Himalayan Languages: Past and Present 9783110898873, 9783110178418

Table of contents :
Introduction
Linguistic synchrony and diachrony on the roof of the world – The study of Himalayan languages
Descriptive linguistics
A grammatical comparison of Shina dialects
Retroflex vowels and other peculiarities in the Kalasha sound system
Direction and differential dative case marking in Magar
Thangmi kinship terminology in comparative perspective
Hidden syntax in Belhare
On the notion of sentence in Classical Tibetan
On discourse functions of the finite verb in Kinnauri narratives
Language change
Preverbal modifiers in Sunwar
Directional prefixes in Kathmandu Newar
Grammaticalization of deictic motion verbs in Seke
“Do” as subordinator in Tshangla
Morphosyntactic transparency in Bantawa
Areal semantics – Is there such a thing?
Shafer’s proto-West Bodish hypothesis and the formation of the Tibetan verb paradigms
Newaric and Mahakiranti
Subject index
Language index

Citation preview

Himalayan Languages

W G DE

Trends in Linguistics Studies and Monographs 149

Editors

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

Mouton de Gruyter Berlin · New York

Himalayan Languages Past and Present

Edited by

Anju Saxena

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 Himalayan languages : past and present / edited by Anju Saxena. p. cm. — (Trends in linguistics. Studies and monographs ; 149) Includes bibliographical references and indexes. ISBN 3-11-017841-9 (cloth : alk. paper) 1. Tibeto-Burman languages - Congresses. 2. Himalaya Mountains Region — Languages — Congresses. I. Saxena, Anju, 1959— II. Himalayan Languages Symposium (7th : 2001 : Uppsala, Sweden) III. Series. PL3781.H56 2004 495'.4-dc22 2004040236

ISBN 3-11-017841-9 Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at . © Copyright 2004 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.

Preface The Himalayan Languages Symposium brings together scholars working on the languages and language communities of the greater Himalayan region: north-western and north-eastern India, Nepal, Bhutan and the Tibetan Plateau, northern Burma and Sichuan, and Nuristan, Baltistan and the Burushaski-speaking area in the west. The Seventh Himalayan Languages Symposium was organized at Uppsala University, Sweden, between August 7-9, 2001. This volume presents a selection of the papers presented at the conference, and also includes three contributions - the articles by Balthasar Bickel, George van Driem, and Anju Saxena - which were not presented at the conference. The conference was funded, in part, by the Swedish Research Council for the Humanities and Social Sciences (Humanistisk-Samhällsvetenskapliga Forskningsrädet - HSFR - now part of Vetenskapsradei). I would like to thank Professor Werner Winter for his keen interest and support during the preparation of this volume. I would also like to thank Birgit Sievert at Mouton de Gruyter for her advice on all aspects of the production of this volume. Anju Saxena

Contents

Introduction Linguistic synchrony and diachrony on the roof of the world - The study of Himalayan languages Anju Saxena Descriptive linguistics A grammatical comparison of Shina dialects Ruth Laila Schmidt

33

Retroflex vowels and other peculiarities in the Kalasha sound system Jan Heegard and Ida Elisabeth Morch

57

Direction and differential dative case marking in Magar Karen Grunow-Härsta

77

Thangmi kinship terminology in comparative perspective Mark Turin

101

Hidden syntax in Belhare Balthasar Bickel

141

On the notion of sentence in Classical Tibetan Claus Oetke

191

On discourse functions of the finite verb in Kinnauri narratives Anju Saxena

213

viii

Contents

Language change Preverbal modifiers in Sunwar Werner Winter

239

Directional prefixes in Kathmandu Newar David Hargreaves

273

Grammaticalization of deictic motion verbs in Seke Isao Honda

285

"Do" as subordinator in Tshangla ErikAndvik

311

Morphosyntactic transparency in Bantawa Jadranka Gvozdanovic

341

Areal semantics - Is there such a thing? James A. Matisqff

347

Shafer's proto-West Bodish hypothesis and the formation of the Tibetan verb paradigms RolandBielmeier

395

Newaric and Mahakiranti George van Driem

413

Subject index Language index

419 430

Introduction

Linguistic synchrony and diachrony on the roof of the world - The study of Himalayan languages* Anju Saxena

1.

The study of Himalayan languages

In the Himalayan region, divergent social patterns are seen on many fronts. While some groups in this region are nomads who move regularly from one place to another with their livestock, others live permanently in villages; while some follow Buddhism as their only religion, others follow Hinduism or Christianity as their only religious faith or follow one or more of these in addition to their local deity. Differences can also be seen in other fields, spanning from the methods of cultivation (while some engage in terraced cultivation, others engage in shifting cultivation), matrimony (polygamy both polyandry and polygyny - in some societies but monogamy in others) to architecture. The Himalayan region presents an exciting field of research to the linguist. It offers a rich smorgasbord of languages belonging to different linguistic stocks and languages representing different typological characteristics. It also has a long tradition of multilingualism. The term "Himalayan languages" usually covers languages and language communities of the Himalayan region, i.e., languages spoken in north-western and north-eastern India, Nepal, Bhutan, the Tibetan Plateau, northern Burma, Sichuan, Nuristan, Baltistan and the Burushaski-speaking area in the west. Languages spoken in this region represent the Indo-European language family (the Indo-Aryan and Iranian subgroups, and the Germanic language English), the Dravidian language family, the Tibeto-Burman language family, and the Austro-Asiatic language family. Additionally, there are language isolates (Burushaski and Kusunda). There are some major practical problems in specifying the number of languages spoken in the region. For one, the information available on languages is not comprehensive. Second, the same language is, at times, referred to by more than one name: the name which its speakers use and the name which outsiders use to refer to that language.1 Matisoff (1999) notes that people of a Naga group call their language Menu, but outsiders know it as Mao or Sopvoma (the name of their main

4

Anju Saxena

village). Furthermore, the same language name is used, at times, to refer to a language but, at other times, it is also used as a cover term to refer to a group of genetically or culturally related languages. Most inhabitants of the Simla state listed Pahaarii (without any further specification) as their mother tongue in the 1901 census (Bailey 1908: ii). Pahaarii (listed as "Pahari" in the Ethnologue), besides being the name of a particular Indo-Aryan language, is also at times used as a cover term to refer to a set of related languages spoken in this region. The Himalayan region has a long history of linguistic diversity and multilingualism, reaching back several millennia. Societal multilingualism is an established tradition, where not all languages which are spoken in one community are employed in all spheres of activity. Instead, in the Himalayan language situation, as in the wider South Asian context, language use is often situationally differentiated, where one language is used, for example, in school or at work, another at home, and possibly yet another for religious purposes. Individuals in such communities may be fluent in all the languages involved or fluent in their mother tongue, but have working knowledge of other languages for use in particular social situations. The choice of language in the religious sphere in Kinnaur is illustrative. Most Kinnauris follow three religions: Hinduism, Buddhism (Lamaism), and worship of their local deity. Sanskrit and Hindi are used while performing Hindu rituals, Tibetan while performing Buddhist rituals, and Kinnauri while performing rituals for the local village deity. The linguistic situation in South Asia differs in this regard from the pattern which migrating social groups in the Western world commonly display, where the second or third generation immigrants give up their mother tongue, and accept the language of the new country (for example, English in the US) as their own language. In such cases language shift is the norm, and language maintenance is an exception. In the South Asian setting, on the other hand, language maintenance is the norm, not the exception. Studies (for example, Gumperz and Wilson 1971 and Subbarao and Arora 1990) have shown that as a result of close long-term language contact, shift from one language to another does not necessarily mean shift from one grammar to another. Instead, the grammars of the languages involved have converged to the point that switching from one language to another often entails nothing more than the exchange of lexical items of one language for those of another in a grammatical framework which remains invariant across languages. Despite this stable multilingualism, language death is not uncommon in the Himalayan region. Here, as elsewhere in the world, languages have died

Introduction: The study of Himalayan languages

5

and are dying at an alarming rate because of the social upheavals brought about by industrialization and urbanization. Indigenous languages with no written tradition and with little or no political and/or economic power at the local and national level are especially likely to fall by the wayside en route to modernity. Speakers of these languages in many cases favor the neighboring dominant language rather than their own language, seeing the latter as more of a liability than as an asset. Some languages completely vanish from the face of the earth, while others are given up in particular contexts. While minority speakers, in many cases, learn the language(s) of the dominant group, the reverse is usually not the case. Many adult Kinnauri speakers, for example, speak Kinnauri as their mother tongue, and many of these speakers living in villages are monolingual. The children, on the other hand, tend to be active bilinguals, with preference for Hindi or the regional Indie variety. Many young people migrate to cities and towns for education and employment, where the lingua franca is not Kinnauri. Such social situations have important linguistic consequences for these lesserknown languages. Asymmetrical situations may lead to the acceptance of the dominant language as the predominant language, which subsequently becomes their only medium of communication. Winter (1993), however, rules out external factors (for instance, large migrating groups) as the prime factors contributing to language death. He presents the Walapai and Bantawa case studies to highlight the important role language attitudes of its speakers play in language preservation vs. language death ("language suicide"; Winter 1993). Irrespective of the factors contributing to this negative trend, it is a fact that a large number of minority languages of this region are threatened with extinction and that very little is known about many of these languages.

2.

Linguistic synchrony in the Himalayan region

2.1.

Descriptive linguistics

There are a handful of languages of the Himalayan region - languages such as Burmese, Nepali and Tibetan - for which a substantial body of data and analyses is available. There are, for example, fairly detailed dictionaries available for Classical Tibetan as well as modem Tibetan (for example, Jäschke [1881] 1987; Hahn 1974; Beyer 1992 for Classical Tibetan; Gold-

6

Anju Saxena

stein 1978, 1984; Mazaudon 1978; Denwood 1999 for modern Tibetan). Further, the works of DeLancey (for example, 1985, 1986, 1990, 1991, 1992), Bielmeier (1985, 1988, 2000), Toumadre (1991) and Tournadre and Jiatso (2001) are other examples of important descriptive works on Tibetan. There is now work being done on dialects other than Central Tibetan, for example by Jackson Sun (1986) and by Roland Bielmeier and his team at Bern. But, for the vast majority of the languages and language groups of this region, there is little or no information available. There is an increasing awareness among linguists about the need to take into consideration data not only from the much-studied Western languages, but also from lesser-known languages. We are, at present, witnessing some positive efforts in documenting and describing Himalayan languages. Here I will name a few. Van Driem (2001) is a recent description of Himalayan languages. This two-volume work provides an introduction to the languages, cultures and the linguistic history of the greater Himalayan region.2 In addition to the broad overview given by van Driem (2001), there have appeared in recent years detailed linguistic descriptions of several smaller languages of the region (for example, van Driem 1987; Genetti 1994; Chelliah 1997). The present volume is also a contribution in this direction. Several papers in the volume (Bickel, Grunow-Härsta, Heegärd and March, Turin, and Schmidt) contribute towards documenting and describing some aspects of lesser-known languages of this region. Ruth Laila Schmidt in this volume compares the Kohistani and Gilgiti dialects of Shina (an Indo-Aryan language belonging to the Dardic subbranch, spoken in Pakistan, Indian Kashmir, and in Ladakh) by describing the development and composition of verb tenses in these dialects. Comparison is also made here with two additional Shina dialects, Drasi and Guresi, in the area of noun and pronoun case-number inflection. This study makes contributions to the determination of the historical development and the exact genetic relationship of these dialects. Jan Heegärd and Ida Elisabeth M0rch provide much-needed information on the sound system of Kalasha (a Dardic language spoken in north-west Pakistan). For a long time, Morgenstierae (1973), which is based on material collected in the 1920s, has been the main source of information on this language. Primarily because of the nature of the data used by Morgenstieme, there is a slight lack of clarity concerning the phonetic system of Kalasha as it emerges from his description. The aim of Heegärd and March's paper is to shed light on some aspects of the Kalasha sound system. Apart from presenting an overview of the segmental system of

Introduction: The study of Himalayan languages

1

Kalasha (with a discussion on the status of the two / sounds found in the language, aspiration, vowel length and nasalization), they also direct our attention also to the peculiar feature of the retroflex vowels in Kalasha and their plausible origin, which crucially involves a language contact situation. Karen Grunow-Härsta discusses the use of the dative case marking in Syangja and Tanahu, two dialects of Magar (a Central Himalayish language of the Bodic subgroup of Tibeto-Burman spoken in central and west-central Nepal). The dative case marker -ke in Magar is, at times, also suffixed to arguments which function as patients (or direct objects, grammatically speaking). As has also been observed in several other languages, the dative case marking on "patients" in Magar occurs in certain specific conditions (for example, when direct objects are animates and higher on the empathy hierarchy). Grunow-Härsta discusses further the development of this phenomenon in Magar: Is it a borrowing from Nepali, the lingua franca of Nepal, or does it reflect traces of the Tibeto-Burman direction system? Her conclusion is that it is not simply a borrowing; rather, it also includes a reanalysis of an earlier direction marking system. Mark Turin's article provides a valuable detailed description of the kinship terms in Thangmi (a Tibeto-Burman language spoken in central eastern Nepal) as a system. For obvious reasons, discussions on kinship terms have concerned both linguists and anthropologists. Beside providing a cautiously optimistic assessment of the possibility of making sociological inferences on the basis of the kinship term systems of a language, he also presents a historical morphological analysis of Thangmi kinship term formation, and puts forward a proposal of possible Tibeto-Burman etyma for a number of Thangmi kinship terms. The aim of Balthasar Bickel's article is to discuss the mismatches which sometimes occur between syntax and morphology. This, according to Bickel, has consequences for current syntactic theories. The basis for his argument involves his observations in Belhare (a Tibeto-Burman language of the Kiranti subbranch spoken in south-eastern Nepal), where there are some syntactic processes which are sensitive to more traditional syntactic "pivots of constructions" - the various conflations of semantic verb argument types forming the basis of nominative-accusative, ergative-absoluti ve, etc. agreement or case-marking systems, which play such prominent roles in modern language typology. Other syntactic processes reflect purely semantic roles (mainly 'actor' and 'undergoer'). In contrast to the situation in many other languages, in Belhare none of the linguistic devices of phrase structure, verbal inflection, or case marking is used to encode any of the

8

Anju Saxena

three pivots discussed by Bickel, the accusative, the ergative, and the restricted ergative pivot. Instead, these devices operate on the basis of semantic participant roles and information structure, leading to a systematic mismatch between syntax and morphology.

2.2.

