burman language family correlates with the mute testimony of any single portion of the archaeological record that happens to have been preserved, discovered and studied by archaeologists. A more general issue is time depth. Archaeological transitions are reconstructed at very different times in the past, e.g. the palaeontologically attested spread of anatomically modem humans, the spread of agriculture, and the sometimes well-defined patterns of dispersal of identifiable cultural assemblages in the Neolithic and Bronze Age. At the same time, many known historical transitions and conquests with linguistic consequences have left little or no clear-cut traces whatsoever in the archaeological record. Therefore, we are free to muse, for example, whether the expansion of early Bodie language communities into the Himalayas was associated with the sudden appearance of colonial exponents of the Majiayao Neolithic in eastern Tibet at mKhar-ro and in Kashmir at Burzahom at the same time that the core area in Gansu shrank during a period of climate change between the Majiayao phase (2700-2300 BC) and the Banshan phase (2200-1900 BC) of the Majiayao sequence. This at least is a scenario which I argued in several earlier publications (van Driem 1998, 2001, 2002). Recent palaeoecological evidence indicates that the vast and once heavily forested Tibetan plateau underwent large-scale deforestation precisely during this period of projected Bodie expansion, and the palaeobotanical evidence indicates that this deforestation occurred at the hands of human settlers (Kaiser et al. 2006, Miehe et al. 2006, Wu et al. 2006, Kaiser et al. 2007). Were these people perhaps Bodie colonists from the Majiayao Neolithic core area who introduced to the Tibetan plateau a new lifestyle with deleterious ecological ramifications? Recently, as alternatives to the scenario outlined here, I have presented several alternative scenarios which differently relate the traceable patterns of dispersal of cultural assemblages in the archaeological record with the present geographical distribution of Tibeto-Burman language communities (van Driem 2006). Linguistic palaeontology, a term introduced by Adolphe Pictet in 1859, is an attempt to understand the ancient material culture of alan-
LOST IN THE SANDS OF TIME
19
guage family on the basis of the lexical items which can be reliably reconstructed for the common ancestral language. Linguistic palaeontology, like any attempt to give an ethnolinguistic interpretation to the archaeological record, invariably raises complex issues. Elsewhere, I have discussed at some length the arguments relating to what we can glean about ancient Tibeto-Burman culture and the role played by broomcorn millet Panicum miliaceum and by foxtail millet Setaria italica, the latter reflected in languages as far flung as Old Chinese .fl btsik in the Yellow River basin and Lhokpu2 ca 7 kto 'foxtail millet' in modem southwestern Bhutan (van Driem 2006). At the present time, the earliest archaeologically attested domestic millet dates from before 6000 BC at J1!13f:?M .x-mglooggou near $illt Chif'"eng, where a Neolithic culture without sickles once flourished (Zhao 2005). Linguistic palaeontology strongly qualifies the ancient Austroasiatics as the most likely candidates for the frrst cultivators of rice. Moreover, Diffloth has shown that the reconstructible Austroasiatic lexicon paints the picture of a fauna, flora and ecology of a tropical humid homeland environment, with three salient isoglosses diagnostic for the faunal ecology of the Proto-Austroasiatic homeland reconstructible all the way to the Austroasiatic level and reflected in all branches of the family, i.e. *mra:k 'peacock Pavo muticus', *tn-kuat 'tree monitor lizard Varanus nebulosus or bengalensis' and *tmyu:? 'binturong' or the 'bear cat Arctitis binturong', a black tropical mammal that is the largest of the civet cats (Diffloth 2005: 78). All of these species are not native to areas that currently lie within China, and, to our present knowledge, these species never were native to the area that is today China. Such linguistic palaeontological evidence therefore appears to render the middle Yangtze homeland hypothesis less likely. More reconstructible Proto-Austroasiatic roots indicative of a tropical or subtropical climate are adduced by Diffloth (2005: 78), i.e. *(b~n)jo:l ~ *j(~rm)o:l 'ant eater, Manis javanica', *d~n 'bamboo rat, Rhizomys sumatrensis' (an Austroasiatic root which has found its way into Malay as a loan), *kac~ 'the Asian elephant, Elephas maxi2 The Lhokpu are an inbred and genetically highly distinct group within the Himalayan region as a whole (Kraaijenbrink et al. 2006a, Parkin et al. 2006a). The impact of matrilocality and cross-cousin endogamy is clearly discernible in the genetic signature of this language community. Many of the ancient Tibeto-Bmman groups may have been matrilineal, matrilocal societies with uxorilocal marriage such as 1he modem Lhokpu and Gongduk of Bhutan.
20
GEORGE VAN DRIEM
mus', *kia9 'mountain goat, Capricornis sumatrensis', *rama:s 'rhinoceros, Dicerorhinus sumatrensis' and *tmriak 'buffalo, Bubalus bubalus'. Additionally, Difiloth (2005: 78) points out a fact long noted by scholars of Austroasiatic linguistics, e.g. Osada (1995), namely that a rich repertoire of reconstructible roots representing ancient rice agriculture is robustly reflected in all branches of Austroasiatic, viz. *~)6a:? 'rice plant', *ragko:? 'rice grain', *caJ]ka:m 'rice outer husk', *lGmu: ge:
3.3. Consonants 3.3.1. Simple initials The palatal stops /ch/, lei and IJI interchange freely with the corresponding palato-alveolar affricate. 5 /ch/ can be found only in the complex initial, /xh/ and /x/ are found in the medial position of the word.
5 Several examples of1he voiced stops can be folmd with high tone as well as with low tone. This phenomenon is attested in o1her Khams Tibetan such as mBathang (sKalbzang 'Gyur-med 1985) and Derge (sKal-bzang 'Gyur-med & sKal-bzang dByangscan 2002: 91-111). The conditions for this split are still unsolved.
60
HIRUYUKI SUZUKI
1'1a ble 5 Consonant exampJ1es high tone 'pig' 'hood'
tY' p b
-pha? 'pa:rje
f'
-f'amu 't:) tY'o
'mother-in-law' 'cloths'
'fa?
'blood' 'get angry' 'snake'
t d
r t
'tu Hi
q_ cf'
' na'} ~e?~a
-~
_\izo
~
-lj& l'lJ.a?
~!:lh
-~!:!he:
'know'
~~
-~~'}
'talk'
bq_ bcb; wd wj wz
-bq,a 'no bcb;o? -wde ba 't~,:o wja? '~,:t;: wzu
-t~,:wfJ:Jo
Table 8 Labia1ised types
Pf' 'Pfl.
't
ptgh
P!:s Pi:~,:
"""tL
'P!:siJ 'Ptso
p(,:h
'uaP!:~,:w 'P~,:he'}
~
-~'}
i.
'i.w
'k:
'p::>? 'k:::>? -~
~s ~(,:
-~o
%
-~san5
-~~,:a?
'bed' 'cloud' 'open' 'cock' 'fifty' 'splash' 'seed' 'wash' 'cover' 'sell' 'cutoff 'idea'
w~ w~
wm wl
'bite' 'at last' 'honest' 'eighteen' 'carpenter' 't~,:ww~ 'fourteen' -w~e:ze 'forget' "1Jga wmo ~lee? po 'wild ox' 'thigh' ~argo
1:able 9 C1usters WI"th gJlideand tripJe 1 c1usters kw gw !:IW xhw b
""kwozu 'wo 'gwofi '!:!we%6 'xhwa OJC'o: 'ba
'ywa do '?amje 'pa: rje
'hard' 'smile' 'forty-three'
yw mj Ij
'turn' 'tired'
~~:)
'participate' 'grandfather' 'hood' 'love'
DIALECfAL PARTICULARITIES OF SOGPHO TIBETAN
63
4. PECULIARITIES OF SOGPHO TIBET AN
In this section I compare the phonological and lexical features of Sogpho Tibetan with Written Tibetan (hereafter WrT) and other Tibetan dialects, especially those spoken in the so-called Ethnic Corridor of West Sichuan, in order to illustrate the peculiar nature of Sogpho Tibetan. My main putpose in this section is not to produce a monocausal historical analysis of Sogpho Tibetan, but to consider the development in a typological perspective. Due to its complex ethnic, historical and linguistic background, the ''Twenty-four villages' patois" wtderwent many influences through language contact and we cannot therefore expect a regular pattern of sowtd changes. In other words, it is necessary to compare the Sogpho Tibetan dialect with other dialects in order to explain its outstanding phonetical and lexical features. Recent research reveals the existence of other typologically peculiar Tibetan dialects, such as Zhongu7 (Sun 2003) and gSerpa8 (Swt 2005). I refer closely to Swt's (2003) analysis in the following discussion. The data presented here are based primarily on my own research and on Swt (2003, 2005). 9 I discuss the consonants, vowels and special word forms of Sogpho Tibetan, but will not consider the tonal patterns, because my analysis differs from previous analyses. 10
4.1. Features shared with other East Tibetan varieties I will not address the most general consonantal features, such as the three-fold opposition of aspirated, non-aspirated, and voiced articulation, concentrating instead on the initial consonant clusters and vowelfmalforms.
7 Spoken in Rewugou district, Songpan Cowrty. As yet not identified with respect to its dialectal affiliation 8 Spoken in Seerl>a district, Seda Cowrty. As yet not identified with respect to its dialectal affiliation 9 See Figure 1 below for the geographical location and distribution of the dialects mentioned here. 10 With respect to the Tibetan varieties, tone is generally treated as a feature of the syllable, not as a feature of the word. Word tone has been established only for Lhasa Tibetan.
64
HIRUYUKI SUZUKI
4.1.1. Consonant clusters Sogpho Tibetan shows an extended set of consonant clusters, which is more complex than in the agricultural varieties of Amdo Tibetan and most Khams Tibetan dialects. In some aspects, Sogpho resembles the nomadic varieties of Amdo Tibetan. Attention is given to the manner of articulation of consonant clusters, 11 whose ftrst element is always pronounced very weakly with the exception of clusters with glides. This trait, as well as the presence of prenasalisation and preaspiration, is standard in almost all of the Khams Tibetan dialects. 12 It is possible to establish the following correspondences between the patterns of consonant clusters in Sogpho and WrT (here and in the following the WrT forms will be given in italics): 1. Prenasalisation typically corresponds to WrT initial m- and 'a-: I_JJgul 'head' mgo
t~EI
~ 'high' mtho
/'lllbm la/ 'worm' 'bu
'rice' 'bras
Exception: /~sho pu/ 'knit' WrT sla 'twist, plait'
2. Labialised preaspiration, initial labial fricatives and stops correspond to WrT b- (another origin for clusters such as /'Ptso/ 'cock' bya po will be discussed in 4.2.3.): word-initially:
l'+t_ml 'wash' bkru /""'¥!: w 'forget' brjed
word-medially: /'t~ wja?./ 'eighteen' bco brgyad
fiJa Pt9ml 'fifty' lnga bcu
3. Preaspiration typically corresponds to WrT initial d-, g-, r-, 1-, and sword-initially:
word-medially:
/_6gml
rt(jlm lirjo/ 'ftfteen' bco lnga
'nine' dgu
t~ 'two' gnyis
/'?.a "til 'monkey' sprel
11 See Suzuki (2005a) for a detailed discussion of the relation between syllable structure and manner of articulation in consonant clusters. 12 Nevertheless, these two characteristics have not been recorded in previous studies, such as Jin (1983), sK.al-bzang 'Gyur-med and sKal-bzang d.Byangs-can (2002), and others. See, however, Causemann (1989) for a discussion of prenasalisation, prelabialisation, and preglottalisation in Nangchen. Prenasalisation and preaspiration are very common in the dataset I have collected.
DIALECfAL PARTICULARITIES OF SOGPHO TIBETAN
65
/"'tsa hkaf 'root' rtsa ba
tt'ka?./ 'language' skad 4. WrT la-btags and sr- correspond to either a complex or a simple initial: complex initial:
simple initial:
laj/ 'flute' gling rwla mol 'monk' bla ma J"'ida f),i5/ 'moon' zla ba rhso?./ 'life' srog
flU/ 'wind' rlung r1o1 'easy' sla13 /'15/ 'bull' glang r~o/ 'hard' sra
/' 6
4.1.2. Vowel-finalforms Similar to other Khams Tibetan dialects spoken in western Sichuan, syllable or word fmals appear to be quite reduced in Sogpho Tibetan.
5. Nasalised vowels typically correspond to nasal finals in WrT, but in some cases the nasals have been dropped leaving no trace: nasalised vowel:
plain vowel:
riO/ 'bull' glang ripe! 'official' dpon rna! 'sky' gnam
/""lea bol 'foot' rkangpa f""Ptl{, fi'tl{ 'cloud' sprin
6. A glottal stop corresponds to occlusive endings in WrT, but at the morpheme boundary, the glottal stop can be omitted: glottal stop:
omitted glottal stop:
rhtva?./ 'iron' lcags rhka?./ 'language' skad rlr!'o?.l 'needle' khab
l'f.u! 'ZpJI 'sixty' drug cu
rpo mal 'Tibetan' bodpa
7. A lengthened vowel corresponds to WrT final continuants, but a WrT final continuant does not in all cases correspond to a lengthened vowel:
13 The correspondences with WrT la-htags are, wi1h the exception ofWrT zl, quite straightforward and the original contrast of the initials is preserved: all combinations lead to high tone, prelabialisation to WrT bl, the simple voiced forrn to WrT gl and rl, and the simple unvoiced form to sl.
66
HIRUYUKI SUZUKI
lengthened vowel: rmo:l 'low' dma'
normal vowel: l_'l).d'Qdi 'rainbow' 'ja'
flicli:J 'snake' sbrul
r~el 'gold'
gser
8. In a few exceptional cases, words with an initial nasal show a nasalised vowel, where the corresponding WrT word has a stop ftnal:
rmil 'eye' mil4 fiJi! 'silver' dngul 4.2. Features not commonly shared with other dialects Sogpho Tibetan shows some peculiar phonetic features that are rarely attested in other Tibetan dialects. 4.2.1. The phonemic opposition /rl and 12(" Sogpho shows a phonological opposition between lrl and I'Z/., which is very rarely found in other Tibetan dialects. In most cases, Sogpho lrl and I'Z/. correspond to WrT r and zh respectively, but not all WrT words with zh correspond to I'Z/. in Sogpho Tibetan:
I_r~ 'mountain' ri I_liz;i 'four' bzhi
r~
l),i
lui
'year before last'
gzhes ning lo Retroflex realisation of WrT zh is also found in Zhongu and Nyishe, 15 but while Zhongu, like Sogpho, allows for a non-retroflex realisation in some words, Nyishe does not: - Zhongu: I~
'mountain' ri
I'Zf'l 'four' bzhi
ls'P::j./ 'field' sa zhing
- Nyishe:
/'ra/ 'goat' ra r'l1i 'four' bzhi f21_e l),i/ 'day after tomorrow' gzhes nyin 14 This example likely originated from dmig/dmyig in Old Tibetan, as in many other dialects. 15 Spoken in Nixi village, Xianggelila (Shangri-La) County, Yurman. Belonging
to Khams Tibetan.
DIALECfAL PARTICULARITIES OF SOGPHO TIBETAN
67
Corresponding to the different realisation of WrT zh as either retroflex or non-retroflex sibilants, the unvoiced counterpart WrT sh may likewise be realised as either retroflex (/~) or alveopalatal/postalveolar sibilant (/9/, IJI) in the above-mentioned dialects of Sogpho and Zhongu, but in Nyishe only as retroflex: - Sogpho: r~Ita/ 'die'
r ¢of 'meat' sha
shi
- Zhongu: /~a/
'deer' shwa ba
!Ji! 'firewood' shing
- Nyishe: r~hal 'meat'
sha
r~he f/U?./ 'east'
shar phyogs
It is observed that in Sogpho Tibetan, one and the same morpheme may possess the variants of I'Z/. and /r/, such as/~~ 'forty' bzhi bcu and 1'11. h~w 'to sit' or 'to dwell', which in Dongwang also f1mctions as a full verb. It is interesting that in many dialects of Tibetan this verb has grammaticised into some sort of evidential marker. 36 In Dongwang, however, hn5 is a visual evidential (most likely from the WrT verb <snang> meaning 'to appear', 'to be visible') and ndo contributes animacy meaning. It is not surprising that a verb meaning 'to sit' might acquire animate meanings. The fact that it is and <snang> have followed. In Sections 2.3.1.1 and 2.3.1.2 below, I illustrate each of these with examples from Dongwang. 2.2.1.1. Conj1mct The existential ndo occurs in declarative clauses with ftrst person animate participants or in interrogative clauses with second person animate participants.
