Linguistics of the Himalayas and Beyond 9783110968996, 9783110198287

Citation preview

Linguistics of the Himalayas and Beyond

W DE G

Trends in Linguistics Studies and Monographs 196

Editors

Walter Bisang Hans Henrich Hock Werner Winter (main editor for this volume)

Mouton de Gruyter Berlin · New York

Linguistics of the Himalayas and Beyond

edited by

Roland Bielmeier Felix Haller

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Linguistics of the Himalayas and beyond / edited by Roland Bielmeier, Felix Haller. p. cm. — (Trends in linguistics. Studies and monographs ; 196) Includes bibliographical references and index. ISBN 978-3-11-019828-7 (hardcover : alk. paper) 1. Tibeto-Burman languages - Himalaya Mountains Region Grammar. I. Bielmeier, Roland. II. Haller, Felix. PL3563.L56 2007 833'.914-dc22

2007035012

ISBN 978-3-11-019828-7 ISSN 1861-4302 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. © Copyright 2007 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Editorial

The present book Linguistics of the Himalayas and Beyond contains a selection of 21 papers from 43 presented to about 60 scholars from 16 countries (Switzerland, Germany, France, Austria, The Netherlands, United Kingdom, Sweden, China/Tibet, Taiwan, Thailand, Japan, Pakistan, India, Nepal, USA, Australia) at the 8th Himalayan Languages Symposium held at the University of Bern, Switzerland, September 19-22, 2002. Beside the presentation of the papers, one full day was dedicated to a parallel workshop on Western Archaic Tibetan, and a public lecture on King Kesar of Baltistan: The Muslimic narrative tradition of the Tibetan national epic was delivered by Klaus Sagaster in German. Finally, three further presentations were given, one by Nicolas Tournadre on Software for instructional materials on Tibetan dialects, one by Roland Bielmeier, Ghulam Hassan Hasni, Felix Haller and Katrin Häsler on the Project of a Balti dictionary, and one by Roland Bielmeier, Felix Haller, Katrin Häsler, Brigitte Huber and Marianne Volkart on the Comparative Dictionary of Tibetan Dialects (CDTD).1 The approximately 250 languages of the Tibeto-Burman family are spoken by 65 million speakers in ten different countries, including Pakistan, India, Nepal, Bhutan, Burma and China/Tibet. They are characterized by a fascinating linguistic, historical and cultural diversity. The languages spoken in the Himalayas, on their southern slopes and on the high Tibetan plateau in the north constitute the core of this diversity. The present book, therefore, deals with an impressive range of languages and linguistic topics, underlines the diversity of the Tibeto-Burman languages in Central and South Asia and highlights their place within present-day linguistic research. For the first time, emphasis is put on the study of the many different Tibetan dialects spoken in linguistic Tibet from the Nanga Parbat and the K2 in Pakistan in the west to the Blue Lake in northern Amdo (Qinghai) and to southern Kham (Yunnan and Sichuan) in the east. Out of the 21 papers, ten are dedicated to Tibetan linguistic studies. Philip Denwood gives four approaches to Tibetan language reconstruction based on script, dialect, lexis and grammar, focusing on the relationship of the language preserved by the orthography of Written Tibetan and the current spoken varieties. Richard K. Sprigg offers a polysystemic phonological

vi

Editorial

analysis of Classical Tibetan orthography and Balti, the westernmost Tibetan dialect in northern Pakistan, based on J. R. Firth's "prosodic approach" of the mid 1930s. Bettina Zeisler shows that the description of Ladakhi, which, alongside Punk, is the Indian neighbour of Balti, as an ergative language needs some qualification, basing her claims on a surprising multitude of basic and marginal sentence patterns and several types of variation within these patterns. Ralf Vollmann also investigates the casemarking system of Tibetan. His research base predominantly consists of modern Central Tibetan which, according to him, can be more adequately described in terms of an active/stative system rather than an ergative/transitive system. Based on natural discourse data, Nancy J. Caplow provides a description of the topographic directional markers, addressing the semantic meaning, syntactic distribution, and discourse function of the various morphological forms in Tokpe Gola, a Central Tibetan dialect spoken in northeastern Nepal. Stephen A. Watters explores phenomena of deixis, evidentiality, and mirativity, finding unexpected results with regard to the typical patterning of the conjunct/disjunct complex in narrative texts in Dzongkha, a southern Tibetan dialect and national language of the kingdom of Bhutan. Case-marking patterns and secondary-verb constructions in Bathang and Rgyalthang Tibetan, two southern Kham Tibetan dialects in Sichuan and Yunnan, are the focus of Krisadawan Hongladarom's contribution who finds that these dialects show significant variation of ergative marking and that secondary verbs need to be distinguished from serial verbs because they behave quite differently. Using data from four Amdo dialects, Karl A. Peet argues that initial labial stops are preserved in the phoneme inventories of these Amdo dialects, despite widespread historical spirantization. Furthermore, he shows that patterns of labial place assimilation in these dialects contradict typologically-predicted and phoneticallygrounded implicational hierarchies, underscoring the cross-linguistic vulnerability of labial obstruents to synchronic and diachronic processes of weakening and change. In Khalong Tibetan, a previously undescribed dialect spoken in a remote corner of Sichuan, where people also speak a variety of Tibeto-Burman Showu rGyalrong as a second language, secondary perfective stems were created out of the new imperfective stems by means of ablaut. As observed by Jackson T.-S. Sun, this is an extraordinary morphological change induced by intimate contact with rGyalrong. Tibetan has often influenced surrounding languages. Interestingly, in this case, the contact-induced linguistic infiltration was directed in the opposite direction. Based on about 600 verbs collected by own fieldwork and illustrated by

Editorial

vii

numerous examples, Felix Haller gives an outline of the different types of stem alternation and verbal valence in Themchen Tibetan, an archaic Amdo nomad dialect enjoying high prestige in Amdo and spoken north-west of the Blue Lake in Qinghai. Among the papers on Tibeto-Burman languages other than Tibetan three are dealing with western Himalayan. Yoshiharu Takahashi investigates the verb inflection (subject, object, and inclusive marking), the deictic pattern of motion verbs, and the case-marking system in Kinnauri, spoken in Himachal Pradesh in northwestern India. On the basis of the same language, Anju Saxena highlights the significance of the notion of context shift in linguistic analyses of narratives, suggesting that the distribution of the referential mention and the ergative marker in Kinnauri narratives cannot be accounted for adequately if we do not take into consideration the notion of context shift, i.e. the perspective chosen in storytelling. Suhnu Ram Sharma's paper on the classification of Bunan provides an updated overview over the Tibeto-Burman languages, including the Tibetan dialects spoken in Lahaul-Spiti district in Himachal Pradesh and in Uttaranchal. Bunan is shown to be heavily influenced by neighbouring Tibetan varieties. Sharma also investigates phonological changes in Bunan, using data from an article by Jaeschke (1865), and completes his paper with a comparison on various levels between Bunan and the closely related Manchad and Byangsi languages. Two papers deal with the Tamangic languages of northern central Nepal. Martine Mazaudon addresses the question of the relationship between vowels and consonants in Marphali. She proposes a revision of the sonority hierarchy concerning vowels and approximants and suggests that thinking of vowels in terms of place of articulation and stricture type would be more enlightening than the usual articulatory definition of vowels. This is underscored by appendices containing comparative data from other Tamangic languages and spectrograms. Isao Honda compares the Thakali plural markers, a Seke enclitic and a Risiangku Tamang topic marker and provides possible cognates of these morphemes. These cognates include demonstratives, 3rd person pronouns and a definite article. Honda suggests that they all originate from demonstratives. Further investigations into this type of change are needed as it does not seem to be very common cross-linguistically. In a paper on two East Bodish languages, George van Driem claims that Dzala spoken in eastern Bhutan and Dakpa spoken in Arunachal and in a few villages in eastern Bhutan appear to form a coherent subgroup within East Bodish. This analysis becomes evident through the large amount of shared core vocabulary, including numerals

viii

Editorial

and personal pronouns, and is corroborated by native speaker intuition. Four papers deal with the Kiranti languages Limbu, Sampang, Chaurasia (consisting of Wambule and Jero), and Sunwar in eastern Nepal listed here from east to west. Boyd Michailovsky discusses a specific construction in Limbu, where the modifier bears the 3d singular "possessive" prefix as a pronominal prefix. It occurs in a common type of colour-term derivative as well as in a number of other words and has East Kiranti parallels as well as an interesting typological convergence in Burmese. Rend Huysmans shows in his paper that the Central Kiranti language Sampang has an accent system which can be perceived only in polysyllabic words in which the accented syllable achieves its prominence mainly by vowel length and the increase of loudness. The word accent is defined as a stress accent that has contrastive function, as it is predictable only for nouns. For other parts of speech like verbs and temporal adverbs, the position of the stress must be lexically specified. In his paper about Chaurasia, Jean Robert Opgenort presents lexical and morphological data illustrating that Wambule and Jero are very closely related. These data and many other shared grammatical features, in addition to the fact that the two varieties are mutually intelligible, plead in favour of the view that Wambule and Jero constitute two sets of dialects of a single language which may well be called "Chaurasia". In her paper Dorte Borchers tries to get closer to solving the question of how to save threatened languages. She summarizes models of language shift and gives examples of threatened languages, including also Sunwar spoken by about 25 000 speakers. Though the reasons for language shift seem to be too individual to be covered by general models she considers language-revitalisation programmes useful if they are implemented early enough to successfully develop consciousness of the own language and combating disastrous language ideologies. Karumuri Venkata Subbarao, Upen Rabha Hakacham and Thokchom Sarju Devi, comparing data from Hindi-Urdu and Telugu, argue in their contribution that PRO has to be Case-marked to account for the data concerning the conjunctive participles in the TibetoBurman languages Rabha and Manipuri in northeastern India. Rabha is a language of the Bodo-Koch group spoken in Goalpara district, while Manipuri or Meithei belongs to the Kuki-Chin group and is the language of the large central plateau of Manipur state. The conference was funded in a very generous manner by the Max and Elsa Beer-Brawand Fund, the Swiss Academy of Humanities and Social Sciences, and the Swiss National Science Foundation. All the participants as well as the organizers owe these institutions a sincere debt of gratitude.

Editorial

ix

Without their financial support, this successful meeting of so many scholars from so many different countries would never have taken place. I also wish to thank Werner Winter, the main editor of this volume, for his active support, Mouton de Gruyter for including this volume in the current series, and Birgit Sievert at Mouton de Gruyter for her help in finally bringing out this volume. Roland Bielmeier

Notes 1.

Roland Bielmeier, Felix Haller, Katrin Hosier, Brigitte Huber, and Marianne Volkart, Comparative Dictionary of Tibetan Dialects (CDTD) in 4 volumes: Introduction, Noun, Verb, Index. Bern (in preparation). For further details cf. Bielmeier, Roland, Comparative Dictionary of Tibetan Dialects (CDTD): A research report. European Bulletin of Himalayan Research 23 (Autumn 2002): 97-101.

Contents

Editorial

v

Reasons for language shift: Theories, myths and counterevidence Done Borchers

1

Directionals in Tokpe Gola Tibetan discourse Nancy J. Caplow

23

The language history of Tibetan Philip Denwood

47

Dzala and Dakpa form a coherent subgroup within East Bodish, and some related thoughts George van Driem Stem alternation and verbal valence in Themchen Tibetan Felix Haller A comparative and historical study of demonstratives and plural markers in Tamangic languages Isao Honda Grammatical peculiarities of two dialects of southern Kham Tibetan Krisadawan Hongladarom The Sampang word accent: Phonetic realisation and phonological function Rene' Huysmans

71

85

97 119

153

A low glide in Marphali Martine Mazaudon

163

Pronominally marked noun determiners in Limbu Boyd Michailovsky

189

xii

Contents

About Chaurasia Jean Robert Opgenort

203

Implications of labial place assimilation in Amdo Tibetan Karl A. Peet

225

Context shift and linguistic coding in Kinnauri narratives Anju Saxena

247

The status of Bunan in the Tibeto-Burman family Suhnu Ram Sharma

265

Tibetan orthography, the Balti dialect, and a contemporary phonological theory Richard K. Sprigg Case-marked PRO: Evidence from Rabha, Manipuri, Hindi-Urdu and Telugu Karumuri Venkata Subbarao, Upen Rabha Hakachant, and Thokchom Sarju Devi

279

291

Perfective stem renovation in Khalong Tibetan Jackson T.-S. Sun

323

On the deictic patterns in Kinnauri (Pangi dialect) Yoshiharu Takahashi

341

Tibetan grammar and the active/stative case-marking type Ralf Vollmann

355

The nature of narrative text in Dzongkha: Evidence from deixis, evidentiality, and mirativity Stephen A. Watters

381

Sentence patterns and pattern variation in Ladakhi: A field report Bettina Zeisler

399

Subject index Language index

427 436

Reasons for language shift: Theories, myths and counterevidence Dörte Borchers Abstract1 Languages currently die increasingly fast; this fact prompted a strong rise in publications discussing probable reasons for and possible counter-measurements to language death. Despite efforts by linguists to maintain the world's linguistic diversity, a reversal of the tendency seems not to be taking place. An attempt to reveal why the proposed remedies to language death do not seem to have the desired effect might bring us at least closer to solving the question of how to save threatened languages. This paper consists of three parts. First, models of language death, language maintenance and language shift reversal are summarised. The theoretical part is followed by descriptions of case studies of threatened languages from different parts of the world. In the third part the applicability of the theories to real life situations of minor linguistic communities is discussed to determine how useful a tool the models are for the description and explanation of minor linguistic community situations and whether they can be used to save a language. 1. Reasons for language shift and language maintenance Reasons for language maintenance and language death are presented in two kinds of linguistic literature. In monographic studies of single linguistic communities probable threats to the maintenance of the particular language under study are often taken to be general, universal reasons for language death. In studies about language shift and language maintenance in general, the aim is not only to name some features that co-occur with language death but indeed to explain language shift as a process and to find universal features of this process on the basis of different case studies. Among the several studies of language shift and language death those by John Edwards (1992), Joshua A. Fishman (1991,1993) and Hans-Jürgen

2

Dörte Borchers

Sasse (1992) are particularly prominent and represent different approaches to the problem. 1.1. Explanations for language shift in case studies of linguistic minorities Explanations of language death are frequently based on a linguist's personal experiences with the situation of one, or perhaps a few, threatened languages. After studying the reasons that seem to be responsible for the death of that particular language, the linguist comes to the conclusion that those reasons might also be responsible for the death of other languages. In case studies of single linguistic communities whose mother tongues are threatened, some factors are repeatedly named as being responsible for language death in general. The reasons asserted are a language's -

low number of speakers (Krauss 1992: 22; Winter 1993: 312) lack of state support (Krauss 1992: 2) or even pressure from the outside to give up the language (Hale and Hinton 2001: 3; Nettle and Romaine 2000:7; Marks Dauenhauer and Dauenhauer 1998: 65,72) speakers' lack of economic power (Day 1985: 164; Hale and Hinton 2001: 3; Nettle and Romaine 2000: 18) speakers' lack of military strength (Day 1985: 164) lack of prestige (Day 1985: 164; Hale and Hinton 2001: 3; Marks Dauenhauer and Dauenhauer 1998: 66; Winter 1993: 312), or its speakers' negative attitude (Day 1985: 173) speakers' bilingualism (Hill 1983: 262; Lanoue 1991: 87-88) speakers' concern for linguistic purism (Hill 1983: 263).

1.2. Sasse: Language shift as a process in three stages Hans-Jürgen Sasse developed a model of language death as a process consisting of three stages. The empirical basis for his model was his own longterm study of Arvanitika, the Albanian language in Greece, and the case of Gaelic. His aim was a model of language shift that might apply to any endangered language, but he is careful to point out that even though his description of the process of language death might apply to many linguistic communities, it might not apply to the situation of every linguistic community (Sasse 1992: 12).

Reasons for language shift

3

According to Sasse, language death is the third and final stage of a process that begins in a bilingual linguistic community with a primary language shift followed by language decay (Sasse 1992: 20). Primary language shift is defined as the switch of LI and L2. After that primary language shift a dominant second language is used in most domains of life, instead of the mother tongue. The switch from one's mother tongue to another language in most domains is, according to Sasse, always triggered by some change in the external setting, in the environment, of a linguistic community. Part of that change in the external setting is a new or changed contact between the linguistic community that shifts from its traditional mother tongue to the language of another linguistic and cultural community. A result of the new culture contact, or new conditions of culture contact, is a change in attitudes of a linguistic community towards their mother tongue. The mother tongue is used in fewer and fewer domains. The dominant language is then used in domains in which the mother tongue was used before. The process goes on until people use the dominant language most of the time and the mother tongue becomes a less important means of communication (Sasse 1992: 10-13). The mother tongue that is spoken only in restricted domains is not properly transmitted to the children of the linguistic community. Children might hear the language at home but are never addressed in their parents1 mother tongue. The children have no practice in using the language actively and will not become competent speakers in their parents' mother tongue. Sasse calls these people, who know their ancestral language passively, who might know a large part of the vocabulary but are not able to produce grammatically correct sentences, according to their parents' and grandparents' norms, semi-speakers (Sasse 1992: 14—17). What happens in the language of semi-speakers, according to Sasse, is different from what occurs in "normal" situations of language contact. The language of semi-speakers loses structures that are present in the language of their parents. In a "normal" language-contact situation the lost structures would be replaced by new ones, but in a language-contact situation that produces semi-speakers, the loss of structures is absolute. The semi-speakers notice their deficiencies, feel embarrassed and try to avoid using their parents' language. The children of the semi-speaker generation might not know their ancestral language at all (Sasse 1992: 17).

4

Dörte Borchers

1.3. Sasse: Language revitalisation Sasse has also formulated a model of language revitalisation. This model involves a reversal of his theory of language death. According to him, a language is threatened as soon as its speakers use another, dominant language in most domains of their daily life. A reversal of the endangerment of the language has to occur quickly after the language has been replaced by the dominant language. Such a quick reversal of the primary language shift is possible only if the speakers change their attitudes towards their mother tongue very soon after the switch has taken place. The speakers have to consciously choose to use their ancestral mother tongue in domains in which they switched to the dominant language. Most important is the continuation of the communication with children. If such a change in attitudes towards the mother tongue and in linguistic behaviour does not happen at all, or happens too late, a revitalisation will no longer be possible, but only a reconstruction (Sasse 1992: 20-22). 1.4. Fishman: Revitalisation While Sasse's models of language shift concentrate on the linguistic community that switches languages, Joshua A. Fishman's model of language revitalisation additionally takes the institutional embedding in the society into account. Fishman's model of language revitalisation presents the transition from a dead language to a living language as consisting of eight stages. Table 1. Fishman (1991: 395, 1993: 71, 2001: 466): Toward a theory of reversing language shift3

1 2 3 4b 4a

Stages of reversing language shift: Severity of integrational dislocation (read from bottom up) Education, worksphere, mass media and governmental operations at higher and nationwide levels Local/regional mass-media and governmental services The local/regional (i.e. non-neighborhood) worksphere, both among Xmen and Ymen Public schools for Xish children, offering some instruction via Xish, but substantially under Yish curricular and staffing control Schools in lieu of compulsory education and substantially under Xish curricular and staffing control II. RLS to transcend diglossia, subsequent to its attainment

Reasons for language shift

5

Schools for literacy acquisition, for the old and for the young, and not in lieu of compulsory education The intergenerational and demographically concentrated home-family-neighborhood: the basis of mother tongue transmission Cultural interaction in Xish primarily involving the community-based older generation Reconstructing Xish and adult acquisition of XSL I. RLS to attain diglossia (assuming prior ideological clarification)

According to Fishman, the order of the eight stages of language revitalisation is fixed. First, at stage eight, a dead language has to be reconstructed by linguists. The reconstructed language should then be taught to adults and should be used in cultural interaction. The most important stage is stage six - the intergenerational transmission (Fishman 1991: 398-399). Whether or not a language is alive or dead can be determined, according to Fishman, by answering the question of whether or not children learn the language at home and use it with friends and neighbours (Fishman 1991: 373-374). This important stage six is followed by the acquisition of writing skills in the language, by making it the medium of instruction in school and a medium used first in the local worksphere, then in the mass media and in official governmental institutions. Ideally, the speakers of the formerly dead language should then at least theoretically be able to lead their lives unimpeded as monolinguals. Fishman's scheme is clearly based on the example of Hebrew in Israel. The revival of many other languages in a similar fashion is unlikely, as for many dead languages even stage eight, the reconstruction, is not possible due to poor documentation. Fishman is certainly correct in pointing out the importance of the transmittance of a language to the next generation. Many languages all over the world are, however, alive without ever having reached beyond stage six. This shows that even though stages one to five might be desirable goals from the linguists' and from the linguistic community's points of view, advancing to stages beyond stage six is not indispensable for a language's survival. A language is not necessarily nearing its extinction because it has never been written down or because it is not represented in mass media.

6

Dörte Borchers

1.5. Edwards' framework for minority language situations John Edwards (1992) developed a preliminary framework for an analysis and a typology of ethnic minority linguistic situations. Edwards is of the opinion that an in-depth analysis of different ethnic minority linguistic situations might allow generalisations and should create an informed presentation of the ethnic minority linguistic situation, from the point of view of the ethnic minority community as well as from the point of view of the ethnic majority community in their vicinity. Based on generalisations about ethnic minority linguistic situations predictions about language maintenance and shift might become possible (Edwards 1992: 38). As a tool for the analysis of situations of linguistic minorities Edwards created a systematic questionnaire that covers the variables "speaker", "language" and "setting" and for each of these variables investigates the different perspectives (cp. table 2), for example, geography, sociology and the media (Edwards 1992: 49). He suggests sample questions for each topic in the questionnaire but warns his readers that these sample questions are not exhaustive. Edwards (1992: 50) suggests that the topics in the questionnaire might be of varying importance for the analysis of different ethnic minority linguistic situations and might not all have the same weight for a general typology of ethnic minority linguistic situations.

Table 2. Edwards (1992: 49, 50): A typological framework for minoritylanguage-situation variables4 Categorization A

Speaker

Categorization B Language (community)

Setting

I Demography

Sociology

Linguistics

1 Numbers and concentration of speakers 4 Socioeconomic status of speakers

2 Extent of the language (see also geography) 5 Degree and type of language transmission

7 Linguistic capabilities of speakers

8 Degree of language standardization

3 Rural-urban nature of setting 6 Nature of previous/current maintenance and revival efforts 9 Nature of inand out-migration

Reasons for language shift

7

Psychology

10 Language attitudes of speakers

11 Aspects of the language-identity relationship

History

13 History and background of the group 16 Rights and recognition of the speakers

14 History of the language

19 Basic facts about geography 22 Speaker's attitudes and involvement regarding education 25 Religion of speakers

20 Basic facts about geography

12 Attitudes of the majority group towards the minority 15 History of the area in which the group resides 18 Degree of autonomy or "special status" of the area 21 Basic facts about geography

23 Type of school support for language

24 State of education in the area

26 Type and strength of association between language and religion 29 Association between language(s) and economic success/mobility 32 Language representation in the media

27 Importance of religion in the area 30 Economic health of the region 33 General public awareness of area

Political

Geography

Education

Religion

Economics

Technology

28 Economic health of the speaker group 31 Group representation in the media

17 Degree and extent of official recognition of the language

1.6. Summary of the models of language shift A summary of the models of language shift under one heading is not easy to achieve. The different approaches emphasise different factors as playing a major role in language death or maintenance. In the case studies a bundle of not necessarily interdependent factors is made responsible for a language's death in general. A small number of speakers and lack of political and economic power of a linguistic community are among the factors named most often. Sasse considers co-occurrence of a sudden and drastic event, together with a contact situation with another linguistic community, to be the major

8

Dörte Horchers

trigger for a community to first change linguistic attitudes and then to stop their language's transmittance. For Fishman languages are threatened especially by the diminution of domains in which they are employed. The most important of these domains is that of the communication at home. Edwards does not make any specific factor responsible for language death, nor a fixed bundle of factors, but tries to cover all factors of importance for the maintenance or death of a language. For each case the trigger to switch languages has to be found individually. The different approaches have in common that their authors share the conviction that a language is not acutely threatened as long as many people use that language in many domains of everyday life. To speak a language and transmit it to the next generation in order to keep it from dying seems logical, and this could also serve as an advisable course of action for any linguistic community. If the maintenance of languages were to be achieved as easily as that, languages would not die. A look at the situation of different linguistic communities might shed some light on where theoretical models match reality and where they fail.

2. Examples of threatened languages The following examples are brief descriptions of minor linguistic communities. The description of the Sun war community is based on the author's fieldwork, the descriptions of the other linguistic communities are cited from recent literature. The portrayed linguistic communities have in common that they are relatively small and that their speakers are bilingual with the language of a major linguistic community of their vicinity. Despite similarities of their situations the linguistic communities do not show the same behaviour towards their respective mother tongues. The criteria that are likely to be responsible for the differences in linguistic behaviour are emphasised in the descriptions. The descriptions are followed by brief discussions of the usefulness and applicability of the models of language shift which were introduced in part one. Most examples presented come from the American continent, because the literature on language death and language revitalisation in America is abundant, while the information on conditions of language death in other parts of the world is rather scarce.

Reasons for language shift

9

2.1. The Arizona Tewa The Arizona Tewa5 are a community of about 300 persons living in Arizona (USA). The Arizona Tewa left a larger group of Tewa living in New Mexico after the pueblo revolts at the end of the 17th century (Kroskrity [1998] 2001: 403). Since that time the Arizona Tewa have lived among Hopi6 with whom they intermarry. The Arizona Tewa maintained their language and in the course of history became trilingual with Arizona Tewa, Hopi and English. The Hopi never exerted any pressure on the Arizona Tewa to give up their culture and language. The language use of the Arizona Tewa is closely connected to their religious practices in which a highly standardised language plays a central role. The texts of prayers and songs are fixed. Texts of other literary genres are presented in a language complying with conventionalised stylistic rules (Kroskrity 2001: 404). The Arizona Tewa do not allow any foreign words to enter into their religious language (Kroskrity 2001: 406). A result of the regular conscious use of the liturgical language seems to be a more conscious use of language in general. The use of foreign words is avoided in everyday conversations, too (Kroskrity 2001: 409). The Arizona Tewa switch between the three languages they have at their disposal depending on the identity of the person they are talking to (Kroskrity 2001: 407) and depending on the topic under discussion (Kroskrity 2001: 405). They pass their language on to their children, and are proud of their trilingualism. In their view their linguistic knowledge makes them culturally superior to the Hopi, who, as a rule, do not speak Tewa (Kroskrity 2001: 407). According to the different models of language shift, Arizona Tewa should have died long ago. The Arizona Tewa are not many, are not very powerful, and they are bilingual. The move from New Mexico to Arizona and their close relationship with the more numerous Hopi is a drastic event that might have altered the Arizona Tewa's language attitudes. As a result of living closely with the Hopi, the Tewa also had to use Hopi in domains which before had been reserved for Tewa. None of these strong triggers to shift to Hopi could compete with the Arizona Tewa's determination to continue using their language. Obviously, their conscious use of the three languages, Tewa, Hopi, and English, preserved their mother tongue.

10

Dörte Horchers

2.2. The Chilcotin The approximately 1800 ethnic Chilcotin7 live in three settlements in British Columbia (Canada). The Chilcotin have experienced an increased contact with native speakers of English since the 1960s when new roads facilitated travelling from and to the Chilcotin area (Pye 1992: 77). Clifton Pye (1992) reports that at the time of his research all Chilcotin adults were competent speakers of their mother tongue. Chilcotin children below the age of ten were considered by their parents to be native speakers of English and tended to be semi-speakers of Chilcotin or not to speak Chilcotin at all (Pye 1992: 77). Seemingly small changes in the traditional linguistic behaviour of Chilcotin speakers and the prestige of English as the language of economic success seem to be the main factors that threaten the maintenance of Chilcotin. In traditional Chilcotin society good and respectful children should have remained silent observers while serious conversations were being conducted by adults (Pye 1992: 78). This tradition, at the time of Pye's study, was observed by not addressing children in Chilcotin at all. Traditionally, adults kept conversations with children brief, but the availability of English as a means of communication made it possible to stop speaking Chilcotin with children altogether. English, not being a traditional language, was used not only for brief conversations with children but also for chatting with them (Pye 1992: 78). The mistakes children made when speaking Chilcotin were taken by their parents as a clear sign that Chilcotin, being a language for serious conversation, was not only an inappropriate language for children but also too difficult for them. At the same time, nobody cared about mistakes children made while speaking English (Pye 1992: 77-78, 80). Some Chilcotin adults realised that not speaking Chilcotin with children would result in the children never learning their ancestral mother tongue. These parents decided to speak Chilcotin with their children from early childhood on and encouraged them to answer in Chilcotin. By consciously conversing with their children in Chilcotin these parents violated what at the time of the study was perceived to be Chilcotin tradition (Pye 1992: 7879). The availability of English made the ongoing language shift from Chilcotin to English possible. The number of speakers and the power of the Chilcotin did not change, as far as can be known from Pye's study. The only event that might have changed language attitudes in the Chilcotin community was the connection to the road system and to Anglo-American

Reasons for language shift

11

culture. The change in attitudes triggered by the availability of English became a threat to the maintenance of the Chilcotin language. The Chilcotin did not value their language less, except as a means for economic progress, but applied their traditions more strictly than before. While the conversation with children was previously a domain in which little Chilcotin was spoken, English made it possible to stop using Chilcotin in this domain altogether. Some features of threatened languages mentioned in single case studies apply to the case of Chilcotin. The arrival of English as a second language can be interpreted as an event that according to Sasse's model triggered a change in attitudes. Chilcotin lost the important domain of intergenerational transmission that according to Fishman is vital for a language's survival. The case of Chilcotin seems to be explainable by different models of language shift. That stricto sensu Chilcotin is threatened because of its prestige as a means of serious conversation among adults is a fact that none of the models of language shift accounts for. 2.3. Sindhi in Malaysia Sindhi is one of the Indo-Iranian languages spoken in Pakistan. At the time of the study by Maya Khemlani David (1998), in the 1990s, 1700 speakers of Sindhi lived in Malaysia, most of them in Kuala Lumpur. The Sindhi speakers in Kuala Lumpur identified very much with being Sindhi. They lived in close contact with each other, visited each other regularly and held close contact with Sindhi communities in Sindh in Pakistan as well as with Sindhi communities all over the world. They tended to marry Sindhis and often "imported" spouses because of the small size of the community in Malaysia (David 1998: 71-72). The Sindhi in Malaysia shared the religion, culture, attitudes, lifestyle and cultural norms. Theoretically they shared the same language, too; but Sindhis in Malaysia spoke English rather than their mother tongue. Only the first generation of immigrants was able at all to speak Sindhi at the time of the research. Even these able speakers would use Sindhi only occasionally (David 1998: 70). English was by far the dominant language of the community. When David asked for the reasons for using English instead of Sindhi, they replied that English was important for the survival and wellbeing of the community while Sindhi was not important in this respect. They argued as if the community could be held together only by a language shift to English (David 1998: 70).

12

Dörte Borchers

The Situation of Sindhi in Malaysia matches the predictions of the models of language shift. A small, but not powerful linguistic community lives, after the drastic event of moving to a new country, in close contact with speakers of other languages and shifts from the mother tongue to a language of the new home country. Noteworthy in this example is the fact that, with regard to other cultural aspects, the Sindhi community in Malaysia behaves conservative. They did not change their attitude towards their inherited culture and identify with the old home rather than with the new one. The only tradition from their home country they give up is the language. 2.4. Sunwar Sunwar8 is a Tibeto-Burman language spoken by about 25 000 ethnic Sunwars, most of whom live in remote villages in Eastern Nepal. Sunwars have their own religion, influenced by Hinduism but void of the concept of a person's ritual purity. The Sunwar community is in a situation similar to that of many minor linguistic communities in the world. Sunwars in the villages live in an environment dominated linguistically by Sunwar. Everybody speaks Sunwar, and the language is passed on to the children. Sunwar villagers are bilingual and expect most other people in the world to be bilingual as well. For Sunwar villagers Nepali is the language they have to know because it is a national language, and Nepali is useful as a lingua franca. Their use of Nepali is limited to instruction in school, where Nepali is the oral and written medium, to conversations with non-Sunwar speakers at the market and with travellers, to listening to the radio, and to conversations with Brahmins who are employed to perform certain rituals. In the towns, especially in the Nepalese capital Kathmandu, the situation is very different. People originating from a village and speaking a mother tongue different from Nepali9 are viewed as uncultured country-bumpkins. The minor languages of Nepal are considered by many mother tongue speakers of Nepali to be mere dialects that are inferior and backward. The dominating language of the towns is Nepali, and the majority of those who dominate the political and economic life in Nepal have Nepali as their mother tongue. Migrants from the villages learn quickly about these attitudes predominant in the city. Their feelings towards their languages change. Sunwar and Nepali are not considered of equal importance anymore. Adults continue using both languages that they have at their disposal. They speak Sunwar with relatives and possibly with Sunwar neighbours.

Reasons for language shift

13

Most parents speak only Nepali with children born in the city. The children, obviously noticing they are not supposed to speak Sunwar, make no attempt to speak it, even though they can daily hear Sunwar at home. The reason for speaking exclusively Nepali with children is the conviction that Sunwar is useless and children should rather become bilingual in Nepali and English. The belief prevails in the Sunwar community in Kathmandu that competence in Nepali and English can be achieved only if the presumably limited space in the children's heads is not used up by Sunwar. English is regarded as modern and considered important for their economic well-being. Few parents raise their children bilingually with Sunwar and Nepali. These parents do not believe the children's capability of learning to be limited and they consciously want to act against the imminent loss of traditional knowledge. Gradually more and more Sunwars in the towns are concerned about the preservation of their traditions, and within the Sunwar Welfare Society, an association for the promotion of Sunwar unity and traditions, the problem of language maintenance has become a topic of discussion. The situation of Sunwar is similar to that of other minor languages in and outside of South Asia. A view concentrating entirely upon the situation of the language in the villages might lead to the conclusion that Sunwar is not a threatened language. Everybody speaks Sunwar and all children learn it at home. A view focussed on the city, though, could imply that Sunwar will rapidly lose speakers and will die in the not too distant future. At the moment most speakers of Sunwar live in the villages, but this does not mean the language will therefore be "safe" forever. Young people are eager to migrate from villages to towns, and attitudes held in the city can easily spread to the villages.10 It is possible that villagers will appreciate the presumably modern idea that the number of languages a child can learn is limited and that English should be learned instead of Sunwar. The evaluation of the situation of Sunwar yields two results, one for the villages and one for the towns. Sunwar is in several villages the language of the majority and of those who are powerful in village politics and economics. People in the village have a positive attitude towards their language and traditions and transmit both to the next generation. The contact between the villages and towns is limited due to poor infrastructure, and therefore people in the village are not much affected by or concerned about the fact that within Nepal their power is extremely limited. In towns, especially in Kathmandu, the speakers of Sunwar are a minority without much political or economic power. The decisive moment for a change in linguistic attitudes is not so much the change of places itself,

14

Done Horchers

which certainly is a drastic event in the life of the migrants, but rather the direct confrontation with the negative attitudes of native speakers of Nepali towards villagers and their traditions. Language ideologies which claim that a person can learn only a limited number of languages, or that certain languages are inferior, discourage parents from using their language with their children. 2.5. Hualapai Hualapai11 is a language of Arizona that, in 1983, had about 1000 speakers. Members of the Hualapai community noticed in the 1970s that English was being used more and more and Hualapai less. As a reaction to the language threat a Bilingual/Bicultural Programme was launched. In the Bilingual/Bicultural Programme ethnic Hualapai, in cooperation with a linguist, worked to overcome the community's negative attitude towards the Hualapai language. The Hualapai were at that time convinced that their language was inadequate for expressing abstract ideas and could therefore not serve as a medium at school (Watahomigie and Yamamoto 1992: 10-11). In 1981 the local school board adopted a Hualapai bilingual programme in which classes were taught by bilingual teachers. Most Hualapai children participating in the programme came from families in which at home adults would use mostly Hualapai and children would speak English (Watahomigie and Yamamoto 1992: 11). The programme did not produce a young generation of speakers of Hualapai, but it was still considered successful. The bilingual education programme infused the Hualapai community with an affirmative opinion of their language and culture. The children learned about their culture, their bilingual teachers learned how to conduct classes in Hualapai, and a linguistic description as well as teaching materials for Hualapai were developed (Watahomigie and Yamamoto 1992: 11). The basis of the programme's success was, according to its organisers, the cooperation of teachers, parents, administrators and academic professionals who functioned as a team (Watahomigie and Yamamoto 1992: 12). 2.6. Rama An example of language revitalisation from Nicaragua's west coast shows results similar to those of the Hualapai Bilingual/Bicultural Programme. According to a survey from 1986, Rama, a Chibchan language, at that time

Reasons for language shift

15

had 24 speakers in Nicaragua (Craig 1992: 19). Rama is one of many minor languages spoken in Nicaragua's Atlantic coast region (Craig 1992: 17). In 1987 the Atlantic coast region obtained a special administrative status, which meant that the local government was encouraged to support the cultural and linguistic traditions of the region. The special status of the Atlantic coast region was obtained in the wake of an initiative of the numerous different linguistic communities living in that region. Before the local government of the Atlantic coast region encouraged the linguistic research of local languages, the Rama community held negative attitudes towards their language. The Rama perceived their traditional language as "ugly", as not even a proper language, and people were ashamed of speaking Rama (Craig 1992: 19). The negative attitudes towards Rama changed when the community got involved in a language-revitalisation programme and professional linguists arrived and began to document the language. The arrival of the linguists was organised by a member of the Rama community, a woman who was especially active in efforts for the maintenance of the Rama language. She spurred on the community to join together in the effort of revitalising the language and also taught Rama in schools (Craig 1992: 20-21). The special status of the Atlantic coast region facilitated the creation of a secure environment. The linguists contributed their skills, organised funds, dealt with the administration and provided computer competence (Craig 1992: 21-22). The programme did not produce a new generation of native speakers of Rama. Similar to the Arizona Hualapai project the Rama project provided the young generation of ethnic Rama with knowledge of their ancestral language. The Rama take pride in their linguistic heritage and their language is at least documented by linguists (Craig 1992: 22-23). The situation of the American languages Rama and Hualapai differs from that of the other languages insofar as the question is not whether these languages are maintained or not but whether they can be revived. The efforts by the respective communities and by linguists led to similar results, namely, the production of descriptive works on the languages and a positive change in attitudes of the speakers of the languages as well as a change in attitudes with those who are ethnic Rama or Hualapai but do not speak the respective languages. The example of the Rama and Hualapai fit the theoretical approaches very well. As other languages described in single case studies, both languages have few speakers and not much political or military power. Whether the negative attitude towards their languages is the result of a sudden and drastic event in the history of the Rama and Huala-

16

Dörte Horchers

pai, as to be expected from the viewpoint of Sasse's model, could only be a matter of reconstruction. Both languages are used, if at all, only to a very limited extent, which according to Fishman is a sign of severe endangerment. 3. How useful are models of language shift? The language shift models cover most features that play a role in language shift. The examples illustrate the very different situations and reactions of minor linguistic communities to possible threats to the survival of their respective mother tongues. The features identified in single case studies do not form a consistent model, of course. Features that figure in the death of one language can play a similar role also in the death of another language, but they do not necessarily do so. Sasse's model covers many cases of language shift. A drastic event in a bilingual minor linguistic community is likely to influence the language attitudes of that community and result in language shift. The examples of the Arizona Tewa and the Sindhi community in Malaysia show that different linguistic communities react to changes differently and that identification with a community's respective traditions does not yield the same linguistic behaviour. Also semi-speakers, a generation with a passive knowledge of the ancestral mother tongue, do not necessarily emerge before a language dies. Sunwar in the towns does not have semi-speakers and neither does Sindhi in Malaysia. Sasse himself also noticed that the generation of semi-speakers might not be the last generation of speakers of a language. Semi-speakers of Arvanitika, the Albanian language in Greece, recently began to use their language as a means of expressing their identity and show a great creativity in producing new grammatical structures. As a result the semi-speakers might keep Arvanitika alive, in a new variety (Sasse 2002). Fishman's model of language revitalisation covers many cases of language shift, too. His emphasis on the necessity that a language has to be used in many domains and most importantly in homes and neighbourhoods to be alive and to stay alive is supported by all examples. However, the revitalisation of languages that first have to be reconstructed, or of languages that only have a few speakers of advanced age left is, in most cases, impossible. Many languages died before they were documented. They cannot be reconstructed. The examples of the revitalisation of the American

Reasons for language shift

17

languages Rama and Hualapai show what is likely to happen if threatened languages are extremely small and all speakers are of advanced age. The success of revitalisation programmes in such cases is the production of descriptive material and a positive change in the ethnic community's attitudes towards its traditions. These results are important, but they do not comprise a revitalisation of languages. Edwards' questionnaire can be a useful tool for a first survey of the situation of a minor linguistic community, as it will help to collect a lot of data that might play a role for the maintenance or death of a language. It is important to keep Edwards' warning in mind that the questions he proposes are only sample questions and that for each linguistic community any further inquiry might have to focus on different features that could be of importance only to the single linguistic community under study. The different approaches to and models on language shift help gain a preliminary general assessment of the situation of a language. Such a general assessment should be followed by an in-depth study of the particular language situation, because the reasons for maintenance or non-maintenance are too individual to be covered by general models. The examples of language revitalisation have shown that real revitalisation has to start much earlier than indicated by Fishman's model and much earlier than practised by many language-revitalisation programmes. If the aim of such a programme is, indeed, to keep a language alive, it has to start when speakers stop transmitting their language. Just as Sasse pointed out, revitalisation is only possible during a short period of time after an event has triggered a change in a linguistic community's linguistic attitudes. The survival of Arizona Tewa against all odds is certainly due not only to the attitude and linguistic behaviour the speakers exhibited at the time of Kroskrity's (2001) study but due to the fact that the attitude and linguistic behaviour were part of their culture before they moved from New Mexico to Arizona. Language-revitalisation programmes for languages with few speakers, all of whom of advanced age, might have the positive effect of giving an ethnic community pride in their heritage and of producing a documentation of their language. Such a programme for a small linguistic community with elderly speakers alone is different from a revitalisation programme that has the realistic aim of revitalising a language and possibly increasing the number of its speakers. The latter will only be successful with linguistic communities that consist of more than a few hundred speakers of all age groups. If these programmes are supposed to save a language, they have to be implemented very early, namely, as soon as the language is no longer

18

Done Borchers

transmitted to the next generation. Ideally, though, such programmes would start even earlier than that with a twofold approach in an environment that is free of any outside pressure to give up the language. In such an environment many minor languages could be saved if their speakers became conscious of their language use. Besides helping linguistic communities to develop such a consciousness, the programmes would have to aim at combating disastrous language ideologies. It is necessary to do away with such misunderstandings as that of the presumably limited storage space for languages in the human brain, or the belief that some languages are objectively better than others.

Notes 1. 2.

3.

4.

5. 6. 7. 8.

9.

This paper would not have attained its present shape without the corrections received from Tatyana Gardner (Göttingen). Even though Krauss considers a low number of speakers a threat to a language's maintenance, he himself mentions as counterexamples Navaho, which had 100 000 speakers a generation ago, and Breton, which within living memory had one million speakers. In the table, Fishman's "X" is the non-dominant language, the endangered mother tongue, and "Y" is the dominant language. The order of 4a and 4b is Fishmans's. RLS is the abbreviation for Reversed Language Shift, XSL stands for a language that underwent shift and has been reconstructed. The questions in the table are sample questions proposed by Edwards (1992: 50) for the corresponding fields. Edwards points out that other and also more questions may be asked in one and the same field. Arizona Tewa belongs to the Kiowa-Tanoan language family (Mithun [1999]2001:443). Hopi belongs to the Uto-Aztecan language family (Mithun 2001: 539). Chilcotin belongs to the Athabaskan-Eyak-Tlingit language family (Mithun 2001:353). Information on the situation of the Sunwar community was collected by the author during four three-month field trips to Nepal in the winter of 1997/98 and in the springs of 1999, 2000 and 2001. The main purpose of the research journeys was to collect linguistic data on the Sunwar language. The field trips in 1999, 2000 and 2001 were made possible by research grants (WR 39-270, WR 39-279, and WR 39-288) from the "Netherlands Foundation of Tropical Research (WOTRO)". Mother tongue speakers of Nepali are here defined as those people who view Nepali as their primary mother tongue. Stricto sensu many people in Nepal have two mother tongues, namely, their ancestral language and

Reasons for language shift

10.

11.

19

Nepali, but they usually consider their ancestral language to be their only mother tongue. Roland Willemyns (1997: 182-183, 191) has shown how in Belgium the attitudes towards Dutch changed due to the economic revival of Flanders. The prestige of Dutch as the language of economic success first increased in Flanders, but then the prestige of the language spread to the capital Brussels, as well. Hualapai belongs to the Yuman-Cochimi language family (Mithun 2001: 577).

References Craig, Colette 1992 A constitutional response to language endangerment: The case of Nicaragua. Language 68 (1): 17-24. David, Maya Khemlani 1998 Language shift, cultural maintenance, and ethnic identity; a study of a minority community: The Sindhis of Malaysia. International Journal of the Sociology of Language 130: 67-76. Day, Richard R. 1985 The ultimate inequality: Linguistic genocide. In The Language of Inequality, Nessa Wolf son, and Joan Manes (eds.), 163-181. (Contributions to the Sociology of Language 36.) Amsterdam: Mouton. Edwards, John 1992 Sociopolitical aspects of language maintenance and loss: Towards a typology of minority language situations. In Maintenance and Loss of Minority Languages, Willem Fase, Koen Jaspaert, and Sjaak Kroon (eds.), 37-54. Amsterdam/Philadelphia: John Benjamins. Fishman, Joshua 1991 Reversing Language Shift: Theoretical and Empirical Foundations of Assistance to Threatened Languages. (Multilingual Matters 76.) Clevedon/Philadelphia/Adelaide: Multilingual Matters. 1993 Reversing language shift: Successes, failures, doubts and dilemmas. In Language Conflict and Language Planning, Ernst Hakon Jahr (ed.), 6981. (Trends in Linguistics. Studies and Monographs 72.) Berlin/New York: Mouton de Gruyter. Fishman, Joshua (ed.) 2001 Can Threatened Languages Be Saved? (Multilingual Matters 116.) Clevedon/Buffalo/Toronto/Sydney: Multilingual Matters. Hale, Ken, and Leanne Hinton (eds.) 2001 The Green Book of Language Revitalization in Practice. San Diego/San Francisco/New York/Boston/London/Sydney/Tokyo: Academic Press.

20

Dörte Borchers

Hill, Jane H. 1983 Language death in Uto-Aztecan. International Journal of American Linguistics 49: 258-276. Krauss, Michael 1992 The world's languages in crisis. Language 68 (1): 4-10. Kroskrity, Paul V. 2001 Reprint. Arizona Tewa Kiva speech. In Linguistic Anthropology: A Reader, Alessandro Duranti (ed.), 402-419. Maiden (USA)/Oxford (UK): Blackwell. First published in Language Ideologies: Practice and Theory, Bambi B. Schieffelin, Kathryn A. Woolard, and Paul V. Kroskrity (eds.), 103-122. New York: Oxford University Press, 1998. Lanoue, Guy 1991 Language loss, language gain: Cultural camouflage and social change among the Sekani of northern British Columbia. Language in Society 20:87-115. Marks Dauenhauer, Nora, and Richard Dauenhauer 1998 Technical, emotional, and ideological issues in reversing language shift: Examples from Southeast Alaska. In Endangered Languages: Language Loss and Community Response, Lenore A. Grenoble, and Lindsay J. Whaley (eds.), 57-98. Cambridge: Cambridge University Press. Mithun, Marianne 2001 Reprint. The Languages of Native North America. (Cambridge Language Surveys). Original edition, Cambridge: Cambridge University Press, 1999. Nettle, David, and Suzanne Romaine 2000 Vanishing Voices: The Extinction of the World's Languages. New York: Oxford University Press. Pye, Clifton 1992 Language loss among the Chilcotin. International Journal of the Sociology of Language 93: 75-86. Sasse, Hans-Jürgen 1992 Theory of language death. In Language Death: Factual and Theoretical Explorations with Special Reference to East Africa, Matthias Brenzinger (ed.), 7-30. (Contributions to the Sociology of Language 64.) Berlin/ New York: Mouton de Gruyter. 2002 Creativity in obsolescent languages. Unpublished handout presented on occasion of the workshop "Multilingualism and language endangerment", 24. Jahrestagung der Deutschen Gesellschaft für Sprachwissenschaft, February 26, 2002, Mannheim (4 pages). Watahomigie, Lucille J., and Akira Y. Yamamoto 1992 Local reactions to perceived language decline. Language 68 (1): 10-17. Willemyns, Roland 1997 Toward a plurilingual urban environment: Language policy and language planning in Brussels. In Language Choices: Conditions, Con-

Reasons for language shift

21

straints and Consequences, Martin Pütz (ed.), 179-193. Amsterdam/ Philadelphia: John Benjamins. Winter, Werner 1993 Some conditions for the survival of small languages. In Language Conflict and Language Planning, Ernst Hakon Jahr (ed.), 299-314. (Trends in Linguistics. Studies and Monographs 72.) Berlin/New York: Mouton de Gruyter.