Discourse

The focus until recently has been on linguistic descriptions of Himalayan languages at the grammatical (morphological and syntactic) level (with some notable exceptions). This was, to some extent, due to the limited amount of data available on these languages. We are at present observing a change in the trend. In recent times, the focus in a number of studies (for example, Bickel 1999; Noonan 2001; Saxena 2001, Genetti and Grain 2003) has been on discourse. As has already been mentioned in the previous section, in his contribution to this volume, Balthasar Bickel shows how syntactic and morphological devices in Belhare cannot be analyzed without recourse to the discourse-level notion of information structure. There has been a growing interest in how the information conveyed in discourse is organized ("packaged"; Noonan 2001) into larger units such as texts, and how the choice of linguistic constructs available in a language contribute to the fulfilment of the overarching goals of the text. According to Slobin's "thinking for speaking" hypothesis, the linguistic categories available in a language influence how speakers of that language think. The term "rhetorical styles" is used in Noonan (2001) to refer to a group of related constructions. The choice of one construction over the other has its effect on the discourse. Consistent with Slobin's hypothesis, Noonan suggests that the presence of a set of related constructions in a language directs its speakers to organize their thoughts along certain lines. Direct quotes in Chantyal (a Tibeto-Burman language of the Tamangic subgroup spoken in Nepal) are encoded by means of a set of constructions which Noonan refers to as "quotatives". They obligatorily contain the verb 'say' as the main verb and direct speech as its complement (the grammaticalized form of the verb 'say' may, but need not necessarily, occur in this construction). Quotatives in Chantyal occur in a number of contexts, e.g. as a purpose and reason marker. Noonan argues that the quotative construction in Chantyal has narrative functions - it yields an effect in discourse. The Deictic Center Theory (Duchan, Bruder, and Hewitt 1995; Li and Zubin 1995) is a cognitive science-based approach for the analysis of lin-

Introduction: The study of Himalayan languages

9

guistic structure in narratives. According to this framework, in a storytelling situation there is a real world where the story-teller (the narrator) and her audience are situated, and a world created in the story (the story world). The narrator decides on the perspective from which the story will be narrated. The origin of the perspective (or "the deictic center"; Mushin 2000) can either be the story world (in the expressive mode of narrating) or the real world (in the descriptive mode), but the object of the perspective is always the story world. The same event can be narrated in the descriptive mode as well as in the expressive mode. Which perspective is chosen is conventionally signalled by a number of linguistic devices (for instance, the selection of lexicalized linguistic forms, such as directional verbs). The ergative marker in Kinnauri occurs only with transitive verbs. Its occurrence is, however, not obligatory. The distribution of the ergative marker in Kinnauri narratives is analyzed in Saxena (2001) using the Deictic Center Theory. Ergative occurs in these narratives in two contexts: (i) It occurs regularly (almost obligatorily) with the subject of the main clause of a direct speech event (the 'he said'-clause) and (ii) it occurs, at times, with subjects of transitive clauses in constructions other than 'he said'-clauses (both inside and outside direct speech). In (ii), the ergative marker occurs regularly in situations which run counter to the expected behavior (including violations of social norms). It comprises a contrastive focus. Ergative, it is suggested, has a "deictic" function - it indicates a shift from the descriptive mode to the expressive mode. This accounts for the almost obligatory occurrence of the ergative in 'he said'-clauses. Ergative in (ii) also has a deictic function - it steers the listeners' "perspective" away from the default expectation mode. In this way, the ergative marker in Kinnauri narratives, which is still governed by transitivity, encodes a shift in the perspective. The relevance of the "sentence" level is well-established in the written language, but its status and relevance in spoken language is debatable. Halliday (1985) and Miller and Weinert (1989), for example, rule out the sentence as a relevant unit in spoken language. On the other hand, Chafe (1980, 1987, 1994) and Genetti and Grain (2003) argue that the sentence is a significant linguistic unit, essential for the organization of "cognitive material" within spoken language. DuBois (1987) proposed the Preferred Argument Structure hypothesis, according to which nominal arguments should occur primarily in the subject position of intransitive clauses (S), in the subject position of transitive clauses (A), and in the direct object position in transitive and ditransitive

10

Anju Saxena

clauses (O). He proposed the following constraints: (i) One Lexical Argument Constraint: Avoid more than one lexical argument per clause, and (ii) Non-Lexical A Constraint: Avoid lexical A's. These two constraints are complemented by the following two constraints: (a) One New Argument Constraint: Avoid more than one new argument per clause, and (b) Given A Constraint: Avoid new A's. Genetti and Grain (2003) followed DuBois (1987) in their investigation of the referential mention in Nepali (an Indie language spoken in Nepal and India). They expanded the scope of their study to also include the distribution of pronominals and zero anaphora. The results of their investigation reveal that DuBois' constraints account for the patterns observed in Nepali both at the clause level and at the sentence level. At the sentence level, a referent is explicitly mentioned once in a sentence. Thus, both DuBois' (1987) and Genetti and Grain's (2003) results suggest that the sentence is a relevant unit in the choice of the referential mention in oral narratives. With a somewhat different focus, Bickel (1999) suggests that languages differ in how integrative their event framing structure is. In more integrative event framing structures NPs are "more central for clause structure" (1999: 27) whereas in languages with less integrative event framing structures, "argumental NPs are relatively marginal additions to clause structure" (1999: 27). He accounts for the observations made in Genetti and Grain (2003) concerning the sentence as a significant syntactic unit in Nepali (see above) by suggesting that the pattern observed in Nepali is a reflection of the language having a more integrative event framing structure. Languages with a less integrative event framing structure, according to Bickel, are not likely to have the sentence as a significant unit for information organization. The relevance and functions of the "sentence" level is also the theme of the articles by Oetke and by Saxena in this volume. Glaus Oetke takes as his point of departure the claim that sentences in Classical Tibetan are marked by a set of three final particles, one of which will appear last in a Classical Tibetan sentence. He then sets out to investigate some of the theoretical assumptions underlying this claim and its alternatives, and the consequences of those assumptions for the notion of sentence in Classical Tibetan, based on empirical evidence from a Classical Tibetan text material. The question Anju Saxena raises in her article in this volume is: What does a finite clause encode in oral Kinnauri narratives? The occurrence of a finite unit is usually regarded as one of the characteristic features of a sen-

Introduction: The study of Himalayan languages

11

tence. Kinnauri being an oral language, Saxena takes the behavior and functions of finite clauses as the point of departure in an investigation of the linguistic status of sentences in this language. She examines the functions of finite and non-finite clauses and the choice of the finite verb morphology in Kinnauri narratives. Traditional Kinnauri narratives have a wellstructured form, where finite verbs occur regularly. The presence of a finite verb in Kinnauri, it is suggested, is a linguistic cue to the listener that the present episode is now finished, and that which follows is a separate discourse unit. The default finite verb system in many traditional Kinnauri narratives has the simple narrative past tense. It is the neutral linguistic device used to encode episode boundaries. One of the linguistic devices encoding digressions from the default structure is the use of other tense and aspect markers. Divergence from the default finite verb system is a linguistic resource for discourse functions which the narrator uses in Kinnauri narratives, and it has a rhetorical function.

3.

Linguistic diachrony in the Himalayan region

3.1.

Language change

Languages change constantly. This simple fact is the raison d'etre of historical linguistics, which has set itself the tasks of describing the ways in which languages develop through time and of trying to explain the mechanisms of language change. All levels of linguistic structure are subject to change but, traditionally, researchers in historical linguistics have concentrated their attention to a large extent on the description of sound change and lexical change, while change in the grammar - especially syntactic change - has received less attention, a situation that has changed somewhat in recent years. Increasingly, it is felt that research on how the grammars of languages develop through time - the mechanisms and causes of grammatical change - can make significant contributions to our understanding of the phenomenon of language in general. The growing interest in grammatical change coincides with a general trend in linguistics where language is seen as inextricably rooted in a social and historical context and not merely as an abstract system, which can be studied just as well, if not better, isolated from that context.

12

Anju Saxena

How, then, do languages acquire new grammatical constructions and categories and discard old ones? Regular sound changes frequently lead to loss of morphological categories, while new bound morphology often arises through the process of grammaticalization, i.e., the development of content words into bound grammatical morphemes via intermediate stages as function words and clitics. Exactly the same developments - grammaticalization clines or pathways - have been attested again and again in languages around the world, for example, the development of nouns into case endings via postpositions, or that of verbs into morphological markers of tense or aspect via an auxiliary stage (Hopper and Traugott 1993: 106ff). The end result of grammaticalization may be zero, i.e., bound grammatical morphemes are ultimately ground into dust by regular sound change. Grammaticalization in many cases is seen as a part of a larger cycle of grammatical change, in which languages move through the successive stages of isolation - agglutination - fusion - isolation. This cycle is fed by the constant mustering of new content words as input to grammaticalization (Givon 1979: 208-209). An important characteristic of grammaticalization is that it is a one-way street, as it were. There is no corresponding - at least no regular - process of "degrammaticalization"; in the overwhelming majority of cases, the development is from a "less grammatical" to a "more grammatical" unit, but not vice versa (Haspelmath 1999). DeLancey (1985) describes the analysis-synthesis-relexification cycle which many Tibeto-Burman languages have gone through in their development of newer directive systems. The three (non-chronologically ordered) stages are: (i) auxiliary, which provides the deictic information, (ii) each of the basic deictic verbs comprises two elements (MOTION and HITHER/ HENCE) and (iii) lexicalization (the 'go/come' distinction, for example). In (ii) the two semantic components (MOTION and HITHER/HENCE) are represented as two separate formal units, and in (iii) the two semantic components are represented as a inseparable complex unit. The motivation for stage (iii) is that even though deictic verbs comprise two semantic components, they are inseparable in the case of an "motional event". The directive marking, according to DeLancey, can not be reconstructed for ProtoTibeto-Burman, suggesting that each Tibeto-Burman language where we find it has gone through these stages independently. The contributions to this volume by Andvik, Gvozdanovic, Hargreaves, Honda and Winter to this volume concern the development of a range of grammatical morphemes.

Introduction: The study of Himalayan languages

13

Werner Winter's paper discusses a set of preverbal modifiers in Sunwar (a northern Kiranti language spoken in eastern Nepal). Sunwar has a rich reportoire of preverbal elements which display a high degree of internal variation. In his paper, Winter describes this variation, sets out to show that it is consistent and systematic, and endeavors to throw light on its origins and historical development. The variation is iconic, where the iconicity consists of certain semantic traits being related to certain phonetic traits of the initial consonant and vowel of the proclitic on a fairly consistent basis. By and large, the iconicity hierarchies found are the ones expected from other languages, and the exceptions (lenis stronger than fortis in the consonants) are plausibly explained by historical developments in Sunwar (*voiceless > voiced and *voiced > voiceless). The consistent use of phonetic iconicity also provides a plausible explanation for certain unique features of the proclitics, as compared to Sunwar core vocabulary, e.g. initial bh-, dh-, gh-. David Hargreaves discusses the morphosyntactic characteristics of directionals and verbal prefixes in modern Kathmandu Newar. A common source of verbal prefixes in Kathmandu Newar is the directional system. kwD, for example, occurs both as a lexical item (kwy-ne 'under/below') and as a verbal prefix (kwo-ka(l)- 'take down' < ka(l) 'take/grab'). There are, however, also a large number of verbal prefixes which are synchronically opaque. He then goes on to discuss semantic and discourse factors contributing to the present state of affairs. A number of root morphemes are Tibeto-Burman cognates, making the results of this study a useful source of information for future work on comparative-historical Tibeto-Burman linguistics. Isao Honda provides a lucid presentation of grammaticalized functions of kha- 'come' in three dialects (Chuksang, Tangbe and Tetang) of Seke (a language of the Tamangic subgroup of Tibeto-Burman, spoken in Nepal). He shows how the range of successively more grammaticalized uses of 'come' correlates both with greater phonetic distance from the original lexical verb kha- 'come', and with greater difficulty for native speakers to identify the element in question as related to the lexical verb. The most grammaticalized use of 'come' is that as a (morphological) marker of future. Based on morphological evidence, Honda shows that the grammaticalization of 'come' has gone further in Tangbe than in Chuksang and Tetang. The grammaticalization of 'come' as future marker is unique to Seke among Tamangic languages, and not very common even in other TibetoBurman languages.

14

Anju Saxena

In his paper, Erik Andvik discusses the grammaticalized functions of 'do' in Tshangla (a Bodic language spoken predominantly in eastern Bhutan) comparing them with the grammaticalized functions of 'say' which is also found in Tshangla. 'do' in Tshangla shows a higher degree of grammaticalization than 'say'. Andvik further discusses the historical development of this grammaticalization and the correlation that is found between the grammaticalized functions of 'do' and negated subordinate clauses. This can be seen, for example, in that Type 2 (the embedded nominalized clause with -\vd) occurs more often as negative than affirmative propositions. Further, in Types 3 and 4 (both representing nominalized clause with -la), only negative subordinate clauses occur. Jadranka Gvozdanovic addresses the time-honored issue of abstract characterizations of linguistic items in her contribution, which deals with the morphosyntactic expression of agentivity in nouns and verbs in the Kiranti language Bantawa. She proposes that the formally coinciding both are expressed with a suffix -a - nominal morphosyntactic categories 'agentive' (which she identifies as 'ergative') and 'instrumental' are one and the same also semantically, and that their interpretation as one or the other is dependent on the semantics of the noun to which they are attached. Furthermore, she proposes that, with a more abstract analysis, this category, together with the formally identical - in that it, too, is expressed with a suffix -a - verbal category 'past tense/deictic inactuality', can be subsumed under an abstract grammatical category 'deictically disfocal', opposed to the verbal category 'agent'. The latter is expressed with a suffix -u but it does not always co-occur with the nominal 'agentive', a circumstance which calls for an explanation, and which in Gvozdanovic's analysis becomes an exponent of the abstract grammatical category 'focus' (pertaining to the goal in the clause, rather than to the agent or instrument).

3.2.