35 Throughout the Tibetan dialects, the coronals in coda position affect the previous vowel in different ways. Frequently, the vowel preceding a coronal is fronted, while retaining roundedness. In Dongwang, however, the vowel is fronted and wtrmmded. 36 Usually a direct evidential, or a 'mirati.ve' marker when used in first person contexts.
•
158
ELLEN BARTEE
(50) Conjunct-Interrogative: 2nd person +an. ndo
S
LOC
¢ss =m 2S home =LOC 'Are you home?' ~ess
8
ss ndo
Q
EX.AN.CONJ
(51) Conjunct-Declarative: 1st person +an. ndo S LOC gai3 ¢ss =m ndo 1S home =LOC EX.AN.CONJ 'I am at home.' Both (50) and (51) above are frequently overheard in telephone conversations. In fact, it is unlikely that these clauses would be uttered in a face-to-face context as it would be implausible for this exchange to take place between speech act participants when they clearly know where the other is. It is difficult to fmd a clause in which the inanimate conjunct existential ze is the main verb of a non-possessive construction. This is because one would obviously not question, nor expect a reply from, an inanimate object. However, there are occasionally utterances such as (52) below, which expresses location rather than existence. (52) Conjunct-Declarative: (Jrd person) -an, ze
S !Ji13
LOC t:J13kjSS t:JSpJJa:53
1S.GEN coat over there 'My coat is over there.'
Ze EX.IN.CONJ
(52) is an interesting example of how the empathy hierarchy can interact with both c/d and animacy in Dongwang. Given the freedom according to the concepts of the 'empathy hierarchy' as discussed above, the speaker could either use the conjunct or the disjunct existential in the clause with the third-personS argument 'my coat'. But it appears that the marking of animacy is more constrained and can only correlate with the actual referent, the inanimate head of the noun phrase 'my coat'. As noted, in possessive clauses the c/d distinction holds for the possessor, but the animacy distinction holds for the possessed. In (53) and (54) below, the possessed arguments are animate.
159
ANIMACY IN DONGWANG TIBET AN
(53) Conjunct-Interrogative: 2nd person posr. +an posd, ndo POSR
POSD
a13ka53 a ndo child Q EX.AN.CONJ 'Do you have a child?' fl855
2S.DAT
(54) Conjunct-Declarative: 1st person posr, +an posd, ndo POSR
POSD
wo5j1a53 J1CU53
=jte
a13ka53
two =DAT child 'We two have many children.' lPL
ma1311XJ55 ndo
many
EX.AN.CONJ
In (53) and (54), conjunct existentials correlate with the second person in interrogatives and with the ftrst person in a declarative clause, but the animate existential correlates with the animate possessed argument a 13ka 53 'child/children'. Conjunct forms are also used in (55) and (56) below, but the inanimate forms are selected based on the inanimacy of the possessed argument (55) Conjunct-Interrogative: 2nd person posr. -an posd, ze POSR
POSD
fl855
zi13gi55 a
Ze
book Q EX.IN.CONJ 'Do you have a book?'
2S.DAT
(56) Conjunct-Declarative: 1st person posr, -an posd, ze POSR
POSD
1]813
a13r155
=ji
bao53zhi
lS.DAT today =GEN newspaper 'I have today's newspaper.'
Ze EX.IN.CONJ
The relevant difference between examples (53) and (54) and examples (55) and (56) is not one of c/d status, as the conjunct form is chosen in all four examples. The difference is in the animacy of the possessed argument In Dongwang this split holds for disjunct environments as well. Disjunct When the single argument of an existential clause is third person, the disjunct form of the existential is used, and the animacy of the S argument determines the form of the disjunct. 2.2.1.2.
160
ELLEN BARTEE
(57) Disjunct-Declarative: Jrdperson +an, ndo d~?
s dess wrssz;, re
kh:Js7xf$i53
= ji
P:Jss
then that
3PL.GEN
=GEN
boy
SUB LOC
t~:JI3wuss
rr fliii3 =n:J
ndo c#i?
s
older RI field =LOC EX.AN.DISJ QUOT 'Then at that time, the older son of the household was in the field.' (Prod058) (58) Disjunct-Declarative: Jrd person -an, ze #? LOC
tc5sswfi53
rc5
¥CI30053
00 s5Ja53
Dongwang
and
rGyalthang between
s
r:J 13
sii53 ze d$i mountain three EX.IN.DISJ 'There are three mountains between Dongwang and rGyalthang.' Examples (57) and (58) are both locational clauses, but the single argument of (57) 'the older son' is animate, while the single argument of (58) is inanimate. This animacy distinction is reflected in the choice of an animate or inanimate form of the disjunct existential. As we have seen, in possessive clauses the choice of c/d is determined by the person of the possessor, but the choice of animate/inanimate forms is determined by the animacy of the possessed argument (59) Disjunct-Interrogative: Jrdperson posr. +an posd, ndo #? POSR
POSD
.khw:J 57xf¥a 53 ~ 53
ndo d$i? a rerc5 dog EX.AN.DISJ Q 'Do they (their household) have a dog/dogs?'
3PL.DAT
ANIMACY IN DONGWANG TIBET AN
161
(60) Disjunct-Interrogative: 1st person posr, +an posd. ndo d:;@? POSR
POSD
ai3kaS3 ndocf¥jP child EX.AN.DISJ 'Do I have a child?'
gai3
are
1S.DAT
Q
The possessor arguments in (59) and (60) trigger the disjwtct forms. Remember that ftrst person questions are somewhat wtusual and there are several ways that such a question could be asked. My consultant said that a question as in (60) above could occur in a context in which someone has told the speaker that s/he has a child (either in joking or accusation), and the speaker responds with Do I (really) have a child? The important thing to note here is that since both possessed arguments are animate, the animate disjwtct existential verb is chosen. The question in (61) below contains an inanimate existential verb, reflecting the inanimate nature of the possessed argument (61) Disj-Interrogative: Jrd person posr. -an posd, ze d:i@? POSR
POSD
zi 13gi55 ze cf¥jP are ro book EX.IN.DISJ Q 'Does s/he have a book/books?'
khua 53 3S.DAT
Both (62) and (63) below contain third person possessor arguments, but the difference in animacy of the possessed arguments triggers different disjwtctforms. (62) Disjunct-Declamtive: Jrdperson posr, +an posd. ndo d:;@? POSR
POSD
]KJ 13 t¢ = jte ]XJ 13 J1CU 53 ndo d¥iP person one =DAT son two EX.AN.DISJ 'A man had two sons.' (ProdOO 1)
(63) Disjunct-Declamtive: Jrd person posr. -an posd. zed:;@? POSR
POSD
zii3giss mai3m:Jss Ze d¥iP book many EX.IN.DISJ 'Sihe has many books.' khuaS3
3S.DAT
162
ELLEN BARTEE
Recall that in reported speech the choice of c/d form can indicate coreferentiality of S or A arguments in the matrix and embedded clauses; when the S or A arguments of both clauses are coreferential, the conjWlct form is used. This holds for existentials as well, with choice further conditioned by the animacy of the possessed argument of the complement clause. The examples below illustrate the animate c/d forms:
(64) Conjunct-Reported speech: Corekrential +an posd, ndo A khuiss
COMPL
khua53
ai3ka53 ndo
3S.ERG 3S.DAT child EX.AN.CONJ 'Slhex said slhex has a child.'
s:J
c4a
say
QUOT
( 65) Disjunct-Reported speech: Non-coreferential +an posd. ndo ~
A khuiss
COMPL
khua53
ai3ka53 ndo ~p
3S.ERG 3S.DAT child EX.AN.DISJ 'Slhex said s/hey has a child.'
s:J
cf7IJ
say
QUOT
Based on my present data set, a preliminary hypothesis is that whatever is capable of volitional movement is animate and everything else is inanimate. The animate category thus includes small and large animals (including insects), demons, 37 children and adults. (66) Conjunct-Interrogative: Joe +an. ndo POSR ~iss
=go
POSD
d.ps5wa53 a53 ndo 2S.GEN =LOC flea Q EX.AN.CONJ 'Do you have (a) flea/s?' (='Is/are there (a) flea/s on you?') Conversely, body parts seem to be treated as inanimates.
TI Tibetans have many classes of gods and demons. I assume that Dongwang speakers would treat all demons and gods as animate, but my data are too sketchy to confirm that. Once I was watching a movie with a friend In the movie one of the characters was seized by a demon. I asked my friend how to express that in Dongwang and she said kh{)55h855dztj ndo @7 'He has a demon', using the animate existential form.
163
ANIMACY IN DONGWANG TIBET AN
(67) Disjunct-Declamtive: Jrd person posr. -an posd, ze @? POSR
POSD
khua 53 kha 55
t~:J 13t~ho
mouth small 'S/he has a small mouth.'
3S.DAT
4;i
zecf¥jP
one
EX.IN.DISJ
(68) Conjunct-Declamtive: Joe -an, ze POSR
yei3
POSD
khi 5~ba53
=go
foot =LOC 'My foot has a wound.'
lS.GEN
m:J 5*:ha 53 t¢
wound
ze
one EX.IN.CONJ
Both (66) and (68) have an overt locative marker, which suggests that these are single argument clauses. The use of the conjunct form is conditioned by the involvement of the second person as the physical location of the 'flea' and the 'wound'. Additionally, the semantics of (67) is more attributive than possessive, but the syntax is as in other possessive clauses (possessor argument is in the dative case; possessed argument is an unmarked NP). In my data, Dongwang speakers treat 'flea' as grammatically animate and 'mouth' and 'wound' as grammatically inanimate. In the following section, I illustrate the distribution of animate and inanimate forms in the auxiliaries.
2.2.2. Animacy split in non-existential clauses In the previous section I have shown how animacy plays a primary role in locational and possessive existential clauses. In this next section, I show how some of the same morphemes reflect animacy in four other aspects of the grammar: in attributive constructions, in nominalised and relative clause constructions, as auxiliaries in intransitive and transitive clauses, and in subordinate clauses.
Attributive constructions In my data, the copulas zl or re are most frequently used when the copula complement is attributive and conveys an inherent quality. Examples of the existential morphemes being used in attributive constructions are infrequent However, when they do occur, the single ar-
164
ELLEN BARTEE
gument of the clause controls the c/d38 and animacy choice. Expressing an attribute construction using the existential form seems to convey a more transitional quality. ( 69) Disjunct-Jrd person attributive -an. w/ existential, ze @? 0 ~IIIJ53
wf55z:s.o53 ma J3m:J55 ma
te53
ni
rr
pigfat
like this
give NI
RI
tHJ55
rr
(f;ha13
much ~55
NEG
rr
f:JJ3m:J55
ze ~'I
then RI eat when RI tasty EX.IN.DISJ 'If (you) don't put too much fat on it, then when (one) eats (pig sausage) (it) is tasty.' (K.il1Pig063) In the context of ( 69) above, the speaker is describing how to make sausages from pig's intestines in just the right way. Although it is not mentioned here, the inanimate S argument, z:1 13w6 53 'intestine', is mentioned several clauses previous to this. In (70) below, the S argument is animate and the animate existential is used as expected. (70) Disjunct-Jrd person attributive +an. w/ existential, ndo c;bi?
s Jo55S:J
ph:Jph:Jph:J
JJa13tiJa53
t¢
ndo ~'l
teacher urn hardworking one EX.AN.DISJ '(The/our) teacher was extremely hardworking.' (MyLife136) The animacy distinction is maintained within the pragmatic nature of the c/d system. That is, speakers are allowed to express various degrees of empathy by using an unexpected c/d form, but they are required to maintain the animacy distinction. Thus in (71) below the speaker is describing school days that took place more than forty years earlier:
38 Taking into account the speaker's potential for pragmatic skewing to convey intimate knowledge or distance as elaborated by the 'empathy hierarchy' mentioned above.
ANIMACY IN DONGWANG TIBET AN
165
(71) Disjunct: 1st person attributive +an. w/ exististential, ndo d¥?
s
t:JI31Wss wo5jJa53 -kl rr then 1PL -PL RI
w
5tse 53 tsh657rii ndo ~p ma re like that smart EX.AN.DISJ NEG COP.DISJ 'At that time, we were not smart like that.' (MyLife047) The single argument of (71) is frrst-person plural, yet the speaker chooses a disjwtct auxiliary. This is a common strategy in which a speaker shifts empathy based on temporal distance. The existential still reflects the animacy of the S argument in spite of the shift in empathy.
Relative clauses Relative clauses in Dongwang are formed when a nominalised verb or clause modify a head nowt. The relative clause occurs before the head nowt. It is interesting that the animacy distinction is reflected when existential verbs ze or ndo are used to form a relative clause. Further, it is the head of the relative clause that triggers the choice of animate or inanimate forms.
(72) Relative Clause: -an head w/ existential ze A P:JI3 t~h:JI3t~h6ss son yowtg.small ZE
n:J
EX.IN
NMLZ
=ji =GEN
0
=ji =ERG
khua 53 3S.DAT
se 5 ]xr53 things
k!Jai3la53 all
tsh:Jss ro53 tht:enss !Wss SUB. WHEN hither gather PFV 'When the yowtger son had gathered all the possessions that he had... ' (Prod007) (73) Relative Clause: +an head w/existential ndo
s khua 53 ndo
n:J
~53
3S.dat EX.AN NMLZ dog 'The dogs/he has is good.'
t:J
DET
a I:poss hn3 good EX.EVI
166
ELLEN BARTEE
In (72) the head of the relative clause is inanimate: se 5]la 53 'things'. This is reflected in the choice of the inanimate existential form ze. (73) is also a relative clause, but the animacy of the head of the relative clause, ~ 53 'dog', in (73) triggers the animate auxiliary. Notice that while the c/d distinction is neutralised, the animacy distinction is maintained. Thus the animacy of the referent within the relative clause controls the choice of animate versus inanimate selection.
Intransitive and transitive clauses In my data, when ndofunctions as an auxiliary in an intransitive clause, it carries additional aspectual meaning, specifically occurring with events, past or present, that are happening within a specified time frame: 39
(74) Con[unct-Interro!J!!.tive: intransitive1 2nd f!E.ISO!JJ. +an
s
re13 ka13 nc4u13 ndo 2s now where go EX.AN.CONJ 'Where are you going now?' ~ess
(75) Con[unct-Declarative: intransitive1 1st f!E.ISon
s
gai3 1s
LOC
n: .. :J
jit(X)t~olu
1996
RI
gai3 tsh:J 55d2;6 1s middle school zjliJcfEe
J1E year
bi 5peP53 jei3 ma graduate do NEG
tlNiJ55 ni PFV
NI
ndo
study EX.AN.CONJ 'I, uh, in 1996 having not yet graduated from middle school, I was studying.' (GetMarOOl/002) In (75) the speaker is taJking about a specific span of time during which her parents urged her to get married. It was during that time that she was studying. I have few examples of ze or ze d$i"Pfunctioning as auxiliaries in intransitive clauses. This supports the cross-referencing function of animacy as one would not expect to fmd inanimate arguments serving 39
Tense and aspect in Dongwang are not discussed in detail in this paper.