Directionals in Tokpe Gola Tibetan discourse1 Nancy J. Caplow

1. Introduction Tokpe Gola (TG) is a dialect of Tibetan spoken primarily in the Taplejung district of northeastern Nepal, in villages along the upper reaches of the Mewa Khola. In this paper I provide a descriptive account of Tokpe Gola's topographic directional markers, addressing the semantic meaning, syntactic distribution, and discourse function of the various morphological forms in which they occur. This descriptive study reveals the preservation, on directionals, of otherwise obsolete distinctions between local case markers. This analysis is based on natural discourse data, consisting of transcriptions of narratives and conversations recorded in the villages of Papung and Donggen, as well as in Kathmandu.2 In the examples, each line represents an intonation unit (IU), after the conventions of DuBois et al. (1993). The transcription is largely phonemic (with some phonetic detail preserved), and follows the 1993 IP A standards, with a few modifications.3 Where I draw comparisons between Tokpe Gola and Written Tibetan, the latter is transliterated according to the Wylie system (Wylie 1959), except that I prefer to use a period V rather than a blank space to represent boundaries between syllables. The semantic basis of the system is described in section 2 below. Sections 3 and 4 illustrate the correlation between the morphological form of directionals and their adverbial and anaphoric functions, respectively. Conclusions are presented in section 5. 2. Semantics of the system The forms of interest here are the three roots ya- 'upgradient', ma- 'downgradient', and ha- On contour'. These are topography-based deictics, referring to locations uphill or upstream, downhill or downstream, and at the same elevation, respectively.4 The location they index may be relative to the position of the speech-act participants at the time of utterance, relative

24

Nancy J. Caplow

to the position of the referents during the event being discussed, or relative to a mentally-constructed position which the speaker evokes through use of the term. These three directional terms comprise a closed class which is finite and complete. That is, with respect to topography, there are no other possible dimensions, nor are any dimensions omitted; any point must lie either upgradient, downgradient, or at the same elevation as another point. Of course, topographic directionals are only a subset of a more extensive deictic system. But it is not surprising that they play an important role in daily conversation: the area in which Tokpe Gola is spoken is dominated by narrow river valleys, with homes and terraced fields sculpted into their steep walls. It is almost impossible to follow a trail from one village to another, or even from one house to another, without going up or down. Furthermore, because the topography is so dramatic, one only rarely refers to a place at the same elevation; thus the term ha- has come to be used more broadly to mean Over there'. Tokpe Gola's topographic directional stems have clear cognates in Classical and modern literary Tibetan: TG ya-, ma-, and ha- correspond to WT (Written Tibetan) ya-, ma-, and pha- respectively (Beyer 1992: 207). I consider Tokpe Gola's stems to be bound. The morphological forms which function adverbially are discussed in section 3, and the forms which function anaphorically are discussed in section 4.

3. The adverbial function When they function in discourse as demonstrative adverbs, directionals occur either with a long vowel, or as part of a larger morphological complex.

3.1. Long vowel forms The long vowel forms ya:, ma:, and ha: are associated with verbs of translational motion, and indicate the direction in which that motion takes place. The verb may be intransitive (most commonly 'come' or 'go') as in (1) and (2), or transitive as in (3) and (4):5

Directionals in Tokpe Gola Tibetan discourse

25

(1) intransitive verb of translational motion (AK31) AK: tbe ya: yog a de. then up come.PST de yar yong.a.'dad6 'Then [we] came up.' (2) intransitive verb of translational motion (AK 19-20) AK: kbä:do kä:-ni ya: f f f'f, load load=NF up go.PST.REDUP khal.do bkal.nas yar phyinphyin AK: ma: f*f f f tf*e ßa. down go.PST.REDUP do.PST mar phyinphyin byas.ba 'Packing the loads, we used to go up (to Tibet) and come back down.' (3) S:

transitive verb of translational motion (BS205) djda nyi-sum-y ya: tfe:nya=ne=si. terrace two-three-four up follow=NF=HS ?.? gnyis.gsum.bzhi yar rjes.?.nas.zer '[He (Khansa)] followed [him (Pema)] up a couple of terraces.'

(4) LP:

transitive verb of translational motion (BS163-164) ha: khu=ni, over.there take=NF phar khw.nas fo:=sa=ni minduk, deposit=PL.NOM=TOP NEG.EX bcug.sa.ni mi. 'dug Taking [it] over there (to Taizom village), there was no safe place to leave it.'

LP:

Aside from false starts and anacoluthon, the directional adverbial expression directly precedes the verb, even when a direct object is lexically overt. In (5) and (6) below, ma: and ya: intervene between the verb and its object. In example (7), though, ma: is apparently separated from the main verb by a subordinate clause.

26

(5)

Nancy J. Caplow

lexically overt object (AK51)

DT: tshä

phö=ni

ma:

tönda

natjade=a\

salt Tibet=ABL down import do.HON-PST=Q tshwa bodnas mar ?.? gnang.a.'dad.a 'So [you] imported salt down from Tibet?' (6)

lexically overt object (AK81)

kbä:do

AK: dzo=la

ya: kyä:

ma:

[ANIMAL]=LOC load

up

take

down take

mdzo.la

yar

bskyal

mar

khaLdo

kyai. bskyal

On the dzos, [we] took loads up, [we] took [them] down.' (7) intervening subordinate clause (AK39) AK: ma: tsha gyak=tu iff. down salt exchange=PURP go.PST mar tshwa rgyag.tu phyin '[We] went down in order to trade salt.' Directional adverbials can also be metaphorically extended for use in the domain of time, as in (8) below: (8) LP:

metaphorical use (BS13-14) gundo=a, in.winter=also dgun.dus yang

LP: phädo

ya:

tfa-ni

far-

gya(y)i si=na,

fifth.month up/beyond arrive/beginning ? fall=COND [NEP] yar bcar.na ? rgyag.?.na 'In winter too, if it hails beyond the start of the fifth month (of the Nepali year) ...' These long-vowel directional adverbs correlate directly with the WT forms yar, mar, and phar, composed of the deictic root followed by the suffix -r, which, in both modern literary Tibetan (Goldstein 1991: 132) and Classical Tibetan (Wilson 1992: 136; Beyer 1992: 248) is added to words ending in open syllables to derive adverbs. WT -ar rhymes are commonly pronounced as -a: in modern spoken Standard Tibetan (ST) (e.g., Goldstein 1991: 33) and often subsequently shortened to a simple open vowel -a.

Directionals in Tokpe Gola Tibetan discourse

27

This is what we observe in Tokpe Gola as well; in faster speech, the longvowel forms ya:, ma:, and ha: usually surface as ya, ma, and ha. Note that while Wilson views -r as deriving adverbs from nouns and adjectives, and Beyer suggests that "[ajdverbs of place are regularly formed from location nouns ...", it is not clear here whether the directional stems ya-, ma-, and ha- should really be categorized as either nouns or adjectives. They could be regarded as deictic roots which do not belong to any lexical category until they are marked morphologically as appropriate for a particular semantic and syntactic context.

3.2. Morphologically complex forms The morphologically complex forms are composed of two directional roots linked by a case marker. Reduplicated forms, as in (9), refer to an extreme or distal point in a particular direction. In other forms, such as those in (10), directionals indicating both the source and the goal of translational motion are linked by an ablative case marker (-ni, -le). These complex forms define not just a direction but a complete trajectory. Before the ablative -m in (lOb, d, e) the stems occur as yei-, mai- and hai-\ I attribute this diphthongization to vowel harmony (rather than, for instance, suggesting that there is an intervening genitive case marker -yi —i). Likewise, I suggest that in (lOc) the ablative marker is -le, which is pronounced [la] due to assimilation with the preceding and succeeding [a].7 (9)

reduplicated directional adverbs a. malama 'way down below' (down.below-LOC-down; lit. 'below down below') b. häihala 'way over there' (over.there-GEN-over.there=LOC/ALL; lit. 'at over there's over there')

(10) trajectory adverbs a. maleya 'up from below' (down-ABL-up) b. mainiya 'up from below' (down-ABL-up) c. yalama 'down from up there' (up-ABL-down) d. ye(y)inima 'down from above' (up-ABL-down) e. halstsu 'here from there' (over.there-ABL-here) f. häinitsu 'here from there' (over.there-ABL-here)

28

Nancy J. Caplow

These forms again function adverbially with transitive and intransitive verbs of translational motion. Their distribution differs from that of the long vowel forms discussed above, though, in that, encoding more semantic information, they have more syntactic freedom. In the examples below they can be observed directly before the verb, separated from the verb by an object or subject, and even intonation unit (lU)-finally and clause-finally.

(11) reduplicated form (BS182) S: thi=la näm=ni häiha=Ia that.side=LOC take.HON=NF over.there.GEN over.there=ALL ?.la bsnams.nas pha.gi pha.la tfon gi duk=si. was.going.HON=HS byon.gyi. 'dug.zer 'Taking [it (that book)] to that side, [he (Lama Khansa)] was going way over there.' (12) directly before the verb (Gl 3) D: tshomgyu(k)=gi ye(y)inima duij=ni, large.pestle=INSTR up.ABL.down beat=NF ?.rgyug.gis ya.nas.mar brdungs.nas 'Beating it down from above with a large pestle ...' (13) IU-7 clause-initial (AK40) AK: mainiya dje: ka:=ni yog. down.below.ABL.up rice load=NF come ma.nas.yar 'bras bkal.nas yong '[We] came up from down there packing rice.' (14) IU-/ clause-medial (BS27) S: tsoq maJeya säb Limbu down.below.ABL.up all X ma.las.yar [NEP] 'All the Limbu people ran up from below

ts5g=ni, run=NF ?.nas ...'

Directionals in Tokpe Gola Tibetan discourse

29

(15) IU-/ clause-final (AK352) AK: dandu nyengf hahtsu. [M.NAME] listening over.there.ABL.here ? nyan.gi pha.las.tshur 'Ha ha. Dandu is listening in from over there.'

It is interesting to note that -le is fossilized as an ablative marker in these lexicalized forms. Elsewhere in Tokpe Gola only -ni is used productively, corresponding to the ablative marker in Standard Tibetan, written nas and pronounced [ne] (Goldstein 1984: 125). Beyer (1992: 269) discusses the use of both nas and las as marking "source" roles in Classical Tibetan (i.e., as ablative case markers). He distinguishes the two by the notion of "boundedness": a nominal source marked with nas refers to a delimited space which contains some component of the motion described, while las is used when the source does not contain any part of the action. Part of this analysis is helpful in understanding the speaker's choice between mainiya and maleya, both meaning 'up from down below', in (13) and (14). From context, it is clear that neither source can be considered bounded; both refer vaguely to villages at lower elevations inhabited by Limbu people. What seems to be consistent with Beyer's interpretation, though, is whether or not some component of the event took place at the source. In (14), where the ablative element is -le, no component of the action of running between one village and another takes place at the source. In (13), on the other hand, where the ablative element is -ni, part of the action of loading sacks of rice onto the pack animals and transporting them did, indeed, take place 'down below' (the packing did). This is also true of (12), in which part of the action of beating the fermented vegetable necessarily takes place up in the air, through which the pestle is moving. Thus we observe that complex directional terms in Tokpe Gola preserve morphological elements and semantic distinctions which are otherwise obsolete in the dialect. In this case, the ablative morpheme -le is preserved in contrast to the productive -ni. In section 4.1 below, we will see the obsolete locative and allative morphemes -na and -ru preserved in contrast to the productive -la.

30

Nancy J. Caplow

4. The anaphoric function In addition to their adverbial function, direction markers can also serve anaphorically; i.e., they can function as demonstrative pronouns referring either to a place or to a person associated with that place. These anaphoric directionals are marked with locative-allative or genitive case clitics, respectively. Again, it is not clear that, because they can be case-marked, Tokpe Gola's directional stems must inherently be nouns; rather, the speaker's use of case-marking indicates that they have referential meaning in the given context.

4.1. Place anaphora Place anaphora is marked by the attachment of locative-allative case markers to the directional roots. Three different case markers are used - -la, -TU, and -na - yielding the nine variants shown in the table below. Interestingly, of these three markers, only -la is used productively and frequently in natural discourse. While they are common in Classical Tibetan, the occurrence of -ru and -na is severely limited in Tokpe Gola; anaphoric directionals seem to preserve these relict endings. Constraints governing the speakers' selection of -la, -ru, and -na forms (and their phonetic variants) are discussed in the sections below.

Table 1. Directionals anaphoric for place directional 'up there' 'down there' Over there'

-la yala mala häla

locative-allative -ru yem maru ham

-na yana mana hana

These nine forms refer not just to a direction, but to a specific place which is recoverable from the discourse context. They also function as clause- or sentence-level place adverbials. Correspondingly, they can occupy different positions in the intonation unit or clause, and need not be close to a verb at all.

Directionals in Tokpe Gola Tibetan discourse

31

4.1. L -ru marked forms With twenty-five tokens of the forms yaru, maru, and haru, directional stems with the -ru suffix occurred far more frequently in the transcripts examined than did forms with the -la or -na ending. In some cases -ru forms have a locative sense, in others an allative sense, and in some cases it is analytically indeterminable. (Of course, in such cases the distinction is not relevant for speech-act participants, either.) In (16) and (17), -ru forms occur with verbs of translational motion. In contrast to (1) and (2), though, they indicate a place rather than a direction. Note that the demonstrative form need not occur directly before the verb. (16)

before the verb (BS 317)

LP: khö ni

yeru

ff=ni,

then up.there.LOC?ALL? kho.nas ya.ru "Then going up there ...' (17) IU-initial (BS 316) LP: /ram over.there.LOC?ALL? pha.ru

ne:bo effigy ?

go=NF phyin.nas

fom=du prepare.offering=PURP gshom.du

ff de. went.l phyin. 'dad '[I] went over there to prepare the effigy./ [I] went to prepare the effigy over there.'8 They also occur more clearly as sentence-level place adverbials occupying various positions in an IU, as in (18) and (19), or even as an independent IU (20). (18) place adverbial; IU-initial (AK174-176) C: dßarj sj=na khö tsh ba=^. djsarj say=COND 3.SG trader=DEF 'bras+X zer.na kho tshong.pa-X 9 '"dearf refers to "that trader"?'

32

Nancy J. Caplow

AK: yaru mi up.there.LOG DD? ya.ru X 'The one up there we were talking about...' AK: yaru tsha:ij sj=ni mi yarj up.there.LOG salt-bazaar keep=NF DD? still ya.ru tshwa.X zin.nas X yang 'Still the one we were talking about keeping the salt market up there (19) place adverbial; IU-final (BS208-210) LP: äfo wä dsuk=gi thaq sik yberu, brother [NAME]=ERG one.time up.there.LOG a.jo dbang.phyug.gis thengs.cig ya.ru LP: pä=la=la, father=HON=DAT pa.lags.la LP: [hükpa fup=ne. scolded 'khrug.pa ?.nas 'One time "Older Brother" Wangjuk scolded father up there.' (20) locative; independent IU (AK236-238) AK: the khö: ma: khü:=ni ya:, then that?they? down take=NF again, still de ? mar khur.nas yang AK: maru, down.there.LOG ma.ru AK: süm=go gyakan de". three=DEF trade(do).PST gs um.X brgya b. mkhan. 'da d 'Then [they] took [it] further down (to Dobhan); down there they traded for three times as much.' -TU forms can also apparently be further suffixed. In (21) and (22) they mean something like 'up there as well' or 'in the vicinity of over there'. I have not yet been able to identify the additional morpheme, but suggest it

Directionak in Tokpe Gola Tibetan discourse

33

may be -rag 'self; emphatic', yarj 'also, again, still', or the nominalizer -pa ~ -wa.

(21) LP:

LP:

vicinity (BS291-292) tha härua, now there.EMPH? da pha.ru.rang? mobe marjbu du:[k], divination=NOM?=GEN? many EX.EP mo.pa 'i mang.po 'dug "'Now there are many [people] around there [who want their] divination [done]." (Lama Pema is quoting what he himself said to Lama Nupu.)'

(22) vicinity IU (JB37) PW: nuruwa sarjgye tha yaruwa here.LOC.EMPH? Buddha and up.there.LOC.EMPH? 'di.ru.rang sangs.rgyas dang ya.ru.rang saggye te. Buddha COP.EP sangs.rgyas 'dad '[They follow] Buddha around here as well as up there.'

It is worth noting that -ru occurs not only on directionals, but also on other deictic stems, as in (23) and (24), and on relator nouns (25) and similar spatial expressions (26). (23) deictic expression (BS250) LP: ya: kyä:-ni gabla t*j furu up taking later near.there.LOC?ALL? yar bskyal.nas 'gab.la ? tha dai taitaild tba now l.SG.HUM.ERG read.REDUP now that.time da bdag.gis bltas.bltas.la da kho'i.sgang.nas 'Later taking [it] up (to his own home) near there (in Simbuk), I was reading it and reading it at that time.'

34

Nancy J. Caplow

(24) indefinite deictic expression (JB133) PW: läma k"aru yötä? lama where.LOC EX.Q bla.ma ga.ru yod.dam? '"The lama, where is [he]?" (The Nepalese army asks this.)' (25) on a relator noun (-orii) (B S141) LP: [* = täggo=ru tha, bed.GEN=DEF on/top=LOC now khri'i=X steng.?.ru da 'So [keeping him] on the bed (26) on other terms (AK337-338) AK: the khoqtsa=o raggio=ru, then inner.pot=DEF outside.handy/nearby.GEN?=LOC de khog.rdza.X ?.?.?.ru AK: fa:r=ni, keep=NF bzhag.nas 'Then keeping the inner pot (of the "still", on the fire) nearby outside

A crucial fact about -ru is where it does not appear: -ru is never used as an allative or locative marker on places which are referenced by a full noun phrase. With full NPs, the marker -la is used exclusively.

4.1.2. -la marked forms

Anaphoric directionals with the -la case marker were less frequent than those with -ru, with only thirteen instantiations in the transcripts examined. In nearly half these cases, the directional terms yala, mala, and häla cooccurred with a full noun phrase specifically identifying the place referred to. As shown in (27) - (29), the lexical NP follows the directional, in the same or a separate intonation unit, providing clarification. Example (28) shows three locatives in apposition, with case concord, as indicated by the square brackets.

Directionals in Tokpe Gola Tibetan discourse

35

(27) lexical place NP in same IU (AK33) AK: khö ni häla PhQ=la f f da de: then there.ALL Tibet=ALL go.PST kho.nas pha.la bod.la phyin.pa.'dad 'Then [we] went over there to Tibet.'

(28) lexical place NP in same IU (AK16) AK: [mi nag=la] DD.(GEN?) house=LOC up.there.LOC X nang.la ya.la [m$kpi nan=la] temba sj=ni:, son-in-law.GEN house=LOC behave.appropriately say=NF mag.pa'i nang.la ?.? zer.nas 'Then in their house, up there, in the son-in-law's house, it was said that [I] behaved appropriately (preserving her family's good name).' (29) lexical place NP in subsequent IU (BS305-306) LP: tsbu yonsaru peja handja pefukide here after.coming book what copy tshur yong.?.ru dpe.cha ?.'dra dpe.bshus.kyi.'dad pesa häla, book there.LOC dpe.cha pha.la LP: täizom=Ia kyä: fak=ni , [PLACE]=LOC bringing keep=NF DM X.la bskyal bzhag.nas X 'After coming here (i.e., "there", to Taizom), why should I copy the book. I brought the book there, to Taizom, and kept it [there].' My data also include several examples in which -la forms are reduplicated to indicate that the distance traveled was considerable or that the journey was difficult. In (30) it is clear that yala (here with vowel assimilation) is used rather than ya: because Jung Bahadur's army is not simply traveling in some direction (upward) but is working its way to a specific destination.

36

Nancy J. Caplow

(30) reduplicated form (JB33-34) PW: ^e(the) ya=le ya=le ya=le ya=Ie ^ = then up.there=ALL REDUP do.PST=NF de ya.las ya.las ya.las ya.las byas.nas PW: Jesa sak gya:=ni {fbo=ni=sf. Lhasa until win=NF go.HON=NF=HS lha.sa X rgyal.nas byon.nas.zer 'Doing that, up up up up they went, winning up to Lhasa.' There were also instances in which -la directionals occurred independently, without an apposed lexical place name, as in (31): (31) -la with locative (BS174) LP: khöTu lö:lötfö=ni häla 3.SG.DAT7LOC? be.relaxed. trust.PST=NF over.there.LOG kho.ru lhod.lhod.bcol.nas pha.la fo: fak pe\ trast.PST.put.PST EP.AUX bcol. bzhag.pa.yin 'Since [I] had complete trust in her [I] kept [it (the book)] over there.' At present, the motivations for selecting -la forms rather than -TU forms in natural discourse are not entirely clear to me. The case-marking distinction is not semantically transparent at the clause level, as both -la and -ru can be locative, allative, or indeterminate. I offer two tentative hypotheses to account for their distribution. Under one interpretation, the fact that -la anaphoric directionals are so frequently accompanied by a full noun phrase suggests that the referents of terms marked with -la may be more remote, in terms of both geography and discourse recoverability, compared to terms marked with -ru, thus sometimes requiring clarification. This seems to be consistent with the impressions of my primary consultant, who adds that speakers may point to a -ru location if it is sufficiently close. An alternative interpretation of these same facts is that -la is appropriate when the location referred to is more specific, thus allowing clarification by a full noun phrase; -ru forms, on the other hand, never co-occur with full NPs, which may be because they refer only to a more vague spatial domain. A speaker may physically point to a -ru location, if it is sufficiently close, in order to narrow down the referent. Further study of both synchronic (Tokpe Gola)

Directionals in Tokpe Gola Tibetan discourse

37

and diachronic (Written Tibetan) uses of these forms may help clarify the issue. While -la is relatively rare on directional terms, it is the only case marker used on full noun phrases referring to place. Some of the examples above illustrate this; there were dozens of other occurrences in the transcripts.

4.1.3. -na marked forms In the transcripts analyzed, there were only five instantiations of the anaphoric directionals yana, mana, and häna in complete clauses which could be evaluated. The low frequency of these forms corresponds to the fact that the -na case marker has the most narrow semantic meaning. While -la and -m have both locative and allative functions, -na is used strictly as a locative, and does not occur with verbs of translational motion. Furthermore, -na is unique in seeming to convey a sense of permanence and continuity. In (32), below, the female elephant's death at Walungchung Gola insures that her presence there is more than temporary. In (33), the point is that these people not only live in a high-elevation region, but that they are adapted genetically to its conditions and thus would suffer physically if they relocated to the semi-tropical plains. (This is, in fact, what happens sometimes when Tibetan refugees emigrate to India.) Thus they have a long and permanent association with this environment. When we reconsider the examples with -la and -ru above, it becomes clear, by comparison, that participants were only temporarily associated with the place referred to. -na is apparently the only marker which is used in existential and locative clauses, as in (34). (32) never to move again ... (JB69) PW: phö ya: leßs mo häna ßyg. male up arrived female over.there.LOG died pho yar sleb.pa mo pha.na shi.ba "The male reached up there (Lhasa), the female died over there (Walungchung Gola).'

38

Nancy J. Caplow

(33) genetic adaptation (JB112-113) PW: tsha- tshepa tsh§=ni the ni na, have.malaria=NF that.ABL l.SG('we') tshad.pa tsha.nas de.nas nga PW: is nai do fl: do sorja tba will.get.sick will.die said then na.yin.do shi.do gsungs.pa da yana mj, up.there.LOG DD? ya.na X 'Then they ("the people up there"?) said "we'll get sick and die of malaria.'" (34) existential/ locative predicate (BS87-88) LP: khöi pesa=o=ra häna, 3.SG.GEN book=DEF=EMPH over.there.LOG kho'i dpe.cha.X.rang pha.na 'His book is there (at Lama Pema's house, not close by) ...' LP: dai lak=la yo. l.SG.HUM.GEN hand=LOC EX bdag.gi lag.la yod 'It's in my possession./1 have it in hand.' The -na case marker does not occur in any other morphological context in my database aside from these directional stems. Like -ru, it never occurs on lexical noun phrases. But its morphological distribution is even more restricted: unlike -ru, it is not used with relator nouns, the stem meaning 'where', or in any other context whatsoever. It is clearly no longer productive in the Tokpe Gola dialect. Thus Tokpe Gola's local case marker -na is interesting in several respects: it is relatively rare, occurring nowhere but on directional stems; it has only a locative and never an allative interpretation; it occurs with existential clauses; the limited data available suggest it may convey a sense of permanence. Several sources were consulted in search of a diachronic basis for these observations and hypotheses. According to Beyer (1992: 267270), in Classical Tibetan the "locus particle" na is used to refer to the place at which an event occurs, but only if it is a bounded space which in some way contains a component of the action. (If the space is unbounded with respect to the event, then la is used.) Considering the data, it is not

Directionals in Tokpe Gola Tibetan discourse

39

clear that this concept is relevant to the distribution of -na in Tokpe Gola. Beyer also provides examples of na used with relator nouns in Classical Tibetan (1992: 270), which we do not observe in Tokpe Gola. Jäschke ([1881] 1958: 298) provides one interpretation of na as an accusative case marker, "added to substantives, in ..., sometimes also to be rendered by on, at, with, to, etc." Chandra Das ([1902] 2000: 725) similarly identifies na as "the locative case sign added to substantives, and to be translated: in, on, at, unto ..."10 But his dictionary also includes a separate entry: "as a symbol na signifies steadiness ..." Though it is not clear how a "symbol" is expressed morphologically, it is possible that some element of this sense may be preserved in Tokpe Gola's -na. An objective of Chandra Das' work was to expand upon Jäschke's by adding the English translation of Sanskrit-Tibetan terms relevant to Buddhist terminology. Sanskrit Buddhist texts were translated into Tibetan "with literal accuracy" in the 7th and 8th centuries, but some of the Sanskrit terminology subsequently became obsolete and obscure. It was hoped that material from early SanskritTibetan translations included in the dictionary could help scholars "elucidate ... the original Sanskrit words, of which they were the equivalent renderings" (Chandra Das 2000: v-vi). Thus the entry for na signifying "steadiness" may reflect a subtle distinction of a Tibetan of 1200 years ago, no longer generally productive in the modern dialects, but still preserved in Tokpe Gola's directional expressions. 4.2. Person anaphora Cross-linguistically, genitive case markers are used to show an association between two entities: between two people, between people and things, between people and places. Among its other uses in Tokpe Gola (and other Tibetan dialects), the genitive case marker is attached to directional stems to show a close association between a person and a location. In some cases the genitive-marked directional can be used independently to anaphorically refer to the person associated with that location. The three forms that serve these functions in Tokpe Gola are: Table 2. Directionals anaphoric for person or object ye(h)i, ye(y)i ma(b)i

Of up there; the one from up there' Of down there; the one from down there' Of over there; the one from over there'

40

Nancy J. Caplow

These genitive-marked directionals have the same function and meaning as genitive-marked full noun phrases. Example (35), provided for comparison, shows an association of a person with a place expressed by a full noun phrase. In (36) a genitive directional is used, with a genitive noun phrase in apposition providing clarification; in (37) and (38) the genitive directional is used independently. The place referred to in these examples is a defining characteristic of the person referred to; in (36), Uncle Gyabu is not simply visiting or staying in Simbuk for a brief period; rather, he lives there, or was born there, and can forever be associated with the place.

(35) association of a person with a full NP place (BS23) S: e fyambi afag=gi khonde oh [PLACE].GEN maternal.uncle=ERG like.that X. 'i a.zhang.gis ? so// so=si.

told.HON=HS gs ungs. song, zer Oh, [I heard that] my maternal uncle from Syamba told about that.' (36) full NP and anaphoric directional term (BS126) LP: a/a/; gyaßu häi Jjgbuk=ki thj, maternal.uncle [NAME] that [PLACE]=GEN DBF a.zhang rgyal.po pha.gi X.gi de 'That Uncle Gyabu, [the guy] from over there, [the one] from Simbuk...' (37) place anaphora; appositive NP construction (AK162) AK: kb5 mi ha(y)i cfe nyo qe=o po ba, that DD? over.there.GEN rice buy.AG.NOM=DEF brought kho X pha.gi 'bras nyo.mkhan.X bor.ba 'That one we were talking about from over there, the rice buyer brought [it]'./ "That one we were talking about, the rice buyer from over there, brought [it]'./ 'That one we were talking about, the one from over there, the rice buyer brought [it]'.11

Directionals in Tokpe Gola Tibetan discourse

41

(38) place anaphora (AK248) AK: ye(h)i tsh gya?-... up.there.GEN salt plains ya.gi tshwa rgya.'Salt from up there (Tibetan salt) ... plains [salt] In (39) and (40) the genitive form occurs without a head, fully anaphoric for the person associated with the understood place 'up there'. (39) person anaphora (BS317-318) LP: kb ni yaru ff=ni, then up.there go=NF kho.nas ya.ru phyin.nas LP: yehi ^opa=o 1έρ=ηΐ, up.there=GEN funeral=DEF arrive=NF (time) ya.gi mchod.pa-X sleb.nas 'Then going up there, at the time of "up there's" funeral...' (40) person anaphora, equational clause (BS207) LP: nyakto ye(h)i hi. [NICKNAME] up.there.GEN COP X ya.gi yin 'It was Nyakto, the one from up there./ Nyakto is the one from up there.' Genitive-marked directionals can function as the head of a relator noun, just as full NP's do, as (41) and (42) illustrate. In (42) h i means 'the one of over there', referring anaphorically to the box. (Note also the use of -TU with a relator noun in this example.) (41) physical association (full NP with relator noun) (BS335) LP: nag=gi fug=la nya:=ni, house=GEN center=LOC sleep=NF nang.gi gzhung.la nyal.nas 'Sleeping in the center of the house ...'

42

Nancy J. Caplow

(42) object anaphora; physical association (with relator noun) (BS147148) LP: nuk taktsa:=ni tb§:sarj(th€saru) ggm, like.this back, behind=ABL taking.out.from.bag box 'dug ltag.rtsa.nas ?.? sgam LP: häi täktsa:=ru hüryuna qei. over.there.GEN back, behind=LOC throw l.SG.ERG pha.gi ltag.rtsa.ru ?.g.yug.? ngas pulled it out of the bag behind my back, like this (gesturing), and threw it behind that over there, the box.' Thus while directional stems with local case markers function to refer anaphorically to places, directional stems with the genitive case marker function to refer anaphorically to people or objects. 5. Conclusions This paper provides a description of the meaning and use of the topographic directionals ya-, ma-, and ha- in Tokpe Gola Tibetan discourse. When they serve an adverbial function, these directional stems occur with a long vowel or as part of a morphologically complex form. When they function anaphorically, they are marked with a case suffix. The local case suffixes -la, -ru, and -na mark forms that refer anaphorically to a place, while the genitive suffix -gi ~ -(y)i marks forms that refer to a person or object associated with a relative spatial position. This descriptive study exposed several interesting historical features. Both adverbial and anaphoric directionals preserve relict morphemes which are documented in Classical Tibetan but are no longer productive in Tokpe Gola. Several of the morphologically complex adverbial directionals include the ablative morpheme -le (las in WT), which does not occur elsewhere in modern Tokpe Gola. The local-case-marked anaphoric directionals similarly preserve distinctions which are not otherwise apparent in the contemporary language. Directionals ending in the locative/allative markers -la, -ru, and -na do not occur with equal frequency, and we observe patterns in the speakers' choice between these forms. Directionals ending in -ru are most common, and this case ending is also observed on relator nouns and other expressions of relative spatial position; -ru thus seems to be the default marker for spatial terms. While -la is used productively and

Directionals in Tokpe Gala Tibetan discourse

43

frequently on full noun phrases, it occurs on directionals less frequently than does -ru. -la may be used to refer to locations which are more distal or more specific; further study will be required to distinguish between these hypotheses. Finally, directionals with -na are exceedingly rare, and seem to have a semantic connotation of permanence, though this hypothesis, too, will require further investigation. The case marker -na does not occur anywhere else in Tokpe Gola. Topographic directionals may have preserved an otherwise obsolete distinction. Further consideration of both elicited and discourse data will be needed to either confirm or reject the hypotheses offered here. If confirmed, what will still remain a puzzle is how these three local case markers came to have the meaning and distribution which they do in Tokpe Gola. It will be interesting to determine whether these patterns can be confirmed in other dialects, or traced to previously unobserved distinctions in Classical Tibetan. It is not surprising that forms such as demonstratives and directionals would be the locus of such fossilization. Given their function, these few terms must necessarily occur far more frequently in conversation than any of the infinite number of full-NP place names to which they refer. While frequency is often associated with phonetic erosion and semantic bleaching, here we find that the continued juxtaposition of directional terms with case markers is actually responsible for preservation of the latter.

44

Nancy J. Caplow

Table 3. Glossing abbreviations

1 3 ABL AG.NOM ALL AUX COND COP DAT

DD DBF

DM EMPH

EP ERG

EX GEN HON

HS

first person third person ablative agent nominalizer allative verbal auxiliary conditional copula dative discourse deictic definite marker discourse marker emphatic epistemic ergative existential genitive honorific hearsay

HUM INSTR LOG [M.NAME] NEG NEP

NF NOM

PL PL.NOM

PST PURP

Q REDUP

SG TOP

=

humilific instrumental locative male name negative Nepali non-finite nominalizer plural place nominalizer past tense purpose question marker reduplicated form singular topic marker precedes suffix precedes clitic

Notes I would like to thank my Tokpe Gola friends for sharing their language with me, and my consultants Dandu Sherpa and Pasang Sherpa for helping me to understand it. My research in Nepal was largely made possible by a Fulbright scholarship from the United States Education Foundation (USEF). I would also like to thank Susanna Gumming for her many helpful comments and suggestions as I analyzed the data. Three of the texts I draw on are lengthy conversations. In Ani Khasang (AK) a village woman in her 50's describes activities of her early married life. In Book of Spells (BS) a local lama relates tales of how he and his uncle and cousins would trick each other to obtain a book of spells used by powerful lamas. Jung Bahadur's War (JB) is a historical narrative about a war between Nepal and Tibet which took place in the mid-1800's. Gundruk (G) is a short procedural narrative describing the process by which gundnik, a fermented vegetable, is prepared. For the palatal approximant I use y rather than / This is consistent with the use of y in the Wylie transliteration of Written Tibetan (Wylie 1959), and thus eliminates the confusion which might arise if I were to use j in transcription and y in transliteration to represent the same sound. For the palatal

Directionals in Tokpe Gola Tibetan discourse

4.

5.

6.

7.

8. 9. 10.

45

nasal, I use fly rather than the IPA/j, a choice which again is consistent with the Wylie transliteration. Palatal stops are transcribed here as ky and gy (rather than the IPA c and j), since c is often used to represent the palatal affricate in Tibetan. I also use ΰ and o, rather than IPA γ and 0 ~ ce, respectively. The terms "upgradient", "downgradient", and "on contour" are commonly used in fields such as geology, hydrology, and geography. They are seldom, if ever, employed by linguists, but I would like to encourage their use, as I define them in the text, as they are succinct and technically precise. The expression "on contour" refers to a point lying on the same abstract topographic contour line. Line 1 of each example provides a phonemic transcription (with selective phonetic detail preserved) preceded by initials indicating the identity of the speaker; " ~ " indicates high tone, while "_" and "_" indicate low tone, the latter with breathy phonation. Line 2 provides a morpheme-by-morpheme gloss, line 3 the Written Tibetan (WT) cognate forms (a "?" indicating that I was not able to determine the cognate for a particular syllable, an "X" indicating that no cognate seems to exist), and line 4 a free translation. Words in square brackets are elided in the Tokpe Gola utterance; words in parentheses provide explanation. Abbreviations such as "AK31" refer to the transcript title (e.g., "Ani Khasang") and the intonation unit number. Glossing abbreviations are provided at the end of this paper. In this example, the past tense is conveyed by the morphemes a de. These correspond to the Standard Tibetan and Written Tibetan sequence pa red. de in Tokpe Gola generally has the same auxiliary, copular, and epistemic functions as red. Here, though, I use a non-standard written form, 'dad, which was suggested to me by Nicolas Tournadre (personal communication 2002). Note, though, that J schke (1958: 539) points out that la can be used to mean not only 'to', etc. but also 'from', adding that "[t]his latter use of la occurs so frequently that it cannot always be looked upon as a misspelling for las, though this would be the more exact word." Thus the ablative morpheme in Tokpe Gola's complex forms may really be -la rather than -le, another example of an obsolete case marker being preserved in directional terms. After someone dies, before the dead body is removed from the house, a cross is made from bamboo and the deceased's clothes are put on it. This is anebo. The word djeag refers to a bazaar at which rice is traded; the speaker, a young woman, wasn't familiar with the term, and is asking the narrator about its meaning. Both dictionaries also include entries for la, of course, Chandra Das listing both na and ru as alternate forms. Neither includes a separate entry for ru.

46 11.

Nancy J. Caplow As much as one might wish to provide a single, decisive free translation, in fact, it is simply not possible at present for me to distinguish between these three plausible offerings. The ambiguity is the result of what I consider to be a gradient relationship between appositive and relative clause constructions in Tokpe Gola and other Tibetan dialects.

References Beyer, Stephan V. 1992 The Classical Tibetan Language. Albany: State University of New York Press. Chandra Das, Sarat 2000 Reprint. Tibetan-English Dictionary. Delhi: Motilal Banarsidass. Original edition, Calcutta: Bengal Secretariat Book Depot, 1902. DuBois, John W., Stephan Schuetze-Coburn, Susanna Gumming, and Danae Paolino 1993 Outline of discourse transcription. In Talking Data: Transcription and Coding in Discourse Research, Jane A. Edwards, and Martin D. Lampert (eds.), 45-89. Hillsdale, N.J.: Lawrence Erlbaum Associates. Goldstein, Melvyn C. 1984 English-Tibetan Dictionary of Modern Tibetan. Dharamsala: Library of Tibetan Works and Archives. 1991 Essentials of Modern Literary Tibetan: A Reading Course and Reference Grammar. Berkeley/Los Angeles: University of California Press. Jäschke, Heinrich August 1958 Reprint. A Tibetan-English Dictionary. With Special Reference to the Prevailing Dialects. Original edition, Routledge & Kegan Paul, 1881. Wilson, Joe B. 1992 Translating Buddhism from Tibetan. Ithaca: Snow Lion. Wylie, Turrell V. 1959 A standard system of Tibetan transliteration. Harvard Journal of Asiatic Studies 22: 261-267.

The language history of Tibetan Philip Denwood

The history of the Tibetan language in all its variant forms is important not only in its own right but also in wider comparative and historical studies. In particular the earliest reconstructible stage of Tibetan, however arrived at, has and will continue to loom large in the comparative and historical study of the Tibeto-Burman and Sino-Tibetan languages, both because of its rather large quantity of linguistic material and because of its alleged antiquity, generally reckoned to be surpassed within these groups only by Chinese. In this paper I offer some thoughts on the problems and possibilities offered by four different approaches to Tibetan language reconstruction: respectively script-based, dialect-based, lexis-based and grammar-based, together with some historical and geographical observations. 1. Script-based approaches If only because of its antiquity, the Tibetan writing system must obviously play a major part at some stage in any attempt at the reconstruction of early Tibetan, and indeed the traditional approach to the problem has been heavily dependent on the written language. So much so, that some scholars have assumed that there is actually no problem at all: "... Shafer would have said that we already know what Proto-Tibetan looks like: it is embalmed in the orthographic forms of Written Tibetan (his Old Bodish' ...)" (Sprigg 1972: 556). Other scholars besides Shafer have tacitly or explicitly assumed that written Tibetan is effectively a phonemic representation of the language spoken at the time it was first written down. H. A. Jäschke in 1860 had already noticed the fact that initial consonant clusters apparently corresponding closely to written spellings were pronounced in some parts of Ladakh and Purik. When discussing his "Phonetic table for comparing the different dialects", he writes (1881: xii): The first column of the Table [in effect a transliteration of the Tibetan script] ... gives the ancient literal pronunciation, as it was in vogue in the

48

Philip Denwood seventh century of our era ... A reference to the Table will amply illustrate the fact, that a pronunciation, adopted on these principles, has actually maintained itself in one or the other provincial dialect, and it is very interesting to notice, that the purest and most striking forms of this survival have their homes in those districts, which are most remote from and least subject to the disintegrating and dissolving influences of the actual centre of Tibetan civilisation, the capital Lhasa.

Variants of the same view have been taken by, among many others, Benedict (1972), Bonnerjea (1936, 1937), Dragunov (1936), Durr (1950), Francke ([1901] 1979), Li (1933), Migot (1956), Sedläcek (1964), Shafer (1974) and Wen Yu (1946). In extreme cases early written and spoken Tibetan are even taken to be one and the same thing: thus "If one desires to know how classical Tibetan is changed into modem dialects ..." (Wen 1946: 1. Further examples inRona-Tas 1966: 106-107). Although they take a much more sophisticated stance, as will be seen below, even R. A. Miller and A. Rona-Tas echo the script-based approach; thus Miller (1955: 284): It has long been generally known that after the Tibetan language was committed to writing, sweeping phonetic changes took place, at least in the central and south-central dialect areas, while the border dialects, those on the east and west, appear to have escaped these changes to a surprising extent; these last, even today, retain a phonetic aspect strikingly like the language preserved for us by the orthography (my italics).

Miller's "central and south-central" and his "border dialects" correspond to Rona-Tas's "non-archaic" and "archaic" dialects respectively. It is worth repeating Rona-Tas's definition here: The living Tibetan dialects can be divided into two major groups: the archaic and non-archaic dialects. I use the term "archaic" for those dialects which do not have pitch as a phonematic suprasegmental feature, and have preserved in a more or less complete form the preradical [initial consonant-cluster] system of Old Tibetan. The term "non-archaic" denotes those dialects which have a phonematic pitch and lost the old preradical system (1966: 21. My italics).

Rona-Tas (1966: 106-108) is careful to distinguish sound from script, keeping their respective sets of terminology distinct (though his term "radical" is common to both sets). His "preradical" and "Old Tibetan" belong to

The language history of Tibetan

49

his phonological set of terms. In practice, however, it is hard to see how his "Old Tibetan" really differs from the written language, since the latter provides his only evidence for it. The script-based hypothesis then is that the spellings of written Tibetan, established at some ancient though often unspecified date, correspond more or less faithfully to the pronunciations of the modem-day border or archaic spoken dialects, thus allowing us to establish with a fair degree of certainty the phonology of Tibetan at the time it was first written down. The hypothesis is mainly concerned with the problems of the syllable-initial consonant clusters which the spellings of written Tibetan appear to be representing. These clusters are certainly many and varied, as a list of the recorded written possibilities will show. Even leaving out those basic letters which might be considered as representing consonant clusters in themselves, (i.e. c- ch- j- ny- ts- tsh- dz-) as well as combinations with subjoined letter -w- and combinations only found in Preclassical documents, the list is as follows: ky- Ar- kl- dk- dky- dkr- bk- bky- bkr- ik- iky- Ik- sk- sky- skr- biky- bskykhy- khr- mkh- mkhy- mkhr- 'kh- 'khy- 'khr- gy- gr- gl- dg- dgy- dgr- bgbgy- bgr- mg- mgy- mgr- 'g- 'gy- 'gr- rg- rgy- lg- Igy- sg- sgy- sgr- brgbrgy- bsg- bsgy- dng- mg- sng- bmg- bsng- gc- be- Ic- mch- 'ch- mj- 'j- rjIj- brj- gny- mny- my- sny- bsny- tr- gt- bt- rt- It- st- brt- bit- bst- thr- mth'th- dr- gd- bd- md- 'd- rd- Id- sd- brd- bsd- gn- mn- m- sn- bm- bsn- py- prdp- dpy- dpr- sp- spy- spr- phy- phr- 'ph- 'phy- 'phr- by- br- bl- db- dby- dbr"b- "by- 'br- ib- Ib- sb- sby- sbr- my- mr- dm- dmy- rm- rmy- sm- smy- smrgts- bts- rts- sts- brts- mtsh- 'tsh- mdz- 'dz- rdz- brdz- gzh- bzh- zl- gz- bzbzl- g.y- rl- brl- shr- gsh- bsh- gs- bs- sr- si- bsr- bsl- hr- Ih- (166 in all).

The reconstruction of early Tibetan then becomes a process whereby the various modern dialects, and particularly those of the far east and far west of the Tibetan language area, are scanned for syllables for which their conventional (or, in some cases, assumed) written Tibetan spellings seem to be straightforward one-for-one representations; e.g. Golok mtsho mtsho, 'lake'; Balti rgjamtsho rgy a. mtsho. 'river'. One then concludes that, for example, the written prefixed letter m- represents m-, and the head letter rrepresents r- in the pronunciation of Tibetan at the time it was first written down. The sounds have been preserved in, in this case, the Golok and Balti dialects, but not in, for example, the Lhasa dialect where we now have ~tsho mtsho. 'lake' and_ojatso or_d?atso rgya.mtsho. 'sea'. Since the great majority of the words of the modern dialects seem to be represented by

50

Philip Denwood

independently known written equivalents from literary works and dictionaries, the whole question of the modern dialects and their relationship to written Tibetan may superficially seem to be unproblematic. 1.1. Problems of script-based approaches I do not wish to condemn this procedure as totally misleading, for it obviously has considerable validity. It is often tempting to agree with J. Sun (1986: 146) that "whatever dialect served as the basis of Written Tibetan, the latter seems quite capable of covering all the modern dialects" - that is to say, that all the modern dialects can be seen as having descended from it, in the way that, say, the multifarious Romance languages are popularly supposed to be descended from "Latin". At the same time, on closer inspection the situation appears to be not so simple. The "original" entity which is the object of reconstruction is seen to be conceived of in a number of different ways by different scholars sometimes even by one and the same scholar. Thus Miller speaks in the passage quoted above of "the language preserved for us by the orthography", but elsewhere says "[Written Tibetan] is utilized here only in default of being able at present to make the ultimate comparison with proto-Tibetan, and is not to be taken as signifying that W[ritten] T[ibetan] is the same as proto-Tibetan, a manifest impossibility ..." (Miller 1956: 349. My italics). Rona-Tas writes: ... I arranged the material according to the structure of the Old Tibetan language hypothetically assuming that all the present-day dialects, as well as the literary language, can be defined in terms of Old Tibetan. This is not assuming ... that all the present Tibetan dialects and the literary language can be genetically connected with Old Tibetan which in its turn can hardly have been a homogeneous language ... The structure of OT is reflected in the texts found in Tun-huang, the Tarim Basin and Central Tibet, but ... the classical or literary language also reflects the same phonemic situation (1966: 107).

Elsewhere (1985: 94-96) he distinguishes "Vortibetisch" from "Urtibetisch", and divides his "Alttibetisch" or "Old Tibetan" into "Friih-Alttibetisch" (650 A.D.), "Mittel-Alttibetisch" (650-814) and "Spät-Alttibetisch" (814-11th century, in which period he would place the great bulk of surviving "Old Tibetan" material).

The language history of Tibetan

51

Even in the works of the five authors referred to above - Shafer, Wen Yu, Miller, Sun and Rona-Tas - we have a number of linguistic entities, variously and sometimes confusingly defined ("Old Tibetan", "Classical Tibetan", "Literary Tibetan", "Written Tibetan", "Old Bodish", "ProtoTibetan", "the language preserved for us by the orthography"): entities which veer back and forth between the written and the spoken, which extend over a time-span of up to 500 years or more, from the mid-7th to the 11th century, and which are only vaguely located in space. Withoutpinning this "original" entity down more precisely, one could argue that the question of its relationship with any particular written or spoken forms becomes virtually meaningless. To give just one example, while Rona-Tas extends his "Old Tibetan" to the llth century, according to Miller (1955: 290-291) already some two centuries earlier: The language of Central Tibet seems to have reached, as far as its phonetic structure is concerned, very nearly the same status in which we find it today, and the distance, phonologically speaking, which separates the language of this inscription from the modern dialects of Central Tibet, is so small as to make all but impossible a comparison with the gulf between the early ninth century language and that reflected in the script.