Language contact

One crucial prerequisite for most kinds of contact-induced language change is an extended period of close contact between the target and the recipient languages. The Himalayan region provides an excellent basis for investigating contact-induced changes. Van Driem (1991) presents a description of the complex linguistic situation of Lohorung and Limbu (Eastern Kiranti languages), where speakers of these languages strive to adopt Nepali, the language of the high-caste people in the region whereas the "low artisanal

Introduction: The study of Himalayan languages

15

castes" whose mother tongue is Nepali adapt their mother tongue (Nepali) to the Kiranti language of the region - illustrating a situation where both sets of languages are affected in one and the same geographical area. Areal linguistics examines issues that arise when a group of languages have been in contact with each other for a long time and the languages in question share structural features which cannot be ascribed to common heritage. The best-known linguistic areas are the Balkans (first described by Sandfeld 1930) and South Asia (Emeneau 1956; Masica 1976). The OV pattern has been suggested as an areal feature of the Indian subcontinent (Masica 1976). Indo-Aryan languages, like other language families in the region, exhibit OV word order and display the "left-branching pattern" (Dryer 1992) whereas other Indo-European languages such as English and Swedish display VO word order and the "right-branching pattern". Matisoff (1990) groups Tibeto-Burman languages into "Indo-sphere" and "Sinosphere" languages, because of the areal influences which these languages have been subjected to. Languages of the Sino-sphere are spoken in the region where Chinese is the dominant language and languages of the Indosphere are spoken in the region where Indie languages are the dominant ones. Tibeto-Burman languages of the Indo-sphere typically display retroflex stop consonants, postsentential relative clauses and the extended grammaticalization of the verb 'say' (Saxena 1988, 1995),3 and typical linguistic features in the Sino-spheric languages are the development of tone, monosyllabicity and "an isolating structure" (LaPolla to appear). Long-standing contact between different language families, as well as among different subbranches of the same language family (for example, within the Tibeto-Burman language family) in the Himalayan region has resulted in intense lexical and grammatical borrowing, generating a situation where the genetic classification of some of these languages (especially those belonging to the Tibeto-Burman language family) is not clear (the genetic classification of Newari, for example, is still under discussion). For languages where we do not have historical data, it is hard, or sometimes impossible, to distinguish similarities between languages due to common heritage from those due to contact in such cases. This is at the root of the so-called Sapir/Boas controversy in the early decades of the last century in American linguistics. The controversy concerned the grouping of the American Indian languages, where Franz Boas tended to see common traits as areal phenomena, i.e., spread through borrowing, while his student Edward Sapir was more inclined to posit genetic relationships on the basis of the same traits (Haas 1976: 66-67; Emeneau [1962] 1980: 56-57). Boas

16

Anju Saxena

argued that in contact situations all levels of grammar can be influenced, and that there is no level of language that can remain unaffected (1911: 48). Sapir, on the other hand, was convinced that it is only the "superficial" aspects of language that can be affected by borrowing: this includes nonbasic lexicon and phonology, but not the "deeper kernel" of language, for example, bound morphology (Sapir [1921] 1971: 192ff). The Sapirian viewpoint became more popular in mainstream - especially American linguistics.4 In this vein, Bickel (1999) argues that some linguistic features are more stable, i.e., tend not to bow down to external influence in contact situations. The basis for his suggestion involves the different event-framing structures observed in Indo-European languages (including Indo-Aryan) and Sino-Tibetan languages. LaPolla (1994) discusses the development of a number of grammatical markers (the "anti-ergative" marker, ergative marker, directive systems, causative marking, person marking and existential verbs) in Tibeto-Burman languages in support of the Boasian viewpoint. He argues (like DeLancey 1985, referred to above, for directive markings) that these developments could not be reconstructed for ProtoTibeto-Burman, suggesting that each of these languages has undergone similar changes independently. The issue of internal vs. external motivations in language change has been the topic of many debates. Gerritsen and Stein (1992) contains a lucid discussion of internal and external factors in language change. Saxena (1997) raises questions concerning motivations and mechanisms of grammaticalization and syntactic change and argues that it is not always easy to distinguish language-internal factors from language-external factors in a contact situation. Rather, in some cases, a more advantageous approach is to recognize the relevance of both (Aitchison 1981; Mithun 1992). Two frequently observed historical sources of tense/aspect are verb serialization and nominalization. Tibeto-Kinnauri languages display examples of both of these sources, where Lhasa Tibetan, Gahri and Tinani grammaticalize nominalizers as perfective aspect markers, whereas Kinnauri and Patani instead reanalyze participle forms as perfective and imperfective aspect markers. The diachronic sources for new aspect morphology are different in the Tibeto-Burman and Indie language groups: Tibeto-Burman - which favors nominalized forms - and Indie - where participles are preferred. The inconsistent pattern in the Tibeto-Burman languages considered here comprise those Tibeto-Burman languages (including Tibeto-Kinnauri languages) which have long been in close contact with Indie languages where Indie languages have long been the dominant languages. In this group of

Introduction: The study of Himalayan languages

17

Tibeto-Kinnauri languages we find new aspect markers that have participle forms as their diachronic source. Yet the very potential for this reinterpretation existed in the structure of the languages themselves, illustrating an interplay of internal and external factors. The articles by Grunow-Härsta, Heegard and M0rch (see above), and by Matisoff raise interesting questions concerning contact-induced changes. Semantic change, in particular change in lexical semantics, has for a long time been among the tools of the trade of comparative linguistics, and more recently, also of areal/contact linguistics (see, e.g., Meillet [19051906] 1958; Paul 1909, chapter IV; Weinreich 1963). At the same time, traditional linguistic accounts of semantic change rarely go beyond the cataloguing of individual instances of such change in a particular language or language family, characterizing the instances - if at all - in terms of a few extremely general processes of semantic change, such as 'narrowing' or 'widening' of the sense of a word, as 'analogy', 'metaphor', etc. (Sturtevant [1917] 1961: 85-130; Collinder 1969: 81-89 are typical examples, picked more or less at random). It could be that we cannot even in principle formulate laws of semantic change (this view is explicitly put forth and defended by Anttila 1972, chapter 7), but we can certainly try to find universal tendencies (Anttila 1972: 147; cf. Heine and Kuteva 2002 on grammaticalization), which is largely an empirical endeavor, even if their uncovering must be guided by theory, and even more so their classification and explanation. James A. Matisoff in his contribution to this volume addresses the issue of semantic change from the point of view of areal linguistics, by posing the two interdependent questions: "[1] Is there such a thing as areal semantics, and if so, [2] can we distinguish between plausible and implausible semantic change/associations in the Southeast Asian linguistic area?" In order to give an answer to these questions, Matisoff looks at a particular component of semantic change, namely semantic associations, first defining what he means by the term and then giving a heuristic logical four-way classification of semantic associations from the point of view of areal linguistics, illustrated with many examples - both single words and compounds/collocations - from a large number of languages. The longer-term goal of his work is of course to sharpen this particular tool of the comparative and areal linguist, making it into less of an art and more of a science: It is an art to decide how much semantic divergence may be tolerated among reflexes of the same etymon. Roots may indeed undergo spectacular semantic changes through time, and the glottochronological dogma against

18

AnjuSaxena accepting semantically shifted cognates when determining degrees of genetic relationship goes much too far. However, the bigger the semantic leap the better the phonological correspondence must be between the putative cognates. Otherwise the phonological and and semantic arguments are like two drunks supporting each other. Crucially, it should not automatically be assumed that semantic associations attested in one linguistic area are universally valid[.] (Matisoff this volume: 381-382)

3.3.

Genetic classification

Genetic classification of languages is intimately tied in with the issues discussed in the preceding section. This is because the basis for genetic grouping of languages are common linguistic features, or rather, recurring and regular correspondences between languages which are thought to result from common ancestry. From the previous section we may conclude that such correspondences could be the result of languages having been in contact for a long time, rather than of their having sprung from some common source. Although genetic classification of languages is in principle beset with this and other kinds of uncertainty, which increase with the time depth involved, we should keep in mind that it is one of the most mature linguistic disciplines, which has accumulated a wealth of knowledge and analytical and methodological tools helping its practitioners separate the wheat from the chaff. In the Himalayan region, the two dominant (in terms of number of languages) language families are the Indo-Aryan branch of Indo-European, and Tibeto-Burman (or the Tibeto-Burman branch of Sino-Tibetan; see below). While there is fairly general consensus for Indo-Aryan as to which languages should be included in the family and how these languages are related to one another, there are several competing theories concerning the nature of the Tibeto-Burman language family (including which languages and language groups should be considered part of it).5 I will summarize here three approaches: the Tibeto-Burman theory, the Sino-Tibetan theory, and the Tibeto-Burman phylogenetic model.

Introduction: The study of Himalayan languages

19

3.3.1. The Tibeto-Burman theory The Tibeto-Burman theory goes back to the eighteenth century. On the basis of common roots, Julius Heinrich Klaproth proposed that Chinese (or Sinitic), Tibetan and Burmese (together with languages which are obviously related to one of these three languages) comprise the Tibeto-Burman language family. This model became an established theory among comparativists and was also included as a subgroup within larger language families. For example, Müller (1855) and Hodgson (1849) propounded the Turanian theory, according to which all the world's languages except Indo-European and AfroAsiatic belonged to one single language family, the Turanian language family. Müller and Hodgson excluded Sinitic from the original TibetoBurman theory. Their version of the Tibeto-Burman theory is described, at times, as the "truncated" version.

3.3.2. The Sino-Tibetan theory From the middle of the nineteenth century to the middle of the twentieth century Sinitic was considered closest to Daic (a linguistic grouping comprising Thai and related languages), quite distinct from the original Tibeto-Burman theory in which Sinitic was a member on the same level as other branches of the Tibeto-Burman family while Daic was not included in the family at all. This view of the division into Sinitic vs. the rest as representing the oldest split of the language family is at the heart of all variants of the Sino-Tibetan theory, although the proponents of SinoTibetan differ in many other particulars (for example Benedict 1942; DeLancey 1987; Matisoff 1999; see Figure 1). Sino-Tibetan is a very large language family comprising languages spoken in China, India, Nepal, Sikkim, Bhutan, Thailand, Laos, Burma and Vietnam. Views as to which specific languages should belong to this family have varied among the proponents of this theory. For instance, Shafer (1966/67) included Thai in Sino-Tibetan whereas Benedict (1942) excluded it (the latter corresponds to the currently prevailing view).

20

Anju Saxena

Sino-Tibetan Chinese

Kamarupan

Himalayish Qiangic

NE India, W Tibet, Burma Nepal, Bhutan, Sikkim

Sichuan, Yunan

JingphoNungishLuish N Burma, Yunan

LoloBurmeseNaxi Thailand, Laos, Vietnam

Karenic

Baic

Burma, Thailand

Yunan

Figure I. The Sino-Tibetan theory (from Matisoff 1999)

There are various proposals for subgrouping the Tibeto-Burman main branch of this language family. Figure 1 represents one such proposal, while Figures 2-5 show another. Note, however, that these two proposals show more similarities than differences. Tibeto-Burman

Bodish

East Himalayan

Kamarupan Kachinic

Rung

Naxi (Moso)

Lolo-Burmese

Figure 2. Higher-order Tibeto-Burman groupings (from DeLancey 1987: 801) Bodic

Bodish

TibetoKannauri

GurungThamangThakali

East Himalayan

Newari

Kham-Magar

Kiranti

Figure 3. Middle-level Tibeto-Burman relationships: Bodic (from DeLancey 1987: 802)

Introduction: The study of Himalayan languages

21

Baric

Kachinic

Luish

Kuki-Chin

Naga

Konyak

Bodo-Garo

Andro, Jinghpo Kadu, etc.

Figure 4. Middle-level Tibeto-Burman relationships: Baric (from DeLancey 1987: 802) Burmic

Rung

Gyarong

Nung

Qiang- ^ / ^„Primi- ~""·-~.._/£-"'"~ Tangut Naxi (Moso)

Burmish

Figure 5. Middle-level Tibeto-Burman relationships: Burmic (from DeLancey 1987: 802) Roland Bielmeier, in his contribution to the present volume, addresses one aspect of the internal structure of the Bodish branch of Tibeto-Burman. The problem is in brief as follows: Are all present-day Tibetan languages referred to as "Bodish" in Shafer (1974) - descended from Classical Tibetan (= Shafer's "Old Bodish"), or should we posit an earlier branchingoff point for some of the languages? Bielmeier accepts Shafer's notion of West Bodish as a separate branch of Bodish (proper), with proto-West Bodish as a separate node on the same level as Old Bodish (see Figure 2 in Bielmeier's article in this volume). On the other hand, Bielmeier sees no reason to accept another of Shafer's Bodish units, proto-East Bodish. Positing West Bodish as a separate unit accounts for certain troublesome verb conjugation correspondences among the Tibetan languages. If they were all posited to be descended from Classical Tibetan, we face the unlikely situation where generally only the perfect stem remains of the (at the most) four stems making up a typical Classical Tibetan verb paradigm. Instead, we can

22

Anju Saxena

assume that verb stem differentiation is a specifically Classical Tibetan development not shared with proto-West Bodish. Bielmeier shows that in fact the formation by suppletion of imperfective verb stems - leading to stem alternation - is still a partly productive process in certain conservative Eastern Tibetan languages.

3.3.3. The Tibeto-Burman phylogenetic model In his article on Tibeto-Burman phylogeny and prehistory, George van Driem (ms.; see also van Driem 2001, chapters 3-6) argues in favor of the Tibeto-Burman phylogenetic model for describing and testing higher-order subgroupings within this language family. Note that for van Driem, 'Tibeto-Burman' means something like the original Tibeto-Burman theory, i.e. a higher-order grouping with Sinitic on a par with the main branches of Tibeto-Burman as defined by the Sino-Tibetan theory (van Driem 2002). On the other hand, language groups are presented here in the form of bubbles and not as well-defined family trees. This, in van Driem's opinion, reflects that we do not, at present, have more specific knowledge concerning the "higher-order branches" and chronological ordering/timing of branching (see van Driem's article in this volume). This should, however, not be taken to argue against the family-tree model for Tibeto-Burman languages. George van Driem's contribution in this volume illustrates how even historical-comparative linguistics very much must be considered a dynamic discipline. In his paper, he revises his own Mahakiranti hypothesis of a decade ago - the notion that Newaric and Kiranti belong together in a distinct subgroup of Tibeto-Burman - in the light of new linguistic field data. The main morphological traits first thought to distinguish Newaric and Kiranti, i.e. Mahakiranti, from other Tibeto-Burman languages have subsequently been unearthed by van Driem also in Gongduk, a language which otherwise does not bear any evidence of standing in a particularly close relationship to Newaric or Kiranti.

Notes I would like to thank Balthasar Bickel and Scott DeLancey for their comments on the first version of this chapter. Further, I am grateful to Stig Eliasson for his feedback regarding the organization of this volume.

Introduction: The study of Himalayan languages 1.

2.

3.

4.

5.