ANIMACY IN DONGWANG TIBET AN
167
as actors in clauses. The only clauses with an inanimate auxiliary form I have found are those in which the argument is a nominalised verb as in (76) below: (76) Disjunct-Declamtive: 1st person, intransitive, -an, ze @?
s ga13
nc4P13
1s go 'I can't go.'
ts:J NMLZ
ze ~?
ma
re
EX.IN.DISJ
NEG
DISJ
In (76), although the S argument is animate and would be expected to take a conjunct form, the auxiliary is inanimate and disjunct This suggests that it is the nominalised verb, rather than the animate actor, which is being indexed. 40 This hypothesis is further supported by the fact that while the verb nc4P 13 'to go' is controllable, the nominalised construction indicates that it is not controllable by the S argument It would be nice if it turned out that in transitive clauses which contain an auxiliary constructed from an existential, the 0 argument controls the auxiliary with respect to animacy. At ftrst glance, this in fact appears to be the case. (77) Conjunct: transitive clause. 1st person A. -an 0. ze A
0
bie 53 ze again one frre up bum EX.IN.CONJ ' ... and again one (of us) kept the frre burning.' (Hardship 110) zc513
f(:i 53 ni 13 z:J 13
(78) Conjunct.· transitive clause, 1st person A. -an 0, ze 0 tsa 5~ba 53
t¢
f(:ha 13
ze
tsampa INDEF eat EX.IN.CONJ ;(We) ate some tsampa. ' 41 (MyLife035) Both (77) and (78) have animate A arguments, but inanimate 0 arguments. It happens that in my data the most frequent auxiliary choice is 40
This could also be analyzed as a possessor clause ('I do not have a go'), but to
the best of my knowledge never occurs with a conjunct auxiliary. More explicit analysis of this construction will be provided in a future paper. 41 Tsampa, a staple for many Tibetans, is made from barley that is roasted and then finely ground It is often mixed together with butter tea to form a cookie-doughlike ball.
168
ELLEN BARTEE
the inanimate form, perhaps for the reason that all such examples have inanimate 0 arguments. The reason for this is related to the animate form being chosen for intransitive clauses: all the examples in my database which co-occur with existential auxiliaries have inanimate 0 arguments.42 Indeed, the fact that there are no counter-examples (e.g., animate 0 arguments with an inanimate existential form) strongly supports the observation that the animacy of the 0 argument possibly controls the choice of auxiliary. However, this is only a tentative analysis.
Subordination The use of an auxiliary in subordinate clauses is optional. When an auxiliary derived from an existential is used in non-complement clauses, only conjunct forms are found, thus neutralizing the c/d distinction, but the animacy distinction is still maintained. This suggests that the conjunct form is the default form as to epistemic source, but that there is no default form where animacy is concerned. (79) Conjunct: sub-clause, exist, Jrd person +an. ndo
s
de 13 IJ:} ndo raJ 55 sit NMLZ EX.AN.CONJ when
}a 13 n:J me ana good NA CONJ.NEG MOD 'When there is someone who lives in Zhongdian (it) is good, right?' (for our business purposest3 (DCWormGrass074) (80) Conjunct: sub-clause1 exist, Jrd 125ZISon -{!g ze (obl) na 53 sky
s
=n::~
=NA
fii53 clouds
ze
rress
EX.IN.CONJ
when
thii353 hn3 moon NEG see EVI.VIS 'When there are clouds in the sky, you can't see the moon.' J::~55Jl:e53
ma
42 While I have many clauses with transitive verbs such as 'kill', 'hit', 'look at', 'listen to', etc., none of them co-occur with the existential auxiliary. This is probably due to the relatively small size of my database (2,200 clauses) rather than any particular restrictions on such co-occmrences. 43 The 'subject' of (79) is the whole nominalised clause .$C 13d055 di 13 dc 13n{) 'live here in Zhongdian (one)'
169
ANIMACY IN DONGWANG TIBET AN
(81) Conjunct· sub-clause, 2nd person posr. -an, posd, ze POSR
POSD
pJxJ13cf¥e53
~55
~i13w0 ss
ze
ne 55
hereafter
2s
question
EX.IN.CONJ
SUB
!]813
tsi53
~
look for IMP 'In the future when/if you have problems, come find me.' 1S.DAT
In (79) through (81) above, there is no c/d distinction in the subordinate clause even though the clauses involve non-frrst-person arguments. However, the animacy of the S argwnent in each of the subordinate clauses is still reflected in the speakers' choice of an animate or inanimate conj1mct form.
Other subsystems I have shown that animacy is a crucial parameter in existentials and in the periphrastic auxiliary forms which are constructed from existentials. This might lead one to propose that animacy may be extended through the language in other subsystems as well. This appears to be the case in a few classifiers. For example, there are two classifiers used for 'pair' or 'two of something': (f:ha 53 -AN, and gu 55+AN. There is also some indication of an animacy distinction in the pronouns, but my examples are too few to say for certain. Given the fact that animacy has not been described as a component in any other c/d system, or for any other Tibetan dialect, the question arises as to why it plays such an important role in Dongwang Tibetan. The conspicuous absence of any mention of an animacy split in descriptions of historical Tibetan grammar indicates that this innovation is most likely not due to genetic inheritance. Tournadre (2001) does not mention this feature of animacy in his overview of Literary Tibetan and contemporary spoken dialects of Tibetan.44 Nor does Beyer (1992) mention this feature in his description of Classical Tibetan. In the next section, I will suggest that this feature is a result of language contact.
44 Tournadre has collected data from a vast mnnber of dialects, but as far as I know, he did not collect data in 1he Sou1hern Kham area of Yunnan Province.
170
ELLEN BARTEE
3. PoSSIBLE ORIGINS OF ANIMACY IN DONGWANG 3.1. Contact-induced change It should not be surprising that Dongwang would have unique characteristics distinct from other varieties of Tibetan. 45 One would expect that languages which are spoken far away from population centers and communication arteries might well develop differently than those within population centers. As part of the southern Khams Tibetan group, Dongwang speakers have lived far away from the 'centers' of Tibetan political and religious influence, yet have been in frequent contact with other language groups in northwest Yunnan. Thomason and Kaufinan (1988: 35) state that it is ''the sociolinguistic history of the speakers, and not the structure of their language, that is the primary determinant of the linguistic outcome of language contact. Purely linguistic considerations are relevant but strictly secondary overall". Dixon ( 1997) argues that languages might have long periods of relatively slow change, punctuated occasionally by dramatic changes due to a complex bundle of factors such as language attitudes, degree of isolation or contact, degree of bilingualism, and structural differences between languages. In order to investigate the 'history' ofDongwang speakers, it might be beneficial to ask certain questions: What possible source languages are there? What social factors might be involved (intermarriage, language attitudes, dominant culture, etc.)? Is the animacy split likely to be a result of borrowing or shift? What other subsystems have evidence ofcontact-induced change? The identification of a source language requires a plausible account of enough contact with another language in order to hypothesise that innovations could have developed through contact. Obviously, the suggested source language should have, or have had, the features that are under consideration. Social factors such as intermarriage, bilingualism and the nature of the contact must be considered as well.
45
Some have suggested that varieties which exhibit great differences may not even
be Tibetan. Sun Hongkai has suggested that Baima, spoken in Amdo, exhibits enough pronounced differences that, in spite oflexical similarity, it should not be considered a dialect of Tibetan. However, as Jackson Sun pointed out, and as is true for Dongwang,
such distinctive characteristics should be expected from a dialect spoken in an area of great linguistic diversity.
ANIMACY IN DONGWANGTIBETAN
171
Although Thomason acknowledges that predictions of language change are probabilistic, not deterministic, she has observed that in most cases of borrowing, there is initially more lexical borrowing than structural borrowing. Contrastively, in most cases of shifting, phonological and syntactic changes are common while vocabulary lags behind (Thomason 2001: 76fl). She concludes: "if we can establish significant structural interference, but there are few or no loanwords, then the interference must have come about via imperfect learning of a target language during shift, not through borrowing" (Thomason 2001: 80). LaPolla discusses the important role that culture and cognition play when speakers begin to learn the language of another culture. When speakers learn some aspect of another language, they may begin to construe the world around them in the same way as the other group, which in turn can lead to the spread of certain constructions or patterns in the speakers' frrst language (LaPolla 1999: 15, 16). Thomason and Kaufman stress that they have found no instances of completely isolated structural interference in only one subsystem of a language (1988: 62). Hopper and Traugott point out that although independent words and morphemes are the most typical items to be borrowed (citing Weinreich 1953), ''very occasionally" whole paradigms can be borrowed (Hopper and Traugott 1993: 210). They cite the Mednyi Aleut case discussed in Thomason and Kaufman (1988: 20) in which Mednyi Aleut had Russian "ftnite verb morphology but with other largely Aleut grammar and vocabulary" (Hopper and Traugott 1993: 210). Thomason agrees that an implicational hierarchy of interference (ftrst borrow words, then grammar) is not the case in instances where ''the people responsible for the innovations are second-language learners of the receiving language: in these cases ... the ftrst and most significant interference features are structural, not lexical" (Thomason 2001: 64). In the following section I argue that it is quite possible that the animacy distinction found in Dongwang is an areal feature which has developed as a result of contact with Naxi speakers. The history of Southern Khams Tibetans is one of frequent and often tumultuous contact primarily with the Chinese and the Naxi. There are Naxi villages in Tibetan areas whose inhabitants are bilingual in Tibetan and Naxi. Crucially, the language of Naxi speakers with whom Tibetans had contact appears to be the only one with the type of animacy features discussed in this paper.
172
ELLEN BARTEE
3.2. Tibetans ofNorthwest Yunnan 3.2.1. Southern Khams While there have been no descriptions of Tibetan dialects which have the animacy split I have described for Dongwang, it is not the case that other related varieties do not exhibit this split. In fact, there are other dialects in the Southern Khams region which distinguish animacy in this way as well. Preliminary research indicates that the animacy distinction I have described for Dongwang is also reflected in the c/d system of Dechen Tibetan,46 with one crucial difference: animate referents can trigger animate or inanimate forms, but inanimate referents can only trigger inanimate forms. 47 This suggests that the split in Dechen Tibetan might be animate versus neutral, rather than animate versus inanimate. While the disjunct morpheme is different from that of Dongwang, the motpheme which distinguishes animacy is from the same historical source. It also is the case that rGyalthang Tibetan auxiliaries reflect an animacy split. Researchers categorise the varieties of Tibetan spoken in NW Yunnan as part of 'Khams', one of the three large dialect groupings of Tibetan spoken in China, but very little research has been conducted to describe this area. 48 In fact, 'Khams' dialects are usually understood to be those varieties spoken in Dege or Chamdo, or maybe the area around Bathang. Due to the lack of descriptive work available, it is impossible to say how far this feature is spread, but it is nearly certain that Eastern Khams, at least those varieties which usually are cited as representative of Khams (in particular Dege, and Chamdo ), do not mark animacy in this way. For example, Gesang Jumian, in Zangyu Fangyan Gailun, 49 includes several Tibetan counties in SW Sichuan province with those spoken in Diqing Prefecture as representing southern Khams dialects (2002: 72). However, he does not mention an animacy distinction and is content to draw examples from Dege dia-
46
Because there has been no sociolinguistic assessment in that area, it is difficult
to ascertain where dialect boundaries may fall. For this reason, I will refer to this dialect as 'Dechen dialect' based on the name of the region. 47 Further study of this dialect may reveal that the forms used for animate referents are not totally interchangeable. 48 Exceptions that I am aware of are Wang Xiaosong and Krisadawan Hongladarom who have written several papers to describe the rGyalthang (Zhongdian) dialect. 49
Jlff9;/j"l!fj618(0verview ofTihetanDialects).
ANIMACY IN DONGWANGTIBETAN
173
lect to typify all of the Khams dialects (2002: 73). Dege distinguishes cld existentials by two forms: }.a and gge (Gesang Jumian 2002: 131fl). There are several possible avenues to explore the origins of animacy indexing in the cld system described for Dongwang Tibetan and other varieties of Southern Khams. If the sociolinguistic history of speakers is the primary determinant of the linguistic outcome in a contact situation as Thomason and Kaufman suggest, then there are questions which must be addressed in order to adequately explore the unique development observed in Dongwang Tibetan, concerning not only the history of contact in NW Yunnan, but also the duration and nature of this contact Crucially, features such as those described here must be (or have been) observed in the language of those with whom contact was made. 3.2.2. Briefhistory ofcontact in NW Yunnan The history of contact in NW Yunnan is fragmentary and tangled, but it is indisputable that Naxi, Chinese and other ethnicities frequently intermingled with Tibetans, often under hostile conditions. Diqing Autonomous Prefecture, located along the historical southern tea route, was the site of many political contests. The tea trade began during the Tang dynasty (618-907) and the 'Tea and Horse Caravan Road' stretched from India and Tibet to SW China, to Lijiang, which has been and still is, a major Naxi city, and up through Diqing Tibetan Autonomous Prefecture (Yang 2004). During the Tang Dynasty, the Yi (Lolo) established the Nan Zhao kingdom which ruled until the Chinese conquered it in the ninth century and the Dali kingdom was established. The Mongols (1206-1367) then conquered the Dali Kingdom with the help of the Naxi king. From then on, the Naxi served as a vassal to help obtain and maintain any ruler's control, all through the Qing dynasty (1644-1911). In 1382, Imperial China used the Naxi king, headed by the famed Mu family, to obtain control of the area and Lijiang was established as the provincial capital. The Mu family continued to rule through the Ming dynasty (1368-1644), occupied rGyalthang for a time, and advanced through NW Yunnan to Tibet (Wang 1995: 55). When the threat of the Mu family to Tibet grew, the Dalai Lama extended Tibet's rule to NW Yunnan (1662-1723) at which time the Mu family sent troops as far north and east as Markham, Bathang and Lithang. Throughout the Qing Dynasty (1644-1911), the Naxi ruled through a
174
ELLEN BARTEE
hereditary system of chieftains. After the Qing Dynasty gained control of the area, Eastern Kham was divided: Bathang and Lithang were annexed to Sichuan and Dechen and rGyalthang were annexed to Ywman. Due to the decaying power of the empire, Chinese control over Tibetan areas had weakened considerably. Many of the Tibetan areas were ruled locally by chieftains, many of whom openly derided Chinese authority (Coales 1919: 230). After a Chinese official was killed on his way to Lhasa, General Chao Erh-feng was sent to help reign in the Kham areas. Van Spengen details General Chao's ruthless attacks on several monasteries and regions of Kham. This, combined with renegade soldiers, Tibetan warlords and Naxi factions, helped to contribute to the making of NW Yunnan into a notorious robber haunt at the beginning of the 20th century. It was during this time that ''the dreaded Tongwa" (Dongwang) terrorised neighbouring populations, raiding Naxi territory, and at one point even controlling Zhongdian (van Spengen 2002: 16-19). There is evidence that the Naxi intermarried and settled in Tibetan areas. There are Naxi villages scattered throughout the Tibetan areas and some of these near the Tibetan border are fully bilingual in Tibetan. I interviewed one Naxi man, a fully bilingual speaker of the Dechen dialect of Tibetan who lives in a predominantly Naxi village in the far NW Yunnan near the Tibet border in Dechen County on the West side of the Lancang Jiang (Mekong River). He told me that there are three villages near him that are known as 'Naxi' villages. 50 Residents of these villages speak Naxi at home, but Tibetan outside of the home. Most are Tibetan Buddhists who live in Tibetan-style houses, wear Tibetan-style clothes and name their children Tibetan names rather than Naxi names. All are fully bilingual, but his impression is that the younger generation seem to be losing Naxi. Finally, when asked if he knew when Naxi immigrants ftrst arrived in his village, he responded that no one knows for sure, but "it is said that perhaps" they came during the period of conquest by the Mu family. According to the brief history just outlined above, this would be sometime in the early 17th century.