1.2. Jäschke and Das In fact, the main bulk of the evidence for the original or ancestral entity, whatever name a particular author chooses to call it, usually turns out to be the lexicon as given in the two most widely used Tibetan-English dictionaries: those of H. A. Jäschke (1881) and S. C. Das (1902) respectively. J. Sun, for example, is quite open about this (1986: 145): "In this thesis Classical Tibetan words as codified in the traditional Tibetan orthography are cited from two dictionaries, Das ... and Jäschke". Various problems arise out of the concentration on this lexicographical corpus:

1.2.1.£tatdelangue Firstly, the dictionaries in question claim neither to be inventories of any particular dtat de langue - that is to say, the Tibetan of any one time or place - nor to concentrate primarily on "Classical", "literary" or "written"

52

Philip Denwood

Tibetan: - in fact rather the reverse. One might wonder how many of the scholars who rely so heavily on the two dictionaries have troubled to read Jäschke's preface (1881: iii-iv; italics mine): The chief motive of all our exertions lay always in the desire to facilitate and to hasten the spread of the Christian religion ... My instrument must be, as in the case of every successful translator of the Bible, so to say, not a technical, but the vulgar tongue. Thus, in contrast to the business of the European philologist, engaged in the same domain, who quite rightly occupies himself with the analysis and commentary of a literary language, the vocabulary and terminology of which he finds mainly deposited in the speculative writings of Buddhist philosophers, it became my duty to embrace every opportunity, with which my presence on the spot favoured me, to trace the living powers of words and of expressions through their consecutive historical applications, till I reached their last significance in their modern equivalents, as these are embodied in the provincial dialects of the native tribes of our own time ... As an inventory of the whole treasure of the language, as a finished key to its literature, this Dictionary, when judged by the high standard of modern lexicography, may seem inadequate; I have, for instance, been unable to consult, much as I could have wished to have done so, all the original and translated treatises in Tibetan which, down to the present, have appeared in Europe, and the reader of a Tibetan work may thus, here and there, look in vain for the assistance he expects.

The dictionary of Das, it is true, contains a large number of entries culled from Tibetan Buddhist and other literature, though many of the works from which they are taken date from much later than, say, RonaTas's "Old Tibetan" period. Das's dictionary itself is in any case heavily indebted to Jäschke, and its editors, in their own words, "have sought to introduce a number of colloquial words and phrases belonging to the Central and Eastern speech". 7.2.2. Spellings The second problem is that a study of the entries in these dictionaries will suggest that some, perhaps many of the spellings found in these dictionaries are of modern forms with no demonstrated traditionally written equivalents and must have been devised by their authors - particularly by Jäschke - or by their Tibetan informants and advisors, often influenced by modern

The language history of Tibetan

53

dialect pronunciation. Thus Jäschke has many examples of Ladakhi dialect words, such as "lding.se. ... L[a]d[akh] quite, very, very much ..." and the game is given away by entries such as "Ida. Ida ... L[a]d[akh] ftq. for kla ..., gla ..., zla ..." (1881: 288-289). This can really only mean that alleged reflexes of items spelt in written Tibetan with initial kl-, gl-, or zl- (and variously pronounced in other dialects) are pronounced beginning with something like Id- in Ladakhi. Thus a local, modern pronunciation is determining the "traditional Tibetan orthography". This procedure of Jäscnke's, which he in no way tries to misrepresent as a reconstruction of "Old Tibetan" or the like, is presumably just a continuation of what native Tibetan lexicographers and writers had been doing for many centuries. The dictionaries of Jäschke and Das have in their turn influenced subsequent lexicography by native speakers. An interesting example of spelling coinage is the Tibetan word for 'English', ~?wdti in Lhasa Tibetan, a loanword so modern as to be in neither Jäschke nor Das. This is spelt dbyin.ji., which would in theory correspond to some pronunciation such as *dbjindfi in "Old Tibetan". It is unlikely, to say the least, that the Tibetan coiner of this written form ever heard such a pronunciation, yet he chose this complex spelling in preference to the simpler "inji. which would have been equally valid. Curiously, the Amdo Khake dialect pronunciation of the first syllable of the word is given as YJ9n- (yj- being the usual Amdo reflex of dby-, though not of "-). This pronunciation must be derived from a reading or spelling-style pronunciation of the written form, incidentally falsely appearing to legitimise the spelling as that of a genuinely "old" syllable. No doubt the coiner had good reasons - though presumably not phonological ones - for his choice of spelling; but it makes one wonder how many other such "artificial" spellings coined over the centuries lurk undetected in the lexicon. It might be argued against this, that regardless of the idiosyncrasies of the spellings of individual words, the spelling system as a whole fairly represents the total range of sound combinations in "Old Tibetan". In reply one may point to the total of some two hundred distinctive syllable initials made available by the writing system of recent centuries, and indeed exemplified in the dictionaries. If one adds combinations with subjoined -w- and obsolete spellings found in Preclassical texts, the total must increase to well over two hundred and fifty. This is at least double the number found in the most initial-rich of the authentic and properly-studied Tibetan dialects (Amdo Khake with about 117). One is tempted to suggest that the total number of initials has been extended to this large figure in order to accom-

54

Philip Denwood

modate ranges of sounds from many different dialects as well as different historical periods, just as, for instance, the International Phonetic Alphabet comprises far more symbols than are required to write any one known language or dialect. (On the other hand, though the total number of initials in written Tibetan is large by the standards of most languages, it pales into insignificance beside Caucasian languages such as Georgian which has well over seven hundred and fifty, so that typologically there is no case for saying that two hundred and fifty is excessive for a natural language.) 7.2.3. Preclassical Tibetan The third problem with the script-based approach is that the oldest dateable stratum of the language, which I have called "Preclassical" (some of it certainly from the 8th-10th centuries if not earlier) often has a distinctly archaic appearance with, among other things, spellings now considered as against the rules. Although it should surely be a starting point for any research into the origins of the language, it is hardly represented in these dictionaries at all, as its texts were not known to their compilers. Those of its spellings which were judged by later Tibetan authorities to be "incorrect" or "impossible" were "tidied up" from about the tenth century onwards. These same authorities may sometimes have invented new items, for example in order to fill in "gaps" in verb paradigms or to regularise items with alternative spellings. It was this tidied-up language which, along with later dialect items, formed the raw material for the foreign lexicographers, including J schke and Das, of modern times, rather than the language which had regularly been written down in the 7th and 8th centuries. I conclude that the canonisation of spellings within "the traditional Tibetan orthography" is the work of many centuries: work which is still continuing as new dialect words and loanwords emerge to demand written spellings, for example in modern literary Tibetan, modern written Dzongkha and written Ladakhi. The "traditional Tibetan orthography" itself is in part a product of this work, and new spelling conventions continue to be proposed in order to meet its current demands. On the one hand the orthography has more than double the number of spellings needed to represent the most "archaic" known dialect (and ten times the number needed for the least archaic), probably because it represents a conflation of many different dialects and historical strata rather than a single έίαί de langue. On the other hand it has discarded many spelling conventions dating from the earliest strata. Much of this effort of adapting

The language history of Tibetan

55

spelling conventions and applying them to actual words did not even start until after the Preclassical period, by which time, if Miller is correct, at least one dialect in Central Tibet had already virtually assumed its modern form. Admittedly the spelling conventions themselves remained pretty stable between say the late thirteenth and the early twentieth centuries, but even so hundreds of words may have first received their conventional spellings during that period, following heaven knows what principles.

1.2.4. Bias towards lexis and phonology The fourth problem is that reliance on dictionaries focuses the attention primarily on vocabulary and phonology, largely ignoring both grammar and actual usage as encountered in texts. Also left out of the question are words in modern dialects for which no independent written equivalent is known, and conversely, for that matter, written words not known from any modern dialect; not to mention words occurring in texts which have not found their way into the dictionaries. All this is neither to dismiss the value of written Tibetan as an aid to language history, nor to belittle the efforts of scholars who have used it so far. Rather does it mean that written material should be used with circumspection, and preferably with a background of some acquaintance with the texts of the various periods of Tibetan literary history. By far the most sophisticated as well as the most comprehensive example of the script-based approach is that of A. Rona-Tas in his remarkable work Tibeto-Mongolica, ostensibly a study of the Tibetan loanwords in the Monguor dialect of Mongol but which incorporates a large body of wellorganised material from (mainly) the "archaic" Tibetan dialects. Rona-Tas is well aware of the methodological problems and pitfalls to which I have drawn attention above; nevertheless he stops short of making his dialect material the effective starting-point of his analysis, writing (1966: 106): "Dealing with a great variety of dialects and with a relatively long period of contacts between Monguor and Tibetan, the most suitable procedure seemed to be to choose a historical point of departure. Therefore I arranged the material according to the structure of the Old Tibetan Language ..." As is apparent from the rest of this passage, already quoted above, "Old Tibetan" is a somewhat broadly defined entity, whose hypothetical forms however, as pointed out by Sprigg (1967: 217), "differ little from the orthographical forms of literary Tibetan."

56

Philip Denwood

2. Dialect-based approaches R. A. Miller in his article on Ladakhi (1956) attempts to reconstruct by "the comparative method", from modem Ladakhi, Balti and Purik data, a "proto-Western Tibetan", which is then compared with "Central Tibetan" and Lhasa Tibetan (all these being modern spoken dialects), before finally bringing in written Tibetan form. The results, he claims, provide "considerable evidence on the theoretical difficulties inherent in any methodology which assumes that the orthography of Written Tibetan may safely be taken as identical with proto-Tibetan." Chang and Shefts Chang (1968: 119-120) firmly base their procedure on internal reconstruction using modem dialect forms, and take even Miller to task for statements of his in a separate work (Miller 1966: 253) where he appears to favour a script-based approach: Neither ... would we recommend ... Miller's curious approach to taking written Tibetan forms as basic ... Internal reconstruction has made a valid claim to recapture historical processes. As a means of gaining historical depth in the absence of written records, and as a means of confirming inferences from writing, this is of unquestioned value. To reverse this procedure, and take historical, written evidence as a clue or [Miller's] "handy guide" to what is readily available synchronically is a strange notion, indeed.

An equally bold attempt to break free of the stranglehold exerted by the written forms of the language on the reconstruction of earlier stages of Tibetan has been made in an article by R. K. Sprigg, who, like Miller, adopts the principle of "... not accepting Tibetan orthographic forms without first testing them against reconstructions based on comparing contemporary [i.e. present-day] spoken forms ..." (1972: 557). Since he confines himself in this article mainly to the same phenomena as most other scholars - syllable-initial consonant clusters, with the associated question of tone Sprigg's approach is, like Miller's and Chang and Shefts Chang's, a thought-provoking alternative to the more common procedure in which written forms play a much more central part. Significant phonetic features found in syllable-initial clusters of cognate lexical items in different dialects (mainly, in this case, Balti, Golok, Sherpa and Lhasa Tibetan) are identified, and the items are classified, as exhaustively as the material will permit, according to various combinations of phonetic features. Each lexical item in a particular dialect can then be allotted a phonological formula (a kind of "phonological spelling"), corresponding to equivalent formulae

The language history of Tibetan

57

for the cognate in the other dialects, which represent reflexes of a "ProtoTibetan" formula. (For the rationale of these formulae, see Sprigg 1963.) Only at the final stage may the resulting Proto-Tibetan formulae be compared with the written reflexes of the items, though the latter may sometimes have been called upon at earlier stages in the analysis to throw light on doubtful or ambiguous cases. The exhaustive and systematic character of this procedure contrasts with the more haphazard and often rather arbitrary style of many of the practitioners of the script-based approach. It enables Sprigg to offer trees of reconstructed forms corresponding to the two syllable-initial spellings zl- and si- which would not only account for the different modern dialect forms, but would also suggest that some of the branches of the dialect tree leading to the modern dialects had already separated before the writing system was devised. Such a suggestion, though already hinted at by Rona-Tas (1966: 143,186), would generally be ruled out ex hypothesl for many exponents of the script-based approach by their implicit assumption that the script came first, the modern dialects later. A valuable by-product of the more sophisticated types of analysis such as those of Sprigg and Miller is the light they may throw on population movements of the speakers of the developing dialects in the early stages of the history of the Tibetan-speaking world. It should be noted, however, that the proponents of dialect-based approaches, like those of script-based approaches, begin with the assumption that all recent forms of Tibetan descend from a single ancestor (Sprigg's "proto-Tibetan"). 3. Lexis-based approaches All the examples of both script- and script- and dialect-based approaches discussed above are heavily oriented towards, or even exclusively concerned, directly or indirectly, with phonology, which of course constitutes only one facet of language along with lexis and grammar, among others. It will be apparent from the above discussion that several criticisms of the standard dictionaries from the point of view of phonological reconstruction apply also to lexical history. While some attempt is made by both Jäschke and Das to distinguish between literary and dialect items, and in the latter case even to distinguish between dialects, this goes hardly any way towards a lexical history of the language. Desiderata for such a history are: 1. a dictionary of Preclassical

58

Philip Denwood

Tibetan; 2. a dictionary of Classical literary Tibetan, and 3. a comparative dictionary of modern dialects. The first - a dictionary of the Preclassical language - is largely available in the vocabularies of Li and Coblin (1987), Richardson (1985), Thomas (1951-1954) and others, though this material needs to be amalgamated, collated and checked. The material for the second, a dictionary of Classical Tibetan, is substantially present in existing dictionaries but needs to be separated from modern dialect items and much more rigorously classified by period and genre. The third, a comparative dictionary of modem dialects, is happily now under way in the Bern project. The sort of results to be expected are hinted at by Bielmeier (1998: 604) in pointing to some "exclusive [lexical] correspondences between Balti and Amdo, the marginal areas of the Tibetan-speaking area."

4. Grammar-based approaches Comparative studies of Preclassical, Classical and dialect grammar are even less advanced than those of lexis. There would seem to be considerable potential here, bearing in mind that the modern dialects, for all their great variability, seem to make in their different ways similar distinctions in the realms of for example tense/aspect, epistemic and judgemental modality, ergativity, and what I have termed "viewpoint". Problems here include an uneven coverage of the different language varieties, and a wide range of methodological stances by different authors, ranging from attempted application of Latin-, Sanskrit- or English-based categories in some earlier studies to various Tibeto-centric approaches more appropriate to modern linguistics. More fundamental perhaps is the sometimes not appreciated difference in genre and/or register applicable to different varieties. Thus the Preclassical documents available to us include annalistic and historical narrative, official charters of a legal nature, some Buddhist literature and some epistolary material and folktales, but little that could truly be called vernacular to set alongside material from the modern dialects. Classical Tibetan texts are mostly of a highly literary nature, sometimes translated from Sanskrit or other languages. The modern languages - Modern Literary Tibetan, the Dzongkha of Bhutan, and written Ladakhi - while related to particular spoken dialects are nonetheless essentially literary idioms. Even much of the modern "spoken" material available is derived by the simple expedient

The language history of Tibetan

59

of asking informants to relate folktales, of which Tibetan-speakers from all areas seem to have an inexhaustible supply. While of great interest in itself the language of such stories frequently shows significant differences from the everyday colloquial, may be somewhat archaic, and at the very least may prove to be an unrepresentative sample of the latter. Compare for example the fairly limited variety of grammatical categories exhibited in the Balti material of Bielmeier (1985), derived mainly from folk literature, with the much wider range in Koshal's grammar of Ladakhi (a variety not too far removed from Balti), which is based partly on the same sort of material, but mainly on everyday conversational language (1979). 4.1. Evidentials One thing that has emerged from the growing body of modern dialect studies is the existence of a complex system of interaction in the verbal phrase between evidential modality; a "intentional/ unintentional" or "controllable/ uncontrollable" dichotomy related to the lexical verb; and a distinction of "person" or, as I and some others prefer to regard it, "viewpoint" or "self-" versus "other-centeredness". From the point of view of language history, this system seems to point in opposite directions. On one hand, the commonality across dialects in the basic distinctions being made and in the operation of the whole system seem to argue for its descent from the putative proto-Tibetan language. On the other hand, there seems to be not much cognacy between the actual lexical items used in the different dialects; also rather little trace of the system in Classical Tibetan, and even less in the Preclassical language. An example discussed by Bielmeier (1998: 600602) will serve to illustrate both of these rather contradictory tendencies. The verb particle or auxiliary pronounced -suk in Balti, -isuk in Ladakhi does seem to have a cognate in several varieties of Amdo Tibetan: Themchen -zap, Dzorge -Z9g, (Bielmeier 1998; Sun 1993 passim; see also Roerich 1958: 46; Kaisang Norbu et al. 2000: 255-257). Moreover it seems to occupy the same position in the verb phrase system in both groups of dialects: it is used typically (though not only) when referring to specific past actions with second and third person subjects which were unwitnessed by the speaker, but for which there is evidence of an indirect kind. Although both Sun and Bielmeier confess to being unable to trace it etymologically, I would identify it with written Tibetan zug., past form of 'dzugs., one of whose meanings is "to put down, to place ... (more genteel than bzhag.pa.)" (Jäschke 1881: 465). This would strengthen one of the possi-

60

Philip Denwood

ble etymologies for the Lhasa Tibetan verb particle _(?a which performs precisely the same function in the verb phrase in that dialect (Denwood 1999: 159; notwithstanding that I there spell it shag., following some modern Tibetan authors.)· The idea may be that the subject has "put down" or "left behind" evidence of the action. In -suk etc. we have an item which as Bielmeier observes, seems to reinforce the impression of phonological commonality between western (Balti) and eastern (Amdo) dialects, with no cognate in central and southern dialects. If I am right in my etymology, -suk (like its Lhasa equivalent) would be a variant of an etymon which is well attested in both Classical and Preclassical Tibetan. However it must be admitted that looking over such evidential particles as a whole over a range of dialects, in agreement with Sun's findings for Dzorge Tibetan vis-ä-vis the Lhasa dialect, "... one is struck by the almost total lack of cognacy of the evidential forms despite the considerable conceptual commensurability between the two systems" (Sun 1993: 994). Furthermore, while both zug. and bzhag. are frequent in both Classical and Preclassical Tibetan, in neither variety do they perform any evidential function. In a previous work I wrote, "... in neither the Preclassical nor the Classical texts is there any evidence of a formal distinction of viewpoint. Nor is there any evidence of a distinction between intentional and unintentional verbs and clauses" (Denwood 1999: 249). To this I would add that I can find little or no trace of the evidential system in these two varieties of the language. If the evidential and related systems are a post-classical development, as might be suggested from the "almost total lack of cognacy of the evidential forms", then one would surely have to postulate a high degree of interaction between dialects from Amdo and Kham to Baltistan to account for the "considerable conceptual commensurability" between them (see below). If on the other hand the systems were present from very early times it is plausible that they might not have been expressed in the non-colloquial Preclassical and Classical languages. Then one would still have to account for the lack of cognacy between dialects. This might not be difficult considering the fact that in other languages (such as English) the use of auxiliaries in the tense/aspect system, and indeed the lexical verbs with which they are cognate, has been extremely variable between dialects, although there is the same degree of "conceptual commensurability" between the latter that we find in Tibetan.

The language history of Tibetan

61

4.2. Future stems The future verb stem as recorded in the dictionaries is in active employment in Preclassical Tibetan, but seems to fall out of use very rapidly in the Classical language and surprisingly is not certainly present in any modern dialect known to me (Kaisang Norbu et al. 2000 list a few separate future stems for Amdo Tibetan but it is not clear whether they regard them as in colloquial use.). Establishing the precise chronology of the future stem's decline in literary varieties and a further search for it in modern dialects would form interesting studies.

5. Geography and history 5.1. Inter-dialect influences The diverse varieties of Tibetan have of course not existed in isolation from one another. Influence of one dialect on another has sometimes been proposed, for example by Roerich (1933: 1-2): The dialects of Central Tibet still exercise a powerful influence on the dialects of Outer Tibet, and each Lahuli hillman, who has visited Central Tibet or made a stay there, represents a channel through which this influence penetrates the local language. Lahuli lamas ... make frequent trips to Bhutan and Tibet proper, and often spend many years in ... some of the great centres of learning in the country. The result is that their speech becomes strongly tinted by that of Central Tibet, and exercises a considerable influence on the everyday speech of the countrymen.

Similar comments are made about the Chone dialect of Amdo (Ekvall 1939) and about Rgyalthang by Hongladarom (1996: 86). Presumably such inter-dialectal influence has been at work since the earliest times, along with interaction between spoken and written varieties since the invention of the script. If Miller is right that by the early 9th century Lhasa Tibetan had already become phonetically almost indistinguishable from its present-day descendant, then perhaps influence from it could have encouraged loss of initial consonant clusters in neighbouring, southwestern and south-eastern areas which were in more regular contact with Lhasa than were the north-western and north-eastern ones.

62

Philip Denwood

5.2. Folk migrations Another factor sometimes overlooked is the possibility of movements of whole populations from one part of the Tibetan-speaking world to another. These are usually not well documented - an exception being the migration of the ancestors of the Sherpas of Solu-Khumbu in Nepal and adjacent regions of Tibet from two districts in Kham in eastern Tibet. There they would have presumably spoken local Kham dialects, and might also have had contact with speakers of non-Tibetan languages. Their ancestors left these areas around 1500 and were settled for some decades in Central Tibet before moving into allegedly virgin territory in the Himalayas where they still live. Whether their present dialect retains any influence from those of their ancestral homelands remains to be investigated. Bielmeier (1998: 605) identifies a number of phonological, lexical and grammatical correspondences between Balti and other western dialects on the one hand, and Themchen and other Amdo dialects on the other. He relates these to the hypothesis of differential change from the "original" language between marginal and central areas, as advanced by M. G. Bartoli (echoing the statement of H. A. Jäschke quoted near the beginning of this paper). There is obviously considerable merit in this idea. However the possibility of continuing direct contacts and folk-movements between the two extremities of the Tibetan language area world should also be considered as an additional factor. The view of a geographical split of dialects as one between eastern and western margins on the one hand and central on the other may have been influenced by a "Lhasa-centric" view of Tibetan civilisation, as well as by the present-day inhospitable nature of the barely inhabited Changthang which lies between the western and eastern dialect areas. The principal "archaic" dialects which can be geographically located with reasonable accuracy are, at the western end, most varieties of Balti and Ladakhi; and at the eastern end, such dialects as Themchen, Huari, Amdo Khake, Tangut, Panaka, Dzorge, Rebkong and some varieties of Golok in Amdo, and Nangchen in Kham. A simple glance at the map serves to show that both the western and the eastern of these dialects in fact lie to the north of about the 33rd parallel. A north versus south dichotomy between them and the bulk of the "modern" dialects is perhaps more valid than an east/west versus central/south one. Just as the most archaic dialects are to the north, so many of the most innovative - at least in phonology - such as Dingri and southern Kham dialects, are to the south: Such a classification suggests the possibility of a continuum of, and/or communication between, speakers of

The language history of Tibetan

63

"archaic" dialects stretching across the whole northern half of the Tibetan plateau. Such few comparative studies of the western and eastern, or better, north-western and north-eastern dialects as have been made would not invalidate the idea of their mutual contact both before and after their purported split from the central/southern dialects. Examination of the si- and zl- clusters by Sprigg (1972) mentioned above suggests that representatives of north-western and north-eastern dialects may preserve a linguistic stage which is actually earlier than the first surviving examples of the Tibetan script, and that the north-western and north-eastern groups may have diverged from one another much later. A known example of an eastern set of dialects which may have connections with the west is that forming the so-called "Wa-ke" (wa.skad.) group. These are associated with nomadic communities known as "Washii" (wa.shul. 'remnants of the Wa', sometimes spelt dba'.shul.). Numbered traditionally at 18, these have at various times in the last millennium been said to have spread to inter-riverine upland tracts (sgang.) in various parts of Amdo and Kham from the Kokonor in the north to as far south as Derge and Lithang. They may be the same as groups known as mLnyag., a name also known as a toponym in several regions between western Sichuan and the old Xixia kingdom east of the Kokonor. No details of anything that can be certainly identified as a Wa-ke dialect have so far been published. However it is possible that the Dzorge dialect is linked to this group, since Dzorge is named as one of the 18 Washii tribes in the dPag.bsam.ljon.bzang. (Stein 1961: fn. 131). Also some material from a Golok dialect published by Sprigg may be relevant. Both can certainly be regarded as "archaic" dialects, like Ladakhi and Balti rich in syllable-initial consonant clusters and lacking lexical tone. The Washü have been associated with several literary and oral traditions, some of which link them with some western part of the Tibetanspeaking world. Tang Chinese sources refer to groups of "slaves" whom they call Wenmo (of which the "mo" should correspond to a Tibetan spelling such as dba',/dba's./dbal., pronounced wa or we in many modern dialects). These people are said to be remnants of the disbanded army of the rebel Tibetan General blon.kong.bzher. (known in Tibetan sources as dba's.kong.bzher.legs.steng.), defeated by the Chinese in 866 A.D. (PelHot 1961: 140). The introduction to the Narthang Kanjur contains a rather similar story of the general dba's.rgod.ldong.btsan., originally stationed on the "road to the west", whose troops nomadised in the Changthang of

64

Philip Denwood

northern Tibet following a campaign against the Bhata Hor (Uighurs of Ganzhou). One community claiming affinity with this wider grouping, the "Washü Sethar" (wa.shul.gser.thar.) now practises nomadic animal herding in the Ganzi Autonomous Prefecture of northwest Sichuan. According to their oral traditions they moved to their present location from Aba (rnga.ba.) just to the east. Before that they had been part of the Golok (mgo.log./ngo.log.) nomadic confederacy to the north around the Bayankara range and in the valley of the Yellow River (Tibetan Machu rma.c/w.), until driven south by encroaching Mongol tribes during the Yuan Dynasty (13th-14th centuries) - indeed they have sometimes been known by the name of "Golok Sethar". The initial affiliation of the Washü Sethar with the Goloks is claimed to be a result of their disbandment, after the collapse of the Yarlung Dynasty in the 9th century, as troops of Tibetan armies which had been recruited from western Tibet (mnga '.ris.) in the 7th century under their leader and eponymous ancestor Wase Kyap (wa.gser.skyabs.) (Gelek 1992). It would be foolish to take either any of this oral tradition or the literary traditions referred to above at face value without further enquiry. Nevertheless there does seem to be a prima facie case for connecting the Washü groups with recruitment of troops in the far west during the Yarlung Dynasty under leaders bearing a name such as wa., dba. or dba's.; for their employment against Uighurs and/or Chinese or rival Tibetans in the northeast; and for their subsequent disbandment and dispersal southwards through Amdo and into Kham. The evidence comes from apparently independent sources: oral, literary, Tibetan and Chinese. We know from inscriptional and documentary evidence that the Yarlung Dynasty's military machine was active in Ladakh and through Baltistan and beyond as well as in northeast Tibet, and that it was in the habit of moving large armies thousands of kilometres from their point of recruitment. 5.3. Archaeology and climate Recent archaeological surveys have confirmed a long-held view that the Tibetan-speaking area, along with many neighbouring regions, has been suffering from progressive climatic desiccation in historical times. It is now clear that in many parts of the northern half of the area, from Ladakh through Ngari and the Changthang to western Amdo, this desiccation has resulted in depopulation for lack of irrigation water, and in deforestation

The language history of Tibetan

65

and desertification. Counterparts of the abandoned settlements, temples and fortifications noted by Tucci in western Tibet have now been identified in many areas across the Changthang: areas which are now virtual deserts but which still show plentiful ruins of agricultural settlements, fortifications and presumed religious monuments. Dating these remains is extremely difficult - they have been termed "pre-Buddhist" in a recent survey by Bellezza, who however acknowledges that this need not mean pre-7th century (Bellezza 2001: 8-10, 16). In the case of, for example, Guge in western Tibet, many centres were thriving in the 10th-12th centuries and were not abandoned till after the 17th. It would seem likely that the agricultural, military and religious activity to which these archaeological remains testify, together presumably with pastoral nomadism, exploitation of minerals and trade, formed the basis of the kingdom of Shangshung and, no doubt, others, extending over most of the northern half of the Tibetan plateau. Against this background the need of the Tibetans to conquer Shangshung and enlist its substantial forces as allies in their onward march towards Amdo in the 7th century makes good sense. At some time after the 12th-13th centuries, however, progressive desiccation led to large-scale abandonment of agricultural settlements and even pastoral nomadism from most areas, transforming them into largely uninhabitable deserts, occupied only by a few bands of nomads and the odd religious hermit. In the words of R. A. Stein: Migrations of tribes or families are possible. The Changthang, a great high northern plateau where nomads roam, must have played a greater role than one might think. We have seen the Sumpa transferred from this plateau to the Chinese frontier, the two "women's countries" at one end and the other of this connecting link ... The identity of certain toponyms and the attribution of a Bonpo land of Shangshung ... to the Sino-Tibetan borderlands are perhaps not merely the results of a confusion (Stein 1961: 54; translation mine). All our analyses prove that the Tibetan tradition is right to attribute a kind of preponderance to the tribes of the Sino-Tibetan borderlands. They certainly played a great role before the creation of the central kingdom. Migrations of peoples or families from the northeast of Tibet contributed greatly to the formation of the ethnic and cultural complex of what is understood as Tibet in historic times. But our analyses show also that a movement in the opposite direction, from west to east, should not be ignored. The northern

66

Philip Denwood plains (byang.thang.) served as a thoroughfare (Stein 1961: 85; translation mine).

There are also political connections between Ladakh and the north-eastern kingdom of Tsongkha in the 10th-12th centuries which may owe something to an earlier ethnic, cultural and perhaps dialectal continuum between the two areas, the exploration of which would take us too far afield for this short article. 6. Conclusion While a century and a quarter of studies have served to confirm the basic validity of Jäschke's bird's eye view of Tibetan dialects, I hope this paper has shown that the situation is not as simple as he suggested. In particular, the identity and nature of "the language preserved for us by the orthography" seems ever more puzzling. Did it correspond to any one spoken variety, and if so, which? Does it lie as it were in dialectal prehistory, or do some modern dialects "pre-date" it in some respects? This and many other questions will keep researchers busy for many years to come. References Bellezza, John Vincent 2001 Antiquities of Northern Tibet: Pre-Buddhist Archaeological Discoveries on the High Plateau (Findings of the Changthang Circuit Expedition, 1999). Delhi: Adroit Publishers. Benedict, Paul K. 1972 Sino-Tibetan: A Conspectus. Cambridge: Cambridge University Press. Bielmeier, Roland 1985 Das Märchen vom Prinzen Cobzan. Eine tibetische Erzählung aus Baltistan. Text, Übersetzung, sprachliche Analyse und Glossar. (Beiträge zur tibetischen Erzählforschung, hrg. von Dieter Schuh, 6.) Sankt Augustin: VGH Wissenschaftsverlag. 1998 Balti Tibetan in its historical linguistic context. In Karakorum - Hindukush - Himalaya: Dynamics of Change, Irmtraud Stellrecht (ed.), 2: 583-610. (Culture Area Karakorum Scientific Studies 4/2.) Köln: Rüdiger Koppe Verlag.

The language history of Tibetan

67

Bonnerjea, Biren 1936 Phonology of some Tibeto-Burman dialects of the Himalayan region. T'oungPao 32: 301-360. 1937 Morphology of some Tibeto-Burman dialects of the Himalayan region. T'oung Pao 33: 238-258. Chang, Kun, and Betty Shefts Chang 1968 Vowel harmony in spoken Lhasa Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 40: 53-124. Das, Sarat Chandra 1902 A Tibetan-English Dictionary with Sanskrit Synonyms. Alipore, Calcutta: West Bengal Government Press. Denwood, Philip 1999 Tibetan. Amsterdam: John Benjamins Publishing Company. Dragunov, Aleksandr A. 1936 Voiced plosives and affricates in ancient Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 7: 165-174. Durr, Jacques A. 1950 Morphologie du verbe tibotain. Heidelberg: Winter. Ekvall, Robert B. 1939 Cultural Relations on the Kansu-Tibetan Border. Chicago: University of Chicago Press. Francke, August Hermann 1979 Reprint. Ladakhi & Tibetan Grammar. Delhi: Seema Publications. Original edition, Sketch of Ladakhi grammar, Journal of the Asiatic Society of Bengal 70 (1, extra no. 2), 1901. Gelek 1992 The Wasu-Sethar: A nomadic community of Eastern Tibet. Text of a talk given at the School of Oriental & African Studies, University of London, 1993. Hongladarom, Krisadawan 1996 Rgyalthang Tibetan of Yunnan: A preliminary report. Linguistics of the Tibeto-Burman Area 19 (2): 69-92. Jäschke, Heinrich August 1881 A Tibetan-English Dictionary: With Special Reference to the Prevailing Dialects: To Which is Added an English-Tibetan Vocabulary. London: Routledge & Kegan Paul. Kalsang Norbu, Karl A. Peet, dPal Idan bKra shis, and Kevin Stuart 2000 Modern Oral Amdo Tibetan: A Language Primer. (Studies in Linguistics and Semiotics 5.) Lewiston: The Edwin Meilen Press. Koshal, Sanyukta 1979 Ladakhi Grammar. Delhi: Motilal Banarsidass. Li, Fang Kuei, and W. South Coblin 1987 A Study of the Old Tibetan Inscriptions. Taipei: Institute of History & Philology, Academia Sinica.

68

Philip Denwood

Li, Fang-kuei 1933 Certain phonetic influences of the Tibetan prefixes upon the root initials. Bulletin of the Institute of History and Philology, Academia Sinica 6 (2): 135-157. Migot, Andris 1956 Recherches sur les dialectes tibe"tains de Si-k'ang (Province de Khams). Bulletin de VEcole Francaise d'Extrlme-Orient 48: 49-55, 417-562. Miller, Roy Andrew 1955 Notes on the Lhasa dialect of the early ninth century. Oriens 8: 284291. 1956 Segmental diachronic phonology of a Ladakh (Tibetan) dialect. Zeitschrift der Deutschen Morgenländischen Gesellschaft 106: 345362. 1966 Early evidence for vowel harmony in Tibetan. Language 42: 252277. Pelliot, Paul 1961 Histoire ancienne du Tibet. Paris: Librairie d'Ame*rique et d'Orient. Richardson, Hugh Edward 1985 A Corpus of Early Tibetan Inscriptions. London: Royal Asiatic Society, de Roerich, Georges N. 1933 Tibetica I. Dialects of Tibet. The Dialect of Lahul. New York/ Naggar: Urusvati Himalayan Research Institute of Roerich Museum. 1958 Le Parier de l'Amdo. Etude d'un dialecte archaique du Tibet. (Serie Orientale Roma 18.) Rome: Istituto Italiano per il Medio ed Estremo Oriente. Rona-Tas, Andras 1966 Tibeto-Mongolica. The Tibetan Loanwords of Monguor and the Development of the Archaic Tibetan Dialects. (Indo-Iranian Monographs VII.) The Hague: Mouton. 1985 Wiener Vorlesungen zur Sprach- und Kulturgeschichte Tibets. (Wiener Studien zur Tibetologie und Buddhismuskunde 13.) Vienna: Arbeitskreis für tibetische und buddhistische Studien, Universität Wien. SedläSek, Kamil 1964 On some Tibetan j-, d- initial clusters and their metathetical forms in Sino-Tibetan lexemes. Orbis 13 (2): 556-567. Shafer, Robert 1974 Introduction to Sino-Tibetan. Wiesbaden: Harrassowitz. Sprigg, Richard Keith 1963 Prosodic analysis and phonological formulae in Tibeto-Burman linguistic comparison. In Linguistic Comparison in South East Asia and

The language history of Tibetan

69

the Pacific, Harry L. Shorto (ed.), 79-108. London: School of Oriental & African Studies. 1967 Review of Rona-Tas (1966). Bulletin of the School of Oriental & African Studies 30: 216-217. 1972 A poly systemic approach in Proto-Tibetan reconstruction to tone and syllable-initial consonant clusters. Bulletin of the School of Oriental & African Studies 35: 546-587. Stein, Rolf Alfred 1961 Les tribus anciennes des marches sino-tibe'taines. Paris: Presses universitaires de France. Sun, Jackson T.-S. 1986 Aspects of the Phonology of Amdo Tibetan: Ndzorge Sterne Xrra Dialect. (Monumenta Serica 16.) Tokyo: Institute for the Study of Languages & Cultures of Asia & Africa. 1993 Evidential in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63: 945-1001. Thomas, Frederick William 1951-1954 Tibetan Literary Texts and Documents Concerning Chinese Turkestan. London: Luzac for the Royal Asiatic Society. Wen, Yu 1946 Phonetic changes in the superadded and prefixed letters in eastern Tibetan dialects. Studia Serica 5: 1-22.

Dzala and Dakpa form a coherent subgroup within East Bodish, and some related thoughts George van Driem

The problem with too many field notes is that there is never enough time to analyse and systematically present all the accumulated linguistic data. Earlier this month in northeastern Bhutan, during the collection of blood samples for the Bhutanese Genome Project, I was confronted with a sorely neglected portion of my field notes. My purpose here is to present some evidence for a statement written down in my field notes in the early 1990s during the First Linguistic Survey of Bhutan. The claim is that Dzala and Dakpa appear to form a coherent subgroup within East Bodish. The intimate proximity of the relationship is manifestly evident in the form of the large amount of shared or nearly identical core vocabulary and is corroborated by the intuitions of native speakers. East Bodish is part of Bodish, a well-defined branch within the Tibeto-Burman language family. The precise contours of Bodic, a presumably larger and still hypothetical superordinate node within the family encompassing at least Bodish, Tamangic and West Himalayish, remain indeterminate. Something about Dzala Dzala is the language of Trashi'yangtse district in the highlands along the upper course of the K'^S' Kholongchu. Dzala is generally known in eastern Bhutan as

4

* loan from the Tukche form Phonetic pitch of the 4 tones in the dialects quoted (in Chao Yuen-ren's [1933] notation) language: village: /'/ I2I I3I I4I

Thakali Marpha 43 45 33/22 51

Syang 43 45 11 33/22

Tukche 54 44/33 11 121

Tamang Risiangku 54 44 33/22 211

Gurung Ghachok 33 54 11 12

Manang Ngawal 33 45 54 31

178

Martine Mazaudon

lv o

αo

lΏ α

i

ω

β

Ά** *β S7

2£

(Μ

Ο

A low glide in Marphali

179

180

Martine Mazaudon

o

•a cd

i is I»

'«i

o

•a•s

!

A low glide in Marphali

181

g

.1 sυ •Β

ΙΒ a,

l "«

j

Ι

182

Martine Mazaudon

O υ

•SΛ

o

l "n

A

t

S

A low glide in Marphali

183

(A 01

Is Ioo

I y> z, o> w and a > gh. The Tamang-Gurung-Thakali-Manang group (or TGTM for short), is a sister branch to the Tibetan branch in the Bodish section of the Bodic division of Sino-Tibetan (Shafer 1955). Marphali, the speech of the village of Marpha, like the dialect of Syang mentioned below, can be tentatively placed in the Thakali sub-group. The TGTM languages are spoken in NorthCentral Nepal. They are closely related, but there is no intercomprehension.

186

13. 14. 15. 16. 17. 18.

19.

Martine Mazaudon For a presentation of the group and a tentative reconstruction see Mazaudon (1978). The spectrograms were realized with Snoori. Thanks to Martine Toda for her help using this program. Obviously "close" when used by Glover next to "low" has to mean "having a certain amount of stricture". It could not have the meaning "high". au is found in a single word: 3kaub 'frog'. Except in some loans from Nepali, which keep a different phonological system. ya is found only once, in the name of Marpha village in Marphali, [3mwa], which is dissyllabic in the neighbouring Syang village [moßag]. Inside the "peripheral narrow approximant" class, a ranking may also exist, as it does e.g. in Tamang (note 3). That ranking is overridden by the left to right rule in Gurung and Marphali, because these languages have developed a strong open or smooth syllable phonotactic constraint (no finals at all in Gurung, only nasals and liquids in Marphali). Data from the author's field notes (1972) for Marpha, Syang and Gopang Thakali, and for Risiangku Tamang. Tukche Thakali is from Hari (1971), Gurung from Glover (1969) and Glover, Glover, and Gurung (1977), Manang from Hoshi (1984).

References Bessell, Nicola J. 1992 Towards a phonetic and phonological typology of post-velar articulation. Ph. D. diss., University of British Columbia. Bowen, J. Donald, and Robert Stockwell 1955 The phonemic interpretation of semivowels in Spanish. Language 31 (2): 236-240. Catford, John C. 1977 Fundamental Problems in Phonetics. Bloomington/London: Indiana University Press. Chao, Yuen-ren 1933 A system of tone letters. Le Mattre Phonttique 30: 24-27. Clements, George N. 1993 Lieu d'articulation des consonnes et des voyelles: une theOrie unified. In Architecture des Representations Phonologiques, Bernard Laks, and Annie Rialland (eds.), 101-145. (Sciences du Langage.) Paris: CNRSEditions. Geiger, Wilhelm 1916 Pali, Literatur und Sprache. (Grundriss der Indo-Arischen Philologie und Altertumskunde 1/7.) Strassburg: Karl J. Triibner.

A low glide in Marphali

187

Georg, Stefan 1996 Marphatan Thakali. Munich/Newcastle: Lincom Europa. Glover, Warren W. 1969 Gurung Phonemic Summary. (Tibeto-Burman Phonemic Summaries 1.) Kathmandu: The Summer Institute of Linguistics Nepal (mimeographed manuscript). Glover, Warren W., Jessie R. Glover, and Deu Bahadur Gurung 1977 Gurung-Nepali-English Dictionary. (Pacific Linguistics C 51.) Canberra: The Australian National University. Hankamer, Jorge, and Judith Aissen 1974 The sonority hierarchy. In Papers from the Parasession on Natural Phonology, Anthony Br ck, Robert A. Fox, and Michael W. La Galy (eds.), 131-145. Chicago: Chicago Linguistic Society. Hari, Maria 1971 A Vocabulary of the Thakali Language. Kathmandu: The Summer Institute of Linguistics Nepal (mimeographed manuscript). Hoshi, Michiyo 1984 A Prakaa vocabulary - a dialect of the Manang language. In Anthropological and Linguistic Studies of the Gandaki Area in Nepal, Musahi Tachikawa, Kiyotomo Mikame, Michiyo Hoshi, and Yasuhiko Nagano (eds.), 133-202. (Monumenta Serindica 12.) Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Hyman, Larry (ed.) 1979 Aghem Grammatical Structure. (Southern California Occasional Papers in Linguistics 7.) Los Angeles: University of Southern California. International Phonetic Association 1949 The Principles of the International Phonetic Association. London: University College. 1999 Handbook of the International Phonetic Association. A Guide to the Use of the International Phonetic Alphabet. Cambridge: Cambridge University Press. Ladefoged, Peter, and Ian Maddieson 1996 The Sounds of the World's Languages. Oxford: Blackwell. Mazaudon, Martine 1972 Field notes on Marpha and Syang Thakali (unpublished). 1973 Phonologic tamang (Νέραΐ). (Langues et civilisations tradition orale 4.) Paris: Soci6t£ d'otudes linguistiques et anthropologiques de France. 1978 Consonantal mutation and tonal split in the Tamang sub-family of Tibeto-Burman. Kailash 6 (3): 157-179. Kathmandu. Payne, David L. 1981 The Phonology and Morphology of Axininca Campa. Arlington, Texas: The Summer Institute of Linguistics, The University of Texas at Arlington.

188

Martine Mazaudon

Prunet, Jean-Frangois 1990 The origin and interpretation of French loans in Carrier. International Journal of American Linguistics 56 (4): 484-502. 1996 Guttural vowels. In Essays on Gurage Language and Culture, to Honour Wolf Leslau, Grover Hudson (ed.), 175-203. Wiesbaden: Harrassowitz. Pulleyblank, Edwin G. 1984 Middle Chinese: A Study in Historical Phonology. Vancouver: University of British Columbia Press. 1994 The Old Chinese origin of type A and type B syllables. Journal of Chinese Linguistics 22 (1): 73-100. 1998 Pharyngeal glides and zero initials in Chinese. In Studia Linguistica Serica, Benjamin K. Tsou (ed.), 1-26. Hong-Kong: City University of Hong Kong. Pullum, Geoffrey K., and William A. Ladusaw 1996 Reprint. Phonetic Symbol Guide. Original edition, Chicago/London: University of Chicago Press, 1986. Reynolds, Elinor, Paula West, and John Coleman 2000 Proto-Indo-European "laryngeals" were vocalic. Diachronica 17 (2): 351-387. Shafer, Robert 1955 Classification of the Sino-Tibetan languages. Word 11(1): 94-111. Vennemann, Theo 1988 Preference Laws for Syllable Structure and the Explanation of Sound Change: With Special Reference to German, Germanic, Italian, and Latin. Berlin/New York: Mouton de Gruyter. Yip, Moira 1983 Some problems of syllable structure in Axininca Campa. Proceedings of the North Eastern Linguistic Society 13: 243-251.

Pronominally marked noun determiners in Limbu Boyd Michailovsky

1. Introduction As in other Kiranti languages, nouns in Limbu are typically modified either by demonstratives, by pronominal prefixes, by other nouns (bearing the genitive suffix, but with a pronominal prefix on the head noun), or by propositions, participles, place-names, time-adverbs, etc., marked by the Limbu "nominalizing" suffix -pa. In a few cases, however, Limbu shows a different construction, in which the modifier bears a pronominal prefix, the 3d singular "possessive" prefix ku-. This type of determination is reflected in a common type of color-term derivative, e.g. kubhwtla 'pale'. The same prefix is found in a number of other words which typically function as modifiers or as predicate adjectives, like kusog 'new'. In compounds, a noun may take the place of the pronominal ku-, e.g. mikphuAla 'European' ('eye-pale'). Similar phenomena are reported from other East Kiranti languages.1

2. Modifier-noun constructions I begin with a review of nominal modifiers in Limbu. Nominals may carry the definite suffix -en or the plural suffix -ha*. 2.1. Demonstrative: DEM.-N (no marking) (1)

kha nese-'n (untimely209)2 that earring-DEF 'that earring'

190

Boyd Michailovsky

2.2. Genitive: (N-GEN)-PRON-N (head- and dependent-marking) The "possessed" head is marked by a pronominal prefix belonging to the paradigm a- '1st person', ke- '2d person', ku- '3d person'. The genitive suffix on the dependent noun is -τεί-ΐε. Limbu independent pronouns do not take the genitive marker. (2)

ke-nrfya-re ku-him-mu (elopeSl) 2-matemal.aunt-GEN 3-house-LOC 'in your maternal aunt's house'

(3)

yeba-ΊΙε ku-dhokk-etmu (untimely!38) shaman-DEF.GEN 3-body-LOC 'into the body of the shaman'

(4)

kha bakss-elle ku-som-dhaq (untimely 271) that box-DEF.GEN 3-top-LOC On (top of) that box'

2.3. Relative clause, place-noun, time-adverb, etc.: X-pa Ν (dependent marking) The dependent-marker here is the "nominalizer" -pa, which in some contexts has a marked feminine -ma (cf. pa 'father', ma 'mother'). The function of -pa as a morphological element - the use which concerns us here is to mark certain elements as nominal modifiers. Like other modifiers, these may serve as nominals in the absence of a head (see below).3 A time-adverb as modifier: (5)

asen-ba before-NOM Old-timers'

khodakmi-re (dance!25) elder-ERG

A subject-relativization using the active participle: (6)

kha ke-si-ba mona-n (untimely270) that ACT-die-NOM man-DBF 'the dead man'

Pronominally marked noun determ iners in Limbu

191

English adjectives are most often translated by Limbu "stative" verbs (whose aspectual character is more complex than their name suggests). Morphologically, these have an unprefixed participial form in place of the active participle of (6): (7)

yom-ba sara Aik (\radcl43) big-NOM wild.boar one 'a big wild boar*

Any modifier, in the absence of an expressed head, can serve as a nominal; in other words, Limbu is a language in which 'big' may mean 'big one'. (8)

a-mba a-mma-re-tmu (marriage39) 1-father 1-mother-GEN-LOC 'to/at my parents' (place)'

(9)

kha ke-si-V-εη (untimely 123) that ACT-die-NOM-DEF 'the dead man'

3. Color-words Color-words form a closed class in Limbu, with particular formal properties (cf. van Driem's "bound colour morphs", 1987: 23-25). The following are found in the Mewa Khola dialect: (10)

Αεί 'red' hi:k 'yellow', 'green' (Tembe, Phedap) let 'raw, unripe, green, light "grue"' mak 'black, dark' om 'yellow' p/u/7 'blue, dark "grue'" pho 'white' rf' gray, pale'

Color- words appear in a number of constructions and characteristic derivations. They occur as predicates with the copulas h'ma 'to be like, to sound like, to make like . . . ' and cokma 'to do, to behave':

192

Boyd Michailovsky

(11) mak h9 pekk-elle bela (elope43) black COP go-SUB time 'when it becomes dark' (12) maikki let lo9 maize green COP 'the maize is green (unripe)' These predicates can serve as modifiers, using the regular relative construction: (13) om ke-lo9-ba yellow ACT-COP-NOM 'yellow flower'

flower

There are a number of derived color-forms, some of them marked by the nominalizing suffix -pa mentioned above. The morpheme -tag 'like' is used to form color adjectives of the form [COLOR]-tagba, e.g. maktagba 'black', phodarjba 'white'. Special suffixes -koqgorjba, -kokkokpa and -yz9yo9ba are used to derive color adjectives expressing degrees of *'[COLOR]ish", e.g. makkoqgoqba 'dark', phiqgokkokpa 'bluish', makyo9yo9ba 'blackish' (A). But the most common and the most characteristic color-form has the prefix ku- and the otherwise unknown suffix -raf-la, for example, kuhetla 'red', kubhora 'white': (14) kubhora phug-ha9(A) white flower-PL 'white flowers' (15) kule-tla sa'-n 'raw meat' (WS268) (16) kuhikla a-phis 'military ("green") office' (D24) (17) kuhetla liq 'reddish thatch-grass' The prefix ku- is clearly the 3d person clitic pronoun, although here it does not commute with the other clitic pronouns (a- '1st person', ke- '2d

Pronominally marked noun determiners in Limbu

193

person'). That is, if I am black, I am kumakla and not *a-makla. Note also that ku- marks the dependent modifier (i.e. the color) and not, as in the genitive construction, the head.4 Like other modifiers, the color-term can function as head: kuhetla (thik) 'a red one', kuhetla-'n 'the red one'. This is the construction that interests us here - or at least its lexicalized trace. Before discussing it further, I will present a few other lexical and morphological uses of ku-.5

4. Other pronominalized modifiers Three other words have the prefix ku- and the suffix -la, although they do not share other color-derivatives: kulo:pla 'cylindrical' (A) (cf. hippet '[manner of] lying stretched out'), kugopla 'round' (A) (cf. kopma 'to surround, to reach around'), kusegla 'dried' (A) (cf. segma 'to dry in the air'). A small group of adjectives has the prefix ku- and the suffix -wet (~ wa in Tembe): kuhigwet 'alive' (cf. hiqma 'to live'), kughogwet 'unripe (of fruits)', kumagwet 'uncooked', kusiwet 'dead' (cf. sima 'to die'). (18)

kusiwet sigbug 'dead tree' (A) kumagwa sa 'raw meat' (Tembe dialect)

Further examples include the word 'new', which has the prefix ku- with no suffix: (19)

kusoqhim 'new house' (cf. sogma 'to sell') kughekya sa 'dried meat' (cf. khekma 'to be dry')

Two pronominalized words express comparisons: (20) thi-n cwait kusiktagba cog-u beer-DEF water like make-3O.pa 'he made the beer like water (i.e. weak)' (21) a-mbhoga-si-re kusiktagba... eaten Aop(dance251) 1-uncle-COLL-GEN like custom not.exist 'there is no custom like that of my uncles' generation'

194

Boyd Michailovsky

(22) J9tnm9 kud^ ba poi(dance62) all same COP 'all (of our body-parts) are the same' In (22), kudorjba is derived from the verb toqma 'to agree'. In (20), kusiktaqba is related to the postposition kusik 'like, in the manner of, which generally has an adverbial sense. In both cases there is an extra argument, the comparandum, which could justify the pronominal ku-\ note, however, that this argument is not in the genitive case.