23

Of course, this phenomenon is by no means unique to the Himalayan region; it is very common for speakers of a language and outsiders to use sometimes completely unrelated names for the language in question. By the greater Himalayan region, van Driem means "[t]he region extend[ing] from the Hindu Kush and Tiänshän mountains in the west to Arunachal Pradesh and upper Burma in the east. Hundreds of different languages [are] spoken in this area, which encompasses the sovereign kingdoms of Nepal and Bhutan, the once independent kingdom of Sikkim, the vastness of Tibet, large tracts of Afghanistan, Kyrgyzstan, Tajikistan, Pakistan, India, Bangladesh, Burma and the Chinese provinces of Sichuan and Yunnan." (van Driem 2001: ix). The use of the verb 'say' is found in some Sino-sphere languages, too (Matisoff 1991), but unlike South Asian languages, the verb 'say' in Sinosphere languages is usually a complementizer and does not normally develop other grammaticalized functions of 'say'. Although probably most linguists - even those of a radically Boasian persuasion - would concede that some linguistic features are more easily and vastly more often borrowed than others, possibly even that there could be a classical implicational scale lurking in the wings, as it were, with open-class lexical items close to one end and fusional inflectional morphology close to the other. Two main factors can be said to jointly determine this state of affairs: (1) the relative maturity of the two fields of inquiry involved. Indo-European comparative linguistics (in the modern sense of the term 'comparative linguistics') has been an active and at times very prestigious branch of linguistics for almost two centuries (Szemerenyi 1989: 6-8), while Tibeto-Burman linguistics is only about fifty years old (Matisoff 1999); (2) the relative accessibility to linguistic scholars of the languages in the respective families: "Indeed, we cannot say for certain how many Tibeto-Burman languages there are or even whether there may not still be a few — possibly in western Nepal, very probably in northern Burma and southeastern Tibet — that are yet to be discovered." (DeLancey 1987: 799)

References Aitchison, Jean 1981 Language Change: Progress or Decay? London: Fontana. Anttila, Raimo 1972 An Introduction to Historical and Comparative Linguistics. New York: Macmillan.

24

Anju Saxena

Bailey, Thomas G. 1908 Languages of the Northern Himalayas. London: The Royal Asiatic Society. Benedict, Paul King 1942 Thai, Kadai, and Indonesia: A new alignment in Southeastern Asia. American Anthropologist 44: 576-601. 1972 Sino-Tibetan: A Conspectus. Contributing editor, James A. Matisoff. (Princeton-Cambridge Series in Chinese Linguistics #2.) New York: Cambridge University Press. Beyer, Stephen 1992 The Classical Tibetan Language. Albany: State University of New York Press. Bickel, Balthasar 1999 Principles of event framing: Genetic stability in grammar and discourse. MS, University of California, Berkeley. Bielmeier, Roland 1985 A survey of the development of western and southwestern Tibetan dialects. In Soundings in Tibetan Civilization. Proceedings of the 1982 Seminar of the International Association for Tibetan Studies held at Columbia university, B. N. Aziz and M. Kapstein (eds.), 3-19. New Delhi: Manohar. 1988 The reconstruction of the stop series and the verbal system in Tibetan. In Languages and History of East Asia. Festschrift for T. Nishida on the Occasion of his 60th Birthday, 15-27. Kyoto: Shokado. 2000 Syntactic, semantic, and pragmatic-epistemic functions of auxiliaries in Western Tibetan. In Person and Evidence in Himalayan Languages, Volume I, Balthasar Bickel (ed.), 79-125. (Special Issue of Linguistics of the Tibeto-Burman Area, 23 (2).) Boas, Franz 1911 Introduction. In Handbook of American Indian Languages, Part 1, Franz Boas (ed.), 1-83. Washington: Government Printing Office. Chafe, Wallace L. 1980 The employment of consciousness in the production of a narrative. In The Pear Stories. Cognitive, Cultural and Linguistic Aspects of Narrative Production, Wallace L. Chafe (ed.), 9-50. Norwood, N.J.: Ablex. 1987 Cognitive constraints on information flow. In R.S. Tomlin (ed.) Coherence and Grounding in Discourse, 21-51. Amsterdam: Benjamins. 1994 Discoi4rse, Consciousness, and Time. Chicago: Chicago University Press. Chelliah, Shobhana 1997 A Grammar ofMeithei. Berlin: Mouton de Gruyter.

Introduction: The study of Himalayan languages

25

Collinder, Björn 1969 Spräket. Meaning till del vetenskapliga sprakstudiet. [Language. Introduction to the scientific study of language.] Stockholm: Natur och Kultur. DeLancey, Scott 1985 The analysis-synthesis-lexis cycle in Tibeto-Burman: A case study in motivated change. In Iconicity in Syntax, John Haiman (ed.), 367-389. Amsterdam/Philadelphia: John Benjamins. 1986 Evidentiality and volitionality in Tibetan. In Evidentiality: The Linguistic Coding of Epistemology, Wallace L. Chafe and Johanna Nichols (eds.), 203-213. Norwood, NJ: Ablex. 1987 Sino-Tibetan languages. In The World's Major Languages, Bernard Comrie (ed.), 798-810. London: Croom Helm. 1990 Ergativity and the cognitive model of event structure in Lhasa Tibetan. Cognitive Linguistics 1 (3): 289-321. 1991 The origin of verb serialization in modern Tibetan. Studies in Languages 15 (l):l-23. Denwood, Philip 1999 Tibetan. (London Oriental and African Languages Library, Vol. 3.) Amsterdam: John Benjamins van Driem, George 1987 A Grammar ofLimbu. Berlin: Mouton de Gruyter. 1991 Taal en identiteit: Indo-Arisch expansionisme in oostelijk Nepal. Bijdragen tot de Taal-, Land- en volkenkunde 147 (1): 61-73. 2001 Languages of the Himalayas, vol. 1-2. Leiden: Brill. 2002 Tibeto-Burman replaces Indo-Chinese in the 1990s: Review of a decade of scholarship. Lingua 111: 79-102. ms. Tibeto-Burman phylogeny and prehistory: Languages, material culture and genes (article). Dryer, Matthew S. 1992 The Greenbergian word order correlations. Language 67: 547-619. DuBois, John W. 1980 Beyond defmiteness: The trace of identity in discourse. In The Pear Stories. Cognitive, Cultural and Linguistic Aspects of Narrative Production, Wallace L. Chafe (ed.), 203-274. Norwood, NJ: Ablex. Duchan, Judith F., Gail A. Bruder and Lynne E. Hewitt (eds.) 1995 Deixis in Narrative. A Cognitive Science Perspective. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Emeneau, Murray B. 1956 India as a linguistic area. Language 32: 3-16. 1980 Bilingualism and structural borrowing. In Language and Linguistic Area, Anwar Dil (ed.), 38-65. Stanford: Stanford University Press. First published 1962.

26

Anju Saxena

Genetti, Carol 1994 A Descriptive and Historical Account of the Dolakha Newari Dialect. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Genetti, Carol and Laura D. Grain 2003 Beyond preferred argument structure: Sentences, pronouns, and given referents in Nepali. In Preferred Argument Structure: Grammar as Architecture for Function, John W. DuBois, Lorraine E. Kumpf and William J. Ashby (eds.), 197-224. Amsterdam: John Benjamins. Gerritsen, Marinel and Dieter Stein (eds.) 1992 Internal and External Factors in Syntactic Change. Berlin: Mouton de Gruyter. Givon, Talmy 1979 From discourse to syntax: Grammar as a processing strategy. In Discourse and Syntax, T. Givon (ed.), 89-112. (Syntax and Semantics 12.) New York: Academic Press. Goldstein, Melvyn C. 1978 Tibetan-English Dictionary of Modem Tibetan. Kathmandu: Ratna Pustak Bhandar. 2nd edition. 1984 English-Tibetan Dictionary of Modern Tibetan. California: The Regents of the University of California. Gumperz, John J., and R. Wilson 1971 Convergence and creolization. A case from the Indo-Aryan/Dravidian border in India. In Pidginization and Creolization of languages, Dell Hymes (ed.), 151-167. Cambridge: Cambridge University Press. Haas, Mary 1976 Boas, Sapir, and Bloomfield. In American Indian Languages and American Linguistics, Wallace L. Chafe (ed.), 59-69. Lisse: Peter de Ridder Press. Hahn, Michael 1974 Lehrbuch der klassischen tibetischen Schriftsprache. Bonn: Michael Hahn. Repr. Swistal-Odendorf: Indica et Tibetica Verlag, 1996. Halliday, M. A. K. 1985 An Introduction to Functional Grammar. Baltimore: University Park Press. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37 (6): 1043-1068. Heine, Bemd and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press.

Introduction: The study of Himalayan languages

27

Hodgson, Brian Houghton 1849 On the aborgines of Nor-Eastern India [sic]. Journal of the Asiatic Society of Bengal CCIII: 451^60. Hopper, Paul J., and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Jäschke, Heinrich August 1987 A Tibetan-English Dictionary with Special Reference to the Prevailing Dialects to which is Added an English-Tibetan Vocabulary. Delhi: Motilal Banarasidass. First published London 1881. 1954 Tibetan Grammar. New York: Frederick Ungar. Masica, Colin P. 1976 Defining a Linguistic Area: South Asia. Chicago: University of Chicago Press. Matisoff, James A. 1990 On megalocomparison. Language 66: 106-120. 1991 Areal and universal dimensions of grammaticalization in Lahu. In Approaches to Grammaticalization, vol II, Elisabeth Traugott and Bemd Heine (eds.), 383^453. Amsterdam/Benjamins: John Benjamins. 1999 Description of the Sino-Tibetan language family. Available on the WWW: . Accessed July 15,2002. Mazaudon, Martine 1978 La formation des propositions relatives en tibetain. Bulletin de la Societe de Linguistique de Paris 73: 401-414. Meillet, Antoine 1958 Comment les mots changent de sens. Linguistique historique et linguistique generate, 230-271. (Collection linguistique publiee par la Societe de Linguistique de Paris VIII.) Paris: Librarie Honore Champion. First published \T\Anneesociologique 1905-1906. Miller, Jim and Regina Weinert 1989 Spontaneous Spoken Language. Syntax and Discourse. Oxford: Clarendon Press. Mithun, Marianne 1992 External triggers and internal guidance in syntactic development: Coordinating conjunction. In Internal and External Factors in Syntactic Change, M. Gerritsen and D. Stein (eds.), 89-129. (Trends in Linguistics. Studies and Monographs 61.) Berlin: Mouton de Gruyter. Morgenstierne, Georg 1973 The Kalasha Language. (Indo-Iranian Frontier Languages, vol. 4.) Oslo: Universitetsforlaget.

28

Anju Saxena

Müller, Fredrich Max 1855 Languages of the Seat of War in the East, with a Survey of the Three Families of Language, Semitic, Arian, and Turanian (frontispiece title: Max Muller's Survey of languages), second edition. London: Williams and Norgate. LaPolla, Randy J. 1994 Parallel grammaticalizations in Tibeto-Burman languages: Evidence of Sapir's 'drift'. Linguistics of the Tibeto-Burman Area 17 (1): 61-80. to appear The role of migration and language contact in the development of the Sino-Tibetan language family. In Areal Diffusion and Genetic Inheritance: Case Studies in Language Change, R. M. W. Dixon and A. Y. Aikhenvald (eds.). Oxford: Oxford Univeristy Press. Li, Naicang and David A. Zubin 1995 Discourse continuity and perspective taking. In Deixis in Narrative. A Cognitive Science Perspective, Judith F. Duchan, Gail A. Bruder and Lynne E. Hewitt (eds.), 287-307. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Mushin, liana 2000 Evidentiality and deixis in narrative retelling. Journal of Pragmatics 32 (7): 927-957. Noonan, Michael 2001 Direct speech as a rhetorical style in Chantyal. Paper presented at the Workshop on Tibeto-Burman Languages, University of California, Santa Barbara, 2001. Paul, Hermann 1909 Principien der Sprachgeschichte. Vierte Auflage. Halle: Niemeyer. Sandfeld, Kristian 1930 Linguistique balkanique: problemes et resultats. Paris: Champion. Sapir, Edward 1971 Language. An Introduction to the Study of Speech. London: Rupert Hart-Davis. First published 1921 New York: Harcourt, Brace & Co. Saxena, Anju 1988 On syntactic convergence: The case of the verb say in Tibeto-Burman. In Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society, Shelley Axmaker, Annie Jaisser and Helen Singmaster (eds.), 375-388. California: University of California, Berkeley. 1995 Unidirectional grammaticalization: Diachronie and cross-linguistic evidence. Sprachtypologie und Universalienforschung (STUF) 350-372 1997 Internal and External Factors in Language Change. Aspect in Tibeto-Kinnauri. Reports from Uppsala University Linguistics 32. Department of Linguistics, Uppsala University.

Introduction: The study of Himalayan languages 2001

29

The role of deictic center in the comprehension of Kinnauri narratives. Paper presented at the Seventh Himalayan Languages Symposium. Uppsala University, 2001. Shafer, Robert 1966/67 Introduction to Sino-Tibetan (Part I: 1966, Part II: 1967). Wiesbaden: Otto Harrassowitz. 1974 Introduction to Sino-Tibetan, part IV. Wiesbaden: Otto Harrassowitz. Slobin, Dan I. 1987 Thinking for speaking. Proceedings of the Thirteenth Berkeley Linguistics Society, 435-444. California: University of California, Berkeley. 1996 From "thought and language" to "thinking for speaking". In Rethinking Linguistic Relativity, John J. Gumperz and Stephen C. Levinson (eds.), 70-96. Cambridge: Cambridge University Press. Subbarao, K. V. and Harbir Arora 1990 On extreme convergence: The case of Dakkhini Hindi-Urdu. In Language Change and Language Contact, Aditi Mukerjee (ed.). Hyderabad: Osmania University. Sun, Jackson T.-S. 1996 Aspects of the Phonology ofAmdo Tibetan. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Szemerenyi, Oswald 1989 Einführung in die vergleichende Sprachwissenschaft. 3., vollständig neu bearbeitete Auflage. Darmstadt: Wissenschaftliche Buchgesellschaft. Tournadre, Nicolas 1991 The rhetorical use of the Tibetan ergativity. Linguistics of the TibetoBurmanArea 14 (1): 93-108. Toumadre, Nicolas and Konchok, Jiatso 2001 Final auxiliary verbs in literary Tibetan and in the dialects. In Person and Evidence in Himalayan Languages, Volume II, Balthasar Bickel (ed.), 49-111. (Special Issue of Linguistics of the Tibeto-Burman Area, 24(1).) Weinreich, Uriel 1963 Languages in Contact. Findings and Problems. The Hague: Mouton. Reprint, first published New York 1953. Winter, Werner 1993 Some conditions for the survival of small languages. In Language Conflict and Language Planning, Ernst Häkon Jahr (ed.), 299-314. Berlin: Mouton.

Descriptive linguistics

A grammatical comparison of Shina dialects Ruth Laila Schmidt

1.