50
Personal interview, March 3, 2005.
ANIMACY IN DONGWANG TIBET AN
175
I examined grammatical descriptions of Yi, Bai, Nu, Lisu, Mongolian and Qiang as possible source languages. 51 None of these have the features that I have described for Dongwang Tibetan. While animacy is one parameter that determines how 0 arguments are treated in Qiang (LaPolla and Huang 2002), the patterns are quite different52 and it is highly unlikely that Qiang would exert influence on Southern Khams while failing to do so on its immediate Eastern Khams neighbours in Sichuan. Naxi and Pumi (Primi), however, make very similar distinctions as those attested in Dongwang. In the following section, I discuss briefly the Naxi and Pumi data.
3.2.3.Naxi Naxi, which is traditionally classified as close to Loloish in the LolaBurmese branch, 53 has about 290,000 speakers who live mainly in the Yulong Naxi Autonomous County to the south of Shangri-la County. Naxi does not have a c/d system, but existential morphemes show a four-way contrast. One contrast correlates to the animacy of the referent, whether in a one-argument clause (correlating with the S argument) or in a possessive clause (correlating with the possessed argument). This is indicated by tonal and consonantal contrasts. The following data comes from the Naxiyu Jianzhi, 54 a sketch of the Naxi language, ( 1985):
51 These are all languages which have had some contact with the Tibetans of Shangri-la. Mandarin is also a possible somce language, but clearly lacks the featmes described in this paper. 52 There are five existential verbs in Qiang. Their use depends on the semantics of where and how the referent exists (e.g., contained, moveable, strength). Animacy is one of the semantic parameters of existential verbs, but the main difference is that the animacy of the 0 argwnent is indicated by casemarking on the NP. 53 I am indebted to Alexis Michaud (personal communication) for directing me to David Bradley's 1975 finding that Naxi is significantly different from Lolo languages and may not even be part ofLoloish, even though they are spoken in roughly the same location It is fair to say, as he points out, that Naxi and Pumi may not be genetically distant. 54 He Jiren and Jiang Zhuyi, eds. The full title is l?ii!ffilf/ifjlf (A sketch of the Naxi language). China has published sketches of all the minority languages. Each one typically contains a brief phonological and grammatical description.
176
ELLEN BARTEE
(82) Naxi: Decl, Jrd person posr. +an posd, ndF 1
s
$i33gy33dy3I ¢33 ni33w?I ggv33 mcu33 nd$Y3I Lijiang person twenty nine thousand EX.AN 'Lijiang County has 290,000 people.' (Naxiyu Jianzhi: 91)
(83) Naxi: Decl, Jrdperson posr. +an posd, ndF 1 POSR
POSD
thai3 Jcul:iyw33 J1i33 phuss ncipr3I 3PL cow two head EX.AN POSR
POSD
ga33 Ja33 dw33 phuss ncipr3I 1s also one head EX. AN 'Their household has two head of cows; our 5 household also has one head.' (Naxiyu Jianzhi: 52)
(84) Naxi: Decl, Jrd person posr. -an posd, POSR
clF3
POSD
thcu33 ba33Ja3I ~ws dcu33 thoss cipr33 3s clothes new one set EX.IN 'She has a set of new clothes.' (Naxiyu Jianzhi: 58)
(85) Naxi: Interrogative, 2nd person posr, -an posd, POSR
drr
3
POSD
u33
the33yw33 ze33tcJ3I cipr33 Je33 2s book many EX.IN Q 'How many books do you have?' (Naxiyu Jianzhi: 102)
In (84) and (85) the existential form ~1 is used when the single argument or the possessed argument (Jcu 13yw33 'cow' and ~i 33 'person') is animate, but in (86) and (87) the existential form ~3 is used when the single argument or the possessed argument ( ba 33la 31 'clothes'
55
First and third person singular pronmms ir1 Naxi are .ya 33and thr.u 33 (Naxiyu Ji-
anzhi: 51).
ANIMACY IN DONGWANG TIBET AN
177
and tkl 3-jr£U33 'book') is inanimate. 56 This parallels the Dongwang Tibetan system presented above. 57 I found no examples of locational statements in the Naxiyu Jianzhi, but apparently the animate form would be used for a sentence such as 'My wife is in the house', but a different verb would be used for a sentence such as 'My bike is at home' For these sentences, one would have to say 'My bike is placed/parked/stored at home'. 58
3.2.4.Pumi Pumi belongs to the Qiangic sub-branch of the Tibeto-Bunnan language family. It is spoken by about 35,000 speakers59 mainly in the SE of Shangri-la County. Little is recorded of their history with Tibetans. The data I have on Pumi indicates that the locational and possessive existentials reflect an animacy split: bbonggf 'to have -AN' and xxiuf'to have +AN'. There is no indication of a c/d distinction. 60 4. DISCUSSION Given the above data, the historical background and the current context, it is quite possible that the animacy distinction observed in Dongwang and certain other Southern Khams Tibetan dialects has developed as a result of language contact The fact that both Naxi, a LolaBurmese language, and Pumi, a Qiangic language, reflect this distinction suggests that this could be an areal feature. The differences in the morphological shape of the animacy markers between Dongwang and Naxi and Pumi indicate that Dongwang developed its animacy markers as a result of contact, but clearly did not borrow the morphological 56 When discussing dialects, the authors of the No:xiyu Jianzhi list the animate existential form for the Eastern variety of Naxi as ti$Y1~ so the animacy distinction is triggered by a tonal contrast alone (p. 112). 57 A finther contrast which Naxi makes, not observed in Tibetan, is between contained and non-contained referents. This seems to be a finther delineation of inanimate arguments, but the authors do not expressly say so. All of the examples in the No:xiyu Jianzhi used the locative particle lo 31 'inside':
(i)
khaJJ
Jd1
fla31
nza33
aJJ
ditch inside water NEG EX.IN.CONTAINED 'There is no water in the ditch.' (N axiyu Jianzhi: 52) 58 Thomas Pinson, personal communication 59 This information comes from the Grimes and Grimes (2004). 60 Jo Chan, personal comrmmication The spelling system is her personal practical orthography which she was using for transcribing her data.
178
ELLEN BARTEE
forms from Naxi. Aikhenvald and Dixon (2003b) suggest that when a significant number of speakers of one language have increasing competence in another language, the languages start to resemble each other. For example, if a language with no noun classes were to come into contact with a language that has noun classes, ''then it is likely to develop its own set of noun classes; most frequently it will achieve this not by borrowing the forms for marking noun classes from a neighboring language, but by developing them from its own internal resources" (Aikhenvald and Dixon 2003b: 2). It appears that the Dongwang animacy distinction may have arisen in this way. The fact that this animacy feature has been incorporated into the c/d system is interesting for several reasons. While animacy in Dongwang is expressed in the c/d system, it is unrelated to typical features of c/d systems. C/d systems typically correlate in some way with the intentionality, volition, control and evidence of S and A arguments. By contrast, Dongwang indicates that speakers are indexing the animacy of S and 0 arguments. This supports the contention that the c/d system in Tibetan is functioning in a cross-referencing capacity; in spite of its pragmatic nature where S and A arguments are concerned, animacy seems to consistently hold for the S or POSD argument. 5. CONCLUSION I began this paper with a descriptive overview of c/d systems in general and in Tibetan in particular. I then outlined how Dongwang is different from other varieties of Tibetan, focusing particularly on the animacy split While animacy is observed mainly in the existential verbs and their functions as auxiliaries, it is also present in other aspects of the language including relative clause construction, subordination and pragmatic skewing. Additionally, there is some indication of animacy spreading to other subsystems, such as classifiers and pronouns. This depth of integration indicates that the animacy split may have developed some time ago. The lack of such a system in other dialects of Tibetan combined with the presence of a similar system in nearby languages suggest that this is an areal feature which may have developed initially as result of contact with Naxi speakers. Finally, the presence of such divergent features as the animacy split in the c/d system of Dongwang points to the possibility that the c/d system is func-
ANIMACY IN DONGWANG TIBET AN
179
tioning in more of a cross-referencing fashion than has been reported for other dialects.
ABBREVIATIONS flrst person second person 3 third person A most agent-like argument in transitive clause AN animate cc complement of copular clause COMPL complement CONJ conjunct COP copula DAT dative DET determiner DISJ disjunct EGO egophorique ERG ergative EVI evidential EX existential FOC focus FUT future HON honorific HS hearsay evidential IN inanimate 1 2
10 IPFV
LOC MOD NEG NMLZ
0
OBL PFV PL POSD POSR PST Q QUOT
s
s VIS
WrT
indirect object imperfective locative modal negative nominaliser most patient-like argument in transitive clause oblique argument perfective plural possessed argument of possessive clause possessor argument of possessive clause past question word or particle quotative single argument in intransitive clause singular visual evidential Written Tibetan
BIBLIOGRAPHY Aikhenvald, Alexandra Y. and R.M.W. Dixon, eds. 2003a. Evidentiality. (Studies in Evidentiality). Amsterdam: John Benjamins. Aikhenvald, Alexandra Y. and R.M.W. Dixon, eds. 2003b. Areal di.f/Usion and genetic Inheritance: problems in comparative linguistics. Amsterdam: John Benjamins. Anderson, Lloyd B. 1986. Evidentials, paths of change, and mental maps: typologically regular asymmetries. In Wallace L. Chafe and JohamJ.a Nichols, eds., Evidentiality: The Linguistic Encoding a/Epistemology. NOIWood, NJ: Ablex, 273-312. Bartee, Ellen. 2007. Dongwang Tibetan. PhD dissertation presented to the Department of Linguistics, UC Santa Barbara. Beyer, Stephen, V. 1992. The Classical Tibetan language. New York: SUNY Press. BracD.ey, David 1975. Nahsi and Proto-Bmmese-Lolo, Linguistics of the TibetoBurmanArea 2.1: 93-150.
180
ELLEN BARTEE
Chafe, Wallace L. and Nichols, Joharma, eds. 1986. Evidentiality: The Linguistic Encoding ofEpistemology. Norwood, NJ: Ablex. Coales, Oliver. 1919. Eastern Tibet. The Geographical Journa/53.1: 228-249. Connie, Bernard. 1998. Rethinking the typology of relative clauses. Language Design 1.1: 59-86. Curnow, Timothy Jowan 2001. Why 'first/non-first person' is not grammaticalized mirativity. In Proceedings ofALS2k, the 2000 conference of the Australian Linguistic Society. http://www.als.asn.au, 1-10. - - . 2002. Evidentiality and me: The interaction of evidentials and first person. In Proceedings of the 2001 Conference of the Australian Linguistic Society. http://www.als.asn.au. 1-10. - - . 2003 Verballogophoricity in African languages. In Proceedings of the 2002 Coriference of the Australian Linguistic Society, 1-8. DeLancey, Scott. 1985. Lhasa Tibetan evidentials and the semantics of causation Berkeley Linguistic Society 11: 65-72. - - . 1986. Evidentiality and volitionality in Tibetan. In Wallace L. Chafe and Johanna Nichols, eds., Evidentiality: the linguistic encoding of epistemology. Norwood, NJ: Ablex, 203-13. - - . 1997. Mirativity: the grammatical marking of unexpected infonnation Linguistic Typology 1: 33-52. - - . 2001. The mirative and evidentiality. Journal ofPragmatics 33: 369-82. Denwood, Philip. 1999. Tibetan. Jolm Benjamins. Dixon, RM.W. 1997. The Rise and Fall ofLanguages. Cambridge University Press. Epstein, Lawrence, ed 2002. Khamspa histories: visions ofpeople, place and authority. Proceedings of the ninth seminar of the IATS. Vol. 4 Leiden: Brill. Garrett, Edward. 2001. Evidentiality and assertion in Tibetan. PhD dissertation presented to the linguistics department at UCLA. Genetti, Carol. 1986. The development of subordinators from postpositions in Bodie languages. Paper presented at The twelfth annual meeting of the Berkeley Linguistic Society, Berkeley. - - . 1994. A descriptive and historical account of the Dolakha Newari dialect. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Gesang Jumian and Gesang Yangjing, eds. 2002. MiH;/J7!fMi:8 Zangyu fangyan gailun ~n overview ofTibetan dialects). Beijing: Minorities Publishing House. GivOn, Talmy. 1982. Evidentiality and epistemic space. Studies inLanguage 6: 23-49. Gordon, Lyrm. 1986. The development of eviden1ials in Maricopa. In Wallace L. Chafe and Johanna Nichols, eds., Evidentiality: the linguistic encoding of epistemology. Norwood, NJ: Ablex, 75-88. Grimes, Barbara F. and Joseph E. Grimes, eds. 2004. Ethnologue: languages of the world 14th edition Dallas: SIL. www.ethnologue.corn Hale, Austin 1980. Person markers: finite conjmtet and di.sjmtet verb forms in Newari. In Ronald Trail, ed, Papers in South-East Asian Linguistics 7, Canberra: Pacific Linguistics, 95-106. Haller, Felix. 2000. Verbal categories of Shigatse Tibetan and Themchen Tibetan. Linguistics ofthe Tibeto-BunnanArea 23.2: 174-188. Hargreaves, David 1990. Indexical ftmctions and grammatical subsystems in KathmanduNewari. Chicago Linguistic Society 26: 179-193. - - . 1991. The conceptual structure of intentional action: data from Ka1hmanduNewari. Proceedings ofthe seventeenth annual Berkeley Linguistics Society 17: 379-389.
ANIMACY IN DONGWANG TIBET AN
181
- - . 2005. Agency and intentional action in Kathmandu Newar. In Himalayan Linguistics 5: 1-48. http://www.lingui.stics.ucsb.edu/HimalayanLinguistics/articles/2005/ HLJ05_Hargreaves_agency. pdf Hasler, Ka1rin Louise. 1999.A grammar of the TihetanDege (Sde dge) dialect. Ziirich: Selbstverlag. _ _ . 2001. An empathy-based approach to the description of the verb system of the Dege dialect ofTibe1an. Balthasar Bickel (ed.). Person and evidence in Himalayan languages. Special issue oflinguistics ofthe Tiheto-Bwman area. 2. 24.1: 1-34. He Jiren, Jiang Zhuyi, eds. 1985. l?fi!fffG{Jfjlf Naxiyu Jianzhi (A sketch of the Naxi language). Beijing: Minorities Publishing House. Hongladarom, Krisadawan. 1996. rGyalthang Tibe1an ofYurman: a preliminary report. Linguisticsofthe Tiheto-BurmanArea 19.2: 69-92. - - . 2000. Indexical categories in Kham Tibe1an and Central Tibe1an. Paper presented at The 6th Himalayan Languages Symposium, University of Wisconsin, Milwaukee, June 15-17. - - . N.d. The Khampas of Tibet's eastern frontiers: language, identity, and etlmohistory. Paper presented at the 2nd Amtual Meeting of the Asian Scholarship Foundation, 1-2 July 2002, Bangkok. Hopper, Paul J. and Elizabeth Cross Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Hu Tan, ed. 1986. Lasa kouyu duhen. (colloquial Lhasa reader) Beijing: Minorities Publishing House. Huber, Brigitte. 2000. Preliminary report on evidential categories in Lende (Kyirong) Tibe1an. Linguistics ofthe Tiheto-BwmanArea 23: 155-174. LaPolla, Randy. 1999. The role of migration and language contact in the Sino-Tibe1an language family. In Aikhenvald, Alexandra Y. and R.M.W. Dixon, eds., Areal diffosion and genetic inheritence: case studies in language change. Oxford: Oxford University Press. LaPolla, Randy and Chenglong Huang. 2002. The copula and existential verbs in Qiang. Paper presented at Workshop on copula clauses and verbless clauses held at the Research Centre for Linguistic Typology. Matisoff, James A. 1993. Sangkong ofYwman: secondary ''verb pronominalization" in Southern Loloish. Linguistics ofthe Tiheto-BurmanArea 16.2: 123-142. Sch.Ottelndryer, Bmkhard. 1980. Person markers in Sherpa. In: Ronald Trail, ed, Papers in South-East Asian Linguistics, No.7. Canberra: Pacific Linguistics, 125-130. Slater, Keith W. 1998. A Grammar ofMangghuer: A Mongolic language of China's Qinghai-Gansu sprachhund. PhD Dissertation presented to the department of linguistics at UC Santa Barbara. Sun, Jackson. T.-S. 1993. Evidentials inAmdo-Tibe1an. Bulletin ofthe Institute afHistory and Philology, Academia Sinica 63-64: 945-1001. - - . 2001. Phonological profile of Zhongu: a new Tibe1an dialect of Nor1hern Sichuan. Paper presented at The Workshop on Tiheto-BwmanLanguages, UC Santa Barbara. Timrgood, Graham. 1986. The nature and origins of the Akha evidential system. In Wallace L. Chafe and Johanna Nichols, eds., Evidentiality: the linguistic encoding ofepistemology. Norwood, NJ: Ablex, 214-221. Timrgood, Graham and Randy J. LaPolla, eds. 2003. The Sino-Tibetan languages. London: Routledge. Thomason, Sarah Grey and Terrance Kaufman 1988. Language contact, creolization, and genetic linguistics. Berkeley: University of California Press.