5. Modifier-compounds In compounds, these modifiers generally appear without the ku- prefix, which can thus be seen as commuting with a compounded noun. There are two types of compounds: the first type occurs with color- words of the type kuhetla 'red', in which the pronominal element ku- may originally have indexed the head of a modifying construction. In compounds, what is logically a head-noun replaces the pronoun: structurally 'meat-fresh' vs. 'itsfresh (meat)'. Note that the usual modifier-head order of Limbu syntax is reversed: (23)

saretia 'fresh meat' (cf. sa 'flesh') gaghekya 'dried fish' (cf. ija 'fish') maikkiretla 'green/unripe maize' (cf. ma±ki 'maize') thegek.hctla 'police (an Assam police force wearing red headgear)' (cf. thegek'head') mikphwtla (1) 'cataract', (2) 'European' ('eye-pale') (cf.

The second type also occurs with color- words, and with 'new'. The pronominal prefix is again lost, but the usual order of determination is preserved: 'black-earth' vs. 'its-black earth'. (The identity of the compounded element is not always clear.) (24)

maklep 'wart' makkham 'black (fertile) soil' (A) (cf. kham 'earth') makkhi 'blood' maksiqro 'charcoal' (cf. sig 'wood') phodiifrek 'Solatium khasianum' (cf. tiq'rekpa 'thorn') hetla.k'yakma 'Polygonum capitatum (?)' (has a reddish inflorescence)6

Pronominally marked noun determiners in Limbu

1 95

he:ttho:kphetia 'rhododendron' (A) (cf. ώοΑρΛεί 'id.') sonkhim 'new house' (cf. him 'house') (A) sonmi 'new man, stranger' (A) (mi is a formative suffix in a number of nouns referring to persons)

This is the order of modifier-head compounds generally in Limbu, with modifiers like cwkpa 'small' (A: cukp ) which never take ku-, e.g. cukphe '(kind of) small knife' (A) (cf. phedza 'Nepalese knife'). 6. Relational nouns and compounds The great majority of ku- entries in the Academy Dictionary (Kainla 2002) are nouns, of which only a few will be cited here. Most have a relational sense, implying a relation to a whole, a source, a material, a complement, etc. In other words, to the extent that these words only appear with the pronominal prefix, they constitute a class of obligatorily possessed nouns.7 Grammatically, the prefix marks the head of a genitive expression, not a modifier as with color-words. (25)

kubherek'half' kube 'edge' kurum 'middle' kumulla 'piece (esp. of meat)' (A) kubuq 'base' (A) kumma:ma^ 'female (of a species)' (A) kusa.p 'root' (A) kusigek 'broken grain' (A)

Many deverbal derivatives fall into this category, distinguished from the source verbs by the pronominal prefix. The prefixed form is either the present stem or the infinitive (suffix -ma). Note that the prefix ku- does not occur in the finite verbal paradigm (see appendix). (26)

kunam 'odor' namma 'to smell' (in kunam nu/phen 'smells good/bad') kudza 'taste' cama 'to eat' (in kudza nu/phen 'is good/bad tasting') kusap 'decoration' sapma 'to write' kudemma 'handle' temma 'to grasp'

196

Boyd Michailovsky kudepma 'lid' kuyaij 'storage place, place to stay' kuyui) 'handle (of a knife, etc.)' kue:p 'suppression' (A) kugeoj 'length' (A) kughe:k 'piece' (A), 'odd member of a pair'

tepma 'to cover' yaqma 'to put inside' (A) yugma 'to set, to fix' epma 'to suppress' keqma 'to be long' kheikma 'to have a nick, a missing tooth'

Relational nouns also enter into compounds. As in the case of modifiercompounds (above), a noun takes the place of the pronominal ku-. However, the semantic structure is different, because the noun that replaces the pronominal is logically a modifier and not the head: 'tree('s)-base' vs. 'itsbase'. (27)

kuduppa 'large intestine' (A) kumulla 'piece (esp. of meat)' (A)

kuburf 'base' (A) kusa:p 'root' (A) kusigek 'broken grain' (A) kuyurf 'handle'

hidwt, hidwppa 'id.' (hi 'excrement') semulla 'piece of meat' (A) (? sa 'flesh') phaksamulla 'a piece of pork' (phaksa 'pork') siqbuq 'tree trunk' (A) (sig 'tree, wood') siqsa.p 'tree-root' (A) ya^sigek, yaasige£ 'broken grain' (A) (ya7 'grain, rice') pheymj 'knife-handle' (A) (phedza 'Nepalese knife')

When the relational noun is itself the modifier, it comes first in the compound, e.g. in piujsammaij 'chief among hunting deities' (A) (cf. sammaq 'deity'), lumso:kma 'middle finger' (A) (cf. soikma 'to point'). Where it is a quantifier, it can apparently come either before or after what is quantified, like NP quantifiers: phrexjdok ('half-pause') 'semicolon' (A), toqbhreiq ('year-half) 'semester' (A). 7. East Kiranti parallels The phenomena described above find an echo in Ebert's description of the closely-related language Athpare (1997: 93-94) (Comments in square brackets and footnotes are mine.):

Pronominally marked noun determiners in Limbu

197

Some adjectives [marked by the (singular) nominalizer -na] alternate with forms that have a prefix lu-l [u-~ o-, identical with the 3d person possessive prefix] and suffixes -pa ~ -ba or -wat. makna I umakpa piqna I upiqba phana I uphaba usikna I usikwat

black green red dry

The following adjectives are attested only with lu-l, although they have the ending -na (rather than -pa or -waf) or no suffix at all. ucaphenna

ugly, bad

ocetnuna ohop

beautiful, nice empty

umaq

raw

[cf. Limbu kudza phen ('its-eating bad')] [cf. nu- 'to become good'] [cf. Limbu hopma 'not to be, not to exist'] [cf. Limbu kumagwet'uncooked']

Ebert remarks (1997: 93, fn. 44): "The origin of the prefix lu-l is obscure. It could be a frozen 3rd person possessive prefix, or the form u- -ba/ -wa could be an old U[ndergoer]/S[ubject] participle. Cf. also o-mok-pa o-yaq-wat o-choq-wat u-pha-ba ~ u-pha-wa

broth broken new money

rice

(< mok-ma 'boil') (< yarj-ma 'beat') (< chog-ma 'sell')8 (< ?)9"

Finally, from Bickel's sketch of another East Kiranti language, Belhare (2003: 563): "One adjective, uchoüat 'new', is attributed without further marking (uchoüat khim or khim uchoüat 'a/the new house') and appears to go back to a possessive construction meaning 'its-newness'. Such constructions are found with a few other concepts: e.g. qatlabu u-sorj [banana 3POSS-ripe] 'the/a ripe banana'". The post-nominal position of the adjective is not necessarily significant, since Belhare is described as head-final with a "post-phrase adjunct position" (Bickel 2003: 562). But if pronominally-marked adjectives are more frequent in this position than other attributes, it might be taken to reflect the headedness of the hypothesized 'house its-newness'.

198

Boyd Michailovsky

8. Burmese Bernot (1971) describes Burmese as a language without a specifically adjectival category, in which demonstratives, possessors, and relatives precede the head. Modifiers which would normally be translated by French or English adjectives are generally drawn from the category of "stative verbs", which can serve as predicates, be followed by modal particles, etc. (I omit consideration of Pali loans, used in the literary language.) These usually follow the noun.10 A few of them - essentially limited to color-words in the spoken language, although exceptional examples with 'new' and 'warm' are also cited - are marked by the prefix 'a-. Thus: lu ^ciTji] 'adult ("man big")' (Bernot 1971: 45) 'ani 'the red flower ("flower PFX-red")' (Bernot 1971: 46)

The prefix in question also serves regularly to form deverbal nouns, as well as being found on a certain number of other nouns. Finally, it is lost in compounds, unless it is in first position. Thus, as in Limbu, we have a nominalizing prefix which marks colorterms in their function as noun-modifiers. This is an interesting typological convergence, but perhaps no more than that. The Burmese prefix is not pronominal, at least not synchronically. The prefixed color-term, like other stative-verb modifiers in Burmese, usually follows the noun. 9. Conclusion Pronominally-marked modifiers constitute a small, closed class in Limbu, consisting essentially of color-terms and the word 'new' - among the usual suspects in closed classes of adjectives (Dixon 1994) - as well as 'living', 'dead', and a few others. They are characterized by the presence of the pronominal prefix ku-, The structure of the modifier, 'its-(quality)', reflects grammaticalization of the notion that the quality is "possessed" by the noun. Syntactically, however, it seems clear that in an expression like kubhora phiuj 'white flower', 'flower' is the head, since head-final order is general in Limbu syntax. The pronominal morphology is frozen and unproductive. The identification of the prefix as the 3d person possessive pronoun is reinforced by the existence of compounds in which the pronoun commutes with a noun. A pronominal form like kuretla 'green, unripe' can be seen as

Pronominally marked noun determiners in Limbu

199

representing any of the possible compounds saretla 'meat + fresh', ma:kkiretJa 'maize + green/unripe', etc. Even in compounds where the internal order corresponds to the usual syntactic modifier-head order (e.g. phodufrek 'white-thorn [Solarium khasianum]'), the prefix is almost always dropped (cf. note 6). We might consider the prefixed adjectives like kuretla as simply grammaticalized pronominalizations derived from saretla, etc., but we would still have to explain the anomalous order of the latter, which falls out naturally from the 'its-(quality)' grammaticalization path. Compounds with the more usual stative-verb-type modifiers - 'big' and 'little', for example have only modifier-head order. Pronominally marked modifiers are reported from two other East Kiranti languages, Athpare and Belhare. They have not been reported elsewhere in the Kiranti group.

Appendix: Syntactic use of ku- with verbs The prefix ku- does not occur as a prefix in the finite verbal conjugation, but it occurs with non-finite verb forms in two constructions. First, it appears with a reduplicated gerund, forming a kind of prepositional nominalization or headless relative with an extentive sense. The original referent of the pronominal is perhaps an implied head (i.e. the amount in [28], the duration in [29], the speech in [30]). (28)

ku-ghambekk-εΐΐε ku-dhon-dhon A?f(elope77) 3-land-DEF.INST 3-support-support only '(a fine amounting to) only as much as his land can bear'

(29)

khemsikpa ku-yuq-yuq-samma njf(tradel55) star 3-stay-stay-until only Only as long as the stars remain (visible)'

(30)

ku-baip-paip pa.p~ma (A) 3-say-say say-INF 'to say whatever comes into one's head'

This gerund is found in a few compounds, e.g. namdhadha 'all day long' ('sunset-set'), namdaittait 'all night long' ('sun-fetch-fetch'). Second, ku- appears with the present stem and the suffix -se in the purpose construction, van Driem's "supine" (1987: 212), where it indexes definite 3d person personal objects (only), commuting with a-' 1st person' and ke- '2d person'.

200

Boyd Michailovsky

(31)

a-himdagm'-εΐΐε ku-la-se peg-agg-ag (untimely 13 3) 1-wife-DEF.GEN 3-fetch-PURP go-lsg.pa-CONJ 'when I went to fetch my wife'

(32)

a-hoq-se pher-ε (road!46) 1-look.for-PURP come-pa 'he came to look for me'

(33)

igga khol-le ku-hog-sε fy-agg-εΐΐε (elope67) 1 DEM-GEN 3-look.for-PURP arrive-Isg.pa-SUB 'when I arrived to look for him'

The purpose construction makes an interesting distinction between definite human and other objects. With indefinite or non-personal objects, neither the pronoun nor genitive marking is used: (34)

hatti tem-se me-dhag-ba (tradelSl) elephant catch-PURP 3pl-go.up-NOM 'where they go up to catch elephants'

(35)

lig hek-se pc Jtor3(untimely217) thatch.grass cut-PURP go.PA if 'if she goes to cut thatch'

(36)

menctiya hog-se pek-muna woman look.for-PURP go-PART '(marriage-brokers) going to look for a young woman'

Notes 1.

2.

Unless otherwise noted, examples are from the author's fieldwork on the Mewa Khola dialect of Taplejung District, Nepal. A few are marked as from the dialect of Tembe, in the Maiwa Khola. Other sources are identified as follows: (A) Academy Dictionary (Kainla 2002) (predominantly Panchthar dialect); (WS) Weidert and Subba (1985) (Panchthar dialect); (D) van Driem (1987) (Phedap dialect). The transcription is IPA-based, except thaty is used for IPA [j] and j for IPA [dzj. Voicing of occlusives is conditioned and not phonological. In the author's transcriptions, vowel length is transcribed only in closed syllables and never on the vowels e, o. Examples from recorded texts are cited with references to the text and the sentence number. The texts (transcription, translation, and recorded sound) are available for browsing on the Lacito archive website: http://lacito.vjf. cnrs.fr/archivage.

Pronominally marked noun determiners in Limbu 3.

4.

5.

6.

7.

8. 9. 10.

201

This element also appears as a formative in a large number of nouns, not necessarily derived, designating persons or animals, e.g. keba 'tiger', thokpa 'weevil', yeba '(kind of) shaman', (feminine: yema), and in some others, e.g. nedeqba 'cheek', including probable loans from Tibetan, e.g. thakpa 'rope', tonba '(kind of) drinking vessel'. The Academy Dictionary cites two examples of such forms with the nominalizer -pa, kumaklaba 'dark-complexioned man' and kumaklama 'darkcomplexioned woman'. It is not clear whether these can occur as modifiers. The precise status of the unaf fixed color morphemes is somewhat ambivalent. Van Driem speaks of "bound colour morphs". This interpretation finds support in the form kuretla 'unripe', because, in the Mewa Khola at least, only dependent elements (suffixes, focus particles, verb stems in certain verbal expressions) show initial in postvocalic contexts (cf. van Driem 1990). Compare the independent noun lay 'leg', kulaq 'his leg'. (This rule is apparently not observed in the standard Panchthar dialect, cf. kule'tla in ex. [15] and in A, and kulaq ~ kurag, cited in A.) Syntactically, however, the color-words appear to be independent in exx. (11) and (12). Note that the suffix -ra is retained in this example, and in hetiaphwj 'poinsettia' ('red-flower') (A). The Academy Dictionary also lists two compounds in which ku- is retained: kubhiglug 'sapphire' ('blue-stone') (A), kuhe'luq 'ruby' ('red-stone') (A) (cf. lug 'stone'). It is not clear to me that this is in fact the case. In my dictionary (2002) I have listed some of them as headwords without the prefix, and A lists most of them both with and without it. It is certain that both kin and body-part terms may appear unprefixed: lagg-elle tor-usige (father-in-law 15) foot-INST attack-lpl.ex-*3 'We two attacked it with (our) feet.' I have applied Ebert's etymology for 'new' to Limbu kusog in (19) above. Athpare 'money' is simply 'red', describing the copper coins that traditionally circulated in Nepal. Limbu money is ke-sem-ba 'cold'. Bernot argues that they are the heads of their NPs, and that the contradiction with the usual Burmese head-final order is only apparent. This question is not discussed here.

References Bernot, Denise 1971 L'e"pithete en birman: contribution ä l'itude des langues sans cate"gorie adjectivale. La Linguistique 7 (1): 41-53. Bickel, Balthasar 2003 Belhare. In The Sino-Tibetan Languages, Graham Thurgood, and Randy J. LaPolla (eds.), 546-570. London/New York: Routledge.

202

Boyd Michailovsky

Dixon, Robert M. W. 1994 Adjectives. In The Encyclopedia of Language and Linguistics, Ronald Eaton Asher (ed.), l: 29-34. Oxford: Pergamon Press, van Driem, George 1987 Limbu. Berlin: Mouton de Gruyter. (D) 1990 The rise and fall of the phoneme M in eastern Kiranti: Sound change in Tibeto-Burman. Bulletin of the School of Oriental and African Studies 53 (1): 83-86. Ebert, Karen 1997 Grammar ofAthpare. München: Lincom Europa. Kainla, Bairagi (chief ed.) 2002 Limbu-Nepali-English Dictionary. Kathmandu: Royal Nepal Academy. (A) Michailovsky, Boyd 2002 Limbu-English Dictionary of the Mewa Khola Dialect. Kathmandu: Mandala Book Point. Weidert, Alfons, and Bikram Subba 1985 Concise Limbu Grammar and Dictionary. Amsterdam: Lobster. (WS)

About Chaurasia1 Jean Robert Opgenort

The Kiranti languages are members of the Tibeto-Burman (or Sino-Tibetan) language family. The ethnolinguistic term "Kiranti" applies to the Limbu and Rai groups of eastern Nepal. The Limbu are the easternmost group. The ethnonym "Rai" is used to denote different groups speaking closely related languages, i.e. Ätfipahariyä, Bahing, Bantawa, Chiling, Chamling, Chintang, Dumi, Dungmali, Hayu, Jero, Khaling, Kohi, Kulung, Lohorung, Mewahang, Nachiring, Puma, Sam, Sampang, Sun war, Thulung, Tilung, Wambule, Yakkha and Yamphu. Hanßon (1991) classifies Wambule ("Umbule") and Jero ("Jerung") as members of the Western group of Kiranti languages. He points out that the differences in phonology and lexicon between Wambule and Jero plead in favour of a definition of the latter as a language of its own. The Wambule and Jero communities also insist on differentiating between two distinct groups, but they acknowledge the intimate relationship between their ethnicities and speeches. The present comparative study, which discusses a number of major phonological and morphological features of Wambule and Jero, aims at quantifying and qualifying the similarities and differences between the two languages. It will become apparent that Wambule and Jero have many grammatical features in common. The additional fact that the two speeches are mutually intelligible pleads in favour of the view that Wambule and Jero constitute two separate sets of dialects of a single language. A suitable name for the language combining the Wambule and Jero dialect groups may well be "Chaurasia". This name is a slight modification of Hodgson's (1857) "Chouras'ya", which is an anglicised version of the historical Nepali name CaurSsiyS or "Eighty-Four" counties. The Chaurasia language is spoken by 7,000 or more people living around the confluence of the SunkosT and DudhkosT rivers in eastern Nepal. The following dialects can be distinguished: (a) The "Wämdyäl" dialect of Wambule is spoken in the Mänebhanjyän Village Development Committee of Okhaldhungä district.

204

Jean Robert Opgenort

(b) The "Udaipure" dialect of Wambule is spoken in Okhaldhungä district in a small area along the upper course of the Rümdü Kholä. (c) The "Hilepäne" dialect of Wambule is spoken in Okhaldhungä and Udaypur districts. (d) The "JhäppälT" dialect of Wambule is spoken in Khotan district. (e) The northern dialect of Jero is spoken around the Maulun Kholä in Okhaldhungä district. (f) The southern dialect of Jero is spoken in Sindhuh district in several villages along the west bank of the Bahadur Kholä.

The data presented here reflect the southern Jero dialect spoken in the village of Mohanfar and the Wambule dialect of Hilepäm. The data were collected during several field trips which I undertook personally in the Chaurasia-speaking area from 1996 to 2000. Jero forms are generally given in the left-hand column, and Wambule forms in the middle column.

1. Phonology The most striking phonological difference between Wambule and Jero is the opposition between the Wambule implosive stops 16, dl and the Jero homorganic nasal consonants Im, n/. These correspondences have turned out to furnish important clues about the genesis of implosive stops in Wambule, many of which can be traced back to clusters of stops plus nasals. However, not every implosive stop in Wambule corresponds to a homorganic nasal in Jero, e.g.' bipcam ~ w 6ipcam 'to suck' and j daqma ~ w (fagma 'self. macam mayaqga maphlem

Bacam ßaöaqge Baphlyam

'to eat, bite' 'egg' 'wing'

me pa cap

ßepacam

'to make cold'

mokom

60

'chicken, fowl'

ni nucam nusum nobu nemki

(fi cfucam cfusum (fwabu dyamki

'name' 'to be well' 'nose' 'ear' 'spittle, saliva'

About Chaurasia

205

Note also that post-consonantal Wambule lyaJ [Jae] and IwaJ [WA] regularly correspond to Jero lei and lol.

2. Nominal morphology This section deals with former elements of noun classification and gender marking, nominal number markers, case markers, discourse markers and the group of nominals.

2.1. Unproductive nominal suffixes Several words in Jero and Wambule incorporate bound elements with a more or less clear meaning. These noun-classifying formatives represent former or non-productive nominal suffixes. Wambule and Jero also show traces of an unproductive grammatical device to mark gender distinctions. The gender distinctions are related to the sex distinction between male and female. < -cu> cacu jomcu

< -co> caco jwamco

'person' 'grandson' 'shaman'

carji garji

carja gar/a

'grain' 'Eleusine coracana' 'rice plant'

Ghku ka:ku

Glwaku ka:ku

'water' 'the Dudhkosr river' 'water'

< -si> kuksi mesi

< -si> kuksi ßyasi

' tree, wood' 'Ficus cunii' 'Elaeagnus parvifolia'

toksi

twaksi

'fruit' 'mango'

< -si ~ -ci> yaksi bremci

< -si ~ -ci> ywaksi bryamci

' small object' 'salt' 'finger'

206

Jean Robert Opgenort

nuso saiso

nusso saiso

'day after tomorrow' 'yesterday'

gumsu gupsu

gumso gupso

'flesh, meat' 'tooth' 'tiger'

< -pa ~ -pu —?u> papa hupu ka?u

papa upo kawa

masculine marker 'father' 'grandfather' 'friend'

mama humu salme

mama umo salme

feminine marker 'mother' 'grandmother' 'young woman'

'day'

2.2. Number markers The Chaurasia nominal number markers are morphologically unbound but syntactically restricted morphemes, i.e. these markers cannot be used on their own as a clause. The number markers appear after the nominal head and precede case suffixes and other nominal markers, with which they show no morphophonological interaction.





< h->



< bak- ~ ba:->

< di-> < lwa->



< bak- ~ ba:->

' to go (come back)' ' to go (away)' 'to die, be dead' 'to come down' 'to lie down' 'to cry, weep' 'to be, sit'

- Transitive verbs show agreement with two arguments, which function as the agent and the patient. The agent initiates the action. The patient is affected by the action.

< tu:t- ~ tu:->





< tiu(s)->





'to eat' 'to dig' 'to drink' 'to see' 'to kill' 'to catch, capture' 'to dig, root' 'to carry' 'to fall asleep (by some agentive outer force)'

- Middle verbs show person and number agreement with one argument, which functions as the middle subject. Self-benefactive, reflexive and passive readings can be ascribed to the meaning of middle verbs. The middle

About Chaurasia

215

subject can be identified as the entity which undergoes the effect of the action expressed by the verb.







'to sit down' reciprocal auxiliary 'to hit oneself 'to wash oneself

Wambule and Jero verbs may have one or more different roots and can be classified in various conjugations on the basis of paradigmatically conditioned root alternation. A conjugation therefore represents a fixed pattern of paradigmatically conditioned root alternation. A distinction is made between the different roots of a single verb morpheme and the different forms of a single root. Different roots are distributed according to a fixed pattern of paradigmatically conditioned root alternation. By contrast, different root forms are forms of a single root which are morphophonemically conditioned by the presence or absence of suffixes which form their environment. The different forms of a verb cannot be accounted for in terms of the modification of a single verbal base. For instance, the second person agreement morpheme is suffixed to the primary verb root in (catch) + (2) + (23s) + (AFF) -* Imak-ni0-mel '(you8) caught (it)', whereas this agreement morpheme is suffixed to the secondary verb root in (catch) + (2) + (ns) + (AFF) -» mainimme '(youp) caught (it)'. The most striking difference between Jero and Wambule transitive morphology is the absence of Jero transitive verbs that end in the post-final segment I si. Wambule transitive verbs that end in the long vowels and in the consonants generally exhibit the presence of a segment I si before the third person non-plural agent agreement morpheme , e.g. tws-u-0-me '(he) drank (it)', but not in other finite forms. By contrast, the Jero cognate form tuü-i-0-me '(he) drank (it)', in which the verb root is followed by the third person non-plural agent agreement morpheme , illustrates that the Jero cognates of these Wambule verbs commonly end in post-final . In Wambule, post-final is only found after root-final . The post-final segments I si and can be traced back to the Tibeto-Burman "causative" suffix * and the "directive" or "applied" suffix *.

216

Jean Robert Opgenort

3.2. The negative



negative prefix

The contradiction of the meaning of a verb is expressed by prefixation of the negative marker to corresponding positive verbs and deverbatives. The negative morpheme is the only prefix found in Jero and Wambule. There is no separated negative marker for the imperative. (8)

J

Uqgu Jero a-no-go. I/we Jero NEG-be-ls 'lam not a Jero.'

3.3. The simplex verb Chaurasia simplex verbs contain a single person and number agreement suffix or a fixed string of suffixes with a maximum of three different elements. The morphemes in the first functional position occur as the initial elements after the verb root, whereas those in the third functional position constitute the final elements of a suffixal string. Cognates in identical functional positions are marked by the symbol (=). The simplex person and number agreement morphemes given in table 1 are listed as their basic morphs. Table L

Simplex person and number agreement morphemes

Jero I



(3s— Is) (3npA) (2-ls) (IpiAS)

Wambule I









(3s— 2s) (ls-2) (dS) (Is) (MID) (IpeAS)







(3s— Is) (3npA) (2-ls) (InsPS) (lpi-3) (3s— 2s) (ls-2) (dS) (Is) (MID) (3-3p) (2s)

About Chaurasia II

(d) (IdeAS) (2)

ΠΙ

(ns) (23s) (3dA)

II



ΠΙ

217

(d) (IdeAS) (2p) (IpeAS) (23s) (3/ns) (InseAS)

Intransitive and middle verbs use the same set of simplex agreement suffixes, which mark the person and number of the grammatical subject. Dual forms require the presence of the dual subject morpheme < -g> (dS). From the point of view of the verbal suffixes, the formal difference between intransitive and middle conjugations involves the presence of the middle marker before the second and third person singular morpheme < -0> in middle forms. The distribution of the simplex agreement morphemes in intransitive and middle verbs is given in table 2.

Table 2. Simplex intransitive and middle morphology Jero Is 2s 3s vi 3s vm ldi/2d/3d Ide Ipi Ipe 2p 3p









Wambule Is 2s 3s vi 3s vm ldi/2d/3d Ide Ipi Ipe 2p 3p









In transitive verbs, by contrast, the simplex agreement suffixes mark the person and number of the agent or the patient, a transitive relationship between the agent and the patient, or a combination of both. Unfortunately, the information on Jero transitive verb morphology has some qualitative and quantitative shortcomings. Only forms of which the morphology is understood are listed in table 3.

218

Jean Robert Opgenort

Table 3. Simplex transitive morphology Jero ls-»>2s Is— 2d Is—2p Is— 3s Is— 3d Is— 3p ldi/2d— 3s Ide— 3s Ipi— 3s Ipe— 3s 2s— Is 2d— Is 2s— le 2s— 3s 2p— 3s 3s— Is 3s— 2s 3s— 3s 3d— 3s 3p— 3s



-ίιη>











[pe.ma] [hjiLpa] [phri.pae] [mdzjok.pa]

'lotus; place name'(PM) 'second' 'ritual dagger' (NG) 'herdsman' (PM)

[za.lak.pa]

'Wednesday'

Second, heterorganic prenasalized segments in rNga-ba provide further evidence against Sun's analysis of prenasalized labial stops as unit phonemes. In mDzod-dge, the nasal segment in historical and clusters is neutralized for place (13a), while in rNga-ba, the place of prenasalization is not predictable from host segment (13b). (13) a.

b.



lakhor.lol lakha.macl (*/mkha.~/) ['khcu.lo] [mkba.mae] (*["kha.-]) [atboq.dfv] [mtbon.po] (*[athon.~])

'wheel'(DZ) 'kidney' 'wheel'(NG) 'kidney' 'beverage' 'high'

Thus, in rNga-ba we are forced either to posit distinct unit phonemes (e.g. akh-/mkh- and "fW-), adding complexity to the phonemic inventory, or to consider these sequences clusters of independent segments. By extension, Sun's claim that labial stops do not occur as independent phonemes following nasal codas in mDzod-dge is further weakened with this crossdialectal comparison. From these alternations, we can conclude that historical labial stops occur as independent phonemes in the synchronic inventories of rNga-ba and Pad-ma, which strengthens the argument for a similar analysis in related Amdo dialects. The data presented in this section are summarized in the following table. The patterns of assimilation of labial stops shown

232

KarlA.Peet

therein suggest that modern Amdo dialects represent various stages along a continuum of diachronic change, with Pad-ma representing the most conservative dialect, and mDzod-dge and rDo-sbis the most innovative dialects. Table 1. Preservation of historical in Amdo dialects8 mDzod-dge rDo-sbis rNga-ba Pad-ma

post-coronal yes yes yes yes

post-vowel no no yes yes

word-initial no no yes yes

post-dorsal no no no yes

2.4. Theoretical arguments for synchronic Ip-l Given the phonetic presence of labial stops in diverse contexts across Amdo dialects, we are presented with two logical possibilities for the derivation of surface [pf^-]. Either phonetic [jfti-] is derived from a phonemic continuant (represented by Iw-l or Ih-f), or phonemic Ipf^-l is weakened in certain contexts to yield [w-] or [h-]. In this section, I consider each of these hypotheses in turn. 2.4.1. /w-, h-/ -* [p(h)-] If we assume that Iff1*-! does not occur in the phonemic inventory, then a rule Iw-, h-l -* [p(l>>-] is necessary to derive the phonetic forms illustrated above. I illustrate here a number of compelling arguments against this hypothesis. First, we observe in the mDzod-dge and rDo-sbis data that while historical labial stops are generally spirantized in word-initial position (14a), in certain cases they are preserved intervocalically (14b). (14) a. lha.maJ b. la.hal

[ha.ma] [a.pha]

'parents' (DB) 'father'

If I h-l is the underlying representation of the surface labial stop in (14b), this segment must undergo fortition intervocalically. But this contradicts the fact that medial intervocalic and word-final position are classic targets

Implications of labial place assimilation in Amdo Tibetan

233

of lenition cross-linguistically, and thus fortition is unlikely in these contexts (Hock 1991: 83). The minimal pair above poses an additional difficulty: if the placeless fricative Ih-l is the underlying representation for each of these forms, then there is no obvious mechanism by which to derive labial place for surface form [pb-J in (14b). Second, it has been observed that labial stops in mDzod-dge and rDosbis are commonly preserved in careful speech and reading pronunciations. If it is the case that colloquial pronunciations are represented by a continuant such as Iw-l or Ih-l (15a), then by the arguments given above, it must also be the case that a separate representation is stored for careful pronunciations (15b). (15) a. /hag/ b. lphagl

[hax] [phax]

'pig (coll.)'(DB) 'pig (reading pron.)'

Although this is conceivable given the significant discrepancy between written Tibetan and spoken dialects, a single phonemic form is certainly preferable to multiple forms for cognitive economy. The clearest way to test this hypothesis in (14) is via speakers who have been minimally influenced by the written language, a consideration which is anticipated by Janhunen and Norbu: "The contrast of the labial stop phonemes can ... easily be demonstrated by referring to the names of the letters pa (pha) vs. ba (ba & pa), well known even to illiterate individuals" (1999: 252). Given this observation, the simplest account is to collapse the two phonemic forms given in (14), thus generalizing mental representations for both literate and illiterate populations. In this case, labial /p(ti>-/ is preferable to placeless Ih-l for the reasons given above. 2.4.2. /p(h)-/ -* [w-, h-] A second logical possibility is to assume that the surface forms in question are derived from underlying labial stops. If we assume that Ιρ*ύ-Ι does occur in the phonemic inventory, then a rule Iff1*-! -*· [w-, h-] is sufficient to derive phonetic forms. Several pieces of evidence argue in favor of adopting this position. First, as captured in table 1 above, ///**-/ -» [w-, h-] alternations are predictable by phonological environment. Thus dialect-specific rules are sufficient to determine whether or not itf^-l is weakened in word-initial,

234

KarlA.Peet

post-vocalic, or post-dorsal contexts. Second, we find that in each of the dialects under consideration, labial stops occur following coronal stop codas (16a), regardless of whether they are preserved after dorsal segments in like contexts (16b). (16) a. b.

lged,pol /"d^og.wael

Old man' (DZ) 'herdsman'

Given these facts, the assumption that ///*V is the underlying form is consistent with the cross-linguistic generalization that coronals are less likely than non-coronals to trigger place assimilation or other coarticulation processes (Jun 1995, 2004; Paradis and Prunet 1991). By contrast, if fw-l is assumed to be the underlying form, we have no basis for justifying fortition after coronal stops and faithfulness after dorsal stops. Thus, we once again have reason to favor the underlying labial stop ltfti-l to a corresponding continuant Iw-, h-l. Third, more evidence for the synchronic presence of labial stops is found in rNga-ba heterorganic obstruent clusters. In the following examples, historical root labial stops that have been marginalized due to a historical process of retroflection surface after resyllabification in coda position.

(17)

[^'Ύγ.ΐχ]

'woolen robe' (NG)



[t^rp.t^y] [(1>)t§&].kbae] [ lap.t§aq]

'puppy' 'chest cavity' 'lama's quarters'

Such alternations can only be derived by assuming Ip-l in underlying forms. 2.5. Summary In conclusion, an analysis which assumes synchronic ///**-/ gives the most economical and least stipulatory account of attested labial stops in each of the four Amdo dialects under consideration. Thus, I will assume for the remainder of this paper that labial stops in the phonemic inventory correspond to historical forms. By extension, the synchronic presence of labial

Implications of labial place assimilation in Amdo Tibetan

235

stops corresponding to , , and reinforces the phonetic value of these graphs at the beginning of the classical period. 3. Place assimilation of labial stops in Amdo Tibetan Having established the synchronic status of labial stops in Amdo, I will now examine the behavior of these segments in the context of place assimilation. I use as a basis for this investigation Jun's (2004) article, which gives a phonetically-grounded account of place assimilation patterns drawn from a wide range of empirical data. In section 3.1,1 give relevant background on place assimilation and describe Jun's implicational hierarchies for this phenomenon as determined by articulatory and perceptual factors. In section 3.2, I show that patterns of place assimilation in Amdo pose a series of contradictions to Jun's hierarchies. Section 3.3 summarizes the findings of this section. 3.1. Phonetic arguments for place assimilation Place assimilation in a consonant cluster is defined as the situation in which a target segment takes on the place of articulation of a trigger segment. For example, in Catalan, a coda nasal segment takes on the place of assimilation of the following obstruent; thus, the nasal segment is the target of assimilation while the obstruent segment is the trigger (Jun 2004: 2). (18) Catalan sofn] amics so[m] pocs SO[TJ] grans

'they are friends' 'they are few' 'they are big'

In the following sections, we explore common cross-linguistic patterns of place assimilation and the phonetic factors that give rise to these patterns. 3.1.1. Typology In Jun (2004), the author conducts a survey of place assimilation in twenty genetically diverse languages.9 In doing so, Jun notes that although it is

236

KarlA.Peet

easy to derive the attested outputs of place assimilation processes in various theoretical frameworks (e.g. generative theory, autosegmental theory, feature geometry), it is more difficult to give a principled account of the crosslinguistic generalizations that emerge from such a study. Jun identifies the following implicational hierarchies describing the effects of segmental manner, place, and position on place assimilation in the languages under consideration. In these hierarchies, "X » Y" indicates that X is more likely than Υ to be a target or trigger of place assimilation, where other phonological conditions are equal. (19)

1. Manner a. Target: nasal » stop » fricative, liquid, glide b. Trigger: stop » nasal, fricative »liquid, glide 2. Place a. Target: coronal »labial» velar b. Trigger: non-coronal » coronal 3. Position a. Target: C,» C2 (for Cfa clusters)

To illustrate the implications of the above hierarchies, we take as an example hierarchy la for target manner (2004: 2-3). This hierarchy states that if fricatives, liquids or glides are targets of place assimilation in a given language, then stops and nasals will necessarily be targets as well (e.g. Japanese, [20a]). Likewise, if stops undergo place assimilation in a given language, then so will nasals, though not necessarily fricatives (e.g. Yakut, [20b]). Finally, hierarchy la predicts the existence of languages in which only nasals are targets of assimilation, while stops and fricatives, liquids, and glides are not (e.g. Malayalam, [20c]). (20) a. Japanese (nasals, stops, and fricatives, liquids, gets) /kam+dal [kan.da] /tob+da/ [ton.da] /kaw+ta/ [kaUa] b. Yakut (only nasals and stops are targets) laan+bitl [aam.mit] lat+kAJ [ak.ka] tyos+kA/ Ixos.ko] (*[%ok.~J) /ufwtal+kA/ [utfwtal.ga] (*[tag.~])

and glides are tar'chew-PAST' 'fly-PAST' 'buy-PAST' Our door' 'horse-DAT' 'room-DAT' 'teacher-DAT'

Implications of labial place assimilation in Amdo Tibetan

c. Malayalam (only nasals are targets)10 Ikamalam+kafajijiul [kamalaq.karappu] lawan+karappul [awaq.karappu] lutkar$amJ [ut.kar$am] (*[uk.~J) Isaptaml [sap.tam] (*[sat.~])

237

'Kamalam cried' 'he cried' 'progress' 'seven'

To sum up these observations, Jun states, "It has been noted in the literature that nasals are the most common target in place assimilation" (2004: 3). Note that there are no languages in Jun's survey which violate the implications of the hierarchy for target manner; thus it is predicted that there are no attested languages in which, for instance, nasals and fricatives are targets of place assimilation, but stops are not. A parallel situation holds for each of the other hierarchies stated in (19). In the following section, I discuss Jun's theoretical account of these typological generalizations.

5.7.2. Phonetics Jun supports the generalizations implicit in these hierarchies with phonetic arguments that consider both articulatory and perceptual factors. He postulates that "... speakers make more [articulatory] effort for those sounds which will produce dividends in terms of enhanced perceptibility" (2004: 8). In addition, he notes that the nature of a particular gesture may favor assimilation of a segment in one context but not in another. To illustrate the consequences of these statements, we take as an example hierarchy 2a for target place. Jun notes that coronal segments have relatively weak perceptual cues because the articulatory gestures created by the rapid movement of the tongue tip are less robust than those created in producing labial and dorsal segments, which involve slower movements of the lips and tongue body. As a result, the latter have a longer period of overlap with adjacent segments than the former, thus making them less vulnerable assimilation targets, as illustrated here (Jun 2004: 13).

238

KarlA.Peet

rapid G, gesture (coronal)

slow Gt gesture (labial, dorsal)

Figure 1. Relative span of articulatory gestures at different places of articulation

Furthermore, Jun points out that dorsal segments are characterized by an additional acoustic cue resulting from the convergence of the second and third formants in neighboring vowels, thus making dorsals more perceptually robust than their labial counterparts. The combination of these phonetic factors is consistent with the order of elements in hierarchy 2a, suggesting that such typological preferences are not mere accidents of the particular languages covered in the survey. Jun makes parallel phonetic arguments to justify the sequences of each of the other hierarchies in (19). In the following section, I show that the implicational hierarchies proposed by Jun are challenged by patterns of labial place assimilation in Amdo Tibetan.

3.2. Amdo Tibetan and place-assimilation hierarchies

Despite Jun's extensive typological and phonetic justification for his implicational hierarchies, they are contradicted in many instances by assimilation patterns of Amdo Tibetan labial stops. In this section, I will examine Jun's hierarchies in the context of Amdo data, limiting my discussion to hierarchies that are not supported by these data. 3.2.1. Target manner First, we examine patterns of Amdo labial place assimilation in the context of hierarchy la for target manner. As described in section 2, Amdo labial stops commonly undergo spirantization following dorsal stops and nasals. (21)

lmt§og.pasl

Γαζογ.γ*^] ^b o ^6 3 )^b o lb o ^C } ' ^« 5 }g c 3» )

CQ

1 a l s ^ j ^ i t j ' * »

i

V) l

0

•*»i l

*f

.a^*i

CQ

m

^

^ SL Ξ •^ -^ 3? < 33 ^

δ

1 % 1^13

-S « 'S S 6 «

έ 1

Ώ

CQ

CQ A

ill

"?* 1 B

II

0 0

32

a

J2

«-H

·*
)

XJiot tet(WT[IMP]) tit (innovated) tet (WT [IMP] )

fkol(WT [PFV] ) fkel (innovated) rkol (WT [IMP] ) ίεοχ (cf. WT [IMPFV] and [PFV] bsags) fsex (innovated) [IMP] )

ablaut: 9 -»i16 (17) a. 'torollsth.' IMPFV t§9l (WT [PFV] < dril>) PFV £//(innovated) IMP t§h9l (innovated) b. 'to pour (into container)' IMPFV

vl9x (WT < blug>)

PFV IMP

vlix (innovated) l9x (innovated)17

These ablaut patterns are summed up as follows:

Perfective stem renovation in Khalong Tibetan

331

Table 1. Perfective-stem ablaut in Khalong Tibetan Input vocalism_ a i

Ablauted vocalism e e

o u(WT«x>)

e e

u e

i i

_

It will be observed that Khalong perfective-stem ablauting consistently yields front unrounded vowels; moreover, after ablaut the input vocalisms i and e become interchanged.

4. Extended applicability of perfective-stem ablaut The ablaut patterns uncovered in the foregoing section are innovative not only in that they created novel stems out of old sources, but also in that their application exceeds the requirements of remedial morphology. As shown above, innovative ablaut served mainly to replace verb forms lost in the paradigm through stem usurpation. However, sometimes ablaut itself became accountable for stem usurpation, causing replacement of readily available perfective stems. Consider for example the stems of the verbs 'to rub' and 'to spin (yarn)', where new ablauted perfective stems took the place of the proper reflexes of the original perfective stems and :

(18) a. 'to rub' IMPFV PFV b. 'to spin (yarn)' IMPFV PFV IMP

a b

p 9r (WT [IMPFV] < 'phur>) p ir (innovated)

n h

D h

k al (WT [IMPFV] < 'khal>) k el (innovated) a b k ol (WT [IMP] < 'khol>)

n b

Although far from being completely productive, innovative verb-stem ablaut has in many cases extended its domain and created multi-stem verbs out of originally invariant ones. Examples of this type include:

332

Jackson T.-S. Sun

(19) a. 'to scatter' IMPFV PFV b. 'to pull' IMPFV PFV IMP c. 'to recount' IMPFV PFV IMP

Xtor (WT ) (innovated) a h

t en (WT < 'then>) "tbin (innovated) a b t on (innovated) ffet (innovated) ffot (innovated)

Moreover, certain Khalong verbs with alternating stems originated from nominal roots. Khalong, for instance, not only verbalized the Written Tibetan nominal roots 'lap; bosom', 'lime or mud for plastering walls', 'claw', 'helper; assistance', and 'door bolt') den (innovated) n don (innovated) a

Perfective stem renovation in Khalong Tibetan

333

As a further manifestation of its vitality, stem ablaut applies even to loanwords, one likely example being χ89Γ 'to stir-fry', a borrowing from a rGyalrong invariant verb with the same meaning: (21) 'to stir-fry' IMPFV PFV

XS9r(< rGyalrong?)18 xsir (innovated)

5. Stem-two ablaut in Showu rGyalrong In two recent studies (J. T.-S. Sun 2000a, 2000b), I proposed that rGyalrong, Lavrung, and Horpa-Shangzhai19 comprise a distinct rGyalrongic subgroup based on evidence of shared idiosyncratic verb-stem formation rules of tone/glottality flip-flop and ablaut. The relevant verbal morphology reaches its acme of richness in the Showu subdialect of Sidaba rGyalrong, where a range of phonological processes is utilized in the formation of verb stems, including a complex system of ablaut. In the Zhongre variety of Showu,20 for instance, many common verbs display stem alternations that involve different vowel grades, as in: (22) a. 'to bake in hot ashes' Stem 1 p§ Stem 2 pbu Stem 3 po b. 'to borrow' Stem 1 τηέ Stem 2 nji Stem 3 rqe Across the rGyalrongic languages that still preserve verb-stem distinctions, stem 1 is the citation and present imperfective stem, stem 2 occurs among other things in perfective and imperfect verb forms,21 whereas stem 3 is restricted to singular transitive nonpast contexts. Thus, the closest functional equivalent of the Khalong perfective stem would be stem 2 in Zhongre rGyalrong. The formation of stem 2 via ablaut in Zhongre involves extensive vocalic alternations, which can be summarized in the following table:

334

Jackson T.-S. Sun

Table 2. Stem 1 - Stem 2 ablaut in Zhongre rGyalrong Stem 1 vocalism ν, Λ; e; a Λ ο; 9 i u v, i; o

Stem 2 vocalism / Β u e o s

Upon close inspection of table 2, two possible affinities between ablaut patterns marking the Khalong perfective stems and the Zhongre second stems become apparent. First, while Khalong perfective-stem ablaut invariably produces front unrounded i or e vowel grades, the front unrounded vowel / is also the most common output in Zhongre stem 2 ablaut, with four input vowels potentially leading to the j-vocalism in stem 2, e.g.: (23) a. 'to yawn' Stem 1 χ&ιη Stem 2 χϊπι b. 'to use' Stem 1 Stem 2 c. 'to catch up with' Stem 1 lchev Stem 2 lchiv d. 'to empty out' Stem 1 c§ Stem 2 c*/ A second point of similarity has to do with the interaction of the vowels i and e such that in Zhongre, as in Khalong Tibetan, the ablauted output of /-grade is e and vice-versa: (24) a. 'to approach' Stem 1 rni Stem 2 me

Perfective stem renovation in Khalong Tibetan

335

b. 'to unload' Stem 1 v//e Stem 2 vlji Furthermore, in the formally richer Zhongre system vocalic alternation by inversion is extended also to the vowel pair u and o, for example: (25) a. 'to entrust' Stem 1 β Stem 2 f o b. 'to carry on back' Stem 1 fkor Stem 2 fkhur 6. Conclusions Khalong presents an interesting case of a Tibetan dialect on the receiving end of deep-rooted influences from its immediate linguistic neighbor Showu rGyalrong. The Showu substratum manifests itself in considerable lexical loans pertaining among other things to native plant and animal life and, above all, in Khalong verbal morphology. Contrary to the general tendency for verb-stem variation to atrophy among modem Tibetan dialects,22 Khalong favors the preservation and even propagation of distinct verb stems; in consequence, the majority of Khalong verbs, particularly transitive volitional ones, participate in stem alternation. A comparison with Written Tibetan indicates that the original perfective stems frequently evolved into modern Khalong imperfective stems. While in the other Tibetan dialects stem replacement of this kind normally led to leveling of stem alternations, Khalong uniquely developed "remedial" ablauting, which replenished the paradigm with secondary perfective stems in -/ or -e vowel grades. Striking affinities are noticeable between Khalong innovative perfective-stem ablaut and Showu rGyalrong stem-building ablaut of comparable form and function. In view of these and other morphosyntactic parallels with rGyalrong,23 a case can be made out for perfective-stem ablaut in Khalong Tibetan as an extraordinary morphological change induced by intimate contact with rGyakong. Contact as a source of morphological change in Tibetan is certainly not limited to the Khalong dialect. Verbal orientation marking in Zhongu24 and Garni25 dialects provides another striking example. Written Tibetan spatial

336

Jackson T.-S. Sun

deictic morphemes acquired the role of verbal prefixes in these dialects. Zhongu Tibetan has grammaticalized two imperative prefixes Z9- and m 9-, respectively from WT 'up' and WT 'down', probably under contact influence from the Qiang language spoken in its vicinity. Sentence (26) illustrates: (26) th9r (m9-/z9-)di downward IMP-sit:IMP 'Sit down!' Though still functionally marginal, Z9- and ma- are on their way of becoming specialized imperative markers. Their erstwhile spatial semantics eroded, they are now interchangeable and the 'downward' meaning must be supplied by the adverbial tho(r9) (WT ). In Garni Tibetan, on the other hand, grammaticalization of spatial morphemes to verbal orientation prefixes has been carried further, in that orientational prefixes are now obligatory with many verb roots and, moreover, are required on all imperative verb forms (Huang 1993a: 145-148), e.g.: (27) JtV5 *(mse13-)tso53 mouth IMF-shut 'Shut up!' Direct importation of Qiang morphological form and function into Tibetan grammar may also be the origin of the hearsay evidential suffix -ji in Ren'entang Tibetan;26 compare the following Ren'entang sentence meaning 'S/he is eating a meal.' and its equivalent in Northern Qiang: (28) a. Ren'entang Tibetan (personal research) t9ht9 khe"gu zv-ji 3S meal eat-HEARSAY b. Northern Qiang (Huang 1993b: 163) qupu stuaxa th9-ji 3S meal drink-HEARSAY The powerful influences Tibetan has exerted on surrounding languages are well appreciated (e.g. Rona-Tas 1966). Contact-induced linguistic infiltration in the opposite direction, however, has not drawn an equal amount of scholarly attention. I hope to have shown that this would indeed be a very important and promising area of research.