Introduction

The Shina language is spoken in the Gilgit and Kohistan Districts of Pakistan, as well as in the Kishanganga and Dras River systems of Indian Kashmir, and in Ladakh. It is classified as a member of the Dardic branch of IndoAryan languages (Morgenstieme 1961). Bailey (1924: xii-xiv) divided it into three main dialect groups: Gilgiti, Kohistani (including Kohistani and Chilasi) and Astori (including Guresi and Drasi). Schmidt (1985, 2001) presents grammatical data suggesting that the Kohistani and Guresi dialects separated from the Gilgiti dialect before they separated from each other. Strand (2001) classifies the Shina dialects into two main groups: The dialect around Chiläs in the east-west portion of the Indus valley above Indus Kohistan is probably the source of the speech that spread upstream along the Indus basin to form the Eastern Shina dialects and downstream to form the kohislyö dialect of Indus Kohistan. Another dialect centers on Gilgit, with an outlying Tibetanized offshoot (Brokskat) in Ladakh. In addition there are dispersed dialect enclaves to the west of the Indus: usuj'u, spoken beside Torwäli in the Chail Valley of upper Swat, the archaic dialects palöla' and Säwi, spoken in enclaves off the Kunar-Chitral River, and perhaps KalkoTi, spoken in one part of KalkoT in Dir Kohistan. In the absence of premodern texts in Shina, we must use comparative linguistic data to trace its development. This paper compares the major processes in the development of verb tenses in the Kohistani and Gilgiti dialects of Shina, and attempts to identify the elements of which these tenses are composed. The paper also compares noun inflections in the Kohistani, Gilgiti, Guresi and Drasi dialects, to identify the case suffixes these dialects have in common.

34

Ruth Laila Schmidt

2.

Attempts to trace the history of the Shina-speaking peoples

The term "Dardic" is derived from the Sanskrit name Darada, which was applied to an ethnic group of antiquity, first mentioned by Strabo and Pliny as Daradae (Allen 1994). Vohra (1983: 53-66) and Jettmar (1977: 411433) provide long lists of classical Greek, Sanskrit, Tibetan and Chinese epigraphic sources mentioning the Daradas, of which I mention only a few here. Alexander invaded the Daedala (Darada?) country after crossing the Kunar River (in modern Chitral?). A commentary on Pänini identifies the DaradJ Sindhu 'the Indus coming from the land of the Daradas'.1 The Kharosthi inscription at Alam Bridge at the mouth of the Gilgit River, dated somewhere between the late 2nd and late 4th century A.D., mentions a Daradaräya ('Darada king'). The RäjätämhginJ, or chronicles of Kashmir, mention a Daraddesa 'country of the Dards' numerous times. Stein (1961, 2: 435) places Daraddesa in the Kishanganga/Neelam River valley north of the Kashmir valley, where modern Gures lies.2 Jettmar (1977: 418) also places Daraddesa in the Kishanganga valley, most likely extending to include the Astor valley and the villages around Chilas on the Indus. Cacopardo and Cacopardo (2001: 90) assemble convincing data to show that the Shina-speaking region (which they call Eastern Peristan) ...was once the seat of a series of literate kingdoms which controlled the routes across the mountains and were a vehicle in the diffusion of Buddhism. As far as Chilas itself is concerned, the petroglyph complexes discovered by Jettmar in the Indus Valley have conclusively shown that this town was not only a part, but probably a centre of such kingdoms for a very long period, that covered most of the first millenium A.D.

Many attempts to trace the origin of Shina connect it with the Darada, postulating a process of expansion, possibly of conquest, from the south, leading to the settlement (or annexation) of Gilgit, followed by further eastward migration of Shina speakers. This is the hypothesis of Jettmar (1975:20-21): At an earlier stage, Shina must have undergone a powerful expansion. It was perhaps the language of the Darada Empire that has been mentioned in the chronicles of Kashmir [the Räjätärahgini]. This empire had conquered the Gilgit Valley during the second half of the first millenium A.D., in the course of which the carriers of the Burushaski language were either

A grammatical comparison ofShina dialects

35

assimilated or pushed back into the valleys of Hunza and Yasin, where they also came under the rule of Dardic dynasties.

Fussman (1989: 55-6) finds at least one innovation linking the present Dardic dialects with the Gandhari language attested in the Ashokan (3rd century B.C.) and later inscriptions, located in the modem North-West Frontier Province of Pakistan. The last surviving language of this family was the now extinct Tirahi. Fussman (1972: 398) finds "les liens entre Tir. et Sh. sont evidents et tres importants." In his 1989 work he concludes: ...we have to suppose that proto-Shina was once spoken much lower south in the Indus valley, maybe as far as Tarbela. Tirahi and Shina came to be separated when they had to retreat before Pashto forced in by the Pathan tribes who conquered these areas...During this process the Southern Dardic area was split; groups migrated and settled in different valleys...the result is the motley linguistic map we now see in Indus Kohistan...It is quite possible that Shina pushed Burushaski further north and replaced it in many parts of the Gilgit area (Fussman 1989: 56).

Unfortunately, no Shina records or inscriptions trace this history. The only Dardic language with a pre-modern literary tradition is Kashmiri.

3.

Lexical similarity among Shina dialects

Koul and Schmidt (1983: 19) counted shared core vocabulary items in a Shina word list of 126 items selected from the 167 items in Fussman (1972), and found that Gilgiti shares 85% of the items with Drasi, 81% with Guresi and 80% with Kohistani. The shared vocabulary items consist almost entirely of first order cognates, that is, words sharing both a common origin and a similar phonological development. Larger vocabulary counts would show a lower percentage of cognates. Radloff (1992) conducted recorded text testing of Shina in six locations, including Gilgit, Koli and Jalkot. The mean score of the Kohistan subjects in understanding the Gilgit text was 66% correct, while the mean score of the Gilgit subjects in understanding a Koli text was 61%. The percent of lexical similarity between Gilgiti and Kohistani Shina is quoted by Radloff at between 65% (for Jalkot) and 64% (for Palas and Koli). The conflict in Kashmir makes it impossible to conduct any systematic study of mutual intelligibility between Pakistani dialects of Shina and those

36

Ruth Laila Schmidt

spoken in India. However when I played a tape recording of speech in the Kohistani dialect to a speaker of the Guresi dialect in 1989, he reported that it sounded "quaint" but that he had no difficulty in understanding it. Schmidt (1985) presented the results of lexical sampling conducted by Vijay Kaul during 1980 in the Gures and Dras valleys and in Garkhon in Ladakh. Like Strand, she found two major dialect divisions in Shina: one spoken in and around Gilgit, with an outlier in Ladakh, and the other consisting of dialects scattered throughout the Indus and Jhelum drainage south and east of Gilgit. The Brokskat outlier of Gilgiti (spoken in Garkhon in Ladakh) shares phonological developments with Gilgiti, rather than with the geographically closer Astori group, but has converged with the Balti dialect of Tibetan to such an extent that it is no longer intelligible to speakers of any other Shina dialect. The folk tradition of the Brokskat speakers, recorded by Francke (1977: 67-73), says that their ancestors migrated to Ladakh from Gilgit. The center of the Shina language is generally considered to be Gilgit, and the Gilgit dialect has some features which do not appear in the other dialects, among which are the formation of perfective tenses of transitive verbs with the absolutive -ee-, -if- in all persons, and the infinitive suffix in -ook. However (as will be seen under (5), only Drasi and Kohistani preserve the old agent case in -e ~ -/, —o. Guresi shares features with both Kohistani and Drasi, although Drasi falls on the Gilgiti side of some isoglosses. As a whole, the Astori group preserves some archaic consonant clusters which have been reduced in the other dialects. Certain phonological features of Shina are common to all dialects studied here. One is the preservation of the Old Indo-Aryan (ΟΙΑ) sibilants /s s s/, as mentioned by Morgenstierne 1961. Another is the development of retracted fricatives /c z/ from the ΟΙΑ consonant clusters tr, ks, dr, bhr etc. A third is the development of contrastive pitch accent, which is attested in the Gilgiti (Radloff 1999), Kohistani (Schmidt and Kohistani 1998), Astori and Guresi dialects, and which can be heard in the Drasi dialect as well. In Brokskat, a falling tone can be heard in long vowels, but there is no trace of a rising tone. Table 1 presents comparative lexical items from five dialects of Shina, together with the most likely ancestral form. For information about the transcription used in this paper, please see "Note on transcription of pitch accent" at the end of the paper, immediately preceding the endnotes.

A grammatical comparison ofShina dialects

37

Table 1. Comparative lexical items in five dialects of Shina3 4

1

Drasi

Brokskat

thaas thoonu goos drist äst tfinu kinu krom ic muuzu s i is tätu sfnu joozi tiki cheec cakku

thaas thoono goos dist äst tuno ki'no krom i? muuzi sii's täto sin o joozi tiki cheec cäko

— this got dit as, ast tmo kyono krum — mtuzi sis täato sir) zozi — — caq

achfi paloo zäa dir dif jamacoo jamco puzo pfizo proono poono sääco säacho

achii paloo zäa dii' jamacoo prnzu proonu säacu

achfi phalaä — — jamo puzo — säaci

deez

dees

achfi paloo zäa — jamacoo prfizo — säaco, saaco dees

did' 'to do' 'house' 'handspan' 'eight' 'sha ' 'black' 'work' 'bear' 'mouse' 'head' 'hot' 'horn' 'birch' 'bread' 'field' 'spinning wheel' 'eye' 'tongue' 'brother' 'daughter' 'son-in-law' 'flea' Old' 'dream'

dis

'day'

Gilgiti

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

*sthäpayya theegas thook gösthägoot dit *dristi-5 astääs tiino tiksnäkino Iqsnäkärmankom flcsaic muuzi müsasfrsäsis taptätaato srngasi'rjo bhürjajoözi *tikkatiki ksetracheec cakräcärko

17 18 19 20 21 22 23 24

äksiphalä-? bhrätrduhitr *jämätra*prisupuränä*sväpra-

25 divasä

Kohistani Guresi thaas thoon goos dis äs tfmo kino kom ich muuzo sis täto sigo joozi tiki — —

achfi phalaä zäa

dees

Items no. 2, 3, 9, 11 and 18 group Gilgiti and Brokskat. Items no. 1, 2, 3 and 11 group Kohistani, Guresi and Drasi. Items no. 3, 4, 7, 10 and 22 group Guresi and Drasi. Kohistani falls on the Gilgiti side in items no. 4 and 7. Brokskat patterns like Guresi and Drasi in items no. 4, 7, 16 and 25. The Guresi and Drasi dialects preserve the largest inventory of final consonant clusters, which is an archaic feature. They also possess initial consonant clusters. However, except in the case of no. 22 (and possibly no. 3), these are secondary developments, due in the case of item no. 7 to metathesis

38

Ruth Laila Schmidt

(Morgenstierne 1947: 152-154), or in the case of item no. 23, to vowel contraction. Gil. carko(no. 16) also shows metathesis.

4.

Comparison of verb tense formation

In Gilgiti, Kohistani and Drasi, the distinction between imperfective and perfective aspect in verb tense has been elaborated through processes of grammaticalization, in which autonomous words have been reduced first to bound auxiliaries and then to assimilated suffixes. The aorist, absolutive and old participles provide the stems, ("aorist" refers to a verb form which states a fact, but does not necessarily express an idea of time. The "absolutive" is an uninflected verbal adverb, sometimes called the "conjunctive participle") The grammaticalized words are the past and present tenses of 'be' and the past tense of 'go' (except in Drasi, where the past tense of 'come' functions as an auxiliary in the imperfect and pluperfect tenses). The stems of intransitive verbs are derived, except in the first person, from old participles, which can be found in proverbs and occasionally in texts, but are no longer productive. The first person stem is the aorist. Transitive verbs, on the other hand, have the aorist as their imperfective stem, and the absolutive as their perfective stem.

4.1.

Intransitive verbs

The aorist, present and past forms of 'walk' are shown in Table 2 below in order to illustrate the inflection of intransitive verbs in Shina. Kohistani forms are shown in the first row, while the Gilgiti forms are shown in the second row. The attempted derivations are shown as well. Asterisks show reconstructed forms. The imperfective stems of 'walk' are Koh. yäaza-, Gil. yaäya- 'walking'. Compare Koh. yäaza Joy'a walking fox' (obsolete). When the present tense of the auxiliary han- 'be' is grammaticalized, the initial h- is lost; the final -a of the stem and -a of the reduced auxiliary are assimilated, and often the final unstressed vowel of the auxiliary is reduced or lost. The extra segment -an ~ -en in the 3rd plural in both dialects seems to have the function of avoiding homophony with the third person singular.

A grammatical comparison ofShina dialects Table 2. Inflection of imperfective forms of 'walk' (yäa-, yaä-) Aorist:

1 2 3

may walk'/'I will walk' etc.

Sg-

PL

yaa-z-am yaa-y-am yäa-z-ee yaa-y-ee yaa-z-ee yaa-y-ey

yaa-z-oon yaa-y-on yaa-z-at yaa-y-at yaa-z-an yaa-y-en

Present tense:

walk, I am walking' etc.

1 .m

yaa-z-m-os *thaaw

Forms in which both dialects preserve the absolutive stem 2.Sg.f.

3.Sg.f 3. Pl.m.f

Gilgiti: Kohistani: Gilgiti: Kohistani: Gilgiti: Kohistani:

thee + giee —> theegiee (the accent in giee falls on a short vowel) thee + gyey —> *theegyey > theegyee thee + *gei -4 *theegei > theegi thee + *gyey —> *theegyey > theegi thee + gie -> *theegie > theegie thee + *gye —> *theege > theega

The accent remained on the first element of the compound in the 2.sg.f, 3.sg.f. and 3.pl. forms in both dialects. In the case of the 2.sg.f. and S.sg.f. we can assume that this happened because the accented vowels in the past forms of 'go' were short in both dialects. The 3.pl. is harder to explain, unless we assume that the original auxiliary was short, as in Gilgiti (*gye), and subsequently became lengthened to the modern gyee. A lengthening of accented vowels, however, is not a Kohistani feature. The absolutive stem is completely lost in the Guresi and Drasi perfective paradigms, as shown in Tables 9 and 10. Since space permits, the Guresi and Drasi paradigms are presented with pronouns.

44

Ruth Laila Schmidt

Table 9. Inflection of past tense of 'do' (/A-) in Guresi did, I was doing' etc.

PI.

Sgl.m l.f 2.m 2.f 3.m 3.f

mey thäas mey thees they tha they thi ses/sesi thaw seso thi

bes thiyees

— thsos thiyeet

— ses thiyee säas thiyee

Table 10. Inflection of past tense of'do' (th-) in Drasi •I did, I was doing' etc.

l.m l.f 2.m 2.f 3.m 3.f

Sg-

PI.

mf thäas mf thees tho tha tho the zesi thäaw zeso thi

äsoja thiyees asoja thiyes tshoja thiyeet tsh«ja thiyet zenoja thiyee zzeja thiye

4.3. Innovations The Drasi dialect appears to present some examples of the past tense of the verb 'come' (äalos came'; äalo 'he came' etc.), instead of the verb 'be' (asilos was'; asilo 'he was' etc.) being used as a past auxiliary. The following samples should suffice. Drasi: l.Sg.m 3.Sg.m 1 .Sg.m 3.Sg.m

rnosu themalus was doing' is not limited to 2nd person nor does -t9 behave as other pronominals do. The position of -id and its allomorphs is directly following the verb stem and is before pronominal agreement markers which fill the verb final slot, as in (27). According to DeLancey (1982), -ta came to mark the inverse relationship 2nd >l st person and subsequently was generalized to encode inverses in all persons. Caughley (1982: 189) calls -tea "contrary-information-flow" marker,19 which alerts the addressee that the flow of information is contrary to the expected direction with respect to rank.