182
ELLEN BARTEE
Thomason, Sarah Grey. 2001. Language contact. Washington, D.C. Georgetown University Press. Toland, Norma R. & Donald F. Toland 1991. Reference grammar of the Karo!Rawa language (Data Papers on Papua New Guinea Language 38). Ukarumpa: Smnmer Institute of Linguistics. Toumadre, Nicolas. 2001. Final auxiliary verbs in Literary Tibetan and in the dialects. Linguistics ofthe Tibeto-BunnanArea 24.1: 49-110. Toumadre, Nicolas and Sangda Dorje. 2003. Manual ofStandard Tibetan. New York: Snow Lion Publications. van Driern, George L, with Karma Tsering (Collaborator). 1998. Dzongkha. Languages ofthe Greater Himalayan Region, 1 Leiden: CNWS. van Spengen, Wim, 2002. Frontier history of Southern Kham: Banditry and war in the multi-ethnic fringe lands ofChatring, Mili, and Gyethang. 1890-1940. In Epstein, Lawrence, ed., Khamspa histories: visions of people, place and authority. Proceedings ofthe ninth seminar ofthe IATS. Leiden: Brill, 4.7-30. Vasquez de Ruiz, Beatriz. 1988. La predicacwn en Guambiano (Lenguas Aborigenes de Colombia, Descripciones 2). Bogota: Centro Colombiano de Estudios en Lenguas Aborigenes, Universi.dad de los Andes. Vesalainen, Olavi and Ma.Ija Vesalainen. 1980. Clause patterns inLhomi. Pacific Linguistics B, 53. Canberra: Australian National University. Wang Hengjie. 1995. i!h&M/i!ti:f
NG
Negative forms are derived from affirmative forms by simple rules. The negative preftx <man-> (NG) negates the lexical verb tuma (vi) 'to be' in all its meanings, i.e. the existential, attributive and locational meanings, and is used for the negative forms of twna (vi) 'to be' in its function as an auxiliary in the perfect tenses. The negative preftx <man-> (NG) is also used for the negation of all non-fmite verb forms. The negative sufftx (NG) negates any other indicative verb form, including indicative forms of the homophonous verb twna (vi) 'to sit', which is never marked by the negative preftx <man-> (NG). 3.2. Suffzxal slot sfl: reflexive voice, the 1---+2 scenario marker and third person patient
1---+2 RFL
3PINPT 3P
The transitive scenario marking motpheme ( 1---+2) indicates a transitive relationship between a ftrst person agent and a second person patient As can be seen in Table 3, second person patient forms constitute a subsystem in the Sampang transitive simplex conjugation where the expression of an agent is limited to forms with a frrst person agent and
THE SAMPANG VERBAL AGREEMENT SYSTEM
197
the 3p---+2s form. The portemanteau motpheme (1---+2) and the second person motpheme (2) are in complementary distribution. The limited expression of an agent in second person patient forms causes homophony of the intransitive 2s forms and the transitive 1e/3np---+2s forms, of the intransitive 2d forms and the transitive n1s---+2d, and of the intransitive 2p forms and the transitive n1s---+2p forms. The motpheme ( 1---+2) shows an asymmetrical distribution in the transitive paradigm. The suffix (1---+2) occurs in all 1s---+2 forms, but scenarios with a ftrst person non-singular agent and a second person dual or plural patient can be expressed by 1---+2d and 1---+2p forms with the sufftx (1---+2) or by n1s---+2d and n1s---+2p forms, which do not contain the sufftx (1---+2). The distinction between a singular agent and a non-singular agent in 1---+2 forms is more important when a singular second person patient is involved than when a non-singular second person patient is involved. Sufftxal slot sfl also contains the frrst nasal element of the discontinuous marker or simulftx (RFL), which expresses reflexivity. The second element of this motpheme can be found in sufftxal slot sf5. As can be seen in Table 1, the reflexive simulftx (RFL) occurs in intransitive forms. In addition to the 1---+2 scenario marker (1---+2) and the frrst part of the reflexive simulftx (RFL), sufftxal slot sfl also contains the non-preterite third person patient motpheme (3PINPT) and the third person patient sufftx (3P). The non-preterite third person patient motpheme (3P/NPT) has the allomotph before /-m/ and the allomotph before the frrst person singular sufftx (1s). After the assimilatory nasalisation of the nonpreterite third person patient motpheme (3PINPT), the frrst person singular sufftx ( 1s) elides, leaving its zero allomotph. In preterite 1di---+3, 1de---+3, 2d---+3, 3d---+3, 2p---+3 and the alternative preterite 1pi---+3 and 1pi---+3ns forms, the third person patient marker (3P) has a zero allomotph before the preterite tense motpheme (PT) through the rule vocal is ante vocalem corripitur.
3.3. Suffzxal slots sf2 and sj3: tense and first person patient or subject sf2:
PT
198 sf3:
RENE HUYSMANS
lsPS lsPINPT lpPS
Time is expressed in Sampang by two tenses, preterite tense and nonpreterite tense, but denoted by a single tense morpheme, the preterite tense morpheme (PT). The existence of a single tense sufftx follows from the fundamental principle against redundant tense marking in the Sampang verbal agreement system. The exact rules for the interpretation of Sampang tense will be discussed in my forthcoming grammar of Sampang. In this article, I concentrate on the morphological expression of tense. The preterite tense morpheme (PT) has a zero allomorph before any vowel, as regulated by the rule vocalis ante vocalem corripitur. The overt allomorph of the preterite tense morpheme (PT) occurs in transitive forms with a ftrst person dual or second person patient and in most intransitive and reflexive forms. In third person patient forms, tense is expressed either by the non-preterite third person patient morpheme (3PINPT) or by the preterite tense marker (PT), as described in Section 2. However, in intransitive and reflexive forms with a ftrst person singular subject, the preterite tense morpheme (PT) is reduced to its zero allomorph before the frrst person singular patient or subject marker ( lsPS), resulting in homophony of the two tenses. The morpheme ( lsPS) is also used to index a frrst person singular patient in the transitive conjugation and this causes homophony of the non-preterite 3s~ 1s forms and the intransitive 1s forms in the preterite and non-preterite tenses. In intransitive forms with a frrst person plural subject, the zero allomorph of the preterite tense morpheme (PI') occurs before the frrst person plural patient or subject morpheme (lpPS) from sufftxal slot sf3, again resulting in homophony of the preterite intransitive lpi and lpe forms on the one hand and of the non-preterite intransitive lpi and lpe forms on the other. These observations also apply to the reflexive conjugation. In transitive froms with a frrst person singular patient, the preterite tense marker (PT) is redundant and therefore not selected because transitive forms with a frrst person singular patient show a morphological differentiation of tense through the non-preterite portemanteau
THE SAMPANG VERBAL AGREEMENT SYSTEM
199
morpheme (lsPINPT) in non-preterite forms and the ftrst person singular patient or subject marker (lsPS) in preterite forms. This strategy is sufficient for distinguishing the two tenses. There is a side-effect related to the tense distinction in frrst person singular patient forms, which consists of expressing the patient or the agent more explicitly in the preterite forms than in the corresponding non-preterite forms. The preterite 2d---+ ls and 3d---+ ls forms are augmented with the frrst person singular copy morpheme ( ls ') after the allomorph of the dual number morpheme (d), the preterite 2p---+ ls form contains the frrst and second person plural agent suffix (12pA) following the allomorph of the second person plural morpheme (2p) and the preterite 3p---+ls form shows the third person non-singular agent marking sufftx (3nsA) after the allomorph of the third person non-singular agent or subject marker (3nsAS). The morphological transitivity of the preterite forms with a frrst person singular patient and a non-singular agent is a sign of the higher level of semantic transitivity associated with preterite tense. From a purely economical point of view, however, increased morphological transitivity is redundant in the frrst person singular patient forms, since the tenses are already distinguished through the portemanteau morpheme (lsP/NPT) and the frrst person singular patient or subject marker (lsPS). The increased morphological transitivity marking a higher level of semantic transitivity in the preterite forms can also be found in the 3p---+lpi and 3p---+lpe forms, where it has developed into a full-blown tense marking strategy. The preterite 3p---+lpi and 3p---+lpe forms are augmented with the third person non-singular agent morpheme (3nsA) after the allomorph of the third person non-singular agent or subject marker (3nsAS) but not the non-preterite 3p---+lpi and 3p---+lpe forms, which end in the allomorph ofthe third person non-singular agent or subject marker (3nsAS). In all other transitive forms with an inclusive or exclusive frrst person plural patient and in intransitive forms with a frrst person plural subject, the preterite tense morpheme (PT) is reduced to its zero allomorph before the frrst person plural patient or subject morpheme (lpPS). Evidently, the expression of tense with a ftrst person plural patient or subject is less important, except when the agent is third person plural. The formal similarity of the non-preterite ftrst person singular patient morpheme ( lsPINPT) and the sufftx
200
RENE HUYSMANS
(lpPS) causes homophony of the non-preterite 3s---+ls form and the 3s---+lpi forms, and the intransitive lpi form in both tenses, and of the non-preterite 3p---+lpi and 3p---+ls forms in both tenses. 3.4. Suffzxal slot sf4: first person
ls lpA'
The morpheme ( ls) indexes a first person singular actant, irrespective of its grammatical role and has a copy, the frrst person singular copy morpheme ( ls '), in sufftxal slot sf8. The copy suffiX (lpA') expresses a first person plural agent and is the echo of the first and second person plural agent marker (12pA) from suffJXal slot sf8. After the frrst person singular patient or subject suffiX (lsPS) and the non-preterite third person patient morpheme (3PINPT), the frrst person singular morpheme ( ls) elides at the surface, leaving a zero allomorph and assimilatory nasalisation of the preceding vocalic sufftx as the only trace of its former presence. The nasal quality of the frrst person singular patient or subject suffiX (lsPS) is explained by the fact that it is always followed by the frrst person singular morpheme ( ls ). 3.5. Suffzxal slots sf5 and sj7: number and reflexive voice sf5: < ... -ici ~ -ci ~ -cu> sf7:
23s d RFL
2p 3nsAS 3nsAP
The morphemes of suffJXal slots sf5 and sf7 are described in a single section because they share many allomorphic rules, and most of them also share function. The second or third person singular number morpheme (23s) elides after any vocalic suffiX and after /g/. The four number morphemes (d), (3nsAP), (2p) and (3nsAS), and the second part of the reflexive simulftx
THE SAMPANG VERBAL AGREEMENT SYSTEM
201
(RFL) have the regular allomorphs , , , and respectively after any suffJX consisting of, or ending in, a vowel and after /g/ or /m/. In this postion, the final vowel /if of these four suffJXes is assimilated to lui before any syllable-fmal bilabial or velar nasal /m/ or /g/, resulting in the allomorphs , , , and respectively. The dual number morpheme (d) has the regular allomorph in intervocalic position. For some native speakers, the third person non-singular agent morpheme (3nsap) has the allomorph after the first person plural agent copy suffJX (1pA') and before the first or second person plural agent suffix (12pA). Other native speakers use the allomorph of the third person non-singular agent morpheme (3nsAP) in this position. The third person non-singular agent or patient morpheme (3nsAP) has the allomorph with a small number of irregular verbs, e.g. the non-preterite 3s---+3ns/3p---+3 form ci-ici (eat-3nsAP) 'he/she/it/theyos will eat himlher/it/themns• of the irregular verb ca,ma (vt) 'to eat'. With these irregular verbs, the morpheme expresses a transitive relationship between a frrst person agent and a third person patient in the non-preterite tense (1---+3/NPT), instead of the regular third person patient meaning in the non-preterite tense (3P/NPT). Singular number of a second or third person actant is indexed by the number morpheme (23s ). The overt allomorph of the second or third person singular number suffJX (23s) is found in the nonpreterite intransitive 2s and 3s forms and in the non-preterite forms with a second person singular patient In accordance with the limited expression of agents in second person patient forms, I shall assume that the second or third person singular number morpheme (23s) can only express singular number of the second person actant here. The dual number morpheme (d) expresses dual number of frrst, second or third person actants. The number suffJX may be inserted only once into suffJXal slot sf5 and can refer to only one actant The presence of only one dual number sufftx in the Sampang simplex template makes ascertaining with which actant the dual number morpheme (d) agrees impossible in transitive forms which express a scenario with a dual agent and a first person dual patient, viz. in the 2ns---+ 1de, 3ns---+ 1di and 3d---+ 1de forms. Agents in forms with a second person patient can only be expressed by the portemanteau morpheme (1---+2) or by the third person non-singular agent or sub-
202
RENE HUYSMANS
ject morpheme (3nsAS), which means that any occurrence of the dual number suffJX (d) in these forms indexes unequivocally the number of the patient. The morpheme (3nsAS) expresses a third person nonsingular agent or subject and is treated in more detail below in Section 3.7, together with the third person non-singular agent morpheme (3nsA). The plural second person number morpheme (2p) expresses a second person plural subject in the intransitive conjugation and a second person plural agent or patient in the transitive conjugation. The second person suffix (2) is redundant in a form marked by the plural second person number morpheme (2p). All forms with a second person actant share a common format, with number being expressed by the second or third person singular morpheme (23s), the dual morpheme (d) and the second person plural morpheme (2p). In the third person patient system, the 2p~3 forms follow the alternative tense marking strategy with the preterite tense suffJX (PT) as in 2d~3 forms because, according to the present analysis, the plural second person number morpheme (2p) blocks selection of the third person non-singular agent or patient suffix (3nsAP), use of which appears to be an integral part of the default tense marking strategy as with 2s~3 and 2s~3ns forms. Whereas the morphology of Sampang simplicia differentiates between dual number and plural number with frrst and second person patients, third person patients can either have no number or non-singular number. The morpheme (3nsAP) expresses a third person nonsingular agent or patient The third person non-singular agent or patient morpheme (3nsAP) supplants the third person non-singular agent or subject morpheme (3nsAS) in the third person patient system, so that there is a single 3p~3/3s~3ns form, and is used in any form where the number of a third person patient is specified, i.e. in the ls~3ns, lpi~3ns and lpe~3ns, 2s~3ns and 3s~3ns forms. Expression of the number of the third person patient by means of the third person non-singular agent or patient morpheme (3nsAP) is in all cases optional. This is analogous to the optional Sampang nominal non-singular number suffJX (ns) suffJXed to a noun (Huysmans, forthcoming). SuffJXal slot sf5 also contains the second element of the reflexive simulfix (RFL). In reflexive forms with a dual subject, I shall assume that the element /ici/ is to be identified with the second
THE SAMPANG VERBAL AGREEMENT SYSTEM
203
-ici> (RFL) and not element of the reflexive simulfJX (3pA), in Sampang reflected by the third person non-singular agent or subject marker (3nsAS). The Sampang third person non-singular agent or subject marker (3nsAS) is cognate with Limbu (3nsAS), Dumi (3ps), Hayu (3p), Thulung (3p~3), Lohorung (3p), Bahing (3p~3/3ps) and (3p/1pe), Wambule (3~3p) and (3ns) and Yamphu (3p). Unlike most Kiranti reflexes of the third person plural agent proto-prefix *<me-> (3pA), the Sampang reflex occupies a rather posterior location in the a:fftxal string. The second Proto-Kiranti morpheme for the expression of a plural agent, the proto-suffix * (12pA), tends to be reserved for first and second person plural agents and is located in a more posterior po-
THE SAMPANG VERBAL AGREEMENT SYSTEM
213