Perfective stem renovation in Khalong Tibetan

337

Abbreviations DAT PUT IMPFV PRO WT

dative future imperfective progressive Written Tibetan

ERG IMP PFV S

ergative imperative perfective singular

Notes *

1. 2.

3. 4. 5.

6. 7.

The data on which this paper is based were gathered in two recent field trips to Sichuan, funded in part by two National Science Council (Taiwan) grants. Thanks are due to the Taiwan Affairs Office of the Sichuan Provincial Government for their kind assistance and hospitality, and to my consultants Minzhen (toqmidqon, from Khalong Hamlet of Khalong Village in Wuyl Township), Jingasong (from Zhongre Village in Ribu Township) and Yaluo (from Mulang Village in Ribu Township) for their devoted cooperation. I also gratefully acknowledge the helpful comments and discussion provided by Roland Bielmeier, Felix Haller, Boyd Michailovsky, and Nicolas Tournadre. Written Tibetan forms will be given in Wylie's standard transliteration, enclosed in angle brackets. Showu (fomi, a Caodeng exo-ethnonym referring to this kind of speech form), which I have previously called Ribu, is a widely distributed subdialect of Sidaba (one of the three major dialects of rGyalrong) spoken at Caodeng (Shazuo and Baoyan Villages only), Ribu (except the Amdospeaking Ruogu Village) and Dawei townships in Ma'erkang County, Rong'an Township in Aba County, as well as Shili and Wuyi townships in Rangtang County. My consultant Minzhen reports that Khalong is quite similar to the speech of N nm d and Rongmuda townships to the north along the Ζέςϋ River. Such as 'to stir-fry', 'to do sth. to excess', and 'to set out'. The Showu words cited here represent the speech of Mulang ("brem) Village in Ma'erkang County, a variety rather close to those spoken in Rangtang County. Showu distinguishes two tones in word-final accented syllables: level (v) and falling (ν); phonemic accent is represented by the acute accent (v). Literally 'earth-dig'. This term is used here rather vaguely to refer to the idealized spoken Tibetan language of the seventh century reflected by the Classical Tibetan orthography.

338

Jackson T.-S. Sun

8.

Written Tibetan vowel clusters were either fused (ft§i 'charm box'). As shown in the sentence below: saji-ki param za-fajdtu ch d-ERG candy eat:IMPFV-PRG 'The child is eating candy.' This set illustrates innovative subtractive morphology which turned perfective stems into corresponding innovative imperative by removing the labial prefix . See also set (8b) below. In place of WT [IMPFV] . Cf. WT 'to do'; the Khalong verb forms may be related variants containing voiceless onsets. The verb means 'to build, to make' in WT. The origin of this verb is unknown. This ablaut pattern produces -e- instead of -i- if the input stem is closed by the codas -η and -χ, in conformity with a phonotactic constraint in Khalong. The Khalong imperative form preserves the original Old Tibetan root *lug unprefixed with the prenasal *N-. Cf. Zhuokeji A^sar, Caodeng xsar, Zhongre xsar, the word seems relatable to WT 'to twist, to spin' (Boyd Michailovsky; personal communication). However, Tibetan ordinarily uses the verb for the meaning 'to stir-fry'. Instead of the cumbersome language name Horpa-Shangzhai, I now propose the simpler term Horpa. Spoken at Zhongre (t§oqre) Village of Ribu Township in Ma'erkang County. Unlike in the other Sidaba subdialect Caodeng, Showu does not oppose the past versus non-past stems straightforwardly by glottality inversion. The differences between the stems are much more complicated, see Sun (2004) for details. In my earlier work, this stem is termed the past stem. Since it also occurs in progressive verb forms and oblique participant nominalizations, it is perhaps less confusing to label it simply as Stem 2. For a useful survey, see Qu (1985). Another suspected rGyalrong structural influence on Khalong verbal morphology is the syncretism of indirect evidential and mirative markings. While Amdo Tibetan generally maintains a formal distinction, Khalong Tibetan, like rGyalrong, expresses both epistemological meanings via a single mediative suffix; contrast: Amdo (Mdzod.dge Byams.me; personal research) α fa -Yd ''men ί^οη-ζ9γ child-ERG medicine drink:PFV-INDIRECT EVIDENTIAL '(I found out/heard that) the child drank the medicine.'

9.

10. 11. 12. 13. 14. 15. 16. 17. 18.

19. 20.

21. 22. 23.

Perfective stem renovation in Khalong Tibetan

339

jod-bk9 1S:DAT money exist-MIRATIVE 'So I have money!' Khalong

jaji-Jb'

24.

25.

26.

man

"fä-fs

chüd-ERG medicine drink:PFV-MEDIATTVE '(I found out/heard that) the child drank the medicine.' -la kormu jo-fa 1S-DAT money exist-MEDIATIVE 4 So I have money!' Zhongu (personal research) is a newly discovered Tibetan dialect spoken at Zhongu Valley in Songpan County of northern Sichuan. For a phonological and lexical synopsis, see J. T.-S. Sun (2003a). Garni is a variety of Tibetan of Muli County in Sichuan Province, where a number of Qiangic languages are also found. Data on Garni are from Huang (1993a). Ren'entang (personal research) is another distinct form of Tibetan found in the Tibetan-Qiang borderland.

References Chang Shefts, Betty and Kun Chang 1982 The persistence of present-tense reflexes in modern spoken Tibetan. Tsing Hua Hsuebao 14 (1/2): 21-30. Huang, Bufan 1993a Zangmianyu dongci de quxiang fanchou [The category of orientation in the Tibeto-Burman verb]. In Zangmianyu Xinlun [Recent Contributions to Tibeto-Burman Studies], Ma Xueliang, Hu Tan, Dai Qingxia, Huang Bufan, and Fu Ailan (eds.), 133-151. Beijing: Central Institute of Nationalities Press. 1993b Zangmianyu de qingtai fanchou [The category of modality in the Tibeto-Burman verb]. In Zangmianyu Xinlun [Recent Contributions to Tibeto-Burman Studies], Ma Xueliang et al. (eds.), 152-165. Beijing: Central Institute of Nationalities Press. Jäschke, Heinrich August 1954 Tibetan Grammar. New York: Frederick Ungar. Lin, You-Jing 2002 Phonological profile of Thewo Tibetan. Paper presented at the 8th Himalayan Languages Symposium, September 19-22, 2002, University of Bern, Switzerland.

340

Jackson T.-S. Sun

Qu, Aitang 1985

Zangyu dongci quzhexingtai de jiegou ji qi yanbian [Structure and evolution of Tibetan verb-stem inflection]. Minzu Yuwen [Nationality Languages and Writing] 1: 1-15. Rona-Tas, Andräs 1966 Tibeto-Mongolica. The Hague: Mouton. Sun, Jackson T.-S. 2000a Parallelisms in the verb morphology of Sidaba rGyalrong and Guanyinqiao in rGyalrongic. Language and Linguistics (Institute of Linguistics, Preparatory Office, Academia Sinica, Taiwan) 1 (1): 161-190. 2000b Stem alternations in Puxi verb inflection. Language and Linguistics 1 (2): 211-232. 2003a Phonological profile of Zhongu: A new Tibetan dialect of Northern Sichuan. Language and Linguistics 4 (4): 769-836. 2003b Qiuji zangyu de yuyin tezheng [Phonological characteristics of Chosrje Tibetan]. Minzu Yuwen 6: 1-6. 2004 Verb-stem variations in Showu rGyalrong. In Studies on Sino-Tibetan Languages: Papers in Honor of Professor Hwang-cherng Gong on His Seventieth Birthday, Ying-chin Lin, Fang-min Hsu, Chun-chih Lee, Jackson T.-S. Sun, Hsiu-fang Yang, and Dah-an Ho (eds.), 269-296. (Language and Linguistics Monograph Series W-4.) Taipei: Institute of Linguistics, Academia Sinica. 2006 Special linguistic features of gSerpa Tibetan. Linguistics of the Tibeto-Burman Area 29 (1): 107-125. Sun, Jackson T.-S., and You-Jing Lin 2002 Pitch, rhyme length, and glottal states in Chosrje Tibetan. Paper presented at the 8th International Symposium on Chinese Languages and Linguistics, November 8-10,2002, Academia Sinica, Taiwan.

On the deictic patterns in Kinnauri (Pangi dialect)* Yoshiharu Takahashi

\. Introduction Kinnauri is one of the Tibeto-Himalayan languages spoken in District Kinnaur in Himachal Pradesh, India (see the maps on page 352). Though it has not been clear in detail for a long time, the actual state of the Kinnauri grammatical system is now understood much better, for some books and papers about it have been published recently. Though we do not have data good enough to know the whole of its grammatical system, some interesting phenomena have been pointed out in those previous studies. The purpose of this paper is to consider the deictic patterns in the Pangi dialect of Kinnauri.1 In section 2, we will discuss the verb inflection, section 3 will be devoted to the motion verbs in Kinnauri, and we will consider the case-marking system in section 4. 2. Verb inflection Kinnauri has been known as a pronominalized language, but we will see that this is not an exact description of the verb inflection of Kinnauri. 2.1. Subject marker on the verb Table 1 gives the inflectional suffixes of subjects.2 Both intransitive and transitive verbs take the same inflectional suffixes.3 Table 1. Inflectional suffixes of subjects in Kinnauri

1 sg. 1-kJ du. pl. excl. /-£/ incl. -0

2 hon. l-fil l-öi

non-hon. 1-nJ

3 hon. /-S/

ordinary -0

342

Yoshiharu Takahashi

Examples of the inflectional suffixes for singular subjects (1) and for plural subjects (2) are given below.

(1)

a. gi kin I

b. ki

kimo

bttok

your house-Loc come.Fut.lS(sg.) will come to your house.'

an kimö

btton

you my house-Loc come.Fut.2S(sg.) 'You will come to my house.'

c. an guruji

an kimö

bttoS

my teacher my house-Loc come.Fut.3S(sg.hon.) 'My teacher will come to my house.' (2)

a. ninä kin kimö bitoc we all (excl.) your house-Loc come.Fut.lS(pl.) 'We all will come to your house.' b. kinä an kimö bitoö you all my house-Loc come.Fut.2S(pl.) 'You all will come to my house.'

Though Kinnauri has been called a pronominalized language, it does not use a pronominal form for all numbers and persons. In fact, it is only the suffixes for first-person and second-person subjects (-k and -n) that can be considered to have the same origin as the first-person and second-person pronouns of Proto-Tibeto-Burman. Rather, in Kinnauri, we should note the fact that the suffixes for the first and second persons are different from each other in the singular, and that the suffixes for the first and second persons are not different from each other in the plural.4 The verb form for the non-honorific second-person subject is the same in the singular as it is in the dual and plural.3 This is also true of both the verb forms for honorific third-person subjects and non-honorific third-person subjects (the suffix of the honorific form is -s, and the non-honorific has no suffix). In other words, the verb is marked in the case of the first or second-person subject in the singular, or the third-person respectful subject. We should notice that the suffix of the first and secondperson plural subject has the same form, that is, -d.6 This means that the suffix of a plural subject distinguishes between first and second persons and third person. That is to say, first and second persons are recognized to be classified in the same group in Kinnauri, that is, speech-act participants (SAPs, hereafter), and the third person to be excluded. Thus, we will say

On the deictic patterns in Kinnauri (Pangi dialect)

343

that SAPs are preferred to the third person in the Kinnauri verb agreement. DeLancey (1980: 47) points out that the important feature of the original TB agreement system is a dominance hierarchy in which SAPs outrank third persons.

2.2. Object marker on the transitive verb In addition to the subject agreement, a transitive verb takes an object person-agreement suffix in Kinnauri if the object is first or second person. This object suffix on a transitive verb has the same form, -c-, whether the object is first or second person, and it is attached to a verb stem before the tense and the subject person suffixes, but there is no suffix for the third-person object. (3)

a. gis nu mJpin sadak I-Erg that man-Dat kill.PUS Ί killed that man.' b. gis kin satek I-Erg youDat kill.l-2O.Pt.lS Ί killed you.'

(4)

a. tonak hit.Pt.lS Ί hit him/her/it.' b. tonoek hit.l-2O.Pt.lS Ί hit you.'

Examples (3a) and (4a) show that the third-person object has no suffix on the verb. On the other hand, in (3b) and (4b), the verbs have the suffix -cfor their objects. In both sentences, because the subject is the first person, the object should be interpreted to be the second person. In (4b), we have no specified pronouns, but the verb itself indicates who hit whom by its suffixes. If we have the suffix of the second-person subject on a verb such as tonten '(you) hit (me)', then the object of the verb is the first person.7 It is interesting that the verb for 'to eat', which usually takes a thirdperson object and no suffix on it as in (5a), assumes the object suffix in the case of a second-person object as in (5b).

344

Yoshiharu Takahashi

(5)

khau dzatok food eat.Fut.lS will eat the food.' b. gi kinü dzäiok I you.Dat eat.l-2O.Fut.lS will eat you.' a. gi I

The object marker on a transitive verb makes no distinction between first and second-person objects. This means that only the arguments within SAPs are marked. We can say that Kinnauri chooses the object suffix on a verb according to a hierarchy in which SAPs outrank the third person. Let us summarize the main points about the verb inflection in the following sentences: 1. As for the subject agreement, first person singular, second person singular, first and second person plural, and third person (no difference between singular and plural) have different suffixes, respectively. Thus, the suffixes for persons within SAPs are differentiated from each other in the singular, and the same is true in the plural. The suffix for the third person has no difference between singular and plural but always differs from first and second persons. 2. A transitive verb assumes the suffix for the first or second-person object, but there is no suffix for the third-person object. 2.3. Inclusives in verb inflection Kinnauri has inclusive forms in the verb inflection.8 The inclusive forms of a verb have not been analyzed morphologically well enough, but can be said to express (i) cohortative and (ii) reciprocal in Kinnauri; example (6a) shows that the speaker invites the hearer to go to Peo together, and in example (6b), the action expressed by the verb is performed reciprocally by the speaker and the hearer. (6)

a. kisanS piö bite weall(incL) PIN go.Pr.lS(incL) 'Let's go to Peo.' b. kiSan (kisanu komö) sase wetwo(incL) our two (incl.) inside-Loc kill.lS(incL) 'We will kill/are killing/killed each other.'

On the deictic patterns in Kinnauri (Pangi dialect)

345

We can say that example (6) shows that Kinnauri can express a reciprocal action between SAPs, that is, a reciprocal action between SAPs is expressed with a special verb form. Generally speaking, inclusives are used for SAPs.

3. Deictic pattern of movement verbs Kinnauri has motion verbs such as bimü 'to go' and binnü 'to come'. This type of verb expresses the movement of something. In Kinnauri, bimü 'to go' expresses movement out of the deictic center or outside the deictic center, and binnü 'to come' expresses movement into the deictic center or inside the deictic center. Figure 1 shows this relationship. 3rd person

1st person

bimü m

„

* 3rd person

binnü * 2nd person

Figure 1. Motion verbs in Kinnauri

Examples of the verb blrnu (7), and examples of the verb binnü (8) are given below. The first person moves out of the deictic center as shown in (7a), and the third person moves outside the deictic center as in (7b). The first or second person moves inside the deictic center as in (8a) and (8b). In contrast to (8a) and (8b), (8c) shows that the verb bimü is used if people inside the deictic center move together out of the deictic center. (7)

a. gi I

nu kimö bitok his house-Loc go.Fut.lS will go to his house.'

346

Yoshiharu Takahashi

b. nugö piö bitoS they PIN go.Fut.3S 'They will go to Recong Peo (from Pangi).' (8)

a. gi

kin

kimö

bitok

I your house-Loc come.Fut.lS 4 1 will come to your house.' b. ki

an kimö

t»ton

you my house-Loc come.Fut.2S 'You will come to my house.' c. kisan kin kimö bite wetwo(incl.) your house-Loc go.lS(incL) 'We two will go to your house.' Because the first-person subject moves inside the deictic center in example (9a), we should use bitok but not bitok, and because the first person moves out of the deictic center in (9b), we use bitok but not bltok. Example (10) indicates that the third person enters the deictic center.9 (9)

a. gi I

kin kimö bitok/* bitok your house-Loc come.Fut.lS/go.Fut.lS will come to your house.' b. gi nü kimö bitok/* bitok I his house-Loc go.Fut.lS/come.Fut.IS will go to his house.'

(10) gis rabindarü kindin bin/*him Setok I-Erg PN-Dat your.place come.Inf/go.Inf send.Fut.lS 4 1 will let Ravinder come to you.' The deictic center is the location of the speech act. In other words, the verbs blmu and binnü are related to the location where the first and second persons perform the speech act. This type of verb is not limited to the verbs describing 'to go' and 'to come' in Kinnauri. Examples (11) through (14) give pairs of verbs such as 'to give' in (11), 'to tell' in (12) and (13), and 'to bring' in (14), and they reveal the same pattern of movement. If we compare example (a) and example (b) of each pair, they show a clear contrast in the movement expressed by those verbs: Example (a) of each pair shows the movement out

On the deictic patterns in Kinnauri (Pangi dialect)

347

of the deictic center, and example (b) the movement into the deictic center. And in Kinnauri, verbs such as 'to give', 'to tell1 and 'to bring' as well as 'to go' and 'to come' have two forms which are different from each other in the root, and do not use the object suffix for expressing the direction of movement, that is, these verbs have a different root form for the person of its indirect object. (11) rannü vs. kemü 'to give' a. gtigis ju khaü nu öhanü rantok I/I-Erg this food that child-Dat give(3O).lS will give this food to that child.' b. gi/gis ju pen kinu ketok I/I-Erg this pen you.Dat give(l-2O).lS will give you this pen.' (12) lonnü vs. rinmu 'to tell' a. gi dopin ju kotbä lötok I he-Dat this story tell.Fut.lS will tell him this story.' b. gi kinu rinak gi candigad bitok I you.Dat tell(l-2O).Pt.lS I PIN go.Fut.lS told you that I would go to Chandigarh.' (13) nu kothäpin anü riain that story-Dat me.Dat tell(l-2O).Imp 'Please tell me the story.' lonnü does not take a first or second-person indirect object, and rinmu does not take a third-person indirect object.10 The next sentences are examples for the verbs p^Tmu and kannu. (14) * vs. kannu 'to bring' a. gis an goenepin me piö I-Erg my wife-Dat yesterday PIN pbeok bring(out-of-S AL) .Pt. l S brought my wife to Peo yesterday.'

348

Yoshiharu Takahashi

b. gis me an konyasü pioc an I-Erg yesterday my friend-Dat PIN my kimö kadak house-Loc bring(into-SAL).Pt.lS brought my friend home from Peo yesterday.' In fact, the difference in meaning between phimu and kannü is clear in (14). It is sometimes difficult to set up the condition that makes the difference clear; however, it can be judged from the explanation of the consultant that these verbs show the same distinction as that of bltnu and binnü. We can observe in this pattern of movement the distinction between SAPs (or within the speech-act location, SAL hereafter) and the third person, and it is obvious that this pattern is the same as the one of the inflectional suffixes of the verbs observed above. We can say that verbs of movement in Kinnauri make a distinction between SAPs (or within the SAL) and the third person by different verb roots, and other verbs distinguish an action towards speech-act participants from one towards the third person (in the case of an object) by using the object suffix on a transitive verb. As seen above, the pattern of movement can be illustrated in the way shown in Figure 1. The verbs of this type we have now are only those which have been observed in examples (7) through (14).11 1. Verbs expressing movement away from the deictic center or out of the deictic center: 67mü, rannü, lonnü, * 2. Verbs expressing movement into the deictic center or within the deictic center: binnü, kemü, rinmü, kannü 4. Case-marking pattern We will examine the case-marking system. Kinnauri has at least six case markers, and has an absolutive case, which takes no case marker, as shown in (15). (15)

Ergative-instrumental: -is, -s Dative: -pin, -u Genitive: -u, -o Locative: -o Comitative: -ran

On the deictic patterns in Kinnauri (Pangi dialect)

349

Ablative: -o Absolutive: no suffix

Some of these case markers have allomorphs. -is and -s for the ergativeinstrumental are allomo hs that are determined phonologically. -is is suffixed to a noun which ends in a consonant, and -s to a noun ending in a vowel. The dative case also has two forms. The choice of the appropriate form of the dative appears to depend on the noun to which the suffix is attached, but there seem to be phonological conditions, too. It is not yet possible to explain the distribution of the dative case marker. The genitive case markers are determined phonologically: -o is used after a noun ending in a non-high vowel, and -u elsewhere. The comitative has only one form, and it may co-occur with a noun in the genitive.12

4.1. Split-ergative pattern As pointed out in Takahashi (1999) and (2001), ergativity in Kinnauri is not an exact description, but Kinnauri has an ergative-like case-marking system. The ergative marker does not occur under some conditions, that is, Kinnauri has a split-ergative pattern. DeLancey (1980) describes the same phenomenon. It is observed that Kinnauri shows a split ergative on SAPs and tends to mark the ergative on the third person. (16) a. gtlgis gasä elk I/I-Erg clothes wash.Pt.lS washed clothes.' b. kilkis gasä ein you/you-Erg clothes wash.Pt.2S 'You washed clothes.' c. nus/*nu gasä öIS he-Erg/he(Abs) clothes wash.Pt.3S(hon.) 'He washed clothes.' The conditions for split ergativity seem to be related to person, aspect, and the active type of the verb, but here we want to pay attention only to person.13

350

Yoshiharu Takahashi

4.2. Dative forms of pronouns Finally, we will consider the dative case of pronouns. In Kinnauri, the pronouns of the first and the second person have only one form for accusative and dative cases. Example (17) shows dative pronouns as direct objects.

(17) a. nus aaO sacoa he-Erg me.Dat kill.l-2O.Fut.3S.QM 'Will he kill me?' b. gis kinü sacek I-Erg you.Dat kill.l-2O.Pt.lS killed you.' c. gis nupin särika I-Erg he-Dat kill.lS.QM 4 Will I kill him?' We need to note that the dative form of the third person is the pronoun nu suffixed by the dative case marker -pin, but as for the first and second persons, their dative forms are used as seen in (17a) and (17b). This phenomenon seems not to be related to the topic discussed above. This condition of dative markings on the direct object has not been analyzed well enough, but we can summarize this phenomenon as follows: 1. Pronouns are usually used in the dative case regardless whether it is a direct or indirect object. 2. It is observed that the first and second-person pronouns have a dative form and their dative forms can be used as the direct object of a transitive verb,14 but as for the third-person pronoun, its dative form is formed with the dative case marker. 3. It is clear that the pronouns of the first and the second person, included in SAPs, are treated differently from the third-person pronoun. What is clear is that the first and second-person pronouns as SAPs have special forms for the dative, which are used as the direct object instead of the absolutive case, and that the third-person pronoun has no special dative case form, but an absolutive form suffixed by the dative marker.15

On the deictic patterns in Kinnauri (Pangi dialect)

351

5. Final comment In this paper, we have observed the verb inflection, the deictic pattern of motion verbs, and the case-marking system in Kinnauri. We should conclude, from what has been observed above, that a hierarchy in which SAPs outrank the third person is dominant in the verb inflection, the deictic pattern of motion verbs, and the case marking in Kinnauri. DeLancey (1980: 11-12) says that the distinction between SAPs and everyone else, and the parallel, or perhaps identical, distinction between SAL and everywhere else, seems to be fundamental to deictic categories which are marked on the verb, and "this is true in Tibeto-Burman". The Kinnauri data which we have observed in this paper illustrate what DeLancey (1980) claims. I think that it is very interesting that Kinnauri has a set of motion verbs which are different in root according to the direction of motion.

Abbreviations IS 1-2O 2S 3O 3S Abl Abs Dat du. Erg excl. Put hon.

1st person subject 1 st or 2nd person object 2nd person subject 3rd person object 3rd person subject Ablative Absolutive Dative dual Ergative exclusive future honorific

Imp incl. Inf Loc pi. PIN PN Pr Pt QM SAL SAP sg.

imperative inclusive infinitive locative plural place name personal name present past question marker speech-act location speech-act participants singular

352

Yoshiharu Takahashi

Maps Himachal Pradesh

..· • 1 \ Pradesh

Notes This paper was read at the 8th Himalayan Languages Symposium in Bern, September 20, 2002. I would like to thank some of the members of the Symposium for their comments on my paper. Many thanks to Mr. Ravinder Singh Negi for kindly teaching me Kinnauri. This study is financially supported by a Grant-in-Aid for Scientific Research on Priority Area (A-2) ("Seinan-chuugoku kara himaraya chiiki no chibetto-biruma kei shousuugengo no kijutsu-kenkyuu [A descriptive study of minor languages from Southwest China through the Himalayan region]", Prof. Yasuhiko Nagano, #12039223), and a Grant-in-Aid for Scientific Research (2000-2003, "Descriptive and morphosyntactic study on Kinnauri", Yoshiharu Takahashi, #12610556), both from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

On the deictic patterns in Kinnauri (Pangi dialect) 1.

2.

3.

4.

5.

6.

7.

8. 9. 10. 11.

12. 13.

353

Data of this paper were collected mainly from Mr. Ravinder Singh Negi from Pangi, one of the villages of Kinnaur. Pangi is ten kilometers away from Recong Peo, which is the administrative center of District Kinnaur. The Pangi dialect seems to belong to the Lower Kinnauri dialect. As for the dialects of Kinnauri, more research is needed. We should mention in passing that this table does not include the full system of the Kinnauri verb inflection, such as imperative, negative and so on. Those inflectional forms will not be discussed here because they are irrelevant to the main subject. The phonemic inventory of the Pangi dialect is: /p, pb, b, t, f* d, f, f* d, δ, h , j, k, kh, g; m, π, ή, ή; is, tsh, dz; S, s, h; r, 1; w, yl. The short vowels are Ii, e, a, o, u, ί7, and the long ones, /f, e, , , , ϊ/, correspondent to the short vowels. It is not important in this discussion, but it should be mentioned that the Pangi dialect has dual and plural forms of pronouns, but no distinction between dual and plural in the verb inflection. The non-honorific second person has only one form for the inflection and its inflectional form is different from the honorific second person in the plural. This may mean that the speaker thinks that if the hearer(s) is/are not respectable, s/he or they should not be included in the speaker's part, that is, the speaker may want to think that unrespectable hearer(s) should be outside of SAPs, like the third person. In fact, this suffix can be used for the non-honorific second-person subject, in which case the degree of deference is a little bit higher than if the suffix -τι is used. It seems that -o cannot express the reflexive, that is, the sentence gis toncek means Ί hit you', but not Ί hit myself, and the sentence gis an toncek is ungrammatical. More research is needed in this area. Kinnauri also has inclusive pronouns. Cf. footnote 11. The verb rinmu is pronounced as rinm by some people in Pangi, but it is not clear why they have different pronunciations for the same word. We have some other verbs used as motion verbs: for example, jarin 'Please come' and paen 'Come with me.' Those two verbs are used in the imperative. The former is used in the situation of a person leaving and being told, 'Please, come again', and the latter can be used with bite 'Let's go'. Cf. Takahashi (2002), for a study of the Kinnauri case markers. In fact, the condition of the person is not definite. Sometimes, the consultant does not allow the absolutive case for the first and second-person subjects with a verb in the past tense. I wrote in my paper "The ergative marker can be used with the first-person subject" (Takahashi 2001: 16), but my consultant tends to answer that he usually uses the ergative marker for the first and second-person subjects in the past tense, so I do not have many examples

354

14.

15.

Yoshiharu Takahashi for the split-ergative pattern. I feel it is necessary to collect more data on ergativity in Kinnauri. In some dialects, the absolutive case forms of first and second persons may seem to be used as the direct object of a transitive verb, so we should take care when collecting data. As pointed out in the Symposium, these types of phenomena in the casemarking system are related to the hierarchy of the noun phrase as a whole. This hierarchy is typologically important, but I would like to leave this problem to another paper. In this paper, I want to pay attention to the forms of the pronouns.

References DeLancey, Scott Cameron 1980 Deictic categories in the Tibeto-Burman verb. Ph. D. diss., Indiana University. Takahashi, Yoshiharu 1999 Kinauru-go no kijutsu-teki kenkyuu [A descriptive study of Kinnauri]. In An Interdisciplinary Field Survey of Bon Culture in Tibet: A Report for a Grant-in-Aid for Scientific Research (The International Scientific Research Program, Prof. Yasuhiko Nagano, #08041040), Yasuhiko Nagano (ed.), 199-213. Osaka: National Museum of Ethnology. 2001 A descriptive study of Kinnauri (Pangi dialect): A preliminary report. In New Research on Zhangzhung and Related Himalayan Languages: Bon Studies 3, Yasuhiko Nagano, and Randy J. LaPolla (eds.), 97-119. (Senri Ethnological Reports 19.) Osaka: National Museum of Ethnology. 2002 A report on case forms in Kinnauri (Pangi dialect). In Report for a Grant-in-Aid for Scientific Research "Reconstruction of the Zhangzhung Language and Formation of Written Tibetan: A Joint Survey of Field Linguistics and Philology" (Prof. Yasuhiko Nagano, #11691050), Yasuhiko Nagano (ed.), 1-13. Osaka: National Museum of Ethnology.

Tibetan grammar and the active/stative casemarking type Ralf Vollmann

1. Case-marking and transitivity In an universal-typological approach, it is meaningful to distinguish these four classes of case systems: 1. nominative/accusative, 2. ergative/ absolutive, 3. active/stative, and 4. direct/inverse. These categories have been invented by linguists in order to be able to give a name to observable differences, but they are obviously not always clear-cut categories. Rather, they may be regarded as maximally contrasting prototypes which are more or less realized in a language. Indeed, combinations of the above-mentioned patterns can be found quite often, such as semantic ergativity (agentive/ergative), split ergativity (ergative/nominative), and also systems with participant (empathy) hierarchies and agent marking (i.e., active + direct types). The variability of phenomena in case-marking typology stems from the fact that case-marking systems have to perform more than one semiotic operation at the same time. (a) Cognitively, a clause is the output of a semiotic process: a speaker gives a profile to a situation, an event construal; the event construal involves a (usually small) number of necessary or stereotypical participants which may be identified by semantic case markers, which shall be termed participant roles. A clear-cut system of central roles in event schemes is presented hierarchically in Langacker (1991: 236): agent, instrument, mover, experiencer, patient; and an absolute participant. (b) On the other hand, at the moment a semiotic sign is created, it does not only have a relation to the external object it stands for (i.e., the signans), but also, inevitably, has a relationship to the other parts of the semiotic system itself; a linguistic sign may thus also have a linguistic entity as its signans (the system thus being self-referential). Case systems may encode more than just semantic participant roles, they can also refer to the structure of the clause instead of the structure of the event. Structural case is a marker which is more concerned with the internal structure of the

356

Ralf Vollmann

clause than with the participant structure of the event. Typically, such types of case-marking have obligatory case-marked roles and thus have to develop transformational regularities which are able to change these patterns (e.g., demoting a participant) for certain needs (i.e., passives, antipassives). (c) Due to the fundamental semiotic principle of figure-ground contrast, an event construal usually distinguishes a main participant and additional participants. Topicalization or prominence can be encoded implicitly in case markers (subject), or it can be implicit in prototypical participants (empathy), and it can of course be reassigned (by [anti]passive, word order, or case marking [irregularities). Similarly, communicative needs can be expressed by the choice of a specific case marker (pragmatic functions). The above-mentioned theoretical needs imposed on case marking can be solved in different ways, a fact which is responsible for the universal-typological variance of case-marking systems. Additionally, the in principle limitless number of different events and of possible participant roles surface in but a small number of case markers and verb classes in any language. Therefore, metaphorization of one prototypical realization of an event construal as a formal representative of many similar cases is one of the most central operations in the development (or historical drift, cf. Moosmüller and Vollmann 2001) of a grammar. This means also that a clause can represent more or less centrally the default relation expressed by that pattern. If one looks only at the formal properties, there seems to be a class of verbs involving exactly two participants, the so-called transitive verbs. This undoubtedly is a concept in many languages which surfaces in a specific case pattern (NOM/ACC, ERG/ABS). The semantic prototype of this relation is said to be an agent/patient, or source/goal relation. But we have to be aware that the notion of "two participants" is a purely formal concept, which (among others) involves sentences such as: (1) (2) (3) (4) (5)

John John John John John

breaks eats up eats beats watches

the cup. the rice. some rice. the dog. the dog.

(6) (7) (8) (9) (10)

John John John John John

sees the dog. feels a pain. knows Tibetan. visits the town. has a lot of money.

Semantically, these event construals are quite different from each other: there is a gradual decrease (downwards) of transitivity which is handled differently in the languages of the world - in English, all these relations are encoded with practically the same structural pattern. Descriptive principles

Tibetan grammar

357

such as transitivity, valency, or, most formally, government are thus but one-sided (reductionist) concepts referring only to an intrasystemic phenomenon, namely to have exactly two obligatory participants marked by a structural case pattern. There are, however, languages which distinguish various levels of transitivity instead of just two (e.g., Hungarian, cf. Reichert 1986). Transitivity is a complex concept which is formed by the evaluation of various parameters (cf. Tsunoda 1985: 392-394). There is, e.g., a different degree of affectedness of the patient between 'break' and 'beat'; there is a difference in the volitionality of the agent between 'watch' and 'see'; and there is practically no defining parameter of transitivity in the event scheme of 'feel', since the subject is an experiencer, and the object is the source of the event. In English, the semantic distinctions seem to play no role for the case system (except for subtle irregularities, restrictions on passivization, etc.). The language has strongly grammaticalized the more formal, or intra-systemic, principle of the necessary distinctiveness of case marking; casemarking systems have only few (formal) cases for many (semantic) event construals, mainly because an event construal usually involves only one, two or, more rarely, three participants at a time, so that for distinctiveness, it suffices to have two or three case markers (or syntactic principles). English, as an example, distinguishes two word-order positions and one preposition in order to encode the most frequent participant relations. There is, of course, always an open class of locatives and secondary relations (comitatives, etc.) which do not fit into the general schemes. Tibetan can distinguish various levels of agent-patient relationships, as was illustrated by Beyer (1992: 266) in his responsibility hierarchy which can express varying degrees of transitivity by the presence or absence of agents and by the choice of verb forms: MOST ASSIGNMENT OF RESPONSIBILITY (11) bla mas sgra chenpos by a spur lama:ERG noise big:INS bird frighten:PFV 'The lama frightened away the birds with a loud noise.' (12) bla mas by a phur lama:ERG bird get.frightened:PFV 'The birds got frightened away by the lama.'

358

Ralf Vollmann

(13) sgra chenpos by a spur noise big:INS bird frighten:PFV 'The loud noise frightened away the birds.' or '(Someone) frightened away the birds with a loud noise.' (14) sgra chenpos bya phur noise big:INS bird get.frightenediPFV "The birds got frightened away by a loud noise.' (15) bya spur bird frighten:PFV '(Someone/something) frightened away the birds.' (16) bya phur bird get.frightened:PFV 'The birds got frightened away.' LEAST ASSIGNMENT OF RESPONSIBILITY 2. Ergativity in Tibetan The Tibetan language can be divided into various spoken dialects and written styles, with no generally accepted norm (except for the written styles). In this contribution, Tibetan is usually synonymous with Modem Spoken Central Tibetan ("Lhasa Tibetan"). The Tibetan ergative (henceforth ERG) is a semantic case marker for volitional agents (cf. DeLancey 1984b: 134, etc.; cf. also Kesang Gyurme' 1992: 251-253 and Tournadre 1996: 279-280, 369). For several reasons, its use is difficult to determine. First, in Tibetan, there is no obligatorily of any participant in the clause (Tournadre 1996: 69), similarly to Pacific and sign languages, so that the (structural) valency of a verb is indefinable: (17) (ngas) (khongla) kha par btang ba (1:ERG) (3:HONALL) telephone LV NS '(I) have called (him) by telephone.'

yin/ CONJ

Although Tibetan ERG applies obligatorily with transitive verbs, this is true only in principle. It turns out that ERG is (more or less) "obligatory" in perfective aspect (PST), can be used in imperfective aspect (PRS), and has

Tibetan grammar

359

emphatic meaning in intentional aspect (PUT; cf., e.g., Den wood 1999: 135; Vollmann 2004: 247; for the aspectual system, see below). Obviously, ERG is not governed by the verb, but is the (semantic) marker of an agent who has acted successfully and willingly (volition). Strictly speaking, this is possible only with perfective aspect, it may be anticipated in imperfective actions, but it is practically impossible in intended (= PUT) events. (18) ngas las ka I:ERG work 'Ihave worked.'

'di by as pa yin/ DEM do:PFV NS CONJ

(19) nga(s) las ka 'di by cd I(:ERG) work DEM do 'lam working.' (20) nga las ka 'di byed I work DEM do Ί will work/ want to work.'

kyi yodl CONN CONJ

kyi yin/ CONN CONJ

Additionally, the ERG interacts with the conjunct/disjunct pattern (a system indicating speaker-hearer relations; cf. Hale 1980 for Newari, DeLancey 1992 for Tibetan) of the auxiliary system, specifically with 1st person participants (where speaker = agent; cf. Chang and Shefts Chang 1980: 17-18). With motion verbs, typically "intransitive", ERG can be used in perfective aspect, implying control over the event (volition) (cf. Chang and Shefts Chang 1980: 17): (21) ngas lha sar phyin pa yin/ 1:ERG Lhasa:ALL gone NS CONJ Ί went to Lhasa (purposely).' (22) nga lha sar phyin pa red/ 1 Lhasa: ALL gone NS DISJ Ί went (= was taken) to Lhasa.' Finally, ERG is used pragmatically, as, e.g., in the following example, indicating the willingness (volition) of the speaker:

360

Ralf Vollmann

(23) ngas togtsam nyos dgos 1:ERG a bit bought must Ί will buy a bit [from you].' (Chang and Shefts Chang 1980: 20) In short, the use of the Tibetan ERG looked rather unclear to many researchers and provoked varying descriptions dealing with split ergativity (cf. Dixon 1979: 95), fluid S-marking (Dixon 1994: 80), the assumed "emergence of a passive construction" (Chang and Shefts Chang 1980: 18), or simply the idea of rules which are "not hard and fast" (cf. Goldstein 1991: 55), or of a "complex set of ... factors" (e.g., Denwood 1999: 194). The use of ERG can therefore be defined as a semantic volitional agentcase marker applied without syntactic obligatorily. This explains why authors dealing with Tibetan ERG tend to define ergativity itself in semantic terms (e.g., Tillemans and Herforth 1989: 106); but ERG systems are also systems with structural, i.e., grammaticalized, case markers. This means that ERG and AB S in ERG systems depend on a predefined verb categorization of "transitivity" to which they apply. Therefore, in ERG systems, one usually finds a type of transformation termed antipassive for its similarity to passivization and its apparent dissimilarity of demoting the PAT role (or object) instead of the AG (or subject). The presence of such (anti)passive operations shows that case marking also has a syntactic level with more formal means for the regular reassignment of role relations; they serve in fact the readjustment of focus and of participant roles in an event construal, it is a shift in perspective. Tibetan has no such syntactic operation (cf. Tournadre 1996: 87-89); since it cannot be assumed, though, that Tibetan is unable to make such distinctions, the recategorization of participant roles in an event construal is expected to surface in another part of the grammar, which logically means, in verb recategorizations. This leads us to consider the language type of active/stative case marking. The assumption of Tibetan being of the active case-marking type has been brought up earlier by DeLancey (1984b), is favoured in Saxena (1991), but refuted by Tournadre (1996: 83-85) on the basis of the characteristics enumerated in Klimov (1979; cf. Tournadre 1996: 33); the entries in this list, however, should have been weighed differently, since some points are more important (e.g., verb classes), and some contain secondary details (e.g., specific examples for CAUS/RES verb pairs ['kill/die']).

Tibetan grammar

361

3. The active/stative language type Purely semantic case marking and the non-involvement of the participants into any form of grammatical concord or grammatical operation, together with non-obligatority of constituents, are usually features found in the active/stative language type which is widespread in the Pacific area and in North America (cf. Mithun 1996: 149). The term "active case" was coined by Sapir (1917) for volitional agents, i.e. subjects of (transitive or intransitive) verbs expressing a volitional activity. That means, such languages do not rely on a grammaticalized concept of transitivity, but rather on the semantic distinction of action vs. state. This can be exemplified in Koasati (Mithun 1996: 149, quoting Kimball 1991: 251): (24) ca-libatli-t !S:PAT-burn-PST Ί got burned.'

(25) nihahci ikbak ca-libatli-t grease hot !S:PAT-burn-PST 'The hot grease burned me.'

AG marking thus depends on the degree of control exercised by the marked participant over the event. The degree of control can usually be modified, so that we find three types of verbs with the following semantic characteristics: (a) inherently controllable, (b) inherently uncontrollable, and (c) unspecified. Thus, we may expect three types of verbs which show some morphological device in order to change the controllability. Indeed, the most important characteristic of active/stative languages is not the semantic content of the AG marker, but the other kind of verb classification that occurs in such languages; they do not rely on a concept of transitivity. In Tagalog, e.g., verbs can distinguish fully controlled and partly controlled actions (Drossard 1986: 12); Lakota verbs distinguish only AG and PAT roles, without regard to transitivity (Drossard 1986: 14). The ACT scheme is thus not the same as in fluid S-marking languages of the ERG type; split ergativity never allows subjects of intransitive verbs to be treated as AG (Dixon 1979: 82, 1994: 70-72). In ACT languages, typically AG-PAT events may well consist of verbs requiring ERG (AG) marking. But less typical transitive verbs, e.g. perception verbs ('watch/see') and experiencer verbs ('feel') may not be considered as events requiring or allowing ERG marking (cf. Tsunoda 1985: 388). As was said in the beginning, the existence of purely semantic systems is theoretically excluded from the universal model of grammars, since every semiotic system inevitably possesses self-referential meanings ("grammar"). Since in active/stative languages, the case markers and the

362

Ralf Vollmann

verb classes are semantic, what is then the grammatical part in these languages? They do not distinguish a concept of transitivity, but the verbs of such a language are oriented towards specific participant roles, primarily towards an agent (AG), towards a patient (PAT), or an absolute, i.e. uninvolved, participant (ABS) - or towards an experiencer (EXP). Therefore, we should expect a relative parallelism between AG, ABS and EXP subjects (for the notion of "EXP subjects", cf. Givon 1984: 100), as is the case in Tibetan: (26) ngas las ka 'di by as pa yin/ 1:AG work DBF do:PFV NS CONJ Ί have done thework.' (27) nga khyedrang mjal ba de hacang dga'po 1:ABS you:ABS meet NS DBF very happy byungl GOAL Ί am very happy to meet you.' (28) nga la yi ge zhig 'byor 1 ALL letter INDEF receive 4 1 have received a letter.'

byungl GOAL

While ergativity could be defined mainly in terms of the presence of a direct object (cf. Givon 1984: 152; i.e., ERG marks the subject of a transitive pattern), an active language has verbs which are oriented towards a participant role (i.e., primarily towards an active or stative [inactive] role); therefore, there is neither syntactic obligatorily nor valency. Since in principle no type of role prototypically has a primary function (focus, topic, theme, subject), the EXP case role can play the role of a subject (or topic), so that the pattern EXP-ABS can be found with experiencer verbs; in Spoken Tibetan, the AUX Ibyungf for EXP orientation typically applies in these contexts: (29) nga la dngul yodl 1 EXP money EX:CONJ Ί have money.' (cf. DeLancey 1990: 296)

Tibetan grammar

363

(30) ngar rmilam de yang bskyar btang byungl 1:EXP dream DBF again LV GOAL Ί dreamt that dream again.' (DeLancey 1990: 307) (31) ngar deb de rnyed byungl 1:EXP book DBF find GOAL Ί found the book.' (DeLancey 1990: 313) (32) ngar cham pa brgyab byungl 1:EXP cold LV GOAL Ί caught a cold.' (cf. DeLancey 1990: 310) Since the distinction of active/stative is one of verb orientation towards AG or PAT (or EXP), respectively, we may expect that any kind of transformational process should be found with respect to verb re-orientation. In Tongan, there are three classes of verbs (Drossard 1986: 18): AG oriented: PAT oriented: neutral:

tamate 'kill' sio 'be seen' kai 'eat/be eaten'

(AG in the ABS case) (PAT in the ABS case) (AG or PAT in the ABS case)

Neutral verbs can be oriented towards both AG and PAT, the optional use of ERG case permitting to disambiguate the meaning:

(33) na'e kai 'ae ik PST eat DEF/ABS fish 'The fish ate. = The fish was eaten.' (34) na'e kai 'e he tangata 'ae ik PST eat ERG/DEF man DEF/ABS fish 'The man ate the fish.' An oriented verb implicitly (i.e., by default) has an orientation, but can be reoriented with morphological means (cf. Drossard 1986: 18-19): (35) na'e tamate 'a Mele PST kill ABS Mary 'Mary killed.'

364

Ralf Vollmann

(36) na'e tamate'i 'a PST kill-TR ABS 'Mary was killed.'

Mele Mary

(37) na'e tamate'i 'e PST kill-TR ERG 'John killed Mary.'

Sione 'a Mele John ABS Mary

Tongan shows both NOM/ACC and ERG/ABS patterns with perception verbs, depending on volition (intention): (38) na'e sio 'a Sione ki PST see ABS John ACC 'John saw the woman.'

he fefine the woman

(39) na'e sio-'i 'e Sione 'ae fefine PST see-TR ERG John ABS woman 'John peered at the woman.' There are also operations to reverse AG orientation: -mia is an intensifier of the action which thereby is putting focus on the action itself; the agent is not allowed in this case (Churchward 1953): (40) na'e tanu'i 'ae kapa ('e Sione) PST bury-TR ABS can ERG John 'The can was buried (by John).' (41) na'e tanu-mia'a e kapa PST bury-MIA ABS can "The can was buried.' Perception verbs, as mentioned above, are only weakly transitive (can the participants be called AG and PAT?) and can therefore be considered a turning point for transitivity. Similarly, in Samoan, one finds NOM/ACC and ERG/AB S patterns with perception verbs (Milner 1973; Seely 1980: 130): (42) na va'ai le tama 'i le i'a PST look ART boy ACC ART fish 'The boy looked at the fish.'

Tibetan grammar

(43) na va'ai-a e le tama PST look-TR ERG ART boy "The boy spotted the fish.'