88

Karen Grunow-H rsta

Example (27) also demonstrates that the verb in Chepang, unlike Jyarong, may agree with only one of the participants rather than with both the agent and patient. Agreement when only one participant is encoded will be with the agent of a direct configuration, as in (26), and the patient of an inverse, as in (27). Chepang further diverges from the Tibeto-Burman prototype represented by Jyarong in one other important respect: in inverse configurations, not only will the reversal of direction from lower agent to higher patient be encoded on the verb as inverse. The primary object will also be marked by -kay. Thompson (1994: 54) observes that, "Inverse construction applies only to arguments which can take the suffix -kay." Caughley (1982: 63) glosses -kay as "goal" and observes that it covers a variety of semantic roles (patients, recipients, benefactives, malefactives). Thompson, in his pragmatic analysis of the Chepang inverse, argues "with hedges" (1990: 410) that -kay is the direct object marker, but states that he is not aware of any evidence that can conclusively place -kay in the direct vs. indirect object category. As with Magar -ke, a primary function of -kay is to encode recipients and benefactives in ditransitive clauses (28) and (29) therefore, I consider -kay a dative marker and gloss it as such. (28)

(29)

Chepang (Cauehlev 1982) Gopal-Ά haw-lem-kay wa?-nis Gopal-ERG child-PL-DAT bird-DL 'Gopal gave the children birds.'

b3y?-?aka-n give-PT-AG

Chepane (Cauehlev. 1982) nay-kay ija le?-laij ?al-na-ij? you-DAT I get-PUR go-NPT-lE Ί will go to get it for you.'

To sum up, Chepang has a clitic -id that marks inverse relationships between participants in accord with person and number hierarchies. Moreover, we have a direction system that redundantly marks deviation from the hierarchies on both verbs and nouns. Such redundancy would allow a language to omit one of the markers, either the verbal or the nominal. Thompson (1990: 420) notes that in Chepang some inverse constructions are marked by -td? alone on the nominal or the verb, some redundantly by both inverse and case, and some by case alone, as demonstrated below (30)-(32): (30)

Chepang (Caughley 1982) [nominal inverse marker, no dative case] ja?-ft-te? gay? ca? tiger-ERG-iNV get IF Ά tiger may get you.'

Differential dative case marking in Magar (31)

Chepang (Caughley 1982) [inverse marker and dative case] niq-kay Ram-i? ?amh je?-te? na-y? you-DAT Ram-ERG food eat-lNV NPT-lPRO 'Ram eats your food.'

(32)

Chepang (Caughley 1982) [dative and no inverse marker] nai]-7i ija-kay qa? prek ?a-ci 2-ERG I-DAT fish split-PT-2 'You split a fish for me.'

89

Thus, it is possible for a direction marking language to mark inverse configurations with the dative case alone. It would, therefore, be possible for a direction marking system to lose redundant verbal marking entirely and to develop into a system in which high-ranking objects are only marked by case. This redundancy in the direction marking system of Chepang would have opened the way for a shift from verbal to nominal marking and for the use of the dative to mark an NP's position on the nominal hierarchy. There is evidence to suggest that this development occurred, in Magar, under the Indo-Aryan influence of Nepali.

4.3.

Direction marking in Magar

Evidence in Magar that dative marking of high ranking objects has evolved from an earlier direction marking system, which will have borne similarities to Chepang, can be found in Magar formatives which are cognate to both the Chepang dative and inverse. Synchronically, this evidence is found primarily in Syangja, the more conservative of the two dialects under study. Syangja preserves pronominal indexing of subjects on the verb, while such indexing is now lost in Tanahu. In Syangja Magar, -ta and its voiced variant -da, is cognate to Chepang -ta, it appears in 2nd person paradigm as in, (33)

Syangja Magar ηαη-e 2-ERG

hos-ke 3-DAT

bfiyak-da-l hit-Jp-IPF

'You hit him.'

-da could be regarded as simply the 2nd person marker, but what argues against this and suggests that -da is a remnant of a direction marker is its anomalous form and its unusual position and distribution within the Magar verbal template and its resemblance to Chepang in these respects.

90

Karen Grunow-Harsta

Magar, like Chepang, has both prefixal and suffixal pronominal agreement. Prefixes occur in the 1st and 2nd person realis:past tense.20 The prefixal pronominals are transparently related to the free pronouns: -ga, -kag, 1st person singular and plural, and -nan, the 2nd person singular. As seen below in (34) -na is the pronominal agreement form for the 2nd person, in (35) -ga is the 1st person singular pronominal prefix, and in (36) -ka marks the 1st person plural prefix . (34)

Syangja Magar natj-e beskarj na-jy-ei 2-ERG bread PRO-eat-PST 'You ate bread.'

(35)

Syangja Magar rja-i beskay ya-jya-y Is-ERG bread PRO-eat-PRO 'late bread.'

(35)

Syangja Magar kaij-ko-i ka-jyae-s 1-PL-ERG

PRO-eat-PST-PL

'We ate bread.'

The 1st person suffix -ag in (35) and (37) also resembles the free pronoun. The 2nd person plural and honorific -nis (38) and its reduced form -5 in (36) is derived from the number two -nfiis. Both 1st and 2nd person suffixes are in the position characteristic of suffixal pronominal agreement formatives in Magar, that is, in the right most and last obligatory element in the verb complex, following tense, mood and aspect markers, as in the following. (37)

Syangja Magar rja-i medherlAkhA p^[[Ayak beskaη jya-l-arj 1-ERG usually everyday bread eat-be-lPRO Ί usually eat bread every day.'

(38)

Syangja Magar na-ko-ko-i beskay jya-mA-le-nis 2-HON-PL-ERG bread eat-NOM-be-2-HON-PRO 'You are eating bread.'

Differential dative case marking in Magar

91

This is the same verb final position as the pronominal markers in direct constructions of Chepang as in, (39)

Chepang (Caughlev. 1982) rja-fi-pay bsy?-ne?-na-ij I-ERG-DIF give-NPT-2PRO-1 PRO Ί will give it to you.'

-da in Magar, on the other hand, like -te in Chepang,21 does not appear in the rightmost slot, but in the leftmost suffixal slot nearest the verb stem. It is never the final element and precedes tense, mood and aspect markers, as in (40) and (41). (40)

(41)

Svangja Magar naq-e beskarj 2-ERG bread 'You eat bread.'

jya-da-l eal-dd-lPF

Chepang (Caughlev. 1982) naij-kay bsy?-ie?-?a you-DAT give-iNV-PST 'He gave it to you.'

Further indication that -da is not a pronominal form is that it co-occurs with and precedes other pronominal agreement formatives, for example, the 2nd person plural/honorific in (42). (42)

Svangja Magar na-ko-i beskar) jya-dd-nis 2s-HON-ERG bread eat-ite-2HON 'You eat bread.'

Moreover, it is -nis that marks 2nd person in the imperative not -da as in, (43)

Syangja and Tanahu Magar kher-nis-na run-iMP-EMPH 'Run!'

Perhaps the strongest evidence that -da is not simply a 2nd person pronominal is that it appears, as does -te in Chepang, as a prefix in (44)-(48). What's more like Chepang, when a prefix it appears in persons other than

92

Karen Grunow-H rsta

2nd as in (45)-(47). -d9 is part of the 1st person irrealis paradigm (45). It shows up in the negative paradigms of all persons (46) in its voiceless form -1(9) or is reduced to glottal stop.22 (44)

Syangja Magar na-ko-i beskaη a-t-jya-nis 2-HON-ERG bread iRR- in Magar does not function as an inverse, as shown by its use in intransitives (48)-(49). (48)

Syangja Magar naij mis-da-l you sleep-i/p-iPF 'You sleep.'

(49)

Svangia Magar jja mi-t-mis-nA 1 NEG-ifo-sleep-iPF Ί may not sleep.'

Nevertheless, its form and distribution strongly suggest that da is a vestige of a direction marking system of which, now, only the dative is left to

Differential dative case marking in Magar

93

mark primary objects. In effect the Tibeto-Burman direction marking system, under the areal influence of Indo-Aryan, has become dative marking of primary direct objects. This is not to argue that differential object marking systems will necessarily develop out of direction making or inverse systems, but instead to recognize that the possibility exists and may have occurred, as it seems to have done, in Magar.

5.

Direction marking and differential dative case marking

What in a direction marking system would lend itself to develop toward differential dative marking of primary objects such as is found in IndoAryan? Differential dative marking of primary direct objects and direction marking are functionally sympathetic; both mark distinctions within and deviations from the nominal hierarchy. According to Comrie (1981: 135), the most natural transitive construction is one arranged from agent to patient and at the same time from more to less animate as well as from more to less definite. Silverstein (1976: 113) described the definiteness parameters of the hierarchy as expressing "semantic naturalness"; it is natural for a lexically specified NP to function as the agent of a transitive verb whereas the patient will be less specified. In other words, the hierarchy moves from a known or definite to an unknown. And, as Givon (1984) has observed, direct objects most often introduce new unknown information. Bickel and Nichols (2002: 59) bring together these various parameters of the hierarchy under one rubric: "indexability". They pinpoint as its basic variable the ease with which a referent can be indexed or identified. An addressee, for example, is more easily identified than a 3rd party, a human is identified before non-human, an animate before non-animate, a single before an aggregate, a definite before an indefinite. As Bickel and Nichols (2002: 59ff.) note not only are person, number, and animacy parameters of the hierarchy but, cross-linguistically, so too are: anaphoricity, gender, kinship, possession, size, discreetness or segmentability. The hierarchy manifests itself in diverse parameters. Comrie (1981: 199) proposes that "the hierarchy cannot be reduced to any single parameter... but reflects a natural human interaction among a number of parameters, which are related but irreducibly individuated." In Indo-Aryan languages, object marking aligns with the hierarchy, in that a low ranking patient-object is natural and expected and therefore unmarked. However, if a construction is out of alignment with the hierarchy,

94

Karen Grunow-Härsta

if, for example, a low ranked NP holds the higher position in the construction, that NP will be marked. Specifically, a configuration with a patientobject ranking high in animacy and/or definiteness will be marked by the dative case rather than the unmarked absolutive. Direction marking in Tibeto-Burman also indicates alignment and non-alignment with the hierarchy by marking the combination higher agent and lower patient as direct and higher patient with lower agent as inverse. Both the Tibeto-Burman and Indo-Aryan systems mark essentially the same thing: compliance with, or deviance from, the hierarchy. The systems do differ insofar as a direction marking system is typically verbal, whereas a primary object marking system is nominal. Nonetheless, in Chepang we find a system which has moved away from the reconstructed proto-Tibeto-Burman verbal direction marking to a system which also marks nominals, and sometimes only nominals. In Magar, verbal marking of agent-patient inversion is no longer operative; -dd has become fossilized in 2nd person paradigm and, in its allomorphic reduced forms, as a clitic in the irrealis and negative paradigms. Only the dative -ke remains to mark patient-objects that are high in rank. Clearly there can be changes in how the hierarchy manifests itself, not only across but within languages. Judging from Chepang and Magar, these changes are gradual; there are different stages at which a single language and members of a language family may find themselves. With changes in manifestation, it appears that languages can also assume new parameters, particularly under the pressure of areal convergence, as Magar has done via assimilation to Indo-Aryan. Along with the shift from direction marking to dative marking of objects has come a shift from marking relational properties to marking inherent properties. Magar no longer marks the rank of patients relative to agents as Chepang does; rather it marks, following the Indo-Aryan model, the patient's inherent rank as it is accorded them by the hierarchy. Likewise, Magar has also assimilated a second non-relative parameter from Indo-Aryan,23 that is, dative marking of definiteness. 6.

Conclusion

In conclusion, in the Magar dative object marking system we find clear evidence of borrowing from Indo-Aryan languages, specifically Nepali, with which it shares analogous patterns and conditions. However, there is also evidence that dative marking of primary objects in Magar is not simply

Differential dative case marking in Magar

95

a caique. It isn't borrowed in toto from an Indo-Aryan source. Rather, it is transposed onto an earlier direction marking system, one shared with Chepang, a related West Central Himalayish language, among others, and one ascribed to proto-Tibeto-Burman. Such a transposition would have been fostered by the sympathetic functions of direction marking and differential dative marking of primary objects.

Notes 1.

2.

3. 4. 5.

6.

Speakers of Magar are found mainly throughout central and west-central Nepal. Data is from two dialects: Syangja and Tanahu, western and eastern dialects respectively. The Syangja dialect is spoken south of Pokhara and Tanahu is spoken west of Kathmandu and east of the Bagmati River. Fieldwork for the study of Magar was funded by NSF Award Number: SBR9728369. The dialects differ with respect to ergativity in case marking. Syangja is more consistently ergative and marks agents as ergative in all tenses and aspects, whereas Tanahu follows the Indo-Aryan pattern of Nepali and ergativity splits along aspect, i.e. only in the perfective aspect are agents ergative. Verbal agreement follows an accusative pattern; subjects only condition verb agreement. Experiencers (so called "experiencer subjects") are usually dative marked in Magar. Tanahu and Syangja, though dialects of one language, differ in important respects (see Grunow-Härsta 2000). The abbreviations used for glosses in this paper are as follows: 1: first person, IE: first person exclusive, 2: second person, 3: third person, ACC: accusative, AG: agent, ATT: attributive, CAUS: causative, COND: conditional, CIR: circumlative, CNF: confirmation particle, DAT: dative, DEM: demonstrative, D.DEM: distal demonstrative, DER: derivational, DIR: direct, DIP: direct information flow particle, DL: dual, EMPH: emphatic, ERG: ergative, GEN: genitive, HAB: habitual, HON: honorific, IF: indefinite future, INV: inverse, INST: instrumental, IPV: imperfective, IRR: irrealis, LOG: locative, NEG: negative, NOM: nominative, NPT: nonpast, PRT: particle, PASS: passive, POSS: possessive, PL: plural, PRO: pronominal agreement, PST: past, RED: reduplication, REP: reported speech, R.DEM: remote demonstrative, SEQ: sequential converb, SIM: simultaneous converb, and TNS: tense. The term "differential case-marking" is used by Comrie (1981) and Bossong (1998) to refer to split case marking patterns of objects.

96

Karen Grunow-Harsta

7.