sition in the Proto-Kiranti morphological template. The present-day Sampang first and second person plural agent sufftx (12pA) can safely be interpreted as a reflex of the Proto-Kiranti first and second person plural agent sufftx * ( 12pA). The third person nonsingular agent suffix (3nsA) may also be a reflex of this protomorpheme. The Sampang first and second person plural agent sufftx (12pA) and perhaps the third person non-singular agent sufftx (3nsA) are cognate with other Kiranti agentive morphemes such as Limbu (pA) and Lohorung (1peAS) and (2p---+3). The Sampang reflexive morpheme (RFL) can be related directly to the Proto-Kiranti reflexive morpheme * (RFL) and is cognate with other reflexive suffixes found in Kiranti languages, such as Limbu (RFL ), Dumi (RFL ), Hayu (RFL), Bahing (RFL), Thangmi (RFL) and Kulung (RFL). In almost all Kiranti languages, the reflexive sufftx fills a single slot, mostly immediately after the verb stem. The distribution over two sufftxal slots of the Sampang simulftx (RFL), namely the first suffixal slot sfl and the number slot sf5, is in all probability a subsequent Sampang reanalysis of the element by analogy with the dual number sufftx (d). In Kiranti, the exclusive proto-sufftx * appears to have been subject to fusion with other sufftxes on more than one occasion. First, the Proto-Kiranti exclusive morpheme * (e) may have fused with the second person proto-morpheme * (2) yielding the ProtoKiranti 1s---+2 morpheme *. This tentative reconstruction by van Driem (1990: 45-6; 1991: 350) is prompted by the close formal resemblance between the Kiranti reflexes of the morpheme * (1s---+2) and the Proto-Kiranti second person morpheme * (2). The Sampang 1---+2 morpheme in sufftxal slot sfl can be related directly to the Proto-Kiranti morpheme * (1s---+2) and is cognate to Limbu (1---+2), Dumi (1s---+2), Hayu (1s---+2), Thulung (1s---+2), Bahing (1s---+2), Wambule (1s---+2), Yamphu (1---+2), Chepang (1---+2) and Kulung (1s---+21NPT). On the basis of semantic criteria, the 1s---+2 proto-morpheme * is located after the tense morphemes in van Driem's model of the Proto-Kiranti verb, in the same slot as the Proto-Kiranti second person sufftx * (2) (van Driem 1991: 350). However, most Kiranti reflexes of the morpheme * ( 1s---+2), including the
214
RENE HUYSMANS
Sampang reflex, are fmmd in a position anterior to the tense slot The Pre-Kiranti origins of the proto-morpheme * (1s---+2) may occupy the second person morpheme * (2) slot (van Driem 1993a: 320), but perhaps it had already been promoted to a position closer to the stem in Proto-Kiranti times. I suggest (with Rutgers 1993: 120) that the 1s---+2 proto-morpheme * in the Proto-Kiranti morphological template be positioned before the tensed auxiliary AUX 1, in the same slot as the reflexive proto-morpheme * (RFL ). The exclusive proto-suffix * (e) may also have fused with the Proto-Kiranti frrst person plural suffix * (1p) (van Driem, 1990, 1991, 1992), yielding the Sampang exclusive patient or subject morpheme (ePS) and exclusive agent marker (eA), and other cognate reflexes, such as Kulung (e) and Lohorung (e). The coexistence of two exclusiveness markers is a subsequent development in Sampang. The third possible candidate for a coalescence of the Proto-Kiranti exclusive marker * (e) and another proto-marker is peculiar to Sampang. The Sampang frrst person plural patient or subject morpheme (1pPS) may be a fusion of the Proto-Kiranti exclusive marker * (e) and the Proto-Kiranti inclusive suffix * (i), as is suggested by Kulung, one of Sampang's closest relatives, which has the reflex (1pPS/NPI') of the exclusive proto-sufftx * (e) in non-preterite frrst person plural patient and subject forms and the reflex (1pPS/Pr) of the Proto-Kiranti inclusiveness suffix * (i) in preterite frrst person plural patient and subject forms (Tolsma 1997: 69). If this analysis is accepted, then the Sampang frrst person plural patient or subject (1pPS) is cognate with the Dumi inclusive marker (i) and other Kiranti (frrst person) plural markers such as Limbu (pPS), Thulung (1pi---+3), Lohorung (1pPS), Wambule (1pi---+3), Yamphu (12p), Thangmi (1pPS) and the Limbu sufftx (ls/NPT). The Sampang non-preterite first person singular patient marker (1sP/NPT) is cognate with Limbu (1s/NPI') and may represent a reflex of the Proto-Kiranti exclusive marker * (e). The Sampang preterite tense sufftx (PT) is cognate with Limbu (PT), Lohorung (PT), Yamphu (PT) and Kulung (PT), and can be readily derived from the vowel of the ProtoKiranti preterite tense morpheme * (PT). The anterior position
THE SAMPANG VERBAL AGREEMENT SYSTEM
215
of the Sampang preterite tense morpheme (PT) is in support of an anterior position of the Proto-Kiranti tense morphemes. The Proto-Kiranti second person marker * (2) is reflected by the Sarnpang second person morpheme (2), which is cognate with Thulung (2), Lohorung (2), Thangmi (2s) and Kulung (2s/PT). In the majority of Kiranti languages, the reflexes of the second person proto-marker * (2) occur in the initial portion of the sufftxal string. The location of the Sampang second person morpheme (2) towards the end of the sufftxal string in sf9 is, from a Proto-Kiranti and a Proto-Tibeto-Burman perspective, highly posterior and most certainly a Sampang innovation. The Sampang reflex of the second person proto-marker * (2) shares its posterior location in the sufftxal string with the Lohorung reflex. The Sampang third person non-singular agent or patient morpheme (3nsAP) is in all probability a reflex of the Proto-Kiranti third person dual patient morpheme * (3dP). After over-generalisation of the original dual meaning of the third person dual patient protomorpheme * (3dP), the proto-sufftx carne to express non-singularity (cf. van Driem 1987: 31-2). The Sampang third person non-singular agent or patient morpheme (3nsAP) is cognate with Limbu (ns), Thulung (3dP), Lohorung (nsP), Bahing (3d), Kulung (3ns) and Yarnphu (3ns). The Sarnpang dual number morpheme (d) is likely to be the reflex of the dual proto-morpheme * (d) and is cognate with numerous Kiranti dual morphemes such as Limbu (d), Thulung (d), Lohorung (d), Hayu (d), Wambule (d), Yamphu (d), Kulung (d) and Chepang fi
j > fi
j >0
Sham
-
-
0-1,e
LEH
-
-
0-i,e
-
PIP HML
? -
MND CEM
? ?
-
? -
-
+
-
SHA
fi-a,o,u
-
GYS
+ +
+ +
ft-a,o,u
ft
-
NYO
+
-
-
J > nr;: ~ > fij
section
2.1
J > r;: _y
fricativisation of clustel initial vel
medial
(sub-) phonernic tone
vcd
cluster
cluster
lU
cluster
-
cluster
y: xd,xb
cluster
X.,
tjJ, but the exact ranking could vary somewhat with the type of articulation. Voiced consonants were perceived as being lower than unvoiced consonants originating from a voiced one, these again lower than aspirated consonants, and those lower than an (originally) unvoiced, non-aspirated consonant (g > k (< ~g) > kh > k). But in this case, the ranking could be completely overthrown by the ranking of vowels, which is more or less binary: a and o being perceived as low, i and u as high. 35 The vowel e, if realised as [e], would be likewise classified as high, but the consultant has a strong tendency to pronounce it as [e] or even [~e] in elicited words and sentences, and this pronunciation is in line with its classification as 'low'. Aspirated consonants as well as unvoiced sibilants were initially classified as 'low', later as 'neutral'. Quite surprisingly, and in contrast to the Shara consultant, the Gya consultant did not describe plain nasals as lower than voiced stops, and this relative height (but still in opposition to high tone nasals resulting from clusters) might be the reason why nasals in the coda position or at a following syllable onset interfered with the consultant's perception of tone: syllable final nasals, in particular 1J and m regularly led to a classification as 'high(er)'. The same effect occurred when an open syllable was followed by a syllable with nasal or high tone onset. To give only one example of the complexity and subtleness of the consultant's perception, the low-tone nasals with 'low' vowel a were ranked in between the 'neutral-tone' aspirated velars with 'high' vowel i or u and the low-tone unvoiced velar with 'high' vowel: ga > gu > kha > ka > ku > gi > khi I khu > ma Ina I pa I ya > ki I tri > kii > kil> kf
With all these subtle phonetic distinctions, the binary phonemic tone distinction seemed to be lost completely, and it was only consistent when the consultant repeatedly classified words such as /kore/ kore 'cup' and /tal rta 'horse' as 'low'. Finally, on understanding the various interacting factors and the relativity of each statement, I used ex35 This, at least, is not only due to the consultant's particular musical awareness. A second consultant, with whom I had the opportunity to work in 2009, got several times confused, misclassifying at first low tone words with the 'high' vowels i or u as having high tone and arriving at the correct description only after a lot of thinking.
254
BETIINA ZEISLER
actly these two words as fixed points of comparison to establish a 'high tone' for a level intonation despite a 'low' vowel, particularly when there was no low tone counterpart to be found with the exact match of vowel and final. 36 While the tonal distinction established by this method roughly corresponds to what a Tibeto-linguist would expect, I observed some anomalies, which could indicate that 'tone', whether defined as pitch or register or defmed as tone contour or as a combination of both, is not yet a true phonemic feature in the dialect of Gya, despite the fact that a few minimal pairs can be found. 37 As we would expect, plain 36 We used the terminology /rombo/ rompo 'fat, stout' for low tone and /trhamo/ phramo 'thin' for high tone, alternatively to phoskad 'male voice' for low and moskad 'female voice' for high tones. The expressions rompo andphramo are common among musicians, but obviously also understood by non-professionals (Rebecca Norman, p.c.). Ths terminology was not without disturbing side effects. When we discussed family terms based on the old family 'prefix' a-, the consultant divided 1hese somewhat arbitrarily in /rombo/ and /trhamo/, changing the classification for some of them from week to week or day to day. When I realised that the majority of the terms for females were classified as 'thin' and the majority of the terms for males as 'fat', and joked that especially the lane! 'aunts' were loosing and gaining weight, the consultant started reclassifying all terms along gender lines. Nevertheless, after the long break between my 2005 and 2006 fieldwork, she eventually declared that the differences were artificial and all terms were 'high' tone, in line with linguistic expectations. TI Quite to the smprise of most Sinologists, Beckwith (2005: 10) challenges the concept of phonemic tone in Chinese, since, according to him, a wrong intonation in "actual connected speech" might not necessarily lead to a different meaning for the listener, whereas any change in consonants and vowels does. He describes tone as "at best, an extra, a redundancy feature" that together with other features could enhance clarity. Beckwith, however, does not seem to be aware that different kinds of minimal pairs have different tolerance features, tlrus the wrong choice of articulation place is typically less tolerable than the wrong choice of articulation manner (particularly when the articulation places are quite distant) and the latter might be less tolerable than mistakes in accentuation or tone (I myself owe this insight to discussions with Thomas Preiswerk). Depending on the language, tonal contrasts may be quite articulate or very subtle. In the latter case, speakers certainly make use of all o1her available cues, especially contextual ones. This makes it very difficult to define the stage where a merely phonetic and gradual differentiation of tones turns into a truly phonemic, that is, discrete opposition Given the fact that a phonemic opposition cannot be gradual, by definition, one could perhaps even argue that what is perceived as low and extra low tone in Gya are both concomittant features of two types of voicing, which might perhaps be called 'semi-voiced' and 'enhanced voicing' The former type may be characterised by its additional aspiration feature. The latter type is found with the additional feature of prenasalisation in various Tibetan varieties, e.g. in Zanskari, Tabo Spiti (Veronika Rein, p.c.), and Kyirong (Huber 2005: 15; cf. also note 41 below for her description), sometimes also with fricativisation and/or aspiration as mentioned in note 26 above.
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
255
nasals and laterals were typically classified as being low tone while those with a pre-radical were classified as high. The same effect should have been found with the palatal approximant y as a radical, but quite a few words that have an Old Tibetan pre-radical and that have high tone in the Shara dialect were qualified as low. As these words are found only with the 'low' vowels a and o, high tone is apparently only perceived when the need for differentiation is high, e.g. /~/ *g.yar 'die, croak, perish',38 in contrast to /ja/ gyaf:t 'feel ticklish' and /jar/ gyarcas 'borrow'. The most striking case is perhaps the contrast between high tone /jok/ *g.yog 'twist (threads)', although not corroborated by any classical written form, and the two compounds based on this verb: /jQkto/ *yogto 'coarse thread' and /jQkJig/ *yogsiy 'stick for twisting' with low tone. In several other cases, the consultant was quite confused about the proper classification. 39 Family tenns, starting with a glottalised a in Old Tibetan, were arbitrarily classified as either low or high (see also note 29), while all were unmistakably classified as high by the Shara consultant Some anomalies in the nasal class were corroborated by the Shara consultant and are thus of historical interest: /mak.pa/ 'husband, sonin-law' was unanimously classified as being high tone, thus the alternative spelling dmagpa and the relation to dmag 'army' mentioned by JAK are obviously correct for Ladakhi.40 Another, quite unexpected case of high tone is GYS /moll for CT mol 'speak (hon)', where other tonal varieties of Ladakh and Western Tibet show low tone realisation, e.g. Trangtse and Man-Merak, both in Upper Ladakh, and Tabo Spiti (CDID). However, the Balti form /hmol/likewise indicates that The Domkhar prommciation /hjar/ points to a pre-radical. E.g. /jiik/ g.yag '(male) yak' and/jOkpo/ g.yogpo, because oftheir 'low' vowel, were repeatedly classified as having low tone, although they were consistently uttered with a level tone contour. Highly surprised, I challenged this from time to time, and after a while, the consultant first conceded that it might perhaps be 'neutral' tone, while in 2008 she finally corrected herself, assigning the same tone to /jiik/ as to /jar/ 'borrow' 40 Note, however that other Western Tibetan dialects have a low tone realisation, e.g. Drokpa (Bawa, Bongba Tshogu), Dingri, Mustang/~ 'bride groom, son-inlaw (living in the wife's family)' (Kretschmar 1986: 420, Herrmann 1989: 360; Kretschmar 1995 IV: 143). Low tone realisation is also reported from Ngari and Kyirong, but o1herwise it seems that a greater part of the Central Tibetan dialects, as well as Kham and Dzongka, show high tone realisation (CDTD). By contrast cr mig 'eye' < OT dmyig is found with high tone or even pre-radical in most modern dialects (cf CDTD). The Gya prommciation /nuK/ is no exception 38 39
256
BETIINA ZEISLER
the word must have had (a variant with) a pre-radical: *dmol, as suggested in the CDTD, or *smol, as fmmd in the writings of an Upper Ladakhi scholar (RN). Interestingly, also the spelling contrast between CT giiis 'two' and iiiSu 'twenty' is reflected in a tonal opposition /pi/ and lpiJul, similarly that between gnam /nMnl 'sky' as used in isolation and nam /nyn/ id. as used in compounds and collocations. By contrast, /m,!b3al 'peacock' was unanimously classified as low tone, making the relation to the classical spelling rmabya somewhat dubious. But low rather than high tone was also observed for gnastshul 'news': GYS /n.!tsull ~ lll.!dzull, but SHA /natsul/ and gnaslugs 'condition': GYS, SHA lll.!dluk/, while the dialects of Shayok and Laga showed the 'regular' high tone: SYK /nazluk/, LAG /naluk/. The word rmilam 'dream' was classified as high: /mttam/ by the Gya consultant, in accordance with linguistic expectation, but as low: /milam/ by the Shara consultant. Unexpected low tone was also observed for GYS, SYK, and LAG ll,!D.gjatllhanrgyas 'together' and other words with initiallh in CT. In general, the Shara consultant, although a singer herseJf, never made as subtle distinctions as her colleague from Gya. Independent of the vowel and the final, the Shara consultant distinguished three main tones: high, neutral, and low, with further graduations of low tone: -
'thramo': the originally voiceless, non-aspirated consonants as well as the nasaL palatal approximant and lateral when combined with a pre-radical
-
'normal' (i.e. neutral): aspirated consonants, voiceless sibilants, and the voiceless laryngal
-
'rompo': devoiced consonants
-
·~saJJ
roma' (lower than that): voiced consonants
-
·~saiJ
roma': plain nasals.