365

le fa ART fish

Again, we find regular morphological operations for changing verb orientation (cf. Milner [1966] 1993: 84, 117): mafai ilo

'be possible, able' 'be aware of

mafai-a iloa

'cope with, succeed' 'spot, notice'

Moreover, there is a causativization device Ifa'a-l (Milner 1993: 43), which transforms static verbs into change-of-state verbs (cf. Drossard 1987: 14,18,19; and others): 1. leaga 'be bad, damaged' fa'a-leaga 'destroy' 'u/na 'be finished' fa'a-uma 'finish' vela 'be hot, cooked' /a'a-ve/a 'cook something' mu'be inflamed' fa'a-mu'burn something' lilo 'be concealed' fa'a-lilo 'hide' pa'u 'fall' /a'a-pa'u 'drop' 2. ofo'besurprised' fa'a-ofo'surprise' 3. sino 'see' fa'a-sino 'show* rnatala 'understand' fa'a-matalä 'explain' 4. masima 'salt (N)' fa'a-masima 'salt(V)' suva'a 'man (N)' fa'a-suva'a 'man (a ship) (V)' suau'u Oil (N)' fa'a-suau'u Oil (V)' /a'a-derivation in Samoan: 1. causativization, 2. psychological effects, 3. experiencer causatives, and 4. ornative formation ('put on y')

And again, an experiencer or directive case (Drossard 1987: 18) plays the role of a third primary-case role: (44) sä ita le tama' He teine IPV angry ART boy DIR girl 'The boy is angry at the girl.' (45) sä fa'a-ita le tama IPV CAUS-angry ART boy 'The girl made the boy angry.'

e le teine ERG girl

366

Ralf Vollmann

The examples from other languages have been given here at some length in order to show the specific characteristics of such systems and the differences to "classical" ergative marking. These examples also show the need for a thorough look at Tibetan verbs.

4. Verbs in Tibetan The Written Tibetan tense system is usually described as having four tense forms: imperfect, future, perfect, and imperative (in Tibetan, Ida Ita bal 'now(ness)', Ima Ongs pal 'not yet arrived', /'das pal 'surpassed, passed away', and Iskul tshigl 'imperative'). Morphologically, verbs of Written Tibetan exhibit a highly opaque morphological pattern (cf. Durr 1950b; Coblin 1976, etc.), which is supported by analytical formation with AUX in both written and spoken varieties (see below).

4: 3a: 3b: 3c: 2a: 2b: 2c: 2d: 1:

PRS klog gtum Ίώτοί bgod Ichrid gtod nyar 'grub gon

PUT bklag gtum dkrol bgo "khnd gtad nyar 'grub gon

PST bklags btums dkrol bgos khrid gtad nyar grub gon

IMP klogs thums khrol bgos khrid gtod nyor gon

REM NONE PRS=FUT PST=FUT PST=IMP PRS=FUT & PST=IMP PRS=IMP & PST=FUT PRS=FUT=PST & IMP PRS=FUT & PST; NO IMP INVAR

The conceptual representation of these forms as "tenses" seems to be inadequate. Tense is a grammaticalised concept of aspect which would even be a typological surprise, after having discovered purely semantic case marking. Tense in Tibetan is rather encoded lexically, such as 'yesterday' or 'tomorrow'. It can be shown from many examples both in syntax and in word formation (cf. Kesang Gyurme" 1992: 268) that the tense forms encode aspects and modes. Specifically, "accomplishment" (PFV/IPV) is a major distinction, and probably "factuality". For example, PUT tense forms have many non-futuric uses (cf. Hahn 1994: 202), which amount to examples such as this one (Losang Thonden 1984: 133; Tournadre 1996: 204):

Tibetan grammar

367

(46) khongtshos khong bkagkyang 3-PL:ERG 3 stop:PFV-CONC ma 'khog pa red/ NEG-stop:IMP-NS-EQU:DISJ 'Though they (tried to) stop him, (he) was not stopped.' Therefore, an aspectual system should be assumed which maps onto the traditional terminology as follows (cf. also Beyer 1992: 261-262): unaccomplished PRESENT FUTURE

fact expectation

accomplished PAST IMPERATIVE

The traditional representation of tense forms does not include a further morphological distinction of Tibetan verbs, which becomes evident in the comparison of the fairly numerous verb pairs of this type: PRS 'grub sgrub 'gum Tujum 'don 'thon 'debs 'thebs 'gyed Tshyed

PUT bsgrub dgum gdon gdab bkye

PST grub bsgrubs bkums Tdiums bton

IMP

btab thebs bgyes khyed

Aob(s)

sgrubs khum(s) thon

translation '(projects which are) realize(d)' '(I) realize (something).' 'kill' 'shrink, be reduced, restricted* 'send out' 'leave' Ί. throw, hit; 2. sow (the seed)' 'be thrown, be hit; etre jet6' 'divide, scatter, disperse' 'be distributed' etc.

We thus find a morphological device distinguishing resultative and causative verb forms in Tibetan; again, the system is morphotactically highly opaque. In spoken varieties, these forms are (if at all) distinguished mainly by a difference in aspiration or by ablaut (cf. Denwood 1999: 108). Since in many cases, a phonotactic distinction can no longer be made, the auxiliary system serves to distinguish the categories in periphrastic constructions with Igyurl for RES and Ibyedl for CAUS verb forms, cf. /'grub par gyurf /sgrub par by edl

'something is realized' 'somebody realizes'

368

Ralf Vollmann

Considering the fact that resultative forms cannot occur in non-factual (expectation) aspect (i.e., they have no PUT and IMP forms, which makes sense only if we are dealing here with aspect forms), we have to merge these two paradigms into one: Table 1. Classical Tibetan verb inflection system unaccomplished PRESENT FUTURE (PRESENT)

fact expectation resultative

accomplished PAST IMPERATIVE (PAST)

(alternative terminology) "differentiative" intentional "undifferentiative"

As mentioned above, due to phonological change (in historical, or prehistorical times), Tibetan verb inflection is opaque and therefore morphotactically weak. Therefore, both Written Tibetan and Spoken Tibetan have developed analytical verb inflections which, among other things, express these same categories: Table 2. Written Tibetan: Analytical verb inflection VERB + par + fact expectation resultative

unaccomplished byed bya 'gyur

accomplished byas shog gyur

(alternative terminology) "differentiative" intentional "undifferentiative"

Table 3. Modern Spoken Tibetan: Analytical verb inflection VERB + fact expectation resultative

unaccomplished giyodl'dug giyinJred gis

accomplished payinlred dang bzhag, byung, song

(alternative terminology) "differentiative" intentional "undifferentiative"

In Spoken Tibetan, there are two more morphosemantic categories in verb morphology, an evidential system (DeLancey 1986), and, as the core of this system, the morphological marking of the speaker/hearer relation ("conjunct/disjunct", cf. Hale 1980, for Tibetan cf. DeLancey 1992). To conclude, Tibetan has a systematic morphological resultative/causative distinction, i.e., a verb-orientation device, which is termed I by a tshig tha dad pal (differentiative verb) and I by a tshig tha mi dad pal (undifferentiative verb) in modern Tibetan grammar (cf. Kesang Gyurme* 1992: 245-247, see above).

Tibetan grammar

369

5. Historiographical approaches Leaving aside earlier missionary attempts, Tibetan has been studied by European scholars since the 19th century (Csoma de Kor s 1834; Schmidt [1839] 1968). Thus, Tibetan ergativity is mentioned in general linguistics in von der Gabelentz ([1901] 1995: 102), together with that of Basque and an Australian language. Thorough knowledge of Tibetan remains with a small number of orientalists, however. These early scholars, such as K r si Csoma Sandor (1834) or, a century later, Jacques Bacot (1946), were usually less influenced by early (and sometimes weird) linguistic theories on ERG (cf. Seely 1980; Dixon 1994: 214-216; Vollmann 2006: 93-95), but instead by the Tibetan indigenous tradition of grammar. The early European grammatical descriptions take great care of the CAUS/RES distinction. Thus, it is interesting to read in Csoma de Kor s (1834) about the distinction between verb forms which have "active/passive" readings (i.e., they behave transitively), and those which rather have a "neutral" meaning - according to the auxiliaries mentioned in this passage, he talks about CAUS/RES: "§171. The two auxiliary verbs that most frequently occur are: Ibyed par/ to do, make; and I'gyur barf to become, grow, wax, change, turn. By the first are formed many active and passive verbs; and by the second the neuters, actives and passives". (Csoma de Kor s 1834: 91). Similar descriptions are found in Schmidt (1968: 143) - which is perhaps not surprising because he has largely followed Csoma's grammar for writing his own (cf. Vollmann 2006: 328). More interestingly, we find a seemingly "philosophical" statement in Bacot (1946), which reads that Tibetan verbs are in principle oriented towards either AG or PAT only in the actual sentence formation (event construal): Par sa forme, le verbe transitif est tout autant Γ action agie par 1'agent que subie par 1'objet. L'action agie par Tagent et cet agent sont dits en tibe*tain /bdag/, personnels ou subjectifs. L*action subie par 1'objet et cet objet sont dit /gzhan/, ext^rieurs ou objectifs. Le verbe a deux aspects ou deux faces, mais il reste le centre autour duquel gravitent les deux termes satellites, agent et objet. II est Γΐαέβ dominante qui e"tablit leur rapport. [By its form, the transitive verb is both the action performed by the agent and undergone by the object. The action performed by an agent and this agent are called in Tibetan /bdag/, personals or subjectives. The action undergone by the object and this object are called /gzhan/, externals or objectives. The verb has two aspects or two faces, but it remains the centre around which the two satellite

370

Ralf Vollmann

terms, agent and object, turn. It is the dominant idea which establishes their relation.] (Bacot 1946: 49-50)

First of all, Bacot states that the Tibetan verb can be oriented towards the AG or the PAT. This refers to the CAUS/RES distinction. Interestingly, the verb seems to be "neutral", and the orientation is realized only when using the (inflected) verb in a clause. Bacot refers here to an indigenous Tibetan grammatical concept, Ibdagl vs. Igzhanl, which has received some attention in the Tibetological literature (cf. Dürr 1950a; Tillemans and Herforth 1989), but is more than difficult to fathom and therefore has been interpreted in various ways.

6. The traditional Tibetan view Tibetan grammar writing generally refers back to Thonmi Sambhota's two treatises which have been saved from destruction during the anti-Buddhist era. In verse 12 of the Irtags kyi 'jug pal (cf. Miller 1993: 91-111; Tillemans 1989: 1; Kesang Gyurme" 1992: 191), Thonmi Sambhota mentions the Ibdagl - IgzhanJ concept, but quite shortly and far from understandable to the outsider; he seems to give one (!) rule for verb morphology with relation to Ibdagl - Igzhanl. The fact that Thonmi's rules are so "terse" (cf. Miller 1993: 14, 63; Herforth 1989: 77) may be due to the fact that his rules don't seem to reflect "Classical Central Tibetan", but perhaps "Central Asian Tibetan" (see below). Whatever the philological problems, the main commentary giving the explanation for this concept seems to be the work of Situ Panchen (16991774). The following passage on verse 12 of Thonmi has been translated by Durr (1950a: 82) into French and by Tillemans and Herforth (1989: 4, 6263) into English; the passage reads according to Durr (1950a: 82) (my translation): (47) las gang zhig byedpa po gzhan dang act what INDEF agent other SOC 'Whatever action, an agent with another (entity) dngos su 'brel ba 'i dbang du by as nasl truly relation:GEN in.terms.of ABL after having entered into a true relation,

Tibetan grammar

371

byedpapo'i dngospo de nyid dang agent:GEN subst:exist DEF ABSTR SOC the actual realization of the agent de'i byedpa dang be as pa la ni DEF:GEN doing SOC established ALL TOP together with the doing bdag ces bya zhingl des bsgrub BDAG so-called GER DEF:ERG accompl:FUT is called BDAG. If this (agent) intends to accomplish (something), par bya ba'i yul gyi dngospo NS:ILL do:FUT NS:GEN object GEN entity the realization of the object/goal of the (intended) action dang de'i bya ba dang bcaspa la ni SOC DEFrGEN do:FUT SOC established ALL TOP together with the intended action gzhan zhes bya 'oil GZHAN so-called:FIN is called GZHAN.' The understanding of this passage depends heavily on the right choice of words for translating Tibetan linguistic terminology into European linguistic concepts. Specifically, the verb forms Ibyed pal (PRS) and /bya bal (PUT) play a decisive role here: the "doing" ("action") and the "intended-doing" ("object"), are distinguished. The text relates only to "doing" and "intention to do", and to the AG and the PAT respectively, and terms these "I" and "OTHER": AG + "doing" = "I"; PAT + "intended doing" = "OTHER". This looks like agreement of PRS forms with the AG, and of PUT with PAT. If the tenses were tenses, this would not make sense; only if these forms (originally) were a verb-orientation device, then this is the description of such a phenomenon: "agreement" between AG and AG-oriented verb form ("I") vs. PAT and PAT-oriented verb form ("OTHER"). In Central Tibetan, however, these verb aspects do not play such a role. Of course, PUT by default has no ERG, but IPV also has optional ERG. The only place in grammar where IPV/FUT still plays a decisive role is in word formation, and there it has to do with orientation, cf. Iston pa sangs rgyasl 'the teacher Buddha' (IPV form, "I") vs. I sangs rgyas kyi bstan pal 'the teachings of the Buddha' (INT form, "OTHER") (cf. Kesang Gyurme' 1992: 271); similarly, /'jigs byedl (PRS), lit. '(somebody who) terrifies' is

372

Ralf Vollmann

person-oriented: 'Bhairava, the destroyer', whereas Ibsgom byal (PUT), lit. '(something which is) meditated (on)' is object-oriented: Object of meditation'. On the other hand, Thonmi most probably did not invent the script, but brought it from (Tibetans in) Central Asia (Khotan, cf. Hoernle 1916), and he may have imported the grammatical rules as well. Therefore, the description probably refers to an earlier, or different verb-orientation device, which in Central Tibet is found mainly in the CAUS/RES distinction. Situ Rinpoche gives examples both of sentences and of verb forms in order to exemplify his explanation, and we find there examples of PUT stems and of CAUS/RES forms (cf. bilingual ed. in Tillemans and Herforth 1989: 6668), cf. with direct AG (BDAG) Icags gser du bsgyur zin iron gold ILL change:PFV ready 'The iron has been changed into gold.' without direct AG (GZHAN) Icags gser du gyurd zin iron gold ILL change :RES ready "The iron has changed into gold.' with direct AG (BDAG) shing bead zin wood cut:PFV ready "The wood has been cut.' without direct AG (GZHAN) shing chad zin wood cufRES ready "The wood has split.'

Thus, Tibetan verbs do not have obligatory nominal constituents, but due to verb orientation, AG or PAT participants are there by implication. In other words, verb orientation rids the language of the need for syntactic valency - in fact, this is why active-type languages do not need syntactic obligatorily. Tibetan causative/factual/accomplished verb forms (which take ERG obligatorily) are not mentioned in the bdaglgzhan theory. But the verb categories where ERG may or may not occur (i.e., imperfective, non-factual) are mentioned: Indeed, Situ talks only about present (i.e., imperfec-

Tibetan grammar

373

tive) and future (i.e., intention) verb forms; imperfective forms are "I", nonfactual forms are "OTHER". It seems from the exemplary verb lists given in Situ (cf. Tillemans and Herforth 1989: 62-63) as if this bdag/gzhan distinction is given on the basis of the distinction of factual acts vs. nonfactual/resultative events/processes. We can therefore try to sum up Situ's model in this way: SOURCE SELF SOURCE SELF

GOAL OTHER GOAL OTHER

FACTUAL ACT SELF NON-FACTUAL ACT OTHER

SOURCE + ACTION GOAL/OBJECT + ACTION

The bdag/gzhan distinction is described in terms of agreement - although there is no agreement morphology - between either verb and SRC or verb and GOAL, respectively, dependent on the aspect forms of the verb. But in this view, Bacot's statements (1946, see above) appear in a new light, and the typological hypothesis brought forth here gets an additional argument from indigenous descriptions. More interestingly, the Tibetan CONJ/DISJ distinction (see above) is often seen as a kind of "agreement" in some specific contexts (i.e., CONJ with 1st person [agentive] subjects) where this agreement with "I" or "OTHER" can make an additional semantic distinction - a phenomenon which has led Chang and Shefts Chang (1980: 18) to hypothesize the emergence of a passive (restricted to first-person subjects); instead, it is another type of verb reorientation (cf. Chang and Shefts Chang 1980: 18 for similar examples): (48) nga(s) dkaryol bcag gi yin I(:AG) cup break CONN CONJ Ί will break the cup (AG: intentionally).' (DeLancey 1990: 307) (49) ngas dkaryol bcag gi red I:AG cup break CONN DISJ Ί will break the cup (emphasis, unintentionally).' (DeLancey 1990: 307) cf. also (50) ngas dkaryol chag song LAG cup break PFV Ί broke the cup (unintentionally).' (Kesang Gyurme" 1992: 255; cf. DeLancey 1990: 300)

374

Ralf Vollmann

In other cases (of non-first person), the CONJ form can implicitly refer to an "I" which is outside the event construal; the first example gives a neutral statement about the builder of the Potala palace; in the second example, however, the CONJ form refers to the ingroup of the writer and the readers as conjunct with (i.e., implicit "owners/builders" of) the Potala: (51) phobrang rnying pa de ni palace old DBF TOP srong btsan sgam pos bzhengs pa red/ Songtsän Gampo:ERG builf.HON NS DISJ "This old palace was built... by Songtsän Gampo.' (quote taken from Tournadre 1996: 93) (52) ... pho brang chen po po ta la ni palace big Potala TOP ... bzhengs yodl built:HON CONJ "The ... big palace Potala was built...' (from a Tibetan newspaper) 7. Conclusion In Tibetan, NPs are not obligatory. The case system has three core cases: ERG, ABS, EXP, which form all patterns of case relations (cf. Tournadre 1996: 75). The ERG case marker has only semantic and pragmatic functions and does therefore not depend on a notion of valency or transitivity, nor does it interact in syntactic transformational processes (e.g. antipassive). It seems to occur rather freely, as a function of semantic needs, being a marker for "volitional agents". On the other hand, default ERG application interacts closely with verb aspects; therefore, a kind of aspectual split seems to be in place (decreasing ERG application along PFV-IPV-INT). But the main, or original, factor for ERG use seems to be the orientation of verbs which is an old morphological category of Tibetan. The phenomenon of verb orientation is described originally in indigenous grammar, under the concept bdag/gzhan ("I'7'OTHER"), for the distinction of IPV/INT verb stems; the main distinction, however, is found in the old morphological category of CAUS/RES verb stems which are also reflected in the auxiliary system of Written Tibetan (/... par byedl vs. /... par gyurl, etc.). The system seems to have further developed another type

Tibetan grammar

375

of orientation, the CONJ/DISJ distinction which mainly specifies the speaker's relation to the event construal; its main distinction could nowadays be termed "I/we-oriented" vs. "other-oriented". Ergative languages are defined on the basis of an ERG case marker; but most ERG languages have a strict relation between a class of transitive verbs with an obligatory structural case role "ERG". They do not have a semantic AG marker, but rather a marker for subjects of what is considered transitive clauses in that language; usually, syntactic transformations, and not merely semantic considerations, apply on the case-marking patterns. The active/stative languages have an AG case marker, i.e., a semantic case marker, which is usually posited in a tripartite case system (AG/ABS/EXP). The case markers "cooperate" with semantic verb categories, which can be changed morphologically ("[dejtransitivizers", etc.). In this way, verbs can change the category, i.e. their orientation: they construe either "actions of somebody" or "events performed on something/somebody". By doing so, they do neither need obligatorily of participants, nor valency, nor structural transformations (e.g., antipassive). The system of Tibetan does not neatly map onto all phenomena observed in "typical" (?) active languages, but it represents a peculiar type of case marking which is more adequately described in terms of "active/stative" than in terms of "ergative/transitive". One could probably consider a macro-areal characteristic of semantic case-marking patterns from (a) the Pacific area (active/stative) (b) across Philippine languages (TOPprominence) (c) towards the (Sino-) Tibetan pattern (semantic ERG case, verb orientation, optional/historical TOP marking) described here. 8. Abbreviations of grammatical terms ABL ABSTR ACT ALL AUX

ablative abstract noun formation active/inactive type allative (+ EXP, DAT) auxiliary verb

ABS ACC AG ART BDAG

CAUS

causative verb form, causativity conjunct dative demonstrative disjunct

CONC

absolute (participant) accusative agent definite article "I" concept of Tibetan grammar concessive

CONN DBF DIR EQU

connector definiteness directive equational AUX

CONJ DAT DEM DISJ

376

Ralf Vollmann

ERG EXP PUT GER GZHAN ILL INDEF INT IPV MIA NOM PAT PL PST RES SOC TOP

ergative experiencer future gerund "other" concept of Tibetan grammar illative indefiniteness intentional imperfective a detransitivizer nominative patient plural past resultative sociative, comitative topicalizer

EX FIN GEN GOAL HON

existential AUX clause-final particle genitive AUX for Istpers. EXP honorific

IMP INS INVAR LV NEG NS PFV PRS REM S SRC TR

imperative instrumental invariable light verb negation nominali zer perfective present remarks subject source transitivizer

References Bacot, Jacques 1946 Grammaire du abstain littfraire. Paris: Librairie d'Amdrique et d'Orient Maisonneuve. Beyer, Stephan V. 1992 The Classical Tibetan Language. Albany: State University of New York Press. Chang, Kun, and Betty Shefts Chang 1980 Ergativity in spoken Tibetan. The Bulletin of the Institute of History and Philology 51 (1): 15-32. Taipei: Academia Sinica. Churchward, C. Maxwell 1953 Tongan Grammar. London: Oxford University Press. Coblin, Weldon South 1976 Notes on Tibetan verbal morphology. T'oung Pao 62:45-70. Csoma de Körös, Alexander (= Korösi Csoma Sändor) 1834 A Grammar of the Tibetan Language. Calcutta: Baptist Mission Press. DeLancey, Scott 1984a Categories of non-volitional actor in Lhasa Tibetan. In Proceedings of the Conference on Participant Roles: South Asia and Adjacent Areas, Arlene R. K. Zide, David Magier, and Eric Schiller (eds.), 5870. Bloomington, Indiana: Indiana University Linguistics Club.

Tibetan grammar 1984b

377

Notes on agentivity and causation. Studies in Language 8 (2): 181213. 1986 Evidentiality and volitionality in Tibetan. In Evidentiality: The Linguistic Coding of Epistemology, Wallace L. Chafe, and Johanna Nichols (eds.), 203-213. (Advances in Discourse Processes 20.) Norwood, N. J.: Ablex. 1990 Ergativity and the cognitive model of event structure in Lhasa Tibetan. Cognitive Linguistics 1 (3): 289-321. 1992 The historical status of the conjunct/disjunct pattern in Tibeto-Burman. Acta Linguistica Hqfiiiensia 25: 39-62. Denwood, Philip 1999 Tibetan. (London Oriental and African Language Library 3.) Amsterdam/Philadelphia: John Benjamins. Dixon, Robert M. W. 1979 Ergativity. Language 55: 59-138. 1994 Ergativity. (Cambridge Studies in Linguistics 69.) Cambridge: Cambridge University Press. Drossard, Werner 1986 Verbklassen, (akup 64.) Köln: Institut für Sprachwissenschaft. 1987 Transitivität (vs. Intransitivität) und Transitivierung (vs. Intransitivierung) unter typologischem Aspekt, (akup 68.) Köln: Institut für Sprachwissenschaft. Dürr, Jacques A. 1950a Deux traitds grammaticaux tibdtains. (Bibliothek der allgemeinen Sprachwissenschaft, dritte Reihe.) Heidelberg: Carl Winter Universitätsverlag. 1950b Morphologie du verbe tib tain. (Bibliothek der allgemeinen Sprachwissenschaft, dritte Reihe.) Heidelberg: Carl Winter Universitätsverlag. Gabelentz, Georg von der 1995 Reprint. Die Sprachwissenschaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. 2nd ed. Leipzig, 1901. Original edition, Leipzig, 1891. Givon, Talmy 1984,1990 Syntax: A Functional-Typological Introduction. (2 vols.) Amsterdam/Philadelphia: John Benjamins. Goldstein, Melvyn C. 1991 Essentials of Modern Literary Tibetan. A Reading Course and Reference Grammar. Berkeley: University of California Press. Hahn, Michael 1994 Lehrbuch der klassischen tibetischen Schriftsprache. 6th ed. (Monographien zu den Sprachen und Literaturen des indo-tibetischen Kulturraumes 10.) Swisttal-Odendorf: Indica et Tibetica.

378

Ralf Vollmann

Hale, Everett Austin 1980 Person markers: Finite conjunct and disjunct verb forms in Newari. In Papers in South-East Asian Linguistics 7, Ronald L. Trail (ed.), 95-106. (Pacific Linguistics Series A-53.) Canberra: Pacific Linguistics, Australian National University. Herforth, Derek D. 1989 Transitivity and voice: The perspective from Tibet. In Agents and Actions in Classical Tibetan. The Indigenous Grammarians on bdag and gzhan and bya byed las gsum, Tom J. F. Tillemans, and Derek D. Herforth (eds.), 75-94. (Wiener Studien zur Tibetologie und Buddhismuskunde 21.) Wien: Arbeitskreis für Tibetische und Buddhistische Studien, Universität Wien. Hoernle, A. F. Rudolf 1916 Manuscript Remains of Buddhist Literature Found in Eastern Turkestan. Facsimiles with Transcripts, Translations and Notes, Edited in Conjunction with Other Scholars. Vol. I. Oxford: Clarendon Press. Kesang Gyurme" (= /Skal bzang 'Gyur med/) 1992 Le clair miroir. Grammaire tibetaine. Traduction, adaptation et commentaires de Heather Stoddard et Nicolas Tournadre. Paris: Editions Prajna. Kimball, Geoffrey D. 1991 Koasati Grammar. (Studies in the Anthropology of North American Indians 1.) Lincoln: University of Nebraska Press. Klimov, Georgij Andreevifc 1979 On the position of the ergative type in typological classification. In Ergativity: Towards a Theory of Grammatical Relations, Frans Plank (ed.), 327-332. London etc.: Academic Press. Langacker, Ronald W. 1990,1991 Foundations of Cognitive Grammar. (2 vols.) Stanford: Stanford University Press. Losang Thonden 1984 Modern Tibetan Language (Deng dus bod kyi skad yig). Dharamsala: Library of Tibetan Works and Archives. Miller, Roy Andrew 1993 Prolegomena to the First Two Tibetan Grammatical Treatises. (Wiener Studien zur Tibetologie und Buddhismuskunde 30.) Wien: Arbeitskreis fur Tibetische und Buddhistische Studien, Universität Wien. Milner, George Bertram 1973 It is aspect (not voice) which is marked in Samoan. Oceanic Linguistics 12:621-639. 1993 Reprint. Samoan Dictionary (Samoan-English, English-Samoan). Original edition, Polynesian Press, 1966.

Tibetan grammar

379

Mithun, Marianne 1996 Overview of general characteristics. In Handbook of North American Indians, vol. 17: Languages, Ives Goddard (ed.), 137-157. (Handbook of North American Indians 17.) Washington: Smithsonian Institution. Moosmüller, Sylvia, and Ralf Vollmann 2001 "Natürliches Driften" im Lautwandel: die Monophthongierung im österreichischen Deutsch. Zeitschrift für Sprachwissenschaft 20 (1): 42-65. Reichert, Christoph 1986 Verteilung und Leistung der Personalaffixe im Ungarischen, (akup 52.) Köln: Institut für Sprachwissenschaft. Sapir, Edward L. 1917 Review of Uhlenbeck 1916. International Journal of American Linguistics 1: 82-86. Saxena, Anju 1991 Pathways of the development of the ergative in Central Tibetan. Linguistics of the Tibeto-Burman Area 14 (1): 109-116. Schmidt, Isaak Jakob 1968 Reprint. Grammatik der tibetischen Sprache. (Herausgegeben von der Kaiserlichen Akademie der Wissenschaften). Original edition, St. Petersburg: W. Graff, Leipzig: Leopold Voss, 1839. Seely, Jonathan Franklin 1980 A Study of Ergativity. (Ph. D. diss., University of Colorado at Boulder 1978.) Ann Arbor/London: University Microfilms International. Tillemans, Tom J. F. 1989 Introduction: Traditional Tibetan grammar on /bdag, gzhan/ and related notions. In Agents and Actions in Classical Tibetan. The Indigenous Grammarians on bdag and gzhan and bya byed las gsum, Tom J. F. Tillemans, and Derek D. Herforth (eds.), 1-36. (Wiener Studien zur Tibetologie und Buddhismuskunde 21.) Wien: Arbeitskreis für Tibetische und Buddhistische Studien, Universität Wien. Tillemans, Tom J. F., and Derek D. Herforth 1989 Agents and Actions in Classical Tibetan. The Indigenous Grammarians on bdag and gzhan and bya byed las gsum. (Wiener Studien zur Tibetologie und Buddhismuskunde 21.) Wien: Arbeitskreis für Tibetische und Buddhistische Studien, Universität Wien. Tournadre, Nicolas 1996 L'ergativite' en tibe'tain. Approche morphosyntaxique de la langue partee. (Bibliotheque de 1'Infonnation grammaticale 33.) Louvain/ Paris: Peelers. Tsunoda, Tasaku 1985 Remarks on transitivity. Journal of Linguistics 21: 385-396.

380

Ralf Vollmann

Uhlenbeck, Christianus Cornelius 1948 Le langage basque et la linguistique generate. Lingua l: 59-76. Vollmann, Ralf 2006 Descriptions of Tibetan ergativity. A historiographical account, (in preparation)

The nature of narrative text in Dzongkha:1 Evidence from deixis, evidentially, and mirativity Stephen A. Watters

1. Introduction Legend narrative texts in Dzongkha are organized, as in any language, according to two major perspectives: the perspective of the narrator and the perspective of the participant characters. The former perspective is the distal view - one that takes place in past time (then), at some other location (there), between third-person protagonists (they). This is the "outer shell" of the narrative. The perspective of the participant characters, on the other hand, is the "near" view and occurs as online dialog in real time (now), between speech-act participants (1st and 2nd persons), at this location (here). Though proximate - distal splits in spatial, temporal, and personal deixis are common in the organization of narrative discourse in many languages, in this paper I want to explore two related phenomena that are more Dzongkha-specific. The first has to do with an extension of the deictic dichotomy to include grammatical connectives. That is, connectives belonging to the outer shell of the narrative are broadly syntactic, having to do mostly with sequence and other temporal notions. These collocate and distribute with the usual markers of distal deixis mentioned above. Connectives within quoted material, on the other hand, are more semantically specific and have to do primarily with speaker opinion. These collocate and distribute with markers of the near view. The other Dzongkha-specific topic I wish to explore comes as an unexpected consequence of the proximate - distal dichotomy in narrative discourse - the ambivalent status of second person. Under normal circumstances, second person combines in a unified category with third person the so-called "disjunct" category. First person is marked differently, the socalled "conjunct" person. This is the expected alignment for Tibetan languages. In "online" material (that part of the discourse that falls within direct quotes), second person combines unexpectedly with first person, the other speech-act participant. Both are marked conjunct. Thus, though (or

382

Stephen A. Watters

because) second person is conclusively neither proximate nor distal, it patterns with first person in proximate contexts. In distal contexts, third person patterns with both conjunct and disjunct verb agreement. Later in the paper, I will offer another possible interpretation of the unexpected alignment. On this interpretation, the perspective of the "omniscient narrator" leaks through to the participant characters in whom he is vesting expression of his own mental attitude (Givon 1998). He is omniscient but not invisible, and his lurking shadow falls across the page. This accords with DeLancey's observation that the conjunct is appropriate for a sentence like "There will be a party on Friday night" if the speaker was on the planning committee. Here, too, the omniscient narrator is privy to second person's reasonings, motives, and expectations, and that knowledge colors his narration. 2. The data The texts analyzed for this paper are taken from Dasho Sherub Thaye's (1987) Rdzong kha 'i gtam rgyud sna tshogs dpe cha gsum pa. This is an anthology of famous Bhutanese short stories and historical legends. The primary texts used in the analysis for this paper are: ο la 'than 'bag mi "The crow who brought a message' and 'dre'i rkyal hag "The leather-bag ghost'. Other texts from the same anthology were consulted to confirm findings. These texts constitute one particular genre of the written tradition: wellknown Bhutanese stories in modern written Dzongkha. It is assumed that what is found here, may not necessarily hold true for other genres of text (e.g., oral text, or written classical text), other text types (i.e., narrative, procedural, and so forth), as well as other authors. As such, the limitations of this study should be noted. 3. Connectives in this text There are nine connectives found in these texts. The addition connectives are: -te "converb connective", -dang 'when ...' and -par '-ing'; the time connectives are -zhingm las 'having ... then', -bar na 'while in the middle of ...', and -khar 'while ...'; contrastive particles are -ein 'if ...' and -rung 'although ...'; and the causality particle is -gis 'because'. Illustrative sentences are given below in (1).

The nature of narrative text in Dzongkha

(1)

383

a. bgola de tshu bltabs tc sgrom nang btsug clothes these fold CON box in stick in bzhag put 'Fold the clothes up and put them away in the chest (box).' b. khong bstun gros rkyab sdodp dang srinm cig they discussion do stay CON demon one Ong-nug came-PAST 'When/ while they were having discussions, a demon came.' c. kho snngm-gis slab inm ma shes par he demon-AG say REL NEG know CON khyim na song-nug house to went 'Not knowing it was the demon who spoke, he went to the house.' d. phar Ihod zhingm las spya dang there come CON monkey and ola-gis thang gi sbu lu song-nug crow-AG field of middle to went 'Having arrived over there, the monkey and the crow went to the middle of the field.' e. yid-ma-ches-par sdodp pa'i barna srinm gi not-trusting stay NOM CON demon of khyim nang Ihod-nug house in came 'While being in a state of not trusting (the demon), they arrived at the demon's house.' f. da nga bcas gnyis lo na rgas pa 'i khar... now we two year in old NOM CON 'Now, while we're growing old in age ...' g. bum de lu gnyen dang grogs cig daughter that to marriage and companion one medp cm ... not CON 'If this daughter doesn't have a companion in marriage ...' h. khyod nyams 'tsher rung ngi a'i dang you lethargy CON my mother and

384

Stephen A. Watters

ap 'badsa song father place go 'Even if it is difficult, go to my father and mother's place.' i. 'dre-gis thob ni 'di gis in ghost-AG get INF this CAUS COP 'It is because the ghost got it.' In table 1, I give the nine connective particles in columns and the semantic relations they encode in rows. Each connective expression is grouped into the four broad categories given by Rudolph (1988) and Halliday and Hasan (1976) (in columns). The number in each cell represents the number of times a particular connective encodes a semantic relation (i.e., -te encodes the relation "sequence" 75 times). The semantic relations in rows are those given by Mann and Thompson (1988). These relations are more narrowly defined and more comprehensive than those of Rudolph (1988). Table 1 illustrates that the overwhelming semantic relation in this text is sequence: 115 of the 177 semantic relations are sequence. The second most numerous relation is one of elaboration. In one relation time is encoded, and in the other additional detail is provided. This is expected given that the inherent nature of narrative text is temporal (Longacre 1983; Labov 1972). Another observation to be made is that some connectives are more closely linked to a particular relation than another. On the one hand, a connective like -te marks the relation between five different semantic relations, and on the other -rung marks only a concessive relation but no others. It turns out that the contrast and causality connectives mark only one semantic relation. The connectives in the addition and time category, however, correlate with a number of different semantic relations (as given by Mann and Thompson).

The nature of narrative text in Dzongkha

385

Table 1.2 Connectives and the semantic relations they encode relation definitions

relation total

connective type contrast causality addition time -fe -dang -par -zhingm -bar -khar -ein -rung -gis

sequence joint circumstance elaboration volitional cause non-volitional cause non-volitional result concession condition purpose total

75 1

las 9

27 1 8

6 1

na 4

115 2 11 16 1

3 10

2

2 1

1

14 12 1 85 37

11

9

4

3

12

14

2 2

14 12 3 177

The Dzongkha narrative illustrates that addition connectives are the most frequent and the least explicit in the semantic relation they encode. As noted above, -fe varies in the semantic relations it links. This suggests that -fe is a connective that is mostly syntactic in function, with little semantic value. As such, it is free to link between almost any two propositions (a converbal construction). The connectives that are linked with one particular semantic relation have, on the other hand, a more explicit semantic content, and as such are not free to encode the relation between just any two propositions. These latter constructions serve to give the speaker's personal views together with information and facts. 4. Distribution of connectives Given that speaker opinions are encoded in contrast and causality connections, it is instructive to determine exactly where these semantic relations occur in a narrative text. Addition/ time and speaker-opinion connectives occur in somewhat complementary distribution. Addition/ time connectives occur mostly outside of direct quotes, and the opinion connectives occur largely inside direct quotes. This pattern is illustrated in table 2. This is not to say that these connective particles occur only in this distribution. First,

386

Stephen A. Watters

there are counter-examples which illustrate that the addition/ time connectives can occur inside quotes, and speaker-opinion particles which can occur outside of quotes, as in the occurrence of -rung outside of quotes. Labov views narrative text as a "... method of recapitulating past experience by matching a verbal sequence of clauses to the sequence of events which (it is inferred) actually occurred" (1972: 359-360). Given this view, the narrator of short-story texts in Dzongkha organizes his text around the temporal sequences which are believed to have occurred. These sequences are somewhat removed from the speaker and recapitulated in past time. These sequences are linked, among other things, through temporal connectives. As such, the addition/ time connectives help encode a kind of distal time view - a time removed from that of the narration of the story. Inside quotes, however, we see a significant lack in temporal connectives, and a marked increase in opinion connectives. These do not necessarily encode the viewpoint of the narrator, but rather some sort of deictic shift, whereby the speaker takes the voice of the participants in the story. As such, the relations encoded in these sequences give the participant's personal views together with information on facts. Table 2. Connectives suggest two pairs of function: Speaker's view point, and speaker's opinions

-fe convert) -dang 'when, and' durative? -par '-ing' somewhat adverbial -zhingm las 'having ..., then ...' -barna 'while in the middle of . . . ' -khar* in the state of being' -rung 'although ...' -cin'if ...' -gis'by ...' causative total

# inside quotes 6 0

# outside % inside quotes quotes 8 72 37 0

ι outside quotes 92 100

semantic category addition addition/time

2

9

18

82

addition

2

7

22

78

addition

0

4

0

100

time

0

3

0

100

time

8 12

5 0

62 100

38

contrast contrast

2

0

100

17

72

causality

The nature of narrative text in Dzongkha

387

5. Distribution of deictic elements 5.1. Pronouns The distribution of connectives as relating to some sort of deictic notion inherent in narrative is confirmed in the distribution of other deictic elements: the distribution of connectives mirrors that of deictic elements in that of person, place, and time-deixis markers. These also tend toward complementary distribution in the same textual environment as connectives. In table 3,1 give the distribution between person and place deixis. 'di 'this' is used mainly in direct quotes; de 'that' is used mainly outside of direct quotes. Similarly, the pronouns nga T and khyod 'you' are used only in direct quotes; khong 'they' is used mostly outside of direct quotes. Table 3. The complementary nature of person and place deixis

'Λ" this' nga Τ khyod 'you' de 'that' khong 'they'

occurrences outside quote 1 52 25

occurrences inside quote 9 32 26 10 1

% outside quotes 10 84 96

% inside quotes 90 100 100 16 4

5.2. Verbal tense The same distribution that exists for connectives, place and person deictic elements exists also for time deixis, as encoded in verbal endings. The pattern is illustrated in table 4 below. The verb particles above the black line in table 4 encode various past tenses and aspects. These verbal particles end finite clauses outside of quotes. The most frequent verbal particle is that of -nug which occurs in twenty-nine verb final positions. This verb particle encodes the historical past or a degree of distance from an event in the evidential hierarchy (van Driem 1998, and Zeisler 2000 have called this "inferred past"). The verbal particles below the dark black line are of a markedly different nature. These are verbal endings which are stative (with no verbal marking); modal (e.g., tshugs 'able', dgo 'must'); commands (with no verbal particle); future tenses (e.g., -in in and -ni in mas)\ present-tense question particles (e.g., -0/p/w/m smo and in nam); present factual tense

388

Stephen A. Watters

(e.g., -0/p/w/m in); and conjunct copulas and existentials (e.g., in, -in pas, yod, and med). Table 4. The complementary nature of time deixis in verbal affixes sentence final

-nug -yi/ci -so nug -pmas -in pas in yod med -niyod 0 (command) 0 (stative) -0/p/w/m smo -0/p/w/m in in nam ga/gam Ong/mi Ong -ni in -ni in mas -do -ge total

29 8 9 5 4 1

1 1

58

inside quotes

2 2 4 9 1 5 1 12 12 7 7 4 3 5 1 3 1 3 82

% outside of quotes 100 80 82 100 50 10

8 8

% inside quotes 20 18 50 90 100 100 100 92 92 100 100 100 100 100 100 100 100 100

The past-tense particles appear in clauses outside of quotes, while the other verbal particles correlate with clauses inside of quotes (in online discourse). This demonstrates the unique nature of quoted speech in Dzongkha narrative text. While the events of the narrative are in past tense, including the final clause in which quoted speech is embedded, the online discourse is indicative of interaction which is decidedly present tense, or lacking in tense (e.g., stative and modal verbs). In summary, then, the distribution of connectives, place and person deixes, and time deixes in verbal affixes parallel one another in encoding the distal time of a legend narrative (offline), and a decided shift in point of view in direct speech (online discourse). This distribution is what we might expect of legend narratives: offline is 'then/there/they', and online is 'here/now/you/me'. However, this results in an unexpected pattern with

The nature of narrative text in Dzongkha

389

regard to the conjunct/ disjunct verbal complex as documented in a number of Tibetan languages, and it is to this issue that I now turn. 6. Evidential and mirative categories in the verb Zeisler (2000) suggests four general evidential distinctions that are made for Dzongkha: immediate perception (new information), other knowledge including self-evident and generic information (old information), inferred information, and hearsay. She suggests the Dzongkha affixes fit into this scheme in the way illustrated in table 5. Table 5.

nonpast

past

Dzongkha verbal affixes and their evidential content (adapted from Zeisler 2000)

immediate perception (new) experience visual 'dug mas pas/ has yilci

other knowledge (old) self-evident generic yod in pa 'i/ba 'i 0

inferred

hearsay

lo

nug

It is beyond the scope of this paper to demonstrate this interaction clearly for Dzongkha (see van Driem 1998 [on which Zeisler presumably based this table] for examples of some of this interaction), but it is instructive to see the extent to which this categorization matches the use and distribution of affixes in the narrative texts studied for this paper. The issue of evidentiality, mirativity, and epistemics has generated much discussion for the Tibetan languages. Häsler reviews some of the main paradigms by which scholars view this system: conjunct/ disjunct, volitionality, and novelty of information. She proposes yet a fourth way of looking at this system that is based on Kuno's (1987) empathy hierarchy. She suggests that the speaker can "choose a different angle from which to look at an action/ event, and thus she can decide to what degree she wants to 'side' with the agent/ subject of her utterance" (2001: 25). In other words, she suggests that the encoding of the prototypical alignments often documented for these languages (i.e., 1st person/ conjunct/ volitional/ old knowledge vs. 2nd and 3rd person/ disjunct/ non-volitional/ new knowledge) can be overridden by other factors that play into the speech-act situation. The patterns which emerge in Dzongkha legend narratives certainly

390

Stephen A. Waiters

point to the fact that the evidential and mirative systems are not obligatorily matched to participants in the sentence. 6.1. Conjunct/ disjunct DeLancey states that "Conjunct contexts are: with 1st person actors in statements and 2nd person in questions, and in complements of verba dicendi, when the actors of the higher and lower verbs are co-referential" (1992: 57). However, this pattern never emerges in the legend narratives studied for this paper. Recall that in Dasho Sherub Thaye's legend narrative texts agent/ subject of offline discourse is 3rd person, and that agent/ subject in online discourse is 1st and 2nd person. In online discourse, the pattern that emerges is that finite verbs and copulas have conjunct/ old knowledge marking even when the subject/ agent is 2nd person. For example in sentence (2), the subject is 'you' (disjunct) and the equative copula is in (conjunct/ old information). (2)

da rung'badrung 'di phyi gnyis-kyi skyabs dang moreover this later two-GEN save and mgon khyod rkyangm cig rang in defender you only EMPH COP-CONJ 'Moreover, the protector and defender of these two humble people, now and forever, only you are.'

While it is easy to understand why a 2nd person subject (disjunct person) is co-referenced to a conjunct copula in this context (it's old information for the "speaker"), the amazing thing about these narrative texts is that 2nd person (disjunct) in online discourse is always co-referenced to the conjunct verbal marker. Another thing to observe is that, in spite of the fact that offline discourse is decidedly past tense and removed from the speaker, the pattern that emerges in finite verbs is an interplay in verbal affixes between immediate perception, e.g. -mas and -yi/ci, and inferred information -nug. The expected pattern is to find 1 st person volitional verbs marked with -yi/ci, and to find finite verbs marked with -nug in other contexts. However, for example, the following two sentences are contiguous to one another in discourse. In sentence (3), the subject is a full noun phrase, and in sentence (4), the agent is a free standing pronoun, but in sentence (3), the finite verb

The nature of narrative text in Dzongkha

391

ends in -ci (immediate perception/ conjunct), and in sentence (4) the finite verb ends in -nug (inferred/ disjunct). (3)

'drei' rkyalltag n'as 'bru bda'-yi ghost Kyeta barley grain follow-PAST.CONJ bda'-yi rang 'gyo-w dang byagphug follow-PAST.CONJ EMPH go-PRES and cliff cave shorn cig 'bad sar lhod-ci big INDEF place arrive-PAST.CONJ 'As the ghost Kyeta went along following (fields) of barley grain, he arrived at the place of a big cliff cave.'

(4)

kho byagphug nang 'dzul-ti haw dang he cliff cave in enter-CON look-PRES and bum cig sngu-sti sdod-sa mthong-nug girl INDEF cry-CON sit-PROG see-INFER 'He entered into the cliff cave, and when looking, he saw a girl sit crying.'

Thus, the immediate-perception past-tense marker -yi/ci alternates with the inferred past-tense marker -nug throughout the story, and the subject/ agent remains 3rd person throughout the texts.

6.2. Mirativity There are two different ways of marking relative novelty of information in Dzongkha in the present tense: one is with the suffix -mas which van Driem (1998) describes as an experienced-perception marker on verbs denoting an activity or feeling, and the other is with the suffix -has (and its morphophonemic variants) which van Driem describes as marking newly acquired knowledge on verbs denoting state or condition. He states that -mas typically patterns with control verbs when the agent is third person, and with first person when the speaker experiences feelings, sensations, and thoughts. It is occasionally used with second person in instances of direct observation. In this section, I explore the distribution of the particles in narrative text.

392

Stephen A. Watters

62.1. Immediate perception and removed information The appearance of -mas at the beginning of most legends (e.g. dang phu lung phyogs cig nang rgyalp cig dang rgyalm cig yodp mas. 'Long ago in a particular place there was a king and queen.') suggests that the function of -mas as an experienced-perception marker is extended in certain contexts. Oddly enough, in Dzongkha, -mas appears to be able to function both as an experienced-perception marker, as well as a marker of an event which is removed from the speaker, and which is not yet fully assimilated as general knowledge. For example in sentence (5), -mas functions as an experienced-perception marker, grounded in the "here" and "now" of the speech act. But in sentence (6) -mas functions as a marker of fact which is not directly observable, which is new, and which is known to be true by the speaker. (5)

khyod ngi gi bolokha sdod-pmas you my side sit-p.DIR.OBS 'You are sitting next to me (and not somewhere else) (flirting).'

(6)

alo men khang nang skye- w mas child hospital in born-w.INFER 'The child was born in a hospital (and not somewhere else).'

It is clear from my gloss, that there is an element of new information built into the function of -mas in sentences (5) and (6). In sentence (5), the speaker is somewhat surprised, and is maybe even flirting. In sentence (6), the function of -mas clearly extends beyond an experienced perception the speaker hasn't been to the hospital, or seen the baby, but the speaker is still certain of the event. The information is new, but not integrated into general knowledge - the child was just born recently. If the speaker wanted to convey that she had observed the birth of the child, or seen the baby, she would say sentence (7). (7)

alo men khang nang skye-yi child hospital in born-PAST.CONJ "The child was born in a hospital. (I was there when it happened.)'