Masica (1976: 30) notes that Persian (Farsi) has striking analogies to Hindi and that Farsi, having been a medieval North Indian prestige language, is a likely source for dative marking of definite objects. Bickel (2002) notes an analogous pattern for dative-marked experiencers; he considers them to be Pan South-Asian and he notes that dative-marked experiencers are not a general feature of Tibeto-Burman, where they would usually be coded by ergatives or absolutives (nominatives) though examples of dative experiencers are found in Newari and Tibetan languages and Himalayish languages. Magar is among those Himalayish languages which have dative experiencers. The term "antidative" is also used by Dryer (1981). Bodish languages frequently have a dative *la/ra, Noonan considers this similarity to Nepali -lai to be accidental. The animacy hierarchy is a continuum. Magar and Nepali differ as to their 'cut-off points', in other words, where on the animacy scale entities will cease to marked dative. As seen Magar marks the feature +human with dative, Nepali marks +animate. In Nepali, possessed objects, though typically known entities, hence definite, are not marked with the dative case,

8.

9. 10. 11.

12.

pello gau-ko jsmindar-le hamrohsntwa that village-POSS landlord-ERG POSS servant 'The landlord ofthat village bought our servant.'

kin-yo buy-PST

The Magar dialects are split with respect to dative marking of possessed entities. Tanahu Magar marks possessed objects with the dative while Syangja, like Nepali does not, as seen below: Tanahu myertuij-e ga-o im-ke tree-ERG IS-GEN house-DAT 'The tree hit my house.' Svan^ia hoca-i qa-o im she-ERG IS-GEN house 'She sees my house.' 13.

thawhak-a collide-PST

wäs-le see-iMPF

Further evidence that -ke is an old and indigenous form and not a borrowing from Indo-Aryan comes from its homophony with the nominative/infinitive. This homophony implies an earlier dative/allative stage. A pathway from allative to infinitive is attested across languages and similarly allative to dative, it is likely that Magar -ke is diachronically an allative marker which took two separate paths: one toward infinitive and another toward dative. Support for this comes from related languages such as Lepcha, in which -ka is an allative. In Chepang a -kay still has allative functions. In addition,

Differential dative case marking in Magar

14.

15.

16. 17.

97

Chepang -kay and -ke in Magar are also used in purposive clauses, in time adverbials. This also suggests that -kay may historically have been an allative which has developed into a dative. In addition, a number of Tibeto-Burman languages, particularly those of the Kirant, have dedicated portmanteau agreement markers for 1st and 2nd person (see Bickel 2002). Direction marking as found in Tibeto-Burman can be considered a specialized type of inverse, one which applies not only to third person, but to 1sl and 2nd as well. This shift has occurred, as DeLancey notes, in Jyarong, Rawang and Limbu. -t3 and its allomorphs are not restricted to this position and are, in fact, enclitic and able to attach suffixally or prefixally and adnominally. Examples below from Caughley (1982: 89) show adnominal use of -te?. ten-ta? ?al ?a today INV go-PST 'Today you WENT.' or

ten ?al-ta??a today go-iNV-past 'TODAY you went.' 18.

Chepang agreement paradigm (DeLancey 1981) ld>2 ld>3 2s-3 3-3 3-2 3-1 2s>ld

19. 20.

21.

-ηααη-c-u -rj-c-u -u -u or (ha -taa-η -ίαα-η-ca

Caughley glosses -te, as C1F (counter information flow), for consistency I gloss this as INV (inverse). Such double marking of pronouns is found among the Kiranti languages, for example, Athpare and Camling, as well as in Rung (Jyarong, Trung), Nungish, and Kuki-Chin (Tiddim Chin, Lakher, Southern Chin). Prefixes are also found in the West Central Himalayish languages: Kham and Chepang. Caughley (1982: 52) makes the same observation for -te, that though it is associated with second person, it is not a 2nd person agreement marker. "It is unlike a true pronominal it always precedes the tense affix."

98

Karen Grunow-Horsta

22.

In its reduced form, a glottal stop, it shows up in Tanahu negative irrealis s well, s in, ηα roti mi?-jy-ai l bread NEG-iRR-eat-lRR Ί may not eat bread.'

23.

Masica (1976: 46) observes that the extension of dative marking to definite objects in Indo-Aryan is likely to owe something to areal influence from Turkic and Farsi.

References Bauman, James 1975 Pronouns and pronominal morphology in Tibeto-Burman. Ph.D dissertation. Berkeley: University of California. Bickel, Balthasar 2002 The syntax of experiencers in the Himalayas. International symposium on non-nominative subjects. Tokyo 18-21, December 2002. (draft) Bickel, Balthasar and Johanna Nichols to appear Inflectional Morphology. In Language Typology and Syntactic Description, Timothy Shopen (ed.). Cambridge: Cambridge University Press. Bossong, Georg 1998 Le marquage de 1'experient dans les langues de l'Europe. In Actance et valance dans les langues de l'Europe, Jack Feulliet (ed.), 259294. Berlin: Mouton de Gruyter. Caughley, Ross Charles 1982 The Syntax and Morphology of the Verb in Chepang. Canberra: Australian National University. Pacific Linguistics Series B, No. 84. Comrie, Bernard 1989 Language Universals and Linguistic Typology. Chicago: University of Chicago Press. DeLancey, Scott 1980 An interpretation of split ergativity and related patterns. University of Indiana Linguistics club press. 1981 The category of direction in Tibeto-Burman. Linguistics of the Tibeto-Burman Area 6(1): 83-101. Dryer, Matthew 1986 Primary object, secondary objects, and antidative. Language 62: 808-845.

Differential dative case marking in Magar

99

Givon, Talmy 1984, 1990 Syntax, Vol. 1, 2. Amsterdam/Philadelphia: John Benjamins. Grunow-Harsta, Karen 2000 A comparative sketch of the verb morphology of two Magar dialects. 6lh Himalayan Languages Symposium. May. 7-9, UWM, Milwaukee, Wisconsin. Kepping, Ksenia 1975 Subject and object agreement in the Tangut verb. Linguistics of the Tibeto-Burman Area 2: 219-232. Masica, Colin 1976 Defining a Linguistic Area: South Asia. Chicago: University of Chicago Press. 1981 Identified object marking in Hindi and other Languages. Topics in Hindi Linguistics 2: 16-50. 1991 The Indo-Aryan Languages. Cambridge: Cambridge University Press. Noonan, Michael to appear Recent language change in the Himalaya. Himalayan Linguistics. Pradhan, Krishna 1982 The structure of the simple clause in Nepali. Unpublished University of Wisconsin-Milwaukee dissertation Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, Robert J. W. Dixon (ed.), 112-171. (Linguistic series 22.) Canberra: Australian Institute of aboriginal studies and humanities press. Thompson, Chad 1990 On the treatment of topical objects in Chepang". Studies in Language 14 (2): 405^27. 1994 Passive and inverse constructions. In Voice and Inversion, Talmy Givon (ed.). Amsterdam/Philadelphia: John Benjamins.

Thangmi kinship terminology in comparative perspective Mark Turin

1.

Introduction

In this article, I describe the largely undocumented kinship terminology of Thangmi (Tharni in Nepali), a Tibeto-Burman language spoken in central eastern Nepal by the ethnic group of the same name.1 Through an analysis of Thangmi kinship terms and the relationships they encode, I draw out a number of structural patterns, which I then compare to other TibetoBurman languages, particularly Kiranti and Newar. There is no shortage of articles dealing with the kinship structures and terminologies of Nepal's various ethnic groups. The 1970s in particular saw scholars focussing their attention on the topic: In 1974, Victor Doherty studied both Brahmin-Chetri and Gurung kinship, in 1975, Richard Burghart described the role of kinship in Janakpurdham pilgrim groups, and in the same year, Nick Allen published an article on Byansi kinship. A year later, in 1976, Allen analysed Sherpa kinship and in 1977, Shresthacharya offered a topology of Newar kinship. Michael Vinding published a report on the kinship terminologies often ethnic groups speaking Bodish languages in 1979, and in 1985, Irene Davids and George van Driem described Limbu kinship terminology. More recently, in 1999, Subhadra Subba Dahal published an incomplete list of Thami kinship terms.2 Alongside anthropologists, many linguists working in the Himalayan region have provided diagrams of the kinship terms found in the languages they describe (see van Driem 1993, 1995; Rutgers 1998; Tolsma 1999). The present article is situated within this descriptive tradition of Himalayan linguistics and anthropology. A word of warning at the outset will remind the reader of the challenges inherent in kinship studies. Bronislaw Malinowski once wrote that "linguistic analysis inevitably leads us into the study of all the subjects covered by Ethnographic field-work" (1945: 302). One such subject is kinship terminology, a topic as much rooted in linguistics as in anthropology. The manner by which people construct and perceive their relationships with others has a direct impact on the words that they use to describe them. The ethno-

102

Mark Turin

centric nature of kinship studies was not lost on Arthur Hocart, who made the pertinent observation that: All our difficulties spring from a preconceived idea that kinship terms everywhere try and express the same thing as they do in Aryan ... languages. (1937: 547)

At best, a cogent linguistic analysis of kinship terminology illuminates specific culturally-defined relationships. At worst, the blending of anthropology and linguistics in kinship studies creates a weak mixture of nebulous social observations combined with the impenetrable abbreviations of linguistic code. Hocart's perfect parody of the latter type of analysis warrants citing in full: A Fijian introduces his wife as wati, so the word is noted as "wife". When it is found that there are hundreds of wati's [sic] who are not his wives, the first translation is not abandoned, but all other uses are explained as extensions: these women, it is explained, are called wives because he might marry any of them if the family so decided; they are wives by anticipation, "potential wives". Upon this muddled lexicography has been built up a whole edifice of primitive promiscuity. (1937: 546)

In an attempt to avoid such difficulties, the present article offers a synchronic analysis of Thangmi kinship terms and then places them in a comparative and historical perspective. 2.

The geographical location of Thangmi dialects and speakers

Two mutually unintelligible dialects of Thangmi exist, spoken in the Dolakhä and Sindhupälcok districts of Nepal respectively. The dialects differ from one another in terms of phonology, nominal morphology, verbal morphology, and also in lexicon. While a number of the lexical differences can be explained by regular morphophonological alternations, these rules do not account for even half of the differences. The Dolakhä dialect of Thangmi exhibits a more complete verbal agreement system while the Sindhupälcok dialect boasts more complex nominal morphology, particularly for locative suffixes and numeral classifiers. In the present paper, I emphasise the differences between kinship terms in the two dialects. Before turning to the details of these kinship terms, however, the geographical distribution of the speakers must be outlined.

Thangmi kinship terminology

103

While the Dolakhä dialect of Thangmi is spoken throughout the central and northern reaches of the district, the Sindhupälcok dialect is spoken only in a handful of villages in the far-eastern valleys of Sindhupälcok. The four villages in which the Sindhupälcok dialect of Thangmi is spoken are flush with the western border of Dolakhä. There is no reason why dialectal differences and isoglosses should necessarily follow the path of political subdivisions and administrative units, but to a greater degree than in many other districts of Nepal, the districts of Dolakhä and Sindhupälcok are separated by a prominent geographical feature: kälincok dadä. This ridge runs north to south, separating the districts from one another, and reaches a maximum elevation of 3,810 metres. While there are many Himalayan ethnic groups for whom such a ridge would pose no obstacle to communication and social interaction, this is not the case for the Thangmi community. There is relatively little intermarriage between the Dolakhä Thangmi and their Sindhupälcok neighbours, and little socio-cultural contact. Generally speaking, it would be fair to say that the Kälincok ridge effectively prohibits cultural and linguistic exchange. 3.

Thangmi ancestors, clans and marriage rules

With regard to marriage, the Thangmi practice group-endogamy and clanexogamy, although adherence to the latter is stricter than to the former. Thus, while it is considered wrong to have romantic relations with an opposite sex relative until seven generations have elapsed between consanguineal relatives, sexual relations culminating in marriage with non-Thangmi partners are not taboo. Ethnic group endogamy is preferential rather than prescriptive, while clan endogamy, including sexual relations with crosscousins, is tantamount to incest. The existence of a set of Thangmi female or women's clans, passed down from mother to daughter, is important. A descent structure for women through female-only clans is rare among the Tibeto-Burman ethnic groups of Nepal, and uncommon elsewhere in Central and South Asia.3 Male clan names in the two Thangmi-speaking areas also vary greatly and no single clan has primacy across all dialect boundaries. According to the Thangmi origin story, the male ancestor, progenitor and forefather of all Thangmi is known as Ya?apati Chuku. His omnipresence and importance may explain the existence of both 'father' and chuku 'fatherin-law' in his name, two highly distinct kinship roles not readily conflated in

104

Mark Turin

Thangmi society. Ya?apati Chuku is, however, both an αρά and a chuku: 'Father' to all Thangmi men and 'father-in-law' to all Thangmi women. The female Thangmi ancestor is known interchangeably as Sunari Ama and Sunari Aji. She fulfils both the role of 'mother' to all Thangmi women and 'mother-in-law' to all Thangmi men, and in so being is the female ancestor and foremother of all Thangmi. The variation in her second name, ama and aji, parallels that of apa and chuku described above: In Thangmi, ama denotes 'mother' and aji 'mother-in-law'. It should be noted that ama also denotes 'mother' in Nepali and that Thangmi aji agrees with with Classical Newar aji 'grandmother, paternal or maternal' (J0rgensen 1936: 18) as well as with Hindi aji or aji 'paternal grandmother' (McGregor 2002: 82).4 Why are these two ancestors interchangeably referred to as 'parent' and 'parent-in-law'? One explanation may be the existence of separate male and female clans. Most Thangmi are well aware of the explicitly incestuous nature of their origin story and it continues to cause shame and embarrassment to them (Turin 1999: 17). The overtly incestuous account of sibling marriages in the Thangmi origin story may be rationalised by reckoning women's clan allegiance solely in terms of descent from their mothers, and men's clan membership as similarly deriving solely from their fathers. Following this logic, even opposite sex siblings with the same biological parents will never be of the same clan since they inherit their clan identity from their same-sex parents. This has the paradoxical effect of "de-incestuising" sibling unions. Whether female clans were an adaptive reaction to the incestuous backdrop of the Thangmi genesis narrative or whether the existence of female clans made it possible for siblings to marry, thereby not contravening the incest taboo, remains a rhetorical question.5 Despite the prohibitions outlined above, most Thangmi men and women appear to be unaware of their clan affiliation. When a wedding is planned, Thangmi ritual practitioners, known locally as guru,6 are called to ratify the alliance. In such moments, guru are required to recall the clan memberships of both bride and groom, and function as archivists of oral history and preventers of incest. 4.