Both the Gya and Shara consultants agreed upon the medium or neutral tonal character of aspirated consonants. This is particularly interesting as it matches almost perfectly the traditional grammatical classification of the four consonant rows or articulation types as pho 'male', that is, 'forced' or 'hard' (k), maniT) 'neutral' (kh), mo 'female', that is, 'soft' (g), and sintu mo 'extremely female/soft' (g). The Changthang dialects of the Durbuk and Nyoma blocks are likewise tonal dialects. However in the case ofNyoma, a clear tonal opposition exists only for words having an voiceless non-aspirated radical
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
257
in Classical Tibetan (high tone) and those having a voiced radical, not preceded by any pre-radical. Outside this opposition tonal descriptions turned out to be more or less arbitrary, independent of vowel properties or syllable codas, and the consultant did not perceive tone as a three or four-way distinction and thus lacked the notion of 'neutral tone'. In fact, voiced consonants were often described as higher than their devoiced counterparts. A similar feature has been described for Derge (Hasler 1999: 257275, for high pitched voiced initials see pp. 267, 269 with further references). Although there is a certain tendency for low pitch realisation of voiced consonants, the pitch may vary considerably from word to word (partially in accordance with the degree to which the consonant has been devoiced), but also in different contexts. Hasler thus argues that pitch only becomes a distinctive (phonemic) feature when the consonant has become completely devoiced. But this neither explains the unexpected high pitch realisation of voiced consonants nor the variations in its realisation. The contrast between the Nyoma and Derge data, on the one hand, and the Gya and Shara data, on the other, clearly demonstrates that the stage where pitch contrasts become phonemic can be reached by two completely different developmental paths. On the frrst path, exemplified by Tibetan grammatical understanding as well as by the dialects of Gya and Shara, the actual realisation of voiced consonants follows the tendency of low realisation, so that the phonemic property merely results from a shift of acoustic awareness, triggered by the devoicing process. As a natural result, the devoiced initials also lose part of their low pitch properties. 41 On the second path, as exemplified by the unrelated dialects of Nyoma and Derge, the actual realisation contradicts the natural tendencies and calls for an explanation. This phenomenon could perhaps be understood as a contrastive reaction vis-a-vis devoiced initials or the neu41 This is corroborated quite nicely by the Kyirong data. According to Huber (2005: 19f), the mwoiced consonants, derived from originally voiced initials show middle tone, those derived from clusters of oral stops and voiced radicals show low tone plus breathy voice, those derived from original clusters of nasal and voiced consonants remained voiced, but display a tonal pattern between middle and low tone. What looks rather like a contradiction, finds its explanation in the fact that voiced consonants are realised with a strong prenasalisation, "exclusively used to reinforce the voiced character" However, the "voiced consonant can be almost devoiced after the prenasalization" (p. 15, emphasis added).
258
BETIINA ZEISLER
tralisation process in general: only those initials that are in the process of losing their voiced quality develop an enhanced and marked low pitch realisation. By contrast, all other initials are interpreted as high and, as in the case of Derge, consequently even produced with a higher pitch. It would be interesting to test whether Nyoma speakers in general, or at least the consultant in question, actually produce voiced initials with a high pitch.
4. PRESERVATION, GENERALISATION, AND EVENTUAL LOSS OF THE OLD TIBETAN SUFFIXES -SI-D (STEM II, IV) AND -DI-s (STEM I)
When one looks only at the verbal stems and not at the complex verb forms that are based on these stems, one may easily get the impression that the Kenhat dialects have lost all Old Tibetan sufftxes, while the Shamskat dialects have well preserved the Old Tibetan sufftx -s (~-d) of stem II and to a certain extent also of stem IV. The picture is, however, much more complex, and one fmds traces of the former suffixes for all three stems in the Kenhat dialects. Even more, these traces are, in certain surroundings (such as the directive and question marker), even more frequent than the corresponding overt Shamskat morpheme. I will begin with stem IV, where the data is obvious and uncomplicated, and will discuss the quite problematic fmdings concerning stem I last.
4.1. Stem IV with directive marker l-ay/ In Shamskat and Leh, the imperative sufftx -s is found only in verbs with an open syllable root. In the Kenhat dialects it is preserved and even generalised before the directive marker /-ag/ - /-sag/ (CT day) in the positive command form. This feature has been systematically checked only for Gya. Here the s-form is always optional, but nevertheless frequently used. According to the consultant, the s-less form indicates that the action is quite urgent and should be performed immediately, whereas the s-form signals that it can also be performed somewhat later. The s-form thus co-occurs with all agentive verbs, independent of their shape. The form /-sag/ appears after open syllable roots as well as after closed syllable roots, including those that do not allow a suffix -s in Old and Classical Tibetan and those where the Old
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
259
or Classical Tibetan stem IV lacks a suffix, e.g. /kok-saJ]!/ kog-sIJa1JISa1J besides lkoal)!/ kog-~a1J 'peel [it] offi' (CT ~gog bkog dgog khog), /kok-sag!/ kog-s-~a1JISa1J or lkoaJ]!/ kog-~a1J 'stop [it!]' (CT ~gegs bkag dgag khog), /kor-SalJ !/ skor-s-~a1JISa1J besides /korag!/ skor-~a1J 'tum[it]!' (CT skor bskor*d bskor: skor*d). In the prohibitive, which is based on stem I, the s-form does not occur. For Nyoma it can be said that the directive marker /-sal)/ is obligatory at least for open syllable roots, in which case it may or may not be accompanied with a fronting of back vowels. A likewise cursory survey of the Pipcha verbs showed that the additional -s is obligatory after vowels, optional after fmal -r and -1, and not permissible after all other fmal consonants. Quite interestingly, in Pipcha a final -s morpheme is commonly realised as a dental fricative /6/, but at least the speaker whom I interviewed also accepted the sibilant realisation for the directive marker as an infrequent option. A similar sound change can also be observed for the past tense suffiX -s (see below). 4.2. Stem]] andpast-tenseforms Generally, the Shamskat dialects (except the Nubra dialects) have preserved the past-tense suffiX -s only for agentive verbs and neutralised it for most non-agentive verbs. 42 But the Shamskat dialects have also generalised the suffiX -s in places where Old and Classical Tibetan do not have any suffiX (e.g. /taJ)sl 'gave', OT gto1): bta1): gta1): tho1J), do not allow a suffiX for phonetic reasons (e.g. /sats/ 'killed', OT gsod: bsad: gsad: sod), or would allow only the suffiX -d (/pans/ 'listened', OT (m)Fian: (m)Fian*d: mflan: non). On the other hand, the past-tense suffiX of agentive verbs is often dropped without leaving any trace in Balti and similarly in the western Sham narrations.43 In northern Nubra, post-final -s is generally dropped, but the pasttense suffiX may be preserved in questions (see below) and in nominalised verb forms, e.g. ARA /rgjaJ]pinl ~ /rgjaJ]spin/ brgya1)-(s)-pin 42 The fonner suffix might have been either dropped or lexicalised A few nonagentive verbs, however, still have two stems, but the distinction of the two fonns might depend on the dialect or even the individual speaker. On the other hand, stem neutralisation can also be found with agentive verbs (Zeisler 2004: 620-622). 43 In such cases, my consultants would say that the suffix is there in principle, but carmot be heard due to the speed of speech. A similar careless prommciati.on or drop can also be observed in the Sharnskat dialects with respect to the -s of the ergative marker.
260
BETIINA ZEISLER
( -payin) '(I) stuffed' Very infrequently, the past-tense suffix shows up in nominalised forms of non-agentive verbs in Shamskat, e.g. OOM
!karl 'become white' (CT dkar '(be) white'): /rgunftk karspa zbear!fik SlJOnl dguncig dkar-s-pa dbyarcig syon 'When the winter had been white, the summer will be green' Other traces of a lost suffJX are not found in the Sham dialects, whereas in Balti, the aspiration of a following nominaliser /-pal - /-pha/ gives evidence of a former suffix, similar to the fricativisation in the Kenhat dialects. In a few cases, this trace can also be found with non-agentive verbs, cf. Kharmang /lfhaq/ chag 'break [--ctr]', /lfhaq-pha/ < chag-s-pa 'broken' and /thop/ thob 'fmd', /thop-pha/ < thob-s-pa 'found' besides IJil si 'die' and I Jispa/ si-s-pa 'dead' (Ghulam Hassan Lobsang 1995: 5-6). By contrast, most of the Kenhat dialects have dropped the pasttense suffJX -s, leaving an obvious trace only in the verb stem with the above-mentioned Kenhat vowel change of open syllables and the deletion offmal clusters with velar fmal and suffJX (-gs, -ys). Nevertheless, traces at morpheme boundaries show that the suffJX must have been generalised even for non-agentive verbs at an earlier stage. Altogether there are four indicators: (a) Vowel change or diphthongisation (b) Loss of final clusters -1JS and -gs (c) Aspiration or fricativisation of a following nominaliser pa -- [pha] - [cpa] (d) Retention in (past-tense) questions, cf. LEH /skjotsa?/, GYS llgotsa?/, ARA /skjora?/- /skjodza?/ skyod-da- skyod-s-sa/IJ,a 'did [you/s/he] (already) come?', NYO /arne nU.m trglsa?/ amatzi snum *grul-s-sa/IJ,a 'Did mother apply oil [on the child's face]?'
As far as the Ladakhi dialects are concerned, the combination of features (a), (b), and (c) is restricted to the Upper Indus, Gya-Miru, and Zanskar dialects, whereas (d) has been observed also in northern Nubra, Leh, and the Changthang dialects. As already mentioned, feature (b) can be found in other Western Tibetan varieties, (c) is also observed in Balti, whereas (a) is typical for most of the modern Tibetan varieties and might be found in the Changthang dialects as well. Features (b) and (c) and their combination can be exemplified with the verb GYS, CEM /lalJ : ta ~ ta tog/ btay btays to1) 'give': /khe tal kho!J-i bta1)S 's/he gave', /ge tafin/ ya!J-i btayspin 'I gave' Depending on the dialect, fricativisation can be found with all agentive verbs plus open syllable non-agentive verbs (Cemre), or with all
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
261
agentive and almost all non-agentive verbs, independent of the syllable structure (Gya). In a few cases of non-agentive verbs, the fricativisation is blocked in Gya, but it is unclear when and why. Perhaps such verbs are mere relicts, as yet unaffected by the generalisation rule. The question marker has been systematically checked only for the Gya dialect. In most cases, the question marker can take only the form /-sal (or at least this would be the clearly preferred form), in few cases both forms. When both forms coexist, the s-form might have an admirative function, expressing surprise, a greater emotional involvement, or indignation, e.g. /matosa?/ masto-s-sa 'did [it] really not matter [to him/her]?' but /matoa?/ mastof:ta 'didn't [something] happen [to him/ her]?' Less frequently, the s-less form might have this function. Otherwise, the s-form might refer to a somewhat more distant past and the s-less form to a more recent or more immediate past. Which form of the question marker is chosen appears to be somewhat unpredictable and conditioned more by the context of the utterance than by the semantics of the verb. 44 In Pipcha, the past tense sufftx left its trace in the optional fricativisation of the (voiced) nominaliser as well as in the loss of fmal clusters with the velar nasal (less clearly with velar stop). Additionally, the past tense suffix is preserved in the question marker, following stem II. As in the case of stem IV, the morpheme is realised as a dental fricative/&, alternatively also as /9/. Due to its unstressed position, vowel a of the question marker (and other morphemes) is commonly realised as lei, but the consultant I interviewed also accepted forms with /a/. 45 On the basis of a cursory survey, it turned out that the past tense sufftx /&is obligatory for agentive verbs (except after fmal -t, where it tends to be omitted) and optional for non-agentive verbs. In the latter case, it may convey a notion of surprise. 44 Due to growing language awareness, the consultant accepted fewer s-less fonns in 2006 than in 2005 and claimed more vehemently that such fonns were only used by Leh speakers. Nevertheless, she still accepted s-less fonns for non-agentive and agentive verbs alike. In 2007, I had to remove about 90% of all remaining s-less fonns from the documentation; they were, if at all, only accepted for non-agentive verbs. In 2008, the speaker, with only one or two exceptions, did no longer accept s-less fonns, even for non-agentive verbs. While I am unable to repeat such checks regularly, it is evident that some linguistic features allow a certain variability, but due to various influences the consultants keep changing their minds about the actual extent of the variability. 45 According to another consultant with whom Thomas Preiswerk worked, 1his might be due to intentional switching to the Lungnak dialect.
262
BETIINA ZEISLER
4.3. Stem I with the foture participle or the causative auxiliary
A trace of an over-generalised suffix -s seems to appear also in stem I, or more precisely in a following future participle morpheme cas~ ces: Purik /-!fa/~ Sham /-!fas/ ~ Leh /-!fes/ ~Upper Ladakhi /-!fel, and in a following causative auxiliary /!fuk/ bcug, according to the all-Ladakhi rule: /s/ + /!f/ - I fl. Again, this feature has only been systematically checked for the Gya dialect. Here, the future participle appears as /-!fe/ after final -k, -t, -n, -p, and, -m. Some verbs are also more or less acceptable with the form /-fe/. 46 After vowels, the regular allomorph is /-3e/, due to intervocalic voicing /-!fe/ - (1--sat/, DRS lrnep-sat/, DOM /mesat/ (younger generation) /mep-sat/ (older generation), GARK /mesat/- /mep-sat/ me-bsad 'fire-extinguishing,
These examples are not attested with the labial stop or are not attested as compounds at all in the other dialects surveyed (for many more examples also with nasal prefix and lexicalised pre-radicals, see Zeisler 2009 and Zeisler in preparation). In some compounds, the same process applies to the (according to a traditional analysis) radical b followed by the post-radical -y- or -r-. The most prominent might be the all-Ladakhi forms /lab-raJ]/ bla-bray 'residence for monks' or /rib-~~ /rib-3al ri-bya 'mountain fowl'. It cannot be precluded that the status of the initial consonant as a preradical or a radical was irrelevant for its shift. But it is also possible that the traditional spellings represent, or lead to, the wrong analysis. 53 In this context, a very particular set of compounds in Gya, some not attested in any other dialect, might be of special interest The second syllable, if isolated, always starts with a simple r, while the first sylla51 E.g. a mentally disturbed person, who has killed several people in fighting. Now the tenn might also be applied to terrorists. 52 lkhi/ khyi 'dog' is often used as an echo word to /rri/ mi 'man', but of course, lkhi(p)sat/ kyhihsad may also mean 'killing or killer of dogs' 53 Apparently, there was no need to differentiate between the sequences preradical+ radical and radical+ post-radical, as in the case of voiced velar plus palatal approximant (where the former sequence is represented by a horizontal and the latter by a vertical combination). The reason might be that exemplars of one of the two sequences were missing or extremely rare. As the early scribes were not trained linguists, it is not necessarily the case that their spellings were always coherent or always represented the correct analysis in tenns of pre-radicals, radicals, and post-radicals. This is quite apparent in the case of the sub-scribed radical-!-, which is reflected in all varieties as (high tone) I, the only exception being the cluster zl, (Huber 2005: 3lf.). But the regular b-prefix for stem II/III in len: hl01Js: blOIJ: lonll01J(s) 'take' has never been recognised as such by Tibetan grammarians. It is 1hus possible that sequences of super-scribed b- and sub-scribed y or r represent original sequences of pre-radical plus radical. The same seems to be true for at least some of the combinations of g- plus -r- discussed below ( cf. also Beyer 1992: 72, note 5, 74-79. Beyer should better not be quoted, since throughout his 'compilation', he 'cites' other somces without giving appropriate references. Unfortunately, in this case, I have not yet come across the original).