In sentence (6), the speaker conveys that she knows the child to have been born, but she was not directly involved in the event. In sentence (7),

The nature of narrative text in Dzongkha

393

the speaker was present at the birth. Thus, the function of -mas as an expression of perception is extended in certain contexts to include an event which was not observed, but which is still new (not fully assimilated knowledge). I suggest it is this function which is being used in a legend narrative when the narrator begins with 'Long long ago ... yodp mas'. This may not be unlike the inferred marker -mis in Turkish which takes on certain pragmatic functions such as the mirative function in narrative texts, as well as irony and satire (Aksu-Κος and Slobin 1986). In Dzongkha, the perception function is extended to include instances where little is known about the inception of the event, but the result of an event is known. This includes observable phenomena, as well as events somewhat removed from the speaker. This is like a legend of the distant past - we were not there to observe what happened, but we know it to be true. However, the speaker doesn't go so far as to suggest that this is fully assimilated general knowledge, as would be encoded in the present-tense ending -p in. As such, the verbal particle -mas can carry a range of functions, even somewhat disparate functions (direct perception and an event removed in time and space from the speaker), which are governed by the pragmatics of the speech act. What both functions have in common, though, is newly acquired knowledge of the event after the fact - whether facts which are newly acquired through direct observation or facts newly acquired through another kind of source. The appearance of -mas in offline discourse and not in online discourse, then, is not surprising and is consistent with our understanding of mirativity, as a distinction for information which is new and not yet assimilated in the speaker's world. 6.2.2. Surprise and contra-expectation In addition to -mas, it is the verbal suffix -has which marks new information on stative and general condition verbs. It is also employed as a suffix on the equative copula -in as in-has. These two copulas can get employed in the tense-aspect system. In the narrative texts studied for this paper, inbas occurs four times in "online" discourse and four times in "offline" discourse. For example, it occurs in the last two sentences of 'dre'i rkyal hag 'the leather-bag ghost' which on the surface is a narrative text, but which is underlying an explanatory text. The reader gets to the end of the story and discovers why it is that humans are plagued with sickness:

394

Stephen A. Watters

(8)

da res nga bcas ra smi tsu na ni dang today we REF people PL sick INF and tsa ni yod mi de nad-kyi sgye'u fever INF EXIS NOM that sickness-GEN leather bag de log sti 'dre-gis thob ni 'di-gis ain that return CON ghost-AG get INF this-AG COP zer slab in-pas3 QUO say COP-MIR 'These days, that there is sickness and fever for people, it is because the leather bag of sickness returned, and the ghost got it, saying, (they) say (or it is said).'

(9)

gal siid nad-kyi sgye'u de 'dre-gis log in the case sickness-GEN leather bag that ghost-AG return ma thob pa ein nga bcas smi-lu NEG get NOM COND we people-DAT n a tsa rtsa las mi 'byung ni in-pas sickness never NEG happen INF COP-MIR 'In the case that the ghost had not got the leather bag of sickness, sickness would never have happened to us people.'

In sentence (8), slab in-pas carries the sense of 'it is said' or 'people say that ...' The narrator isn't entirely convinced that what is said about the matter is true. Similarly in the sentence (9), the narrator isn't convinced that the leather bag of sickness is the cause of people's illness. In this way, then, -pas is acting less as an evidential than as a mirative particle - marking information which is new, but more to the point, somewhat dubious in it's truth content. A further example of in-pas is that found in ο la 'than 'bag mi 'The crow who brought a message' where the verbs of speech previous and after end in -nug, but sandwiched in the middle is the sentence slab slab in-pas 'had said (it seems)'. In this sentence the parents of the girl agree to give their daughter in marriage to a demon, and are reported as saying so with slab slab in-pas. It sounds somewhat like centra-expectation - a degree of disbelief that the parents could have been so stupid as to give their daughter in marriage to a demon. Again, the particle -pas would appear to convey a certain amount of disbelief on the part of the speaker, and not really a statement of the distance of the source of information. Understanding the

The nature of narrative text in Dzongkha

395

particle -pas to be a mirative particle also helps explain its ability to appear equally inside and outside of quote material. It is tied less to person agreement and data source and more to the relative novelty to the speaker (and/ or narrator). To summarize, then, the prototypical conjunct/ disjunct complex found in some Tibetan languages is not found in Dzongkha in Dasho Sherub Thaye's legend narrative texts: in online discourse conjunct copulas and verb affixes can be co-referenced to both 1st (conjunct) and 2nd (disjunct) person pronouns; in offline discourse 3rd person pronoun (disjunct) can be co-referenced to both conjunct and disjunct verbal particles. The function of -mas as an immediate-perception marker can be extended to contexts removed in time and space. The verbal particle -pas is the only particle which appears equally in both "online" and "offline" discourse. Both particles function to encode the relative novelty of the information, including events which were not expected, and came as a surprise to the narrator. What appears to happen in these narrative texts, then, is that the perspective of the omniscient narrator (the story teller who is first-person with respect to the outside narrative) is leaking through to the participant characters in whom he is vesting expression of his own mental attitude. He encodes surprise, new and old information, and inferred information, but not necessarily in ways which are syntactically tied to the viewpoint of the participants of the story, but rather to his own omniscient viewpoint. 7. Conclusion In summary, then, in Dzongkha the distribution of connectives, place and person deixes, and time deixes in verbal affixes parallel one another in encoding the distal time of a legend narrative (offline), and a decided shift in point of view in direct speech (online discourse). This distribution is what we might expect of legend narratives: offline is 'then/there/they', and online is 'here/now/you/me'. In Dzongkha, the perspective of the omniscient narrator leaks through to the participant characters in whom he is vesting expression of his own mental attitude. This has unexpected results with regard to the typical patterning of the conjunct/ disjunct complex. In "online" material (that part of the discourse that falls within direct quotes), second person combines unexpectedly with first person. Both are marked conjunct. Thus, though (or because) second person is conclusively neither proximate nor distal, it patterns with first person in proximate contexts. In distal contexts, third person

396

Stephen A. Watters

patterns with both conjunct and disjunct. In addition, the particle -has which marks mirativity appears equally in both proximate and distal contexts. The particle -mas occurs only in distal contexts, even though in some conversational contexts it is used as a marker of direct perception. This patterning points to the fact that the evidential and mirative systems in Dzongkha are controlled not only by the semantics of the verb and syntactic patterns, but are controlled as much by pragmatics as anything else. It remains to be seen how much this is true for other Tibetan dialects, and whether or not there is a cline in the pragmatic/ syntactic nature of the evidential and mirative systems that distinguishes the use of these systems for each dialect. Notes 1.

2.

3.

Dzongkha is a southern Tibetan speech variety, and as such fits in the South Bodish Unit of the Bodish Branch of Shafer's (1955) classification. It is spoken as a mother-tongue by some one-hundred thousand speakers in Bhutan, as well as a few communities in West Bengal, India. There are relations which are not encoded by any kind of connective, as well. These are semantic relations which are implicit from the context (i.e., motivation). The relations given in table 1 are only those which are joined by a connective particle. Other relations are not included in the table. In speech this is pronounced as [Jme], and as such is often written as in mas.

References Aksu-Ko9, Ayhan, and Dan Slobin 1986 A psychological account of the developments and use of evidentials in Turkish. In Evidentiality: The Linguistic Coding of Epistemology, Wallace Chafe, and Johanna Nichols (eds.), 159-167. Norwood, NJ: Ablex. Dasho Sherub Thaye 1987 Rdzong kha'i gtam rgyud sna tshogs dpe cha gsum pa [Dzongkha Short Stories Book Three]. 2nd ed. Thimpu: Kangwa Jangsa. DeLancey, Scott 1992 The historical status of the conjunct/ disjunct pattern in Tibeto-Burman. Ac/a Linguistica Hafniensia 25: 39-62.

The nature of narrative text in Dzongkha

397

van Driem, George, with the collaboration of Karma Tshering of Gaselo 1998 Dzongkha. (Languages of the Greater Himalayan Region 1.) Leiden: Research School CNWS. Givon, Talmy 1998 The Usual Suspects: The Grammar of Perspective in Narrative Fiction. (Institute of Cognitive & Decision Sciences, Technical Report 98-06.) Eugene: University of Oregon. Halliday, Michael, and Ruqaiya Hasan 1976 Cohesion in English. ("Introduction" and "The analysis of cohesion" 1-30, 329-355.) (English Language Series 9.) London: Longman. Häsler, Katrin 2001 An empathy-based approach to the description of the verb system of the Dege dialect of Tibetan. Linguistics of the Tibeto-Burman Area 24 (1): 1-34. Kuno, Susumu 1987 Functional Syntax: Anaphora, Discourse and Empathy. Chicago/ London: The University of Chicago Press. Labov, William 1972 Language in the Inner City: Studies in the Black English Vernacular. ("The transformation of experience in narrative syntax" 1: 354—370, 2: 370-396.) Philadelphia: University of Pennsylvania Press. Longacre, Robert E. 1983 The Grammar of Discourse. ("Combinations of predications" 77149.) (Topics in Language and Linguistics.) New York: Plenum Press. Mann, William, and Sandra A. Thompson 1988 Rhetorical structure theory: Toward a functional theory of text organization. Text 8 (3): 243-281. Rudolph, Elisabeth 1988 Connective relations - connective expressions - connective structures. In Text and Discourse Constitution: Empirical Aspects, Theoretical Approaches, Jänos S. Petöfi (ed.), 97-133. Berlin/New York: de Gruyter. Shafer, Robert 1955 Classification of the Sino-Tibetan languages. Word 11(1): 94-111. Zeisler, Bettina 2000 Narrative conventions in Tibetan languages: The issue of mirativity. Linguistics of the Tibeto-Burman Area 23 (2): 39-77.

Sentence patterns and pattern variation in Ladakhi: A field report1 Bettina Zeisler

1. Starting point: Tibetan languages as ergative languages Ergative languages seem to be devoid of the syntactic categories Subject and Object because the semantic or logical "subject" of a transitive sentence is treated differently from that of an intransitive sentence, which is treated like the semantic "object" of a transitive sentence. The transitive "subject" receives an oblique case marker (or marker of crossreference), while the "object" and the intransitive "subject" remain unmarked (Absolutive). Following this definition, Tibetan languages are generally described as ergative languages (with temporally or pragmatically conditioned split patterns), because the (mostly transitive) EFFECTING2 AGENT receives the same case marker as the INSTRUMENT, while the (mostly intransitive) NONEFFECTING AGENT or UNDERGOER and the PATIENT are in the Absolutive. Grammars or descriptive studies of Tibetan varieties usually treat the function of the case markers in detail, but a systematic classification of sentence patterns is quite rare. Even the three classifications given below present only between four and six different sentence patterns, leaving aside most sentence patterns with arguments other than "subject" or "object", particularly sentences with a valency higher than 2. On the other hand, these classifications are also meant to account for the semantic distinction of [±control] verbs: A [+control] verb presupposes an animate and intentional AGENT, which not only intends the action and its result but controls the action up to the point of transformation, a [-control] verb does not presuppose, it may even preclude such an AGENT. However, with the exception of Ladakhi and Balti, this distinction does not affect the basic case marking. The [±control] distinction thus leads to an unnecessary multiplication of sentence patterns (these doublets will be represented with shades in the following tables).

400

Bettina Zeisler

Table 1. Five basic sentence patterns LT/CT (Toumadre 1996: 214)3 valence

±volitif

type_

[cas] Abs (Erg +) Abs [rather: Abs (+ Instr)] Obi + Abs Abs + Obi Erg + Abs Erg + Obi Erg + Abs

2 benefactif 2 affectif 2 ergatif 2 + mixte jL + ergatif Obi: Oblique = Dative/Locative

Table 2. Five (CT) or six (WT) basic sentence patterns (Zeisler 1999)4 "subject"

intransitive):

Abs

intr.: trans(itive) affection: trans, experience: trans, directional: trans. 2 arguments: trans. > 2 arguments:

Abs/(Erg) Abs » Γ"Γ WT Instr Aeswi Erg Erg/(Abs) Erg

goal

other

"object"

(Instr) D/L / Instr Abs D/L D/L

(Instr) (Instr) (Instr)

Abs Abs

verb

(auxiliary)

[-ctr]

l-*3

[+ctr] [-ctr] r 1 [-ctr] [+ctr] [+ctr] [-t-ctr]

1/3 1 -*3 l-*3 1/3 1/3 1/3

Table 3. Four basic sentence patterns EAT/CtrT (Haller 2006: 65-67) key basic pattern additional arguments ncA V (patient^,,^,) (D, Com, E) ncDA V (benefactive/recipientja^,,, patient.^^,.) prohibitive not possible ncEA V (experiences^, patient,^^) (D,Com,E) prohibitive possible cA V (agentabsolutive) (D, Com, E) cED V (agentergative, patient^,) cEA V (agentereative. patientabsointive) (D, Com, E) nc: [-ctr], c: [+ctr]; A: absolutive, D: dative, E: ergative (including instrumental), V: verb

Tournadre and Haller are probably on safer ground in avoiding a classification in terms of transitivity, but they do not explicitly state that a valency of 2 does not necessarily correspond to a transitive sentence. Sentences with an ergative first argument and a dative(Tlocative) second argu-

Sentence patterns and pattern variation in Ladakhi

401

ment ("mixte", "transitive directional", "cED") as well as sentences with an absolutive first argument and a dative(/locative) second argument ("affectif', "transitive affection") are treated on the same level as transitive sentences of the prototypical ergative pattern. The classification thus does not differ substantially from that of Zeisler (1999). However, an alternative classification of the affective and directional patterns as extended intransitives in the sense of Dixon (1995: 122-123) might be more appropriate.5 Tournadre's seemingly ambivalent "1/2" verbs are basically [-control] verbs with a valency of 1, which do not admit an AGENT argument. It is, nevertheless, possible to add an additional, non-obligatory CAUSE argument that takes the Instrumental. If one allows for a rise in valency in this case, one should also treat movement verbs as ambivalent verbs, since they allow an optional GOAL or SOURCE argument (and sometimes both arguments appear, leading thus to a valency of 1/2/3). Most verbs would allow for an additional INSTRUMENT, CAUSE, or MEDIUM argument in the Instrumental as shown in the above charts from Zeisler (1999) and Haller (2006) and for additional LOCATION, BENEFICIARY, or PURPOSE arguments. It thus seems to be preferable to restrict the indication of valency to (obligatory or optional) core arguments, i.e. arguments that are licensed by the semantics of the particular verb (e.g. GOAL and SOURCE for movement verbs). Other optional arguments will be discussed in this paper only in so far as they exhibit an unpredictable case marking.

2. A closer look at Ladakhi sentence patterns It turned out that none of the above classifications is sufficient and that the description of Ladakhi as an ergative language needs some qualification. This finding holds also for Tibetan in general, although Ladakhi exhibits some peculiarities that are probably due to intensive language contact with Indo-Iranian languages, e.g. the experiencer construction with the Aesthetive (dative/locative case marking).

2.1. Weak Subjects Ladakhi shows evidence of a weak syntactic category of Subject. Each sentence provides a subject slot, which must be filled explicitly or implicitly. In the absence of an (explicit or implicit) animate EFFECTING AGENT or EXPERIENCER, the INSTRUMENT (usually taking the Comitative), will be

402

Bettina Zeisler

necessarily shifted to the EFFECTING AGENT or EXPERIENCER position, taking thus the Ergative (1) or Aesthetive (2). In contrast to accusative languages, the argument in the subject slot can take one of three case markers: Ergative, Aesthetive, or Absolutive. (1) a. rdiujma-s SAS beam-Erg

INSTRUMENT -»EFFECTING AGENT

thokpo-φ roof-df-Abs

khurenok. carry-Pr.GInf

PATIENT

Subject Non-Subject "The beam carries the roof.' b. rdiujma-nag thokpo-φ rtsikpa-s beam-Corn roof-df-Abs wall-Erg

khurenok. carry-Pr.GInf

INSTRUMENT

INSTRUMENT PATIENT

-»EFFECTING AGENT

Non-Subject Non-Subject Subject 'With (the help of) a beam, the wall carries the roof.' Note the little dialogue with the Saspol informant: B.Z.: Is it possible to say: c. rdurjma-natj thokpo-φ khurenok. beam-Corn roof-df-Abs carry-Pr.GInf INSTRUMENT PATIENT Non-Subject Non-Subject ? 'The roof is carried with (the help of) a beam.' SAS (spontaneously): su-s? who-Erg

EFFECTING AGENT Subject 'By whom?'

Sentence patterns and pattern variation in Ladakhi

(2)

thog-a

DOM roof-Aes

trak-φ

theakcasminuk.

truck-Abs

be.able.to.carry-Ng.Ft

403

INSTRUMENT PATIENT -»EXPERIENCER Subject Non-Subject 'The roof won't be able to carry the truck'. (Someone wanted to put a truck on the roof.)

Despite this apparent plurality of case marking in Tibetan, there are some syntactic restrictions for the subject slot: (a) The case marker for the subject slot (as well as for a few salient arguments, such as RECIPIENT and TARGET) cannot be substituted by a synonymous case marker or postposition, in contrast to case markers for less salient arguments (e.g. INSTRUMENT, BENEFICIARY, LOCATION, or SOURCE), (b) Only the subject slot allows pragmatic case neutralisations besides contrastive marking.6 (c) The subject slot corresponds to the discourse topic or theme in neutral word order, (d) Shifts in position of the Subject depend on pragmatic motivations and are informationally marked in contrast to other arguments, which exchange their positions more freely. Together with other semantic-pragmatic criteria such as NP-accessibility, saliency hierarchy, etc., these features constitute a weak syntactic Subject. At least, the so-called "semantic" roles AGENT and PATIENT, introduced for the ergative languages, cannot be defined by purely semantic criteria; they also have syntactic properties. Shamskat Ladakhi additionally shows a functional narrowing of the ancient ergative/instrumental case marker, which is no longer used for a typical INSTRUMENT, MEDIUM, or CAUSE argument, but only for the EFFECTING AGENT argument. The Instrumental as a peripheral argument marker is replaced by the Comitative, except for some rather marginal and obviously frozen patterns. In the Kenhat dialects, the EFFECTING AGENT argument is in the Genitive, and the Genitive is also found in the frozen patterns, but the INSTRUMENT is typically in the Comitative. The contrast of Subject and Non-Subject is thus quite evident in both Ladakhi varieties, while it is blurred by the formal identity of the case markers for the EFFECTING AGENT and the INSTRUMENT in other Tibetan languages. It might be a promising approach to classify case markers according to their function, rather than their form when formal identity goes along with different syntactic behaviour for different functions. I will thus distinguish case markers according to their subject function ("Ergative", "Aesthetive") and non-subject function ("Instrumental", "Genitive", "Dative/Locative").7

404

Bettina Zeisler

2.2. An unexpected multitude of basic and marginal sentence patterns The small set of standard sentence patterns, as used for the typological classification, usually accounts for less than half of the available verb frames of a given language. Due to my experiences with Ladakhi, I am not quite convinced that this is sufficient for the understanding of a particular language and that one can leave the greater "rest" of the more semantically motivated cases and prepositional or postpositional phrases for the lexical entries. In the Domkhar dialect, verbs or readings with only one prototypical sentence pattern (01 or 08) make up merely 35% of all verbs and readings (41% with the ditransitive pattern 09), but many verbs allow pattern variation so that the respective patterns occur in 40% (52%) of all (1798) attested patterns.8 There are eight additional main sentence patterns in Tibetan and Ladakhi, each of which is shared by several verbs (03-07, 10, 11) and/ or occurs with a certain frequency (02, 03, 06,07). Patterns 03 and 07 have been mentioned in section 1. Pattern 06 is restricted to the expression of possession and some related verbs in most Tibetan varieties, but has been generalised in Ladakhi for all "transitive" [-control] verbs of perception, cognition, possession, change of possession, and modality. The position of the comitative noun in pattern 11 is crucial. At the third position, it expresses a co-"object", but if the Comitative is shifted one position to the left, the second and third position collapse to a collective PATIENT. At the second position, the comitative argument expresses a co"subject", but if the Comitative is shifted one position to the left, the first and second position collapse to a collective AGENT.

Table 4. Examples of clause types - main patterns 1-place predicates 01 LLV

02 SAS

change, movement kho-0 sog. s/he-Abs went-Pa 'S/he went.'

| Abs

2-place predicates predication, transformation Abs kho-0 gergan-φ in I gyur. s/he-Abs teacher-Abs is-Pr / become-Pa 'S/he is / became a teacher.'

| Abs

Sentence patterns and pattern variation in Ladakhi

03 TTR GYS

Abs ~Loc affection, goal oriented movement (coming into) existence ~Loc Abs kho-0 kuSu-a thadet. s/he-Abs apple-DatLoc like-Pr 'S/he likes apples/ the apple.' (βογ-e tshe)-a mentok-φ khyapteduk. (we.incl-Gen garden)-DatLoc flower-Abs be.full-Perf Our garden is full of flowers.'

04 ARA

get out, through, move away Abs Abl (di baspo)-0 la-na khelemiaq. (this bus-df)-Abs pass-Abl be.able.to.cross-Ng.Perf 'This bus could not cross the pass.'

05 SAS

contact, separation, atypical fill [-ctr] Abs Com" Tserig-0 Αηιηο-naq igyaqsok. Tsering-Abs Angmo-Com distance-Pa.Inf 'Tsering must have distanced himself from (lit, with) Angmo.' experience, possession, modality Acs Abs khi-a bila-φ thoqse... dog-Aes cat-Abs see-cc "The dog saw a cat and ...' directional activity Erg ~Loc aba-s coktse-a rdiujs. father-Erg table-DatLoc beat-Pa 'Father beat on the table.' non-directional activity, transformation Erg Abs khoze (di s/7. $rui)s-e kitap)-0 s/he-Erg (this story-Gen read-Pa book)- Abs 'S/he read this story book.' kho-ze ralukSik-0 sat. s/he-Erg goat.sheep-lq-Abs kill-Pa 'S/he killed some goats and sheep.'

06 SAS 07 SAS

08

ra TIR

09 SAS SAS

405

-

3-place predicates bring-type, deposit10 Erg ~Loc Abs transformation (zgyur), division Erg Abs ~Loc gergan-is thrugugun-la IqliS-0 fabanok. teacher-Erg child-pl-DatLoc English-Abs teach-Pr.GInf 'The teacher teaches the children English.' kho-s (luk tshaqma)-0 pene-a zgyurs. s/he-Erg (sheep all)-Abs money-DatLoc transform-Pa 'S/he made all the sheep into money.'

406 10 TIR

11 DOM

GYS

Bettina Zeisler take away-type Erg tAbs Αοΐφ t npo-ze miun-ikana watjmo-0 kokstekher. minister-Erg man-pl-PPosAbl power-df-Abs deprive. away-Pa 'The minister(s) deprived the men of their power.' connection, mixture, atypical fill [+ctr] Erg Com9 Abs compete with Abs Erg Com kho-s samar-φ chu-naq Sresok. s/he-Erg kerosene-Abs water-Corn mix.Palnf 'S/he mixed kerosene with water.' (R igzin Palbar)-e (Rigzin Jom)-rag tshanrik-0 thrap. (Rigzin Palbar)-Erg (Rigzin Jora)-Com science-Abs compete-Pa 'Rigzin Palbar competed with Rigzin Jora in science.'

There are thus as many as eleven main sentence patterns in Classical Tibetan and Ladakhi alike. In Domkhar Ladakhi they occur in 89% of all frames. The remaining 11% consist of about 40 additional marginal patterns. Some of them allow the same case to appear in several obligatory positions of three-place predicates, e.g. two or three times Absolutive (20, 21, 30) or two times Dative/Locative (25), something one would not expect on the basis of Classical Tibetan where only the instrumental case shows up twice in a sentence: as a Subject marker (Ergative) and for optional INSTRUMENT arguments (Instrumental). Double Absolutive is restricted to two-place predicates of predication sentences (02), case neutralisation (pragmatic ergative split, see note 6), and may perhaps be found in some derived verb constructions. Otherwise, doubling of cases for obligatory arguments seems to be avoided, something that cannot be said for Ladakhi. Of special interest are patterns 12-14, where the sole argument of a oneplace predicate receives an oblique case (Ablative, Aesthetive, and Ergative). Pattern 12 is attested so far only for a single impersonal expression, whereas pattern 13 is quite common as an alternative for the regular pattern 01 of (mental) state verbs, possibly indicating a greater affectedness. Pattern 14 again violates the definition of ergativity. It appears with the verbs muk 'bark' and bos 'howl' of dogs, wolves, and foxes, perhaps because these sounds are perceived as acts of communication or as affecting those who happen to perceive them.11 Pattern 14 does not seem to be specific to Ladakhi, cf. LT khyi: ch qgi: Ά dog-Erg is barking' (Goldstein and Narkyid 1984 sub 'bark') and LT kh : qwsu 'He-Erg cried' (Goldstein and Narkyid 1984 sub 'cry'). Communication verbs typically have a RECIPIENT argument (the addressee) as well as a CONTENT argument (the preceding or following speech). The CONTENT argument might even be filled with the

Sentence patterns and pattern variation in Ladakhi

407

help of afigura etymologica. Rather infrequently, these verbs appear without any of these arguments as descriptions of intransitive activities. In such cases, the (still) EFFECTING AGENT remains in the Ergative. Thus one could think of 'barking' and 'howling' as communication verbs stripped of their RECIPIENT and CONTENT arguments. They would then correspond to the second meaning of /ia, 1. 'look at', 2. 'have a look', which is likewise stripped of the TARGET argument, see (13c), section 3.2.1. 'Barking' and 'having a look' could perhaps be understood as intransitive derivations of basically transitive verbs, preserving the original case marking of the remaining argument. Equally surprising is the use of the Genitive besides the expected Absolutive or Comitative (CT Instrumental) for the MEDIUM of [±ctr] 'fill' verbs (15 and 16 for [-ctr], 32 and 33 for [-»-ctr]) in Shamskat. The pattern is attested in principle for all inagentive and causative 'fill' verbs and verbs with related semantics, such as skor 'spread, scatter around' and tram 'distribute, spread'. No other Shamskat verb shows genitive case marking for any of its arguments, and the MEDIUM argument of other verbs, such as 'suffice' and 'repay' is always in the Instrumental.12 The only explanation to me is that the construction must have been borrowed for the 'fill' verbs from the Kenhat dialects, where the distinction between Genitive and Instrumental is radically neutralised (if it were ever existent). Still, the elliptical combination of LOCATION and MEDIUM argument (patterns 16 and 33), signifying that every single spot of the particular CONTAINER has been filled, seems to be somewhat surprising and is not accepted by all informants equally for all verbs.13 Generally, one can observe an increase in dialectal or individual variation with the increase of marginality of the situation described. At the extreme end, expressions of exchange or barter cannot be said to show any pattern at all. This is due to the fact that the linear order of semantically symmetric arguments undergoing or performing the exchange is arbitrary. Symmetric arguments are thus preferentially represented by collective expressions (cf. section 3.3.2 below). Table 5.

Marginal patterns

12

1-place predicates impersonal expression for have pain | Abl

DOM

(η-i

trotp)-eaqna

zerekhyongenak.

(I-Gen stomach)-PPosAbl ache.bring-Pr.Aud '(It) aches/ burps out of my stomach.'

408

Bettina Zeisler states (with strong affection) Aes (rinboche kundun)-a jaluga? (chief .priest title)-Aes be.able.to.see.hon-Pr.Q 'Is the chief priest able to see?'

-

-

14

some animal sounds, non-focussing use of sense organ (cf. example 13c)

-

-

TIR

khi-ze

13 GYS

Erg

muak.

dog-Erg bark-Pr.Aud Ά dog is barking.' 2-place predicates Instr [-ctr] events and cause, media, etc. Abs fiU (with) [-ctr] Abs Gen DOM kho-0 jigri-s Srante dar. s/he-Abs fear-Instr very tremble-Pa 'S/he trembled badly with/ because of fear.' DOM dziq-0 chu-i gaqseduk. pond-Abs water-Gen fill-Perf "The pond filled with water.' 16 fill into (with) [-ctr] ~Loc Gen DOM daruqaq zaqzb-iaq chu-i maqgaqsok. still pot-PPosLoc water-Gen fill-Ng.Pa.Inf '[Every single space] inside the pot is not yet filled with water.'

15

20 SAS

21 SAS

25 TIR SAS

-

3-place predicates reflexive transformation (co 'do, make') Abs Abs Abs kho-0 khoraq-naq-khoraq-0 igyapo-φ cos. s/he-Abs s/he. self-Corn- s/he.self-Abs king-Abs make-Pa 'He made himself king.' reflexive transformation (zgyw) \ Abs | Abs | ~Loc kho-0 khoraq-naq-khoraq-0 rgyapo-a zgyurs. s/he-Abs s/he.self-Com-s/he.self-Abs king-DatLoc transform-Pa 'He transformed himself into a king.' Abs | ~Loc | ~Loc labour-force exchange skortsana, kh -φ khuis-0 skortsana, khoq-0 khuyu-0 go.round-sim threshing -Abs they-Abs balaq-iphia reaq. khimtsespaun-la khimtsespagun-Ia balaq-a reruk. neighbour-pl-DatLoc cow-PPos / -DatLoc depend-Pr 'At the time of threshing, they depend on their neighbours for the (exchange of) cows.'

Sentence patterns and pattern variation in Ladakhi 30 TTR

32 33 09 DOM

SAS

34 SAS

08 09 lla lib 10 40 44 43 SAS

SAS

409

transformation (co 'do, make') Erg Abs Abs gyapo-ze kho-0 yokpo-0 cos. king-Erg s/he-Abs servant-Abs make-Pa 'The king made him (his) servant.' fill with [+ctr] Erg Abs Gen fill (into) with [+ctr] Erg ~Loc Gen Erg -Loci tAbs fill into churpon-is rdzigpo-0 ~ rdzigp-iag chu-i skileduk. water.master-Erg pond-df-Abs ~-PPosLoc water-Gen fill-Pr 'The overseer for the water filled (every single space inside) the pond with water.' ga-s (zagsbu tshagm)-inaga chu-0 ~ -i skagspin. I-Erg (pot all)-PPosLoc water-Abs ~-Gen fill-Pa Ί filled water into all the pots / [the] inside [of] all the pots with water.' protect, chase away -Locj: | jAbl Erg Srugmagun-is pagdzotpo-a Skunm-ekana Srugs. guardian-pl-Erg treasure-df-DatLoc thief-PPosAbl protect-Pa 'The guardians protected the treasure from thieves.' exchange, barter Erg (collective) Abs (collective) -Loci Erg (collective) iAbs Erg (collective) Abs Com Erg Com Abs (collective) Erg Abl Abs (collective) Erg Abl -Loci iAbs Erg Com -Loci tAbs Erg Com Abs Com ($ama-nag Cagpa)-s (Lowlander-Com Northener)-Erg (culi-nag bal)-0 rjepsl zdeps. (apricot-Com wool)-Abs barter-Pa (08) culi-a bal-0 rjeps/ zdeps. apricot-DatLoc wool-Abs barter-Pa (09) 'The Sham people (Lowlanders) and Changthang people (Northerners) bartered apricots and wool / wool for apricots.' Sama-s Cagpa-na ~ -ekana Lowlander-Erg Northerner-Abl ~ PPosAbl (culi-nag bal)-0 zdeps. (apricot-Com wool)-Abs barter-Pa (10) culi-a bal-0 zdeps. apricot-DatLoc wool-Abs barter-Pa (40) "The Sham people bartered with (lit. from) the Changthang people apricots and wool / wool for apricots.'

410 TIR

Bettina Zeisler Cagtaqpa-ze Sama-nampo Changthang.people-Erg Lowlander-PPosCom bal-0 naz-la ~ -n rjeb. wool-Abs barley-DatLoc - -Com barter-Pa (44,43) "The Changthang people bartered with the Sham people wool for ~ with barley.'

Another surprise was the fact that additional INSTRUMENT arguments of [±control] sense perception may appear in the old Instrumental instead of the expected Comitative (the latter is attested only in the LLV and possibly in Domkhar14). The Ablative, however, seems to be more common, and not all perception verbs allow the Instrumental. Except for the instrumental arguments, the sentence patterns are not restricted to perception verbs and additional arguments, but may be found with other verbs and obligatory or optional core arguments: pattern 18 is attested so far only for the Classical Tibetan verb rtsom 'treat of, pattern 19 only for the Ladakhi verb kan 'press (while writing)', pattern 26 is quite common in Ladakhi, especially for experiencer derivations of patterns 03 and 06 (cf. section 3.3.5), pattern 27 is found with some verbs of [-ctr] acquisition. Note that the obligatory first argument is typically omitted in imperative constructions. Table 6. Exotic additional arguments

18 19 47 S AS

LLV

26 27 50 SAS

SAS

use of sense organ, non-focussing

Erg Erg

Abl Com Instr

-

Abl

Abs Abs

(Erg) mig-is ~ -na itos! lce-na ffog! eye-Instr ~ -Abl look-Imp tongue-Abl taste-Imp 'Look by ~ from (the perspective of) your eyes! Taste from (the perspective of) your tongue!' (lce-mo-φ dar-gyi mdud-pa)-daq ma-myaq-ba (tongue-Abs silk-Gen knot)-Com taste-Ng.cc 'Without taking a taste with the silk knot of the tongue ...' sense perception through sense organ Acs ~Loc Abs

Aes Aes

Instr

kho-a lc-eka qarmo-0 tshor. s/he-Aes tongue-PPosLoc sweet-Abs perceive-Pa 'S/he perceived the sweet(ness) on the tongue.' kho-a mig-is ~ -na micik-φ thoqanuk. s/he-Aes eye-Instr ~-Abl man-lq-Abs see-Pr.GInf 'S/he would see a man by ~ from (the perspective of) (his/her [own]) eyes.'

Sentence patterns and pattern variation in Ladakhi

411

3. Variation of sentence patterns Variation of sentence patterns, particularly variation of focus and telicity (the spray/load type, see section 3.1.3 below) have been discussed in detail under various labels for Indo-European languages, but so far this phenomenon is rarely discussed for Tibetan languages, except for case neutralisations conditioned by discourse pragmatics (see note 6). Skalbzaq Hgyurmed (1981: 46, 1992: 26-27) further mentions some semantically conditioned variations between patterns 07 and 08. Besides pragmatic case neutralisations and dialectal variance, Ladakhi shows several types of variation with and without semantic shift. 3.1. Variation without change of basic meaning 3.1.L Variation motivated by (emotional) distance Marked cases, in particular the Ergative might be replaced with the Absolutive in neutral statements. Although the replacement of the Ergative seems to follow a general tendency (cf. Bielmeier 1985: 141-143, also for the replacement of the Genitive), it occurred only from time to time and unpredictably in my interviews, without obvious semantic reason. Most often, the Absolutive was first used by the informants and the marked case was given then as alternative, spontaneously or on request. Sometimes the marked case was used first and then the Absolutive was given as alternative, spontaneously or on request. Quite frequently, however, the informants rejected an alternative with the Absolutive. Typically, the marked cases correspond to the classical pattern. As could also be observed, some speakers would show less variation, at least in the interviews. This kind of variation is much more common in the Kenhat dialects (or at least in Gya-Sasoma). Here, the marked case indicates that the event is temporally or spatially distant, while the Absolutive is preferred when the event takes place in the present or a near past or close to the place of the speaker (and listener). This is especially true for the Ergative-Absolutive alternation, and as the second Domkhar informant admitted, this motivation might play a role for the Shamskat speakers as well, but unlike the GyaSasoma informant, none of the Shamskat informants described or conceptualised it on their own. In some cases, the marked case also indicates a kind of modal distance, namely some surprise or a stronger affectedness. The latter is especially true for the Absolutive-Aesthetive alternations.

412

Bettina Zeisler

(3) butsa-φ ~ -5 bumo-a SAS boy-Abs ~ -Erg girl-DatLoc 'The boy waited for the girl.' (4) (i nanbo)-0 ~ -na SAS (this house-df)-Abs ~ -Abl 'Having passed this house

zguks. wait-Pa

gomse pass-cc

3.7.2. Variation depending on arguments The use of the marked case, here the Comitative, corresponds to the classical pattern with the Instrumental for the MEDIUM of [±ctr] 'fill'. The older pattern was definitely rejected for typical media for the [-ctr] filling of pots, such as water, milk, curd, or grain, but appeared spontaneously with less typical media such as gold, pearls, or turquoise. Cf. also the example from the LLV with the [+ctr] verb and a likewise quite atypical MEDIUM of filling. (5) a. zaqb-in a chu-φ gaqseduk. TIR pot-PPosLoc water-Abs get.filled-Perf 'The pot is filled with water (lit. into the pot water has filled).' b. zagbuo-0 ser-na gaqseduk. pot-df-Abs gold-Corn get.filled-Perf 'The pot is filled with gold.' LLV c. (khyi-phrug de)-s (khyi-rkyag-dag khyi-lcin)-darj (puppy that)-Erg (dog.faeces-Com dog.urine)-Com (nag tshaq-ma)-0 bskaqs- btaqs I (house all)-Abs fill.give-Pa 'The puppy, with faeces and urine, filled up the whole house.'

3.1.3. Inversion, variation of focus, and additional arguments (6) a. kho-0 zdugzqal-na drol. SAS s/he-Abs suffering-Abl get.free-Pa 'S/he succeeded to get free from suffering.'

Sentence patterns and pattern variation in Ladakhi

413

b. kho-a zdugzqal-0 ατοί. s/he-Aes suffering-Abs get.free-Pa 'S/he happened to get free from suffering.' (Informant's comment: the second event seems to be more accidental than the first. - With tsonkhaq 'prison', (6b) could only be used if the prison had collapsed.) (7) a. kho-0 kusu-a tshimseduk. SAS s/he-Abs apple-DatLoc be.satiated-Perf 'S/he has got/ eaten enough apples.' b. kho-a kusu-0 tshimseduk. s/he-Aes apple-Abs be.satiated-Perf 'S/he has got enough (of) apples.' (Informant's comment: very difficult to differentiate, perhaps different importance on eater or apple.) In my opinion, the difference lies in the degree of accidentalness, like in the preceding example. Note, however, that not all informants agreed to the second alternative for this verb. (8) a. garwa-s ser-la kagu-0 zos. SAS smith-Erg gold-DatLoc amulet-Abs produce-Pa "The smith produced an amulet (working) on gold.' b. garwa-s ser-na kagu-0 zos. smith-Erg gold-Abl amulet-Abs produce-Pa 'The smith produced an amulet out of gold.' c. garwa-s ser-φ kagu-a zos. smith-Erg gold-Abs amulet-DatLoc produce-Pa 'The smith shaped the gold into an amulet.' Verbs of the '(un)load' type15 frequently show the following pattern: (9) a. khog-is stakhur-φ poks. SAS they-Erg horse.burden-Abs unload-Pa "They unloaded the horse burden.' b. khoq-is stagun-0 poks. they-Erg horse-pl-Abs unload-Pa 'They unloaded the horses.'

414

Bettina Zeisler

c. khorj-is st-ekana stakhur-0 they-Erg horse-PPosAbl horse.burden-Abs 'They unloaded the burden from the horses.'

poks. unload-Pa

3.1,4. Variation in movement and direction (10) a. kho-s tri-0 rdoa-(:) truts. SAS s/he-Erg knife-Abs stone-DatLoc sharpen-Pa 'S/he sharpened the knife on a stone (moving the knife over a stone).' b. kho-s tri-a sakdar-0 truts. s/he-Erg knife-DatLoc rasp-Abs sharpen-Pa 'S/he sharpened the knife with a rasp (moving the rasp over the knife).' (11) a. chumaun-0 chu-a gor. TIR water.maid-pl-Abs water-DatLoc be.late-Pa 'The water maids were late (to go/ start) for the water.' b. chumaun-la chu-φ gor. water.maid-pl-Aes water-Abs be.late-Pa "The water maids were late (to come back) with the water.'16 3.2. Variation for particular meanings 3.2.7. Polysemy and homonymy (12) a. Arjmo-s Tserirj-a boseduk. SAS Angmo-Erg Tsering-DatLoc call-Perf 'Angmo has called Tsering.' b. Aqmo-s Tserirj-φ dron-la Angmo-Erg Tsering-Abs party-DatLoc 'Angmo has invited Tsering for a party.'

boseduk. call-Perf

With possible contrast of specific and unspecific TARGETS: (13) a. kho-s tiwi-a has. SAS s/he-Erg T.V.-DatLoc look-Pa 'S/he watched (the) T.V.'

Sentence patterns and pattern variation in Ladakhi

415

b. ga-s kitap-0 I Jammu-φ haspin. I-Erg book-Abs / Jammu-Abs look-Pa Ί looked for a book (in the shop) / went sightseeing in Jammu.' c. kho-s has. s/he-Erg look-Pa 'S/he hod α look: 3.2.2. [±control] variation (14) a. ga-(:) dzawo-φ thuk. SAS I-Aes friend-Abs meet-Pa [-ctr] Ί met my friend(s) (by accident).' b. ga-0 kho-nag thukpin. I-Abs s/he-Com meet-Pa [+ctr] Ί met with him/her (by appointment).' 3.2.3. TARGET "objectification" The verb rdurj 'beat, strike, hit (against, on)' expresses an activity directed towards a thing or animate being. Hence, the second argument, the TARGET, is in the Dative/Locative. In some derived verb constructions, such as 'to beat = play a drum' or 'to beat = knock at a door', the TARGET is semantically incorporated and treated as a PATIENT (15b and d). Neither [±animacy] nor [±definiteness] are responsible for the shift in case marking. If the Dative/Locative is applied on 'drum' or 'door', the original independent verb meaning reappears (15c and e). (15) a. zigbatpa-s bugbu-a rdugs. SAS farmer-Erg donkey-DatLoc beat-Pa 'The farmer beat (on) the donkey.' b. (de monbo)-s (draman soma)-0 rdugs. (that mon-df)-Erg (drum new)-Abs beat-Pa "That mon (musician) played a/ the new drum.' c. kho-s (g-e draman)-la rdugs. s/he-Erg (I-Gen drum)-DatLoc beat-Pa 'S/he beat on/ hit against my drum (accidentally or in order to make trouble).'

416

Bettina Zeisler

d. kho-s zgo-0 rdugs. s/he-Erg door-Abs beat-Pa 'S/he knocked at the door (in order to get in).' e. kho-s zgo-a rdugs. s/he-Erg door-DatLoc beat-Pa 'S/he hit against the door (accidentally or in order to make trouble).'

3.3. Variation of valency 3.3.1. Ambitransitives (16) a. kho-φ skyuks. WAK s/he-Abs vomit-Pa 'S/he vomited [-ctr].' b. kho-s (culi magpo)-0 s/he-Erg [!] (apricot many)-Abs 'S/he vomited many apricots [-ctr].'

skyuks. vomit-Pa

5.3.2. Reduction of symmetric arguments Comitative arguments typically reflect a semantically symmetric relation with the subject (contact and exchange) or patient (connection and exchange) argument. The asymmetric case marking, however, shifts one of the arguments into the focus position. The choice depends on pragmatic features. In the neutral context of elicitation, the informants prefer a collective or enumerative expression, such as khog ffiska 'they both' or (17b). Collective expressions are also typical of reciprocal events, and can be used in order to get rid of a second asymmetric argument (bearing a comitative or directional marker). (17) a. Tsenq-0 Agmo-nag rgyags. SAS Tsering-Abs Angmo-Com distance-Pa 'Tsering distanced (himself) from Angmo.' b. (Tseng-nag Αητηο)-φ rgyags. (Tsering-Com Angmo)-Abs distance-Pa 'Tsering and Angmo became distanced (from each other).'

Sentence patterns and pattern variation in Ladakhi

417

3.3.3. Some reflexive actions (18) a. a-ze gustiq-φ trhus. ARA mother-Erg clothes-Abs wash-Pa 'Mother washed the/ her clothes.' b. 773-0 lakpa-0 trhuspin. I-Abs hand-Abs wash-Pa Ί washed my hand.' Reflexivity does not necessarily lead to an absolutive construction, and it is not yet clear when it does and when not. 3.3.4. Argument suppression (possessor constructions) (19) a. ionpo-ze miun-ikana waqmo-φ TIR minister-Erg people-pl-PPosAbl power-Abs koxsekher. snatch.away-Pa "The minister(s) snatched away the power from the people.' b. ionpo-ze (yiilmi tshaqm-e waqmo)-0 minister-Erg (villager all-Gen power)-Abs koxsekher. snatch.away-Pa 'The minister(s) snatched away the villagers' power.' (20) kho-ei~kho-s papu-φ SAS s/he-Gen ~ s/he-Erg woollen.shoe-Abs '(S/he) put on his/her woollen shoes.'

gons. dress-Pa

Again, the informants often prefer the possessor construction, and in some cases, the more explicit construction seems to be blocked: (21) a. zigbatpas-s (lug-i bal)-0 draks. SAS farmer-Erg (sheep-Gen wool)-Abs shear-Pa 'The farmer sheared the wool of the sheep.' b. ziqbatpas-s ??lug-ikana bal-φ draks. farmer-Erg sheep-PPosAbl wool-Abs shear-Pa 'The farmer sheared the wool-Abs ??from the sheep.'

418

Bettina Zeisler

3.3.5. Experiencer derivation Most [-control] verbs allow the addition of an EXPERIENCER argument. The construction expresses that something happened to a person without his or her control, although not necessarily without his or her intention. The experiencer construction can thus be used in place of an expression of ability. The closest equivalent seems to be German gelingen, as used both with a normal and a sarcastic undertone. The construction thus differs from Tournadre's (1996) ambivalent "1/2" verbs mentioned in section 1 above. I would take this construction as a sort of valency or argument raising derivation (in a few cases the valency remains the same), although the derivation is not marked on the verb. The alternative approach, namely to hold that the underlying verbs were ambivalent, faces the problem that there is no syntactic restriction for the derivation. In principle, the EXPERIENCER argument can be added to any verb of any valency. In the case of [+control] verbs, however, the derivation often needs stronger support from the context, and it might be semantically blocked, e.g. by the availability of a beneficiary reading. This is especially the case with verbs of higher valency. (22) a. (rj-i pakspa)-0 gokseyoqenuk. DOM (I-Gen skin)-Abs peel.come-Pr 'My skin is peeling off.' b. kho-a (yay-i pakspa)-0 marjgok. s/he-Aes (yak-Gen skin)-Abs peel-Ng.Pa 'S/he did/ could not get the skin of the yak peel off.' Notes Fieldwork in Ladakh is part of the research project on "Semantic roles, case relations, and cross-clausal reference in Tibetan" within the Special Research Program 441, supported by the DFG (see http://www.sfb441.unituebingen.de/bll/). The purpose of the first field trip to Leh from July to September 2002 was to transcribe recordings of the Kesar epic with the help of Ladakhi informants. However, after only two weeks at 3.500 m above sea level, the electronic notebook showed severe symptoms of high altitude sickness and eventually broke down. I had thus to switch to unrecorded interviews about sentence patterns in Ladakhi, beginning with the transitive verbs of the LLV, a written version of the epic, and then going through the first quarter of Jäschke's ([1881] 1992) CT dictionary. The informants were

Sentence patterns and pattern variation in Ladakhi

2. 3.

419

Phuntsok Dolma from Saspol in eastern Sham, Tundup Wanggyal from Tirit in central Nubra, Tsering Yudon from Aranuk in western Nubra, and Tsering Padma from Wakka in northern Purik. All these dialects can be subsumed under one group, which I would like to call "Shamskat" (dialects of Lower Ladakh). At the time of the first submission, all in all 362 or about 40% of the Ladakhi simple verbs (and their readings) were systematically tested. Due to the quite unexpected results, the following field phases (20032007) were used for further exploration. With the help of the main informant Tshewang Tharchin, as well as Thrinlas Chosphel a (nearly) full set of 797 main entries (with 517 additional readings for 280 of the entries) could be established for the dialect of Domkhar hi western Sham. It also turned out that the "Kenhat" dialects of Upper Ladakh differ quite substantially, not only in their phonetics, but also in their lexicon and grammar, thus the collection of data from Gya-Sasoma started in 2005, the informant being Tshomo Minggyur. The paper focuses on simple verbs. Derived or light verb constructions as well as collocations behave in a quite unpredictable way, but have not yet been tested systematically. The informants were asked not to delete obvious arguments (as they would do in natural speech) and to think of the verbs hi question as stage plays, the actors of which have to be specified. Whenever it appeared promising, the sentence context was varied. In several cases, different types of arguments were tested. Doubtful cases and marginal patterns were discussed with as many informants and speakers as possible. I would like to thank all informants for their patience, additionally also Rebecca Norman (SECMOL, Phe) for all the day- (and night-) long discussions on Ladakhi issues. I would also like to thank the editors for providing an opportunity to present the fruits of our joint efforts to a greater public; I am particularly indebted to Felix Haller for working hi all actualisations that occurred since the first submission. As an approximation for the Tibetan term thadadpa (cf. Zeisler 2006: 65, n. 12 and note 5 below). Abbreviations 1. Tibetan varieties and sources ARA Aranuk, western Nubra CtrT Central Tibetan CT Classical Tibetan DOM Domkhar, western Sham EAT Eastern Amdo Tibetan GYS Gya-Sasoma, Upper Ladakh KHAL Kesar story, recorded 1996 hi Khalatse LEH Central Ladakhi, Leh standard

420

Bettina Zeisler LLV LT RGYA SAS TIR WAK WT

Lower Ladakhl Version of the Kesar Epic (Francke [19051941]1981) "Lhasa" (Central and Exile) Tibetan Das Märchen vom Prinzen Cobzan (Bielmeier 1985) Saspol, eastern Sham Tint, central Nubra Wakka, northern Purik Western Tibetan (Ladakhi and Balti)

2. Grammatical terms Abl Ablative (including postpositions in the tables) Abs Absolutive (zero) Acs Aesthetive (Dative/Locative as Subject case) Aud auditive knowledge cc clause chaining Com Comitative ctr control DatLoc, D/L Dative/Locative df definiteness marker Erg Ergative (Instrumental or Genitive as Subject case) Ft Future (unspecified) Gen Genitive GInf generic or inferred knowledge Imp Imperative incl inclusive plural Inf inferential knowledge Instr Instrumental ~Loc locative case variable (including postpositions in the tables) Iq limiting quantifier: One', 'some' Ng Negation Pa Past (unspecified) Perf Perfect (unspecified) pi Plural PPos Postposition Pr Present (unspecified) Q question final marker sim expression of simultaneity () complex NPs on which a case marker operates as a whole preferred order, change of position possible For the sake of simplicity, morphemes other than case will not be separated from the lexeme and the description will be kept as unspecific as possible.