Family structure and the spacing of settlements

Victor Doherty noted the Brahmin-Chetri custom of "spacing homesteads widely rather than grouping houses into a nucleated village with the fields of all the residents surrounding this" (1974: 27) and concluded that "the settlement pattern scatters the members of the society, so do the marriage and kinship rules" (1974: 37). A similar pattern is observable in the Thangmi-

Thangmi kinship terminology

105

speaking area: Most houses are built far apart from one another, and each dwelling is surrounded by fields. In Thangmi villages, it is rare to see houses packed close together with arable lands surrounding the entire settlement. This scattering may be explained in a number of ways. An economically-motivated argument might suggest that Thangmi land holdings are predominantly small (due in part to generations of land exploitation) and that villagers build their homes within the boundaries of their land. This explanation is not sufficient, however, and does not explain why even wealthy Thangmi families with substantial land holdings dispersed across a wide area choose to live far apart from one another. A more convincing explanation lies in the observation that Thangmi households are predominantly nuclear rather than joint: Children leave the parental home to establish independent homes early, and grandparents do not live with their grandchildren. While this is uncommon in hill Nepal, more uncommon still is the practice that when one grandparent dies, the other grandparent continues to live alone and does not move in with one of their children. In the Thangmi-speaking villages of Dolakhä and Sindhupälcok, the majority of Thangmi couples above the age of 55 are self-sufficient and live alone. It is a common sight to see 60-year old men and women carrying wood, fetching water, tending animals and cooking for themselves. The few immigrants to the area from Nepal's other ethnic groups and castes vocally express their outrage at this cultural practice. In particular, Newars living in otherwise predominantly Thangmi villages are none too impressed by what they see as a flagrant disregard for seniority and age. This opprobrium is of little concern to Thangmi couples, who view nuclear rather than joint families as both a cultural norm and a marker of their ethnic identity. While it is tempting to map elements of the cultural practices outlined above onto the structures of Thangmi kinship terminology, we must be wary of reductive or causative logic. Nevertheless, I believe that elements of the cultural context of Thangmi familial life may be represented in the linguistic terms used to describe social relationships. As the remainder of this article will show, the Thangmi language does indeed encode a number of kinship relationships in culturally revealing ways.

5.

Representing kinship

If the aim is to understand the structure of a kinship system through an analysis of the categories denoted by the indigenous terms, a clear crosscultural metalanguage for kinship is needed.

106

Mark Turin

Over time, certain modes of representing kinship relationships have emerged as de facto standards, ranging from the taxonomical tree-like model with circles and triangles to a list or glossary full of abbreviations (see Vinding 1979: 208-220). In this article, I have opted to represent kinship relationships in both ways, with kinship charts offered as an appendix. First, I present the kinship terminology of Thangmi. Second, I offer a morphemic analysis of elements and components of Thangmi kinship terms. Third and finally, I relate these terms to their Tibeto-Burman etyma and draw some concluding remarks. Table 1 below shows the common abbreviations used for referring to kinship relationships. Abbreviations may be combined to indicate complex relationships. Table 1. Standard abbreviations used for kinship relationships mother sister daughter wife elder e.g. 6.

(M) (Z) (D) (W) (e)

father brother son husband younger

(F) (B) (S) (H)

(y)

(FyZ) = father's younger sister

Thangmi kinship terms

The Thangmi language differentiates kin on the basis of generation, age within a specific generation, gender, in-law relationships and to a limited extent, kin through sibling relationship versus kin through one's spouse. The gender of the speaker can be a differentiating factor, but not in all situations. There is no single or specific word for 'spouse'. In addition to distinctions on the basis of generation, Thangmi differentiates on the basis of age within generations by marking the relative age of kin with respect to the speaker. Table 2 below presents the Thangmi sibling terms from the Dolakhä dialect. Table 2. Sibling kinship terms from the Dolakhä dialect of Thangmi elder brother elder sister

(eB) (eZ)

biibu fete

younger brother younger sister

(yB) (yZ)

hi/ humi

Thangmi kinship terminology

107

Apart from one notable exception, sibling kinship terms are common to both dialects. In the Sindhupälcok dialect of Thangmi, the 'sex principle' is not strictly upheld: somewhat akin to Limbu (Davids and van Driem 1985: 122-123), both younger brother and younger sister are referred to by the same gender-indifferent term hu 'younger sibling'. Aware of the confusion this can generate, speakers of the Sindhupälcok dialect may qualify the term hu 'sibling' by prefacing it with either calaca 'male, man' or camaica 'female, woman', as in Table 3 below. Table 3. Younger sibling kinship terms from the Sindhupälcok dialect younger brother younger sister

(yB) (yZ)

(calaca) hu (camaica) hu

Age in generation also functions distinctively in kinship terms referring to members of the speaker's parents' generation. Thangmi kinship nomenclature distinguishes between uncles and aunts by blood and uncles and aunts by marriage, somewhat like Gurung (Pignede 1966), but unlike Limbu (Davids and van Driem 1985: 124). In the Dolakhä dialect, the age distinction is upheld for father's brothers only and not for father's sisters, as illustrated in Table 4. Table 4. Father's siblings kinship terms from the Dolakhä dialect father's elder brother father's younger brother father's sister (elder or younger)

(FeB) (FyB) (FZ)

jekhapa ucyapa nini

In the Sindhupälcok dialect, the age distinction is upheld across all of father's siblings, as illustrated in Table 5. Table 5. Father's siblings kinship terms from the Sindhupälcok dialect father's elder brother father's younger brother

(FeB) (FyZ)

jhya?apa pacyu

father's elder sister father's younger sister

(FeZ) (FyZ)

jhya?ama nini

108

Mark Turin

In Sindhupälcok, Thangmi speakers do not differentiate for the gender of a younger sibling, but do differentiate on the basis of age for a father's sister. On the other hand, speakers of the Dolakhä variety of Thangmi distinguish the gender of a younger sibling, but do not differentiate between father's elder sister and father's younger sister. In the Sindhupälcok dialect, age in generation relative to the speaker's father is differentiated. In the Dolakhä dialect, parallel uncles, of the same gender as the parent, are distinct on the father's side, while cross aunts, of different gender to the parent, are not. This is reminiscent of Limbu kinship terminology in which there is only one term for 'father's sister' (FZ), nya?, and one for 'mother's brother' (MB), nwa?, indeterminate of age (Davids and van Driem 1985: 123). The age in generation principle also functions distinctively in kinship terms for mother's siblings in both dialects. In both Dolakhä and Sindhupälcok, the Thangmi language upholds the age distinction for all four siblings of ego's mother, as shown in Tables 6 and 7. Table 6. Mother's siblings kinship terms from the Dolakhä dialect mother's elder brother mother's younger brother

(MeB) (MyZ)

palam malam

mother's elder sister mother's younger sister

(MeZ) (MyZ)

jekhama macyu

Table 7. Mother's siblings kinship terms from the Sindhupälcok dialect mother's elder brother mother's younger brother

(MeB) (MyZ)

palam mou

mother's elder sister mother's younger sister

(MeZ) (MyZ)

jhya?ama phus?ama

In the Dolakhä dialect, the age in generation principle also applies to aunts and uncles by marriage on the father's side. Aunts by marriage on the father's side (FBW) are differentiated for age, while uncles by marriage on the father's side (FZH) are not, following the pattern outlined above. Table 8 illustrates the kinship terms for the spouses of father's siblings in the Dolakhä dialect of Thangmi.

Thangmi kinship terminology

109

Table 8. Kinship terms for the spouses of father's siblings from the Dolakhä dialect father's elder brother's wife father's younger brother's wife

(FeBW) (FyBW)

jekhama ucyama

father's sister's husband (elder or younger)

(FZH)

mama

It is worth noting that the kinship term for father's sister's husband (FZH) in the Dolakhä dialect, mama, is homophonous with the Nepali kinship term for mother's younger brother (MyB), mama. I believe these terms to be etymologically unrelated. In the Sindhupalcok dialect, the age distinction is upheld across all of father's siblings' spouses, as illustrated in Table 9. Table 9. Kinship terms for the spouses of father's siblings from the Sindhupalcok dialect father's elder brother's wife father's younger brother's wife

(FeBW) (FyBW)

jhya?ama macyu

father's elder sister's husband father's younger sister's husband

(FeZH) (FyZH)

jhya?apa mou

In the Dolakhä dialect, the age in generation principle similarly applies to aunts and uncles by marriage on the mother's side. Uncles by marriage on the mother's side (MZH) are differentiated for age, while aunts by marriage on the mother's side (MBW) are not, following the pattern outlined above. Table 10 illustrates the kinship terms from the Dolakhä dialect used for spouses of the mother's siblings. Table 10. Kinship terms for spouses of the mother's siblings from the Dolakhä dialect mother's elder sister's husband mother's younger sister's husband

(MeZH) (MyZH)

jekhapa pacyu

mother's brother's wife (elder or younger)

(MBW)

nini

In the Sindhupalcok dialect, the age distinction is upheld across all of the mother's siblings' spouses, as shown in Table 11.

110

Mark Turin

Table 11. Kinship terms for spouses of the mother's siblings from the Sindhupalcok dialect mother's elder sister's husband mother's younger sister's husband

(MeZH) (MyZH)

jhya?apa phus?apa

mother's elder brother's wife mother's younger brother's wife

(MeBW) (MyBW)

malam nini

There are some crucial differences between the Dolakha and Sindhupalcok dialects regarding uncles and aunts by blood and by marriage, as illustrated in the examples above. In the Dolakha dialect, cross aunts (father's sisters) and their spouses are not differentiated for age, while in the Sindhupalcok dialect they are. Other Tibeto-Burman languages configure these kinship relationships differently. In Limbu, for example, kinship nomenclature makes no distinction between uncles and aunts by blood versus uncles and aunts by marriage (Davids and van Driem 1985: 123-124), while in Gurung, the terms for aunt and uncle by blood are distinct from those for aunt and uncle by marriage, as in the Sindhupalcok dialect of Thangmi. With the exceptions noted above, then, uncles and aunts by marriage are terminologically differentiated in Thangmi on the basis of age within generation of spouse relative to ego's parent. In both dialects of Thangmi, both cross-cousins and parallel cousins are classified as siblings, and a marriage taboo exists between ego and any cousin, cross or parallel. Unlike other Tibeto-Burman groups in Nepal, such as the Thakali, Tamang or Gurung, the Thangmi do not practice preferential cross-cousin marriage (i.e., ego with MBD or FZS). In the Thangmi language, the age of ego determines whether a male cousin is called 'elder brother' or 'younger brother', not the birth order of their parents. In other words, the son of ego's paiam (MeB) is termed hu (yB) when younger than ego, even though this individual's father is older than ego's mother. As with younger siblings, the Sindhupalcok dialect does not differentiate between gender for cousins: Both younger male cousin and younger female cousin are termed hu. This is illustrated in Table 12. Table 12. Kinship terms for cousins in both dialects elder male cousin elder female cousin

bubu tete

younger male cousin younger female cousin younger cousin (male or female)

hu (Dolakha) humi (Dolakha) hu (Sindhupalcok)

Thangmi kinship terminology

111

While the kinship terms for cousins are not differentiated by the gender of their parents, the terms for nephews and nieces are. In Dolakhä as in Sindhupälcok, the children of ego's brothers are classificatory children and responsibilities to them are akin to responsibilities for one's own children. On the other hand, sister's children are regarded as nephews and nieces, and the main social responsibility for them lies with the siblings of the sister's husband (ZHB and ZHZ). The kinship terms for ego's children and the children of ego's siblings are the same in both dialects of Thangmi, and are given in Table 13 below. Table 13. Kinship terms for ego's children and children of ego's siblings in both dialects son, brother's son daughter, brother's daughter

(S), (BS) (D), (BD)

ca camai

sister's son sister's daughter

(ZS) (ZD)

baine bini

In the Sindhupälcok dialect, neither the gender of a grandchild nor the gender of his or her parent is expressed in the kinship term. Daughters' sons and daughters, as well as sons' sons and daughters are all referred to as cacha (SS, SD, DS, DD). The same does not hold for the Dolakhä dialect, in which the gender of a grandchild is made explicit, but the gender of his or her parent remains immaterial. This is demonstrated in Table 14. Table 14. Kinship terms for grandchildren in the Dolakhä dialect son's son, daughter's son son's daughter, daughter's daughter

(SS), (DS) (SD), (DD)

cacha cachi

A handful of speakers of the Sindhupälcok dialect, particularly from the village of CokatT, insist that Thangmi kinship terms for great-grandchildren, great-great-grandchildren and even great-great-great-grandchildren exist and are in common use. While older speakers from this village use these terms, they have not been corroborated by other native speakers beyond this village. It is interesting to note that while grandchildren are not differentiated for gender, as both grandson and granddaughter are termed cacha, gender differentiation does exist for the following generations, as shown in Table 15 below. It is likely that the element /-ni/, present in the kinship terms for greatgranddaughter and great-great-granddaughter, is of Nepali origin.

112

Mark Turin

Table 15. Kinship terms for four generations of grandchildren in the Sindhupälcok dialect (CokatT village) grandson granddaughter great-grandson great-granddaughter great-great-grandson great-great-granddaughter great-great-great-grandson great-great-great-granddaughter

cache cacha cayä cayäni cuyu cuyüni Oyu Oyuni

Similar to Thangmi kinship terms for siblings, the terms for brothers-inlaw and sisters-in-law distinguish both for sex of referent and for relative age of sibling. Aside from a regular phonological variation, these terms are the same in both dialects, as shown in Table 16. Table 16. Kinship terms for siblings-in-law in both dialects elder brother's wife

(eBW)

pairi ~ poiri

elder sister's husband

(eZH)

jarphu

younger brother's wife younger sister's husband younger sister's husband

(yBW) (yZH) (yZH)

hu wari (humi) damari (Dolakha) jyamai(Sindhupälcok)

The Thangmi term wari is rarely used without a preceding kinship term to qualify it. On its own, with no qualification, wari occupies a lexical domain identical to Nepali buhäri. Both Thangmi wari and Nepali buhäri have a range of meanings which fall within the category of 'wife of male kin younger than ego'. The lexical item wari may be a loan from Nepali buhäri, since the intervocalic [h] in buhäri elides in allegro Nepali speech, and the initial [b] in Nepali is often realised as [w] in Thangmi. In combination with a prefixed qualifier Thangmi wari covers a range of meanings including 'younger brother's wife' (yBW), 'brother's son's wife' (BSW), 'son's wife' (SW), 'son's son's wife' (SSW), 'daughter's son's wife' (DSW) and 'sister's son's wife' (ZSW), as illustrated in Table 17.

Thangmi kinship terminology

113

Table 17. The range of meanings for Thangmi wari in both dialects younger brother's wife

(yBW)

hu wari

brother's son's wife

(BSW)

ca wari

son's wife son's son's wife daughter's son's wife sister's son's wife

(SW) (SSW) (DSW) (