266
BETIINA ZEISLER
ble, if isolated, would end in a vowel, but is, at least optionally, closed by a labial stop in the compound. In one case, the second element is attested as an independent word with a labial cluster onset in the western Sham dialects, e.g. GYS 1-b-rak/ or 1-rak/ '-lining' corresponding to DOM /brak : braks : brok/ or, with metathesis, lrbak : rbaks : rbok/, also lrak raks rok/ 'join together, attach, add'; cf. GYS lftk-rak/ gcig-(X)rag, lpib-rak/ giiis-Xrag, lsum-rakl gsum-(X)rag, l3ib-rak/ bziXrag 'having one, two, three, or four panels or grids (said of a window)',54 cf. DOM lpir-bakl, lsum-brak/, l3ir-bak/, lyar-bak/ ~ lrroarbak/ (younger generation) 11)a-Xrag, ltruk-brak/ drog-Xrag, lrdunbrak/ rdun-Xrag, lgjat-brak/ brgyad-Xrag, lrgur-bak/ dgu-Xrag, llfurbak/ bcu-Xrag. The Domkhar metathesis points to an original form sbrag, 55 and thus we might substitute sb and not only b for X The triple cluster would regularly yield lrak/ in most Ladakhi dialects (including present-day western Sham). 56 We can fmd a similar compound in all-Ladakhi /khib-ragl, KHAL, DOM lkhizbagl 'dog-fly', ultimately derived from khyi-sbra1J (cf. JAK). There are also other compounds where the second element does not appear as an independent word in the western Sham dialects, such as GYS ldorukl ~ ldob-rukl ~ ldom-ruk/, DOM lrdoaruk/ ~ lrdoab-ruk/ rdo-Xrug 'small stones'; GYS, DOM IJab-ruk/, ARA IJaruk/ ~ IJabruk/ (neither form is very common in ARA) sa-Xrug 'little pieces of meat', or where the second element is not attested with labial cluster 54 The consultant could also think of using /J]ii.b-rak/ 11JaXrag, /gub-rak/ dguXrag, and /!jlib-rak/ bcuXrag 'having five, nine, ten grids' for the multigrid windows of modern houses in Leh, but so far the new fashion has not yet fmmd its expression in the Gya dialect. 55 The consultant, from whom the present data is obtained, would accept the metathesis only in case of triple clusters. But some Domkhar and Takmacik speakers of 1he younger generation apply the metathesis also to the double cluster hr. For the moment, I am unable to say whether this is an innovative overgeneralisation or rather typical for the two villages. Apart from Balti, the feature seems to be somewhat exceptional with respect to all other western dialects, where the only form attested so far is Purik /rbi, rbis/ for bbri 'write' 56 Cf cr sbrag sbrags 'lay, put one thing above another' WT /rak/ (JAK; cf also his wr compounds with numerals!), Balti (Skardo) /rbaql 'pile up, join (in activity), follow, come along', Purik (Kargil, Ciktan) /zbraql, eastern Sham (Nurla) /rak/ 'join together' (CDTD), western Sham (DOM) /rbak/ - /brak/ - /rak/ 'join together, attach, add, give dowry' (according to a second consultant /rak/ is used only in connection with the goats to differentiate the verb from /brak/ 'sheer', but for the main consultant there would be usually no need for such dissimilation). For the regular Balti metathesis in triple clusters, cf also Bielmeier (1985: 212).
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
267
onset, such as GYS /doril/ ~ /dob-ril/, DOM /rdoaril/ ~ /rdoab-rill, ARA/rdoril/ ~(less frequently) /rdob-ril/ rdo-Xril 'round stone'. The pronunciation /rill might go back to a verb sgril bsgril with the meaning 'make round'. Similarly, we fmd the compounds GYS ljlirikl ~ /pi"b-rik/ gnis-Xrig 'two rows (of turquoise on the perak, the women's traditional headgear)' and /garik/ ~(as heard) /gab-rik/ l1JaXrig 'five rows, not attested in the other dialects. Here again, the form /rik/ may relate to the verb sgrig bsgrigs bsgrig sgrigs 'put in order, arrange, etc.'. The conclusion would thus be that /ruk/ 'small piece' ultimately relates to either *(b)sgrug or *sbrog. In fact, we do fmd a Classical Tibetan verb sgrog bsgrugs bsgrug sgrugs 'collect, gather, pluck, pick up', an activity one may perform with 'small pieces' in particular. 57 The Gya dialect provides us with another, even longer, series of compounds with the second syllable /-rag/ or /-b-rag/, indicating that the entity described by the frrst syllable is 'alone, single, nothing but itself, not going together with the prototypical counterpart or accessory'. These compounds are only rarely attested in other dialects. Most people would use expressions with /ftkftk/ gcigcig, /ftkpo/ gcigpo 'single' or /khorag/ khora1J 'itself instead. I can offer only a few examples (the full set will be given in Zeisler in preparation) :58 -
ltriratJI- ftrib-rat]/ gri-(X)rat} 'knife without meat'~ 9
-
/tfhub-fat)/ chu-Xrat} 'nothing but water'
-
/p,!!b-rat]/ Pfa-Xrat} 'a single fish'
-
/tiirai]I- /tab-fat)/ rta-Xrat} 'horse without companion, foaL saddle'
-
/daral)/- (rarely) /dab-rat]/ mda}:l-ra1J 'arrow without bow'
-
/m,!!b-rai]I ma-Xrat} 'mother whose child(ren) went to another place'
57 Cf. Balti (Skardo) /rgik/ 'put in order', Purik (Kargil, Ciktan, and Tshangra) /zgrik/ 'fit in (trans.)' and 'arrange, line up, row, repair' as well as Tshangra /'ZJ!,flllhet/?T 'discrimination'< ?yaphes
-
SHA /j.!!Jgjat/T, GYS (elder people) /j.!!Jgjat/T - (younger generation) /j.!!Jgjaslr, SYK /j.!!f&ias/L; LAG /ji!fgjat/T- (less frequently) /j.!!f&ia/0 'development, progress' < yarrgyas, also as verb: GYS /gja/ - /gjat/ rgyas 'increase, spread, grow, develop [--ctr]'
-
GYS, SYK (only grandparents) /rudbal/T 'tortoise, turtle'< rossbal
(grandfather)
64 The first form apparently implies a final dental stop. The second form could be due to regular inteiVocalic voicing of simple consonants, or due to a secondary development of the first form. 65 The word bhras is replaced by lbato/ (from Shina or K.asluniri) in SHY and other Changthang dialects. In the comp01Dlds /drastuk/, /drasil/, and /S8J]dras/, the word is clearly a loan from Leh.
272
BETIINA ZEISLER
-
SHA /ret/T, GYS (elder people) /ret/T - (younger generation) /res/L> SYK, LAG /re/0 'turn, alternation'
kelak
elder sister's dress(es) ? dress
?Aymo 's elder sister ?
?Aymo's elder sister's dress(es) khimtsea
tfol.
neighbour-DatLoc
entrusted
'Agmo entrusted [her] elder sister's dresses to the neighbours.' ? 'Agmo 1 's elder sister2 entrusted [her2] dresses to the neighbours.' ?'The dresses of Agmo 1's elder sister2 were entrusted to the neighbours [by her2].'
b .Agme dress
Aymo 's elder sister khimtsea
tfol.
neighbour-DatLoc entrusted 'Agmo1's elder sister2 entrusted [her 112/their] dresses to the neighbours.'
9. MARKING OF TENSE AND EVIDENTIALITY The two dialect groups also differ in the selection of mmphemes or complementary verbs for complex tense forms, the most prominent being: -
complementary verb /dati (Sham skat and Leh /duk/) as marker of continuous or repeated events
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
291
-
use of the non-contracted combination of nominaliser /-ba/ ba plus auxiliary /-fi.ot/ yod (Shamskat /-et/ or {-bat} and Leh /-at/), at least in combinations with evidential markers, cf. e.g. GYS /saon tab(b)afi.ot(t)ro/ sabon btabbayodJ:zgro (LEH ltab_!Lt(t)ro/) 'might sow' or /saon tab(b)afi.okanak/ btabbayodmldzan*IJ-og 'must be/have been sowing' (LEH ltabetkjak/)
-
future tense {-kan} (kan, ran, han) < mkhan92 for speaker-prominent events (Shamskat and Leh /-in/), 93 also used in Cemre and formerly in Gya for (speaker-prominent) present tense (Shamskat /-et/ and Leh /-at/)
-
generic and inferential marker {-kak}, negation {-kamanak} ?< mkhan *~Jog or rag (Leh /-anok/, Sham skat /-intsuk/- /intsok/)94
-
distance marker {-kanak} (Leh /-kG)ak/) < mkhan /(k(h)antsuk/- /(k(h)antsok/)
-
regular use of inferential past marker {-tok} (tok, dok, rok), not common in Shamskat
+
+ *!Jog (Shamskat
The evidential motphemes /-kan/ and /-ak/ ~ 1-kakl are again found in the neighbouring varieties, cf. the speaker-prominent future I -(k)anl (Pin Spiti), /-kan/ ~/-ken/ (Tot), /-kajinl > /-ken/ (Tabo Spiti) vs. nonspeaker-prominent future /-(k)ak/ (Pin and Tabo Spiti) (Sharma 1989: 315f, 1992: 80-83, Veronika Hein, p.c.). The ftrst motpheme is also found in Kyirong, where it is used in the form {-k~} for habitual events in general and in the form {-l&(ji:)} for future acts of the speaker (Huber 2005: 110, 124). An inferential future marker I -kak/ is also found in Mustang, as well as a future motpheme /-ka/ + /-rak/ or /-nak/ ?< mkhan +rag, expressing certain or generic knowledge (Kretschmar 1995: 145, 149f). The Mustang inferential past marker {tuk} (tuk, ruk; Kretschmar 1995: 156) might likewise be related to the above-mentioned {tok}. This, in turn, might be related to the auxiliary f:ulug, which in many Tibetan varieties has an experiential (eye-witness) function, but is not attested as such in Mustang. Although this does not seem to be immediately intuitive, the experiential function of the auxiliary f:tdug can be conceptually connected Signalling certain knowledge (cf also Zeisler 2004: 652, note 243). ln Nyoma only with 1he awilli.ary /-(J)in/ added to 1he future verbal nolDl. 94 A seemingly related inferential (future) marker is /-(fi.)ak/ *bog; Leh /ok/, westem Sham and Ciktan Purik /-uk/ -l-oki or /-pok/, but northern Nubra {-suk}- {-sok}. Perhaps the marker spread from the Kenhat to the Sharnskat dialects. In the Nyoma dialect, however, the form {-kak} conveys the notion of definiteness and implies intimate knowledge or responsibility for acts of third persons, replacing in this ftmction the morpheme /-in/. 92
93
292
BETIINA ZEISLER
with an inferential marker. In contrast to generally shared knowledge, knowledge based on a singular (visual) experience is less certain. Its content merely 'seems' to be the way it was perceived. The experiential auxiliary f:tdug could thus be interpreted as a reservation on the part of the speaker concerning the truth-value of the statement: 'apparently' or 'as far as I could observe'. This strategy of reservation would explain the possible mirative use of the auxiliary as described by DeLancey (1997).
10. CONCLUSION
As the Kenhat dialects differ in many ways and quite substantially from the Shamskat dialects, it is not appropriate to treat these differences as merely dialectal variations. The differences manifest themselves most obviously at the phonetic level (fricativisation and emerging tone vs. clusters) and at the grammatical level (genitive vs. ergative agent marking, verbal auxiliaries), but also at the semantic level (many Shamskat words are not used or have a different connotation in Kenhat and vice versa). Other differences are less obvious, but are nevertheless important, such as the traces of the past-tense and imperative -s suffix, and the differences in the argument frames of verbs (as in the case of the medium argument). From an exclusively merely phonetic perspective, the differences between the various Ladakhi dialects appear to be gradual, and it may be justified to group the Leh dialect with the phonetically conservative Shamskat dialects. However, this approach does not account for the essential difference at the level of grammar, due to which the Leh dialect can only possibly be grouped with the phonetically innovative Kenhat dialects. The somewhat unexpected mixed character of the Leh dialect itself can be explained by historical facts (Leh as an important point of commercial exchange, repeated settlement of Balti speakers around Leh). Interestingly, it is the historically 'younger' dialects that have exerted the greater phonetic influence on the historically 'older' one (although one can observe some grammatical influences also in the opposite direction). The above fmdings not only show that a classification of dialects cannot be achieved solely on the basis of surface phonetics, which may be more readily influenced by external factors than the grammatical layer, but also show that the terms 'conservative' vs. 'innovative' are
KENHAT, THE DIALECfS OF UPPER LADAKH AND ZANSKAR
293
quite misleading when based exclusively on phonetics. After all, the Kenhat dialects seem to have retained more lexical and grammatical archaisms than the Shamskat dialects. The two dialect groups reflect separate historical developments, and, in fact, different linguistic sub- and adstrates: While the whole of Ladakh and adjacent regions were originally populated by speakers of Eastern Iranian (Scythian), Lower Ladakh (as well as Baltistan) was also subject to several immigration waves oflndoaryan (Dardic) speakers and other groups from Central Asia. Upper Ladakh and the neighbouring regions to the east, by contrast, seem to have been populated additionally by speakers of a non-Tibetan Tibeto-Burman language, namely West Himalayan (Old Zhangzhung; for the complex historical background cf. Zeisler forthcoming). The original ethnic differences between these populations continue to be reflected to this day in mentality and culture. While the gradual Tibetanisation (i.e. the shift to Tibetan as Ll) might not have started in the north-western areas before the end of the lOth or the beginning of the 11th century, when the (possibly merely pretending) descendants of the Old Tibetan imperial dynasty established themselves as kings in Guge and Ladakh (incidentally, this is also the time of the second spread of Buddhism), the interesting development of case marking in the Western Tibetan varieties could be indicative of a very early assimilation process between early Old Tibetan and the languages spoken in Zhangzhung. It cannot be precluded, therefore, that the process of Tibetanisation already took place in (some parts of) this region during the imperial period. One could at least expect that the lingua franca of the empire would have been more dominant in this region.
294
BETIINA ZEISLER
APPENDIX: EXAMPLES OF INITIAL AND MEDIAL FRICATIVISATION
CEMRE
r, s + k--+ [hH:X]-[~] (depending on vowel) [hi],
[~i]