Sentence patterns and pattern variation in Ladakhi

421

Table 7. Realisation of case markers

Com

CT kyisl gist gyisl -s dag

Gen

•H kyil gil

Instr, Erg

Sham/Punk Nubra -s/ -is~ [s] -zel -tse

Leh -el -i

Upper Indus -e ~ -se

-nag (-fiampo) -el -i

-tag/ -dag/ -rag -na(g) (-ffampo) (-fiampo) -e ~ -se -eiJ -i~[d]

-a ~ (:)/ -la (-ru)

-a ~ (:)/ -la

-a ~ ( :)/ -la

-ne (~ -nesu) (-ru)

(Purik: -na) (-ru) (-ru)

se-Gen + naga, ha -ne ~ -nesu se-Gen + nagne, hane

Gen + Gen + (n)ag(a), ka fn^a"a, ka -na -na ~ -nas Gen + Gen + (n)agna, (n)a(n)a, kana kana

-n {-fiampo) -eil -i

gyj D/L,Aes

la

Loc Loc./ Purposive PPosLoc

na tu! du/ τυΐ -if su Gen + nag + du, na, la nas, las Gen + nag-nas

Abl PPosAbl

4. 5.

Gen + aaga, ka -ne Gen + nagne, kane

-a ~ ( :)/ -la

A revised version with eleven basic patterns is given in Zeisler (2004: 254). Cf. Zeisler (2004: 254; for further discussion of the directional pattern see also Zeisler 2006: 84-87). The modern Tibetan grammar describes the directional pattern as thadadpa 'with difference' referring to the traditional distinction of bdag 'self (EFFECTING AGENT, INSTRUMENT) and gian Other' (PATIENT, TARGET). The term thadadpa is widely taken to mean 'transitive', but according to the Tibetan mainstream, inagentive transitive verbs, such as perceive something, are excluded (Zeisler 2006: 65, n. 12). Nevertheless, since the first argument of a thadadpa verb always takes the ergative marker, ergativity and "difference" or agentive transitivity are obviously defined in circular dependency. Whether seen as "transitive" or "intransitive", the directional pattern violates the standard definition of ergativity. It should be emphasised that the TARGET argument (i.e. the thing or person towards which an activity is directed) cannot be understood as a PATIENT that merely takes a somewhat idiosyncratic case marker. Even in that case, the directional pattern would not be the standard pattern for transitive sentences. LaPolla's claim (1992: 3-4) that Ladakhi and the dialect of Purik show primary object marking and thus generally (Balti only partly) dative/locative marking of the PATIENT in transitive, but absolutive marking in bitransitive sentences is contrary to evidence. The relevant data are already presented in Rangan (1979: 73-75) and Koshal (1979: 65). LaPolla seems to have mistaken Francke's ([1901] 1979: 12) and Bailey's (1920: 5)

422

6.

7.

8.

9. 10.

11.

12.

13.

Bettina Zeisler rather superficial statements that "datives" might be found where they would have expected "accusatives" (i.e. Absolutives). For Balti see Read (1934: 7-8) and Bielmeier (1985: 138-140). See e.g. Tournadre (1991) for the contrastive use of the Ergative for intransitive actors and the de-emphasising use of the Absolutive for transitive actors. Contrastive marking can also be found in the patient slot (cf. Tournadre 1994: 645; Zeisler 2006: 73-80). I will not distinguish between the subject and non-subject functions of the absolutive, for the very reason that it is not a case marker in its own right, but a non-marker. The sign '-0' will be added in the examples merely for the convenience of the reader. It does not represent any entity. According to a counting in 2004 the latter number comes close to the total percentage of the corresponding frames in German (48% out of 17381 frames), but the distribution differs fundamentally: the monovalent intransitive pattern supplies 21% of all frames in Domkhar, but only 8% in German, the bivalent transitive pattern 19% in Domkhar, but 32% in German, and the trivalent pattern of the 'give' type 12% in Domkhar, 8% in German (the German data was derived from GermaNet with the help of Claudia Kunze, GermaNet). Comitative arguments are often replaced by collective expressions, cf. 3.3.2. The locational and the absolutive arguments of 'deposit' verbs may exchange their position freely, the second slot being reserved for the more specific and given argument. The Tirit informant suggested that there could be some reason for barking (e.g. somebody is coming). He also suggested a directional and/ or communication reading: khize (Erg) tnia I khisaa (DatLoc) muak "The dog is barking at a man / the dogs'. Cf. the difference between English 'The dog is barking at me' and 'The cow is mooing *at me' (Sam Featherston, SFB 441, University of Tuebingen). The Domkhar informant would use this pattern only for a concrete event and a specific animal, not for statements about dogs or foxes in general. He would subsume under the verb bos also a particular kind of meowing, which sounds like the crying of a child and might be taken, particularly at night time, for the crying of a ghost: jiks-m jiks! ijaci bilas (Erg) boseinak 'Don't be afraid! (It is only) our cat (that) cries'. Cf. also Khapalu Balti ta khwe, rgyalu Cobzaq, τηίγϊη (PPosLoc) chugwe (Gen) cal skagseyotpa. 'Now tears had been filled completely into his, prince Cobzang['s], eyes' RGYA (36/14, 4) besides 01/7/17 (PPosLoc) churjo (Abs) cal skagse 'tears having been filled completely into [his] eyes' RGYA (36/14,7-8). cal seems to be an intensifying and/ or onomatopoetic expression, indicating i.a. completeness or surprise. Most probably it can only be used with skag and similar verbs (for similar collocations, cf. Zeisler to appear). E.g., patterns 16 and 33 seem to be the preferred ones for the Saspol informant for the verbs gaij and skag; she apparently also interpreted KHAL

Sentence patterns and pattern variation in Ladakhi

14.

15. 16.

423

luqba (Abs) ie (Gen) skaqs (32) '(He) filled the valley with the flesh (of the animals)' as luqba: (DatLoc) £e (Gen) skaqs (33). But the main Domkhar informant originally rejected pattern 33 totally for skaq and said that he would not use pattern 32 although it seemed correct. He also objected to the use of pattern 15 for gag with the two arguments zaqsbu 'pot' and chu 'water', but had no objection with respect to other combinations of CONTAINER and MEDIUM. By contrast, the second Domkhar informant accepted patterns 15 and 16 equally for the verbs gag and khil, patterns 32 and 33 for the verb skil and skaq. In Domkhar, the Comitative is typically realized as -na, only infrequently as -nag, merging thus with the Ablative and rare uses of the Locative. The informant opted for a comitative interpretation 'with, by means of, alternatively for a locative interpretation On', rather than for an ablative interpretation 'from (the perspective of)'. A similar pattern is found with [-ctr] verbs as 'to be filled' or 'to be blocked'. Informant's comment: "Chumas always go to the river and lose their time by playing etc. and then they are late". Shepherds, by contrast, could only be late to start with the goats, pattern (11 a), but, according to a somewhat androcentric logic, they would not be late in bringing them home.

References Bailey, Thomas Graham 1920 Linguistic Studies from the Himalayas. Being Studies in the Grammar of Fifteen Himalayan Dialects. (Asiatic Society Monographs 18.) London: Royal Asiatic Society. Bielmeier, Roland 1985 Das Märchen vom Prinzen (?obzan. Eine tibetische Erzählung aus Baltistan. Text, Übersetzung, Grammatik und westtibetisch vergleichendes Glossar. (Beiträge zur tibetischen Erzählforschung 6.) St. Augustin: VGH Wissenschaftsverlag. Dixon, Robert M. W. 1995 Ergativity. Cambridge: Cambridge University Press. Francke, August Hermann 1979 Reprint. Ladakhi and Tibetan Grammar. Original edition, Sketch of Ladakhi grammar, Journal of the Asiatic Society of Bengal 70: 1-63, 1901. 1981 Reprint. The Epic of Gesar. (ed. by Kunzang Tobgyel, and Mani Dorje) (Vol. 29.) Thimphu, Bhutan: Druk Sherig Press. Original edition, GSamyulna bSadpahi Kesargyi sgruqs bzugs. A Lower Ladakhi Version of the Kesar Saga. (Bibliotheca Indica 168.), Calcutta: Royal Asiatic Society of Bengal, 1905-1909 (Fasc. l~4), 1941 (Fasc. 5).

424

Beitina Zeisler

Goldstein, Melvin C, and Ngawangthondup Narkyid 1984 English-Tibetan Dictionary of Modern Tibetan. Dharamsala: Library of Tibetan Works and Archives. Haller, Felix 2006 Verbal valence in Shigatse Tibetan and Themchen Tibetan. Linguistics of the Tibeto-Burman Area 29 (2): 63-77. Jäschke, Heinrich August 1992 Reprint. A Tibetan-English Dictionary. With Special Reference to the Prevailing Dialects. Original edition, Routledge & Kegan Paul, 1881. Koshal, Sanyukta 1979 Ladakhi Grammar. Delhi etc.: Motilal Banarsidass. LaPolla, Randy J. 1992 Anti-ergative marking in Tibeto-Burman. Linguistics of the TibetoBurman Area 15 (1): 1-9. Rangan, Krishnasamy 1979 Purki Grammar. (Central Institute of Indian Languages Grammar Series 5.) Mysore: Central Institute of Indian Languages. Read, Alfred Frank Charles 1934 Balti Grammar. (James G. Forlong Fund 15.) London: The Royal Asiatic Society. Skalbzarj Hgyurmed 1981 Bodkyi brdasprod rigpahi khridrgyun rabgsal melor] [The very clear mirror, a guide to the understanding of Tibetan grammar]. Khrerjtuhu: Sikhron Mirigs Dpeskrunkharj [Chengtu: Szechuan National Publishing House]. 1992 Bodkyi brdasprod rigpahi khridrgyun rabgsal melorj: Le clair miroir. Enseignement de la grammaire tib taine. (Traduit, adapte" et comment^ par Heather Stoddard et Nicolas Tournadre.) Arvillard: Editions Prajfia, Collection Tib^taine. Tournadre, Nicolas 1991 The rhetorical use of the Tibetan ergative. Linguistics of the TibetoBurman Area 14 (1): 93-107. 1994 Tibetan ergativity and the trajectory model. In Current Issues in Sino-Tibetan Linguistics, Hajime Kitamura, Tatsuo Nishida, and Yasuhiko Nagano (eds.), 637-648. Osaka: The Organising Committee. 1996 L'ergativite' en tibotain. Approche morphosyntaxique de la langue parUe. (Bibliotheque de l'Information grammaticale 33.) Louvain/ Paris: Peelers. Zeisler, Bettina 1999 Verbalkonstruktion im Ladakhi. / Relative tense and aspectual values in Tibetan languages. A comparative study. Ph. D. diss., Freie Universität Berlin.

Sentence patterns and pattern variation in Ladakhi 2004

2006

to appear

425

Relative Tense and Aspectual Values in Tibetan Languages. A Comparative Study. (Trends in Linguistics, Studies and Monographs 150.) Berlin/New York: Mouton de Gruyter. The Tibetan understanding of karman: Some problems of Tibetan case marking. In Medieval Tibeto-Burman languages II (PIATS 2003: Tibetan Studies: Proceedings of the Tenth Seminar of the International Association for Tibetan Studies, Oxford 2003), Christopher I. Beckwith (ed.), 57-101. (Brills Tibetan Studies Library 10.) Leiden/Boston: Brill. "Wenn du deine Mühle gemahlen hast, womit mahlst du dann dein Mehl?" - Idiomatische Wendungen im Ladakischen.

Subject index ablative, 27, 29,42,45, 125, 207, 208,349,406,410,421,423 ablaut, 86, 87-88,90, 324,328-335, 338 absolutive, 88-89,95,104-105,125, 127,128,129,131,145,348349,350,353, 354,355, 399, 400-401,402,406,407,411, 417,421-422 accent, 155-156,160,337 domain of, 155 accent language, 155,156 Accessibility Hierarchy, 249 accusative (ACC), 39, 350,355, 356, 364,421-422 active/stative case marking, 355-380 active/stative language type, 125, 361-366, 372, 375 actor, 390,422 addition connective, 382, 384-386 adjective, 27,189,191,192,193, 197,198,199,212,220,300 adverb, 26, 27,293 adverbial directional, 42 adverbializer, 294 aesthetive case, 401-402,403, 406, 411,421 affective pattern, 401 affirmative verb form, 221-222 agent (AG), 88, 124, 126, 127, 128, 131,138,145,146,148,207, 214,217,355,356,357,358, 359,360,361,362,363,364, 369-371,372,374,375,389, 390-391, 399,400,401-402, 403,404,407,421 agreement, 214, 215, 219, 292, 309, 314-315,317 allative,29,31,34, 36,38

ambitransitive, 416 anacoluthon, 25 anaphora, 110, 249 anaphoric demonstrative, 104 anaphoric directional, 30, 34, 36, 37, 40,42 animacy,415 antecedent, 248,249 antipassive, 356,360, 374, 375 appositive clause, 46 approximant, 44, 163,164, 165-167, 168,171,172,173,174,185, 186,239,240,242,275 argument suppression, 417 article, 110 articulatory vowel space, 166-167 aspect, 58,60,120,123, 124,126, 127,132,133,138,140, 142, 146,149,221,258,349,358359,366-367,368,371,373, 374,387,393 aspiration, 271-272 asymmetric case marking, 416

background information, 247,248, 259 backward control, 317 bdag/gzhan, 370, 372-373, 374,421 benefactive, 89,400 beneficiary, 401,403,418 bilabial, 154,170, 275, 285 bilingualism, 2, 3, 8, 9, 12, 13, 14, 16,323 Binding Theory, 301, 307, 319 breathiness, 45, 113,182,183 see also murmuring

428

Subject index

case-marked PRO, 291-321 case marking, 120, 123,124-131, 145-146,205,206,207-208, 213,220,341,348-350,351, 354, 355-358,360,361-362, 366,375 case neutralisation, 403,406,411 causality connective, 384-385 causativity, 128, 133,137-138,365, 367,368,372 cause, 401,403 c-command, 307 close vowel, 163,165 cluster, 47,48,49,56,61,63, 204, 228,231,234,235,236, 241,245 loss of, 272 voice in a, 95 coda, 227,228,229,231,234,235, 242 cohortative, 344 collective expression, 407, 416,422 collocation, 419,422 color-word, 191-193,194,195,198, 201 see also word of colour comitative, 207, 208, 348, 349,357, 401^K)2,403,404,407,410, 412,416,421,422,423 common ground, 250 comparative marker, 294 concessive, 294-295 conditional clause, 136 conjugation, 214-215, 217 conjunct/disjunct (CONJ/DISJ), 359, 368,373,374-375,381-382, 388-389,390-391,395-396 conjunctive participle (CP), 291, 292-296, 297,299-309, 311, 315,317 connective, 381, 382-385,387,396 distribution of, 385-386, 387, 388, 395 semantic relation encoded by, 384-385,396

consonant system, 275,276 consonantal strength, 164 contact-induced morphological change, 324,335 container, 407,423 content argument, 406-407 context shift, 247-263 contour, on, 23,45 contra-expectation, 393-395 contrast connective, 384-385 contrastive focus marker, 248 control, 59,132, 359,361,399 control verb (+ctr), 88,89, 90-91, 93,130,144, 361,391, 399,400, 406,407,409,412,415,418 derivation of, 94 core argument, 401,410 coronal, 227-228,232, 234,236, 237-238,240,243,245 dative (DAT), 88,89,120,125,127, 128,129,130,145-146,149, 291,297,298,299,300-301, 303,305,308-309,310,311312,314,315,316,319,348, 349,350,400-401,403,406, 415,421-422 dative-subject construction, 291, 299,313,315 debuccalization, 227 default expectation, 248,260 definite article, 97,103-104,105, 106,107,110,111 definite marker, 106,110 definiteness, 104,106,110,415 deictic, 110,387-389 deictic adverb, 211 Deictic Center Theory, 250,251 deictic centre, 134,247, 248,250251,253,256,258,260, 345347, 348 deictic marker, 258-260 deictic pattern, 341-354 deictic root, 26, 27

Subject index deictic verb, 250 deixis,24,381-397 demonstrative, 31,43, 97-118, 210 demonstrative adverb, 24 demonstrative adverb of direction, 207,208,210-211 demonstrative adverb of manner, 208,211 demonstrative adverb of place, 210211 demonstrative pronoun, 30, 210 deontic modality, 142 dependent-marking, 190 descriptive mode, 247,248,250251,255,256,259,260,261 determiner, 110, 111 deverbal noun, 198 deverbative, 212,216, 222 devoicing, 271 direct speech, 247-263 directional, 23^46 directional pattern, 401,421 directional stem, 24, 27,30, 31,38, 39,42 directional term, 24, 29, 34,37, 43, 45 directive, 207, 211 discourse, 23-46, 247-263 discourse function, 23 discourse marker, 205,208 discourse recoverability, 36 discourse topic, 403 discourse unit, 248-249, 255 discourse world, 249 dorsal, 227-228,229,231,232, 233-234, 237-238,240,241 double demonstrative, 101,105,107, 111 double determination, 107 double possessive, 123,146 downgradient, 23,24,45 empathy hierarchy, 389 episode, 248,251

429

epistemic,45,58,389 equational clause, 41 ergative, 88,120, 123,124,125129,130,131,145,146, 148149,247-263,314,348,349, 353-354, 355-380,400-402, 403,406,407,411,421,422 ergative language, 125,126,309, 375,399-401,403 ergativity, 58,125,126,127,146, 349,354,358-360, 362,369, 406,421 evidential, 59-60 evidentiality, 381-397 exclusive, 209 exhortative, 218 existential/locative clause, 37-38 experiencer (EXP), 88,291,297309,315,318-319,355,357, 362, 363, 374, 375,400, 401402,403,418 experiencer derivation, 410,418 expressive mode, 247,248,250251,252,255,256,257,259, 260,261 extended intransitive, 401 factual verbal adjective, 220-221 feature percolation, 313 finite verb, 213-222 fluid S-marking, 360, 361 folk migration, 62-64 fortition, 232-233,234 fricative, 76,122,163,164,165167,168,185,233,236,237, 239,244,272,275,284 future stem, 61 lack of, 85 future tense, 387 gender marking, 205 general knowledge, 392,393 genitive (GEN), 27, 30, 39,40, 41, 42,124-125,189,190,194,195,

430

Subject index

200,207,208,291, 297, 298, 299,300,301,303,304,305, 306-307, 308-309,310, 311312,314,318-319,348,349, 403,407,411,421 genitive-marked directional, 39-40, 41-^2 genitive-subject construction, 298 gerund, 222 glide, 164,165,168,169,170-171, 172,173,174,175,176,177, 236,239 globally prominent participant, 253, 256,257 goal, 27, 356,400,401 grammar, 47, 55,57, 58-61,62 grammaticalization, 97,104-108, 123,133,146,149,198, 199,336 Head Feature Convention, 313 head-marking, 190 hearsay, 336,389 high vowel, 163,164,165,166,175, 185 homonymy, 414—415 honorific, 123,148 immediate perception, 389,390391,392-393,395 imperative, 213,216, 219-220,325326,336,338,353,410 imperative stem, 94,148,325,327, 338,366 lack of, 91 imperfective aspect, 141,220,358359 imperfective stem, 324, 325-328, 333,335 implicational hierarchy, 225, 226, 235,236,238,239,240-241, 242,243,245 implosive stop, 204 genesis of, 204

inceptive and terminative boundedness, 221 inchoative aspect, 133 inclusive, 209,218,344-345,353 indefinite word, 212 indirect evidence, 59-60,338 inference, 387, 389, 390-391, 393, 395 infinitival marker, 158,159 infinitive, 222,313-314,315,317, 318 inflectional category, 214 initial change of, 86-87, 89-90 instrument, 124,207, 355,399,401402,403,406,410,421 instrumental (Instr), 124-125,348, 349,400,401,403,406,407, 410,412,421 intentional/unintentional, 59,60 interference factor, 247, 252-253, 255,256-258 International Phonetic Alphabet (IPA), 53-54 International Phonetic Alphabet symbol, lack of, 164,167 interrogative word, 212 intonation unit (IU), 23, 28, 30, 31, 34,249 intransitive subject, 125, 214,361, 399 intransitive verb, 24-25,28, 214, 217,219,341,359,361 judgemental modality, 58 kinship term, 159 labial, 171,225-246,338 labial nasal, 228, 229,239,241-242 labial obstruent, 225,228,243-244, 245 labial stop, 225-246

Subject index language attitude, 7,9,10-11,16 language contact, 3,119,146,401 language death, 1-3,4,7, 8 reason for, 2 language decay, 3 language history, 47-69 language ideology, 14,18 language maintenance, 1-8,13,122 language reconstruction, 5,47,49, 50,53,56 language revitalisation, 4-5, 8,14, 16-17 language shift, 1-21 language shift reversal, 1 language threat, 14 length, 155-156,276,325 lenition, 232-233 lexical suffix, 207,208,222 lexicography, 52,53 lexicon, 51,53 lexis, 47,55,57-58, 62 light verb construction, 419 lingua franca, 12,120 linguistic coding, 247-263 linguistic purism, 2 loanword, 53, 54,55, 225, 227,230, 280,333 local case marker, 23, 38,42,43 locally prominent participant, 253 location argument, 401,403,407 locative (Loc), 29,31, 32,34-35,36, 37,38,39,103,125,207,210, 222,297,309,310,311-312, 314,348,357,40(M01,403, 406,415,421,423 locative-allative case, 30 locative demonstrative, 102,103 long-distance agreement, 291, 313 long vowel, 24-27,28,42 loudness, 155-157 low approximant, 165, 167-169, 170,178 low glide, 163-188 low vowel, 164,175

431

manner adverb, 292-293 maximal constriction, point of, 165167 medium argument, 401,403,407, 412,423 middle marker, 217 middle subject, 214-215 middle verb, 214,217, 219 mirative, 338-339 mirativity, 381-397 modal verb, 133,143 modality, 120,123,133,137,142, 146 mode, 366 modifier-compound, 194-195,196 modifier-head order, 194,199 modifier-noun construction, 189-191 monosystemic analysis, 279-280 mood, 148 motion verb, 132, 133,134, 341, 345,351,353,359 movement verb, 345-348,401 murmuring, 276 see also breathiness mutual intelligibility, 203,222 narrative, 247-263, 381-397 narrative world, 249-250,255 nasalization, 272,325 natural class, 173,175 natural discourse, 23, 30,36 negative imperative, 220 negative marker, 216,220 nominal, 205,209,210,212-213 nominalisation, 212 nominative (NOM), 295,296,297, 298, 299, 300, 301, 302, 304, 305,308-309,310,311,314, 315,316,319,355,356,364 nominative-accusative language, 309 non-control verb (-ctr), 88-89, 91, 94, 399,400,401,404, 405,407, 408,410,412,415,416,418,423 non-identical subject, 295-296

432

Subject index

non-nominative case, 309, 314 non-productive nominal suffix, 205206 noun classification, 205 noun-classifying formative, 205 NP-accessibility, 403 nucleus, 164,171,172,185 nucleus peak, 164 number, 206,209, 213, 217, 273, 274 number agreement, 214, 216, 218, 219 number marker, 205, 206, 208,209 numeral, 212 numeral system, 76,78-79 object, 25-26,28,127,129,149, 199,309,314,315,343,344, 347,348,350,354, 357, 360, 362,369-370,371-372,373, 399,400 object marking, 120,123, 128,129131,145-146, 343-344,350,421 obligatorily possessed noun, 195 oblique case, 399,400,406 obstruent, 122,228,234,235 obstruent instability, 244 offline discourse, 388, 390,393, 395 "Old Bhutanese" phonetic features, 75-76 old/new knowledge, 389,390, 391, 393, 395 "omniscient narrator", 382,395 online discourse, 381,388,390, 393, 395 onset, 227, 239,242, 245 orientational prefix, 336 orthographic form, 47,55,56, 226 orthography, 48,50,51,53,54, 56, 66,226,228,279-289 paradigmatically conditioned root alternation, 215 participant prominence, 247,258

participant role, 355-356, 360, 362 passive, 214, 356, 360, 369, 373 past tense, 45,140, 288,295, 315, 353-354,387,388,390,391 patient (PAT), 88-89,126,130,131, 145,146,214,217,309, 314, 315,355,356,357,360,361, 362,363,364,369,370,371, 372,399,400,402,403,404, 415^16,421,422 pattern variation, 399-425 perfective aspect, 149,358-359 perfective stem, 324,325-326, 327331,333-334,335,338 perfective stem renovation, 323-340 peripheral vowel, 175 person, 59,126, 209,213,217, 258, 273,274, 342, 344,347, 349, 353,381-382,389,390,395 person agreement, 214,215,216, 218,219,395 person anaphora, 39-42 person deixis, 381, 387, 388, 395 personal pronoun, 76, 79,110, 112, 209,273 phoneme theory, 279, 288 phonological change, 270-272 phonological rule, 270 phonological system, 275,276 phonology, 49,53, 55, 57, 62,154155,156 phrasal affix, 207 pitch, 48, 155-156,177, 280-281, 284,285,286, 287, 325 pitch accent, 155-156 pitch accent language, 156 place anaphora, 30-39,40-41 place assimilation, 225-246 place deixis, 387,388,395 place-noun, 190-191 place of articulation, 164,167,172, 174 plural marker, 97-118 pluriformity of referents, 206

Subject index polar co-ordinate vowel diagram, 173-174 polysemy, 414-415 polysystemic analysis, 279-280, 287,288 possessive pronoun, 209 possessor, 209,307-308,315,318319,417 predication sentence, 406 prenasalization, 228-229,231 present tense, 387-388,391, 393 prestige, 10,11,19,85 lack of, 2 pro, 303,306,309, 319 pro-drop, 302-303 progressive aspect, 133,140,338 progressive assimilation, 241-242, 243,245 prominence, 155-156,160 prominence factor, 252-256,259 prominence hierarchy, 253,261 prominent participant, 247-248,252, 254,256 pronominal prefix, 189,190,194, 195-196,198 pronominalization, 267, 272-274, 276 pronominalized language, 267,268, 269,272,274,341,342 pronominalized modifier, 193-194 pronorninally marked noun determiner, 189-202 pronoun, 97,98,100,103-104,105, 106,109-110,114,115,206, 209,248,249,251,259,301, 305-306,307,317,319,350, 387,395 prosodic analysis, 279, 286 purpose argument, 401 purpose construction, 135-136,137, 199-200

433

quantifier, 196 quoted speech, 381,385-386,387, 388,395 reader involvement, 250 reading-style pronunciation, 53, 227, 230,233,245,285 real world, 249-251,260 reason adverb, 293 recipient, 89,400,403,406-^07 reciprocal, 302, 344-345 reduplicated gerund, 199 reduplication, 27,28, 35-36 referential competition, 249 referential distance, 249,252, 255 referential mention, 247-263 reflexive, 126, 214, 302,353 reflexivity, 417 regressive assimilation, 241-242, 243 relocation, 212-213 relational noun, 195-196 relative clause, 46, 190-191,199 relator noun, 33-34,38, 39,41-42 remedial morphology, 324,328,331 removed information, 392-393 resonant, 165-167,175,185 responsibility hierarchy, 357-358 resultative (RES), 360,367-368, 369,370,372, 373,374 resyllabification, 234 retroflection, 234 retroflex, 239, 240, 242,269, 275, 276,281 Reversed Language Shift (RLS), 45,18 saliency hierarchy, 403 script, 47-55, 56,57, 61, 63 secondary verb, 120, 123, 132-145, 146,148,149 self-benefactive reading, 214 "self-" vs. "other-centeredness", 59 semantic bleaching, 43, 105

434

Subject index

semantic ergativity, 355 semi-speaker, 3,10,16 semivowel, 163,165-166,167,170, 171,172,173,185 sentence pattern, 399-425 variation of, 411-^418 sequential action, 292 serial verb, 132,133,146 similaritive case, 207,208,211 simplex verb, 213,216-218,220 sonority hierarchy, 163-164,171, 173,175,185 source, 27,29, 356,357,401,403 source case, 207,222 spatial deictic morpheme, 335-336 speech-act location (SAL), 348, 351 speech-act participant (SAP), 23-24, 31, 342-343, 344,345, 348,349, 350,351,353 spelling-style pronunciation, 53, 280-281,284,285,287 spirantization, 225,226,227-228, 229-230,232,238,240,243-244 split accusativity, 131 split ergativity, 126,131,349, 353354,355,360,361 stative verb, 191,198 stem alternation, 85-95,324,325326,333,335 borrowing of whole paradigm, 85-86 distribution of types of, 89-90 elimination of, 85,91 expression of, 86-88,89-90 new development of, 85 number of verbs with, 89 percentage of verbs with, 89,90 type of, 86-88,89-90 verb class and, 89-90,91 stem usurpation, 328,331 story world, 250-251, 260 stress, 155-160,240,242, 325 stress accent, 156, 158,159,160 stress language, 156

stricture type, 164,172,173-174 subject, 28,59,60,128,130,131, 133,149,190,197,217,219, 253,258-259,291-321,341, 342,343,346,353-354,356, 357,360,361,362,373,375, 389,390-391,399,400,401, 402,403,416,422 subject agreement, 343,344 subject marking, 120,125-129,131, 145,146,341-343,406 substratum, 272,298,323, 324,335 subtractive morphology, 338 superstratum, 298 supine, 199,222 syllable structure, 163,164,239,240 symmetric argument, 407 reduction of, 416 syntagmatic analysis, 287-288 systemic value, 286-287,288 target argument, 403,407,414-416, 421 target manner, 225,236, 237,238239,242,245 target place, 225,237, 240-241,242, 243,245 target position, 225, 241-242,243, 245 telicity, 149,411 temporal adverb, 157,159-160 tense, 58,60,123, 124,148,258, 366-367,371,387-389,393 tense marker, 292, 295,309, 315 text analysis, 248,255,260 text world, 249-250 tha dad pa, 368,419,421 theme, 309,313,314-315,362 Tibeto-Burman (TB) agreement system, 343 Tibeto-Burman "causative" suffix, 215 Tibeto-Burman "directive" or "applied" suffix, 215

Subject index time-adverb, 189,190-191 time connective, 382,384-386 time deixis, 387-389,395 timing, 166,170,171,172,173 tone, 45,56,63,113,155-156,175, 177,275,276, 325,337 domain of, 155 tone language, 155, 156 tone letter, 72 topic, 252,255-256, 362 topic marker, 97,98-100,102,114, 148 topicalization, 356 topographic directional, 23, 24,42, 43 topography-based deictic, 23 toponym, 74 traditional linear analysis (text-mapping), 249,252,253 trajectory adverb, 27 transitive subject, 125,128,131, 145,149,253,258-259,399 transitive verb, 24, 25,28,129,214, 215,217,219,248,258,335, 341, 343-344, 348, 350, 354, 356,358,361,364,369,375, 404,407,418,421 transitivity, 127,132,214,355-358, 360,361,362,364,374,400,421 trilingualism, 9 undergoer, 197,399 unit phoneme, 228, 229, 231 upgradient, 23,24,45 valency, 356-357, 358, 362, 372, 374,375,399,400,401,418 variation of, 416—418 verb agreement, 342-343, 382 verb class, 88-89 change of, 90-91 verb classification, 88 verb inflection, 341-345,351,353, 368

435

verb of translational motion, 24-25, 28,31,37 verb paradigm, 54,273-274 verbal morphology, 85 verbal orientation marking, 335-336 verbal valence, 85-95 verbal valence pattern, 88-89 viewpoint, 58, 59,60, 386 vocabulary, 52, 55 vocoid, 170,171 volitionality, 128, 357,389 vowel assimilation, 35 vowel harmony, 27 vowel system, 275,276 weak subject, 401-403 weakening, 225, 226,239,240, 243244,245 Whitehead's "prehension", 286-287 word accent, 153-162 word of colour, 212 see also color-word writing system, 47,53,57 zero anaphora, 247-248,252,253, 255-256,257 1-place predicate, 404,406,407-408 2-place predicate, 404-405,406,408 3rd person pronoun, 97,104,110, 111,113 3-place predicate, 405-406,408-410

Language index Aghem, 168 Alttibetisch, 50 Amdo Khake, 53, 62 Amdo nomad dialects, 85,148,225 Amdo sedentary dialects, 225 Amdo Tibetan, Amdo dialects, Eastern Amdo Tibetan (EAT), 53,58, 59,60,61,62,147,225-246, 323,324,337,338,400 Amharic, 243-244 Angami, 297 Aranuk dialect (ARA), 405,417 archaic Tibetan dialects, 48,49,54, 55,62-63 Arizona Tewa, 9, 17,18 Arvanitika, 2,16 Assamese, 293, 294,295-296,297, 298,299, 308 Athabaskan-Eyak-Tlingit language family, 18 Athapaskan languages, 168 Äthpahariyä", 203 see also Athpare Athpare, 196-197,199,201 see also Äthpahariyä Axininca Campa, 168 Bade, 110 Bahing, 203 Balti dialect, 49,56,58,59,60,62, 63,115,266,279-289,399,421422 Bantawa, 203 Bantu languages, 106 Baragaonle Tibetan, 106-107,109110,111 Basque, 110, 369 Bathang Tibetan, 119-152 Belhare, 160,197,199

Bengali, 295, 297,298, 299, 319 Bhotia, 267 Black Mountain language, 74,7576,82 Bodic,71 Bodish.71,74 Boo-nan, 268-269 see also Bunan, Gahri Braga Manangba, 107,108,109,113 see also Praka Manang Breton, 18 Bumthang, 74, 75,81-82 Bumthang group, 74 Bunan, 265-278 see also Boo-nan, Gahri Burmese, 198,201 Byangsi, 266, 267, 268,269, 272, 274 Camba LahujT, 267 see also Chamba Lahauli, Chamba Lahuli Cambodian, 132-133 Caodeng rGyakong, 337, 338 Carrier, 168 Castilian Spanish, 244 Catalan, 235 Caucasian languages, 54 Cauräsiyä, 203 see also Chaurasia, Chouras'ya Central Asian Tibetan, 370 Central Bodish, 76 Central Tibetan (CtrT), 51,52,55, 56,61,85,119,122,123,147, 358,371,400 Chabcha dialect, 245 Chadic languages, 97,110,111-112, 241 Chaile, 99

Language index Chamba Lahauli, 266, 267 see also Camba Lahuh, Chamba Lahuli Chamba Lahuli, 266-267 see also Camba Lahuli, Chamba Lahauli Chamling, 203 Chantyal, 97,100-101,102-103, 104,105,107,109,111,113 Chaudangsi, 266,268 Chaurasia, 203-223 see also Cauräsiyä, Chouras'ya Chibchan languages, 14-15 Chilcotin, 10-11,18 Chiling, 203 Chinese, 47,119,120,122,168 Chintang, 203 Chone dialect, 61 Chorbat sub-dialect, 281 Chouras'ya, 203 see also Cauräsiyä, Chaurasia Chuksang Seke, 99,107-108, 111, 112,113 Classical Central Tibetan, 370 Classical Tibetan (CT), 24, 26, 29, 30, 38-39,42,43,48, 50, 51-52, 58,59,60,61,74,85,93,94, 125,148,225,226,227, 245, 280-286,287-288, 337,368, 400,406,407,410,421 Cushitic languages, 241 Dakkhini Hindi-Urdu, 296 Dakpa, 71-84 see also Täkpa Dakpo dialect, 73-74 Dänjongkä, 107,115 Darma, 266 see also Darmiya Darmiya, 266,268 see also Darma Dartsedo Tibetan, Dartsedo dialect, 124,144,147-148

437

Derge dialect, 120,123, 125, 130131,133,135,136,147,148 Dingri dialect, 62, 109 Dolakha Newar, 104-105 Domkhar dialect (DOM), 403,404, 406,407,408,409,410,418, 419,422,423 Dravidian languages, 291, 299 Dumi, 160, 203 Dungmali, 203 Dutch, 19, 156 Dzala, 71-84 Dzongkha, 54, 58,71, 73-74,75, 105,110,115,147,149, 381-397 see also Garke Dzorge Tibetan, Dzorge dialect, 59, 60,62,63 see also mDzod-dge dialect, Mdzod.dge dialect EastBodish,71-84 East Kiranü languages, 189,196197,199 Eastern Tamang, 101,107,114 English, 9,10-11,13,14, 58,60, 156,168,185,191,198,280, 294,297,356,357 French, 198, 297 Friih-Alttibetisch, 50 Gaelic, 2 Gahri, 266 see also Boo-nan, Bunan Garni Tibetan, Garni dialect, 335, 336,339 Garhwal dialect, 266,267 Garke, 71 see also Dzongkha Georgian, 54 German, 156,418,422 Ghachok Gurung, 102,103, 104, 105,109,113,114,175,176-177 Ghale, 113

438

Language index

Gidar, 110 Golok dialect, 49, 56,62,63 Gondhla, 266 see also Titian, Tinani Gongduk, 75-76 Gopang Thakali, 176,186 Gurage, 167-168 Gurma, 107,114 Gurung, 97,164,170-171,186 Gyakar, 99 Gyalthang Tibetan, 147 see also Gyethang Tibetan, Rgyalthang Tibetan Gya-Sasoma dialect (GYS), 405, 406,408,411 Gyethang Tibetan, 147 see also Gyalthang Tibetan, Rgyalthang Tibetan

Hausa, 110 Hawaiian, 110 Hayu, 203 Hebrew, 5 Hilepäne dialect, 204 Hindi-Urdu, 291-321 Hmar, 299 Hona, 110 Hopi, 9, 18 Horpa, 338 see also Horpa-Shangzhai Horpa-Shangzhai, 333, 338 see also Horpa Hualapai, 14,15-17,19 Huari dialect, 62 Hungarian, 357 Icelandic, 291, 317 Indo-Aryan languages, 291,293,309 Indo-Chinese, 265 Indo-European languages, 411 Indo-Iranian languages, 11,401

Jad dialect, 266,267 Janggali, 266,267 see also Raji, Raute Japanese, 156,236,244 Jero, 203-223 see also Jerung Jerung, 203 see also Jero JhäppälT dialect, 204 Johari, 266 see also Rangkas, Saukiya Khun Kagate, 107 Kaike, 113 Kambata, 241 Kanashi, 266 Kanawari, 266 see also Kinnauri Kandze dialect, 123,124 Kannada, 295,297 Kashmiri, 292 Katike Tamang, 101,105-106,107, 109,111,113,114 Kenhat dialects, 403,407,411,419 Kera, 110 Khaling, 155,203 Khalong Tibetan, 323-340 Kham, 105 KhamKoind, 120 Kham Tibetan, Kham dialects, 60, 62,119-121,122-124,129,130, 131,132,136,145,146,147,148 Khapalu Balti, Khapalu sub-dialect, 281,284,287,422 Khasi, 297 Kheng,74,81-82 Khoksar dialect, 266 Kinnauri, 247-263,266,269,341354 see also Kanawari Kiowa-Tanoan language family, 18 Kiranti languages, 153,155,189, 199,203 Koasati, 361

Language index Kohi, 203 Kokborok,318-319 Kolong dialect, 266 see also Tod dialect, Todkad Konyak, 297 Kuki-Chin languages, 265 Kukish, 265 Kulung, 203 Kurtop,74,81-82 Kyirong Tibetan, 105,106,107,109, 110-111 Labrang dialect, 147 Ladakhi, 53,54,56, 58-59,62,63, 107,125,266, 399-425 Lahul dialect, 266 see also Lahuli Lahuli, 266 see also Lahul dialect Lai, 107,114 Lakota, 361 Latin, 50,58,244 Lavrung, 333 Leh dialect, 421 Lhasa Tibetan (LT), Lhasa dialect, 49, 53,56, 59-60, 61,119,120, 122,123,124,125-126,133, 135,137,142,148,149,281, 284,286,328, 358,400, 406 Lhoke, 115 Lhokpu, 75 Limbu, 189-202,203 Lisu, 119 Literary Tibetan, 51-52,55 Logone, 110 Lohorung, 203 Lower Kinnauri dialect, 353 Malayalam, 236-237,297,301 Manang, 186 see also Manangba Manangba, 97 see also Manang

439

Manchad, 266,267,269, 272-276 see also Manchati, Patni, Pattani Manchati, 266 see also Manchad, Patni, Pattani Mandarin Chinese, 120 Mangde, 74, 82 Manipuri, 291-321 Marathi, 279-280 MarphaThakali, 98,99,100-104, 108,109,110,111,112,113, 115,169,170,175,176-177, 179,180,181,182,183,184,186 see also Marphali, Mawatan Thakali Marphali, 163-188 see also Marpha Thakali, Mawatan Thakali Mawatan Thakali, 98 see also Marpha Thakali, Marphali mDzod-dge dialect (DZ), 225,226, 227-229,230,231-232,233, 234,239 see also Dzorge Tibetan, Mdzod. dge dialect Mdzod.dge dialect, 338-339 see also Dzorge Tibetan, mDzoddge dialect Mewahang, 203 Mewa Khola dialect, 191,200,201 Mina, 110 Minyak, 119 Mittel-Alttibetisch, 50 Mizo, 299 Modern Literary Tibetan, 24,26,54, 58 Modern Spoken Tibetan, 132, 269, 270,368 Mongol, 55 Monguor dialect, 55 Mönke, 72 Mon-Khmer languages, 297 Mupun, 110 Musey, 241

440

Language index

Nachiring, 203 Nakh-Daghestanian languages, 317 Nangchen dialect, 62 Nar-Phu, 97,103-104,105,108, 109,111,113,114,115 Navaho, 18 Naxi, 119,146 Nepali, 12-14,18-19,99,186,203, 212 Ngawal Manang, 177 see also Ngawal Manangba Ngawal Manangba, 106,109,113, 114 see also Ngawal Manang Niger-Congo languages, 107 non-archaic Tibetan dialects, 48 Northern Qiang, 336 Nubra dialect, 421 Nup, 74,81-82 Nyamkat, 266 OldBodish.47,51 Old Msh, 244 Old Tibetan, 48^t9,50-51,52, 53, 55,125,148,269,272,324,338 Oriya, 295

Pa'a, 110 Pacific languages, 358 Pad-ma dialect (PM), 225,231-232 Paite, 299 Pali, 198 Panaka, 62 Panchthar dialect, 200,201 Pangi dialect, 341-354 Patnam dialect, 266 Patni, 266 see also Manchad, Manchati, Pattani Pattani, 266 see also Manchad, Manchati, Patni Phedap dialect, 191,200 Philippine languages, 375

Praka Manang, 176-177 see also Braga Manangba Preclassical Tibetan, 49,53,54-55, 57-58,59,60,61 Proto-Indo-European, 168,244 Proto-Tamangic, 109 Proto-Tibetan, 47, 50,51,56-57,59 Proto-Tibeto-Burman (PTB), 101, 107,272, 276,342 Proto-Western Tibetan, 56 Puma, 203 Pumi, 119,146 Punjabi, 292,297, 313 Purik dialect, 56,266,268,421 Pwo Karen, 114 Qiang, 119,336 Qiangic languages, 324-325, 339 Rabha, 291-321 Rai, 203 Raji, 266,267 see also Janggali, Raute Rama, 14-17 Rangkas, 266,268 see also Johari, Saukiya Khun Ran Po Bhäsa, 267 see also Rongpo Raute, 266 see also Janggali, Raji rDo-sbis dialect (DB), 225, 226, 229-230,232,233,240,245 Rebkong dialect, 62,245 Ren'entang Tibetan, 336,339 rGyalrong, 119,324, 333, 335, 337, 338 rGyalrongic subgroup, 333 Rgyalthang Tibetan, Rgyalthang dialect, 61,119-152 see also Gyalthang Tibetan, Gyethang Tibetan Ribu, 337 see also Showu rGyalrong

Language index Risiangku Tamang, 97,98-100,101, 102,103,104,105,107,109, 113,114,175,176-177,186 rNga-ba dialect (NG), 225,230, 231, 232,234,238-239,240,241-242 Romance languages, 50 Rongpo, 267,274 see also Ran Po Bhasa Sahu Tamang, 107,108,113,114 Sam, 203 Samoan, 364-365 Sampang, 153-162,203 see also Sängpäng Sängpäng, 154 see also Sampang Sanskrit, 39, 58,280,281,284 Saspol dialect (SAS), 402,404,405, 408,409,410,412-416,417 Saukiya Khun, 266,268 see also Johari, Rangkas Sayanci, 110 Seke, 97,98-104,105,107,108, 109,111,112,114 Sema, 297,299 Shamskat Ladakhi, 403,407,419, 421 Sherpa dialect, 56,115 Shigatse dialect, 125, 147 Showu rGyalrong, 323-324,333335,337,338 see also Ribu Sidaba rGyalrong, 333, 337,338 sign languages, 358 Sindhi, 11-12,16 Sino-Bodic, 265 Sino-Tibetan, 47,185,203, 265,268 Skardu sub-dialect, 284, 287 South Asian languages, 292,293, 295,297 Southeast Asian languages, 132-133 southern Kham Tibetan, southern Kham dialects, 62,119-152 Spanish, 167-168,243-244

441

Spät-Alttibetisch, 50 Spiti dialect, 266,274 Standard Spoken Tibetan, 119,129, 133,149 Sunwar, 8, 12-14,16,18, 203 Swahili, 106 Swedish, 107 Syang Thakali, 98-99,102,108, 111,112,113,169,175,176177,185,186 see also Yhulkasompaimhi Thakali Tagalog,361 TaTcpa, 73-74 see also Dakpa Tamang, 97,101,114,164,185,186 Tamang-Gurung-Thakali-Manang (TGTM) group, 168, 185-186 Tamang Thakali, 98 see also Tukche Thakali Tamangic languages, 71, 97-118 Tamil, 295,297 Tangbe,99,111,112,113 Tangut dialect, 62 Telugu, 291-321 Tembe dialect, 191,193,200 Tetang Seke, 99,102, 103,105, 111, 112,113 Thai, 132-133 Thakali, 97, 98-100,101,105,109, 110,112,114,169,185 Thebor, 268 see also Tibar-skad Themchen Tibetan, Themchen dialect, 59,62, 85-95,147 Thulung, 155,203 Tibar-skad, 268, 269 see also Thebor Tibetan, 23, 39,45,46,47-69,85, 105,106,107,109,110,114, 122,124,125,126,129, 132, 133,134,135,147,149,201, 226,230,266,267,268-270,

442

Language index

272,274,276-277,279-289, 323,324-325,326,335,336, 337,338, 339, 355-380, 399401,403,404,411,419,421 Tibeto-Burman languages, 12,47, 71,97,107,112,114,126,153, 163,164,168,203,260, 265278,291,297,298,299, 309, 318,351 Tibeto-Himalayan languages, 341 Tibeto-Kinnauri, 268 Tilung, 203 Tinan, 266,267 see also Gondhla, Tinani Tinani, 266 see also Gondhla, Tinan Tirit dialect (TIR), 405,406,408409,410,412,414,417 Tod dialect, 266, 274 see also Kolong dialect, Todkad Todkad, 266 see also Kolong dialect, Tod dialect Tokpe Gola Tibetan, 23-46 Tongan, 363-364 Tsez, 317 Tshangla, 72,74, 75 Tukche Thakali, 98,99,100-104, 108,109,110,112,113,175, 176-177, 186 see also Tamang Thakali Turanian, 265 Turkish, 393 Udaipure dialect, 204 Umbule, 203 see also Wambule Urtibetisch, 50 Ute, 106 Uto-Aztecan language family, 18, 106 Vietnamese, 132-133 Vortibetisch, 50

Wa-ke, 63 Wakka dialect (WAK), 416 Wambule, 160,203-223 see also Umbule Wämdyäl dialect, 203 West Himalayish languages, 71,265 Western Drokpa dialect, 109,110 Western Himalayan languages, 265 Western Tibetan (WT), western dialects, 62,149,400 Written Kham, 123, 148 Written Tibetan (WT), 23,24, 26, 36-37,42,44,45,47,48,49,50, 51-52,53,54,55,56,59,104105,106,107,110,111,115, 120,122,123,124,125,128, 140,148,177,233,269,270, 324,326-328,332,335-336, 337,338,366,368,374

Xdi, 110 Yakkha, 203 Yakut, 236 Yamphu, 203 Yhulkasompaimhi Thakali, 98 see also Syang Thakali Yi, 119 Yuman-Cochimi language family, 19 Zhongre rGyalrong, 333-335,338 Zhongu Tibetan, 335-336,339 Zhuokeji rGyalrong, 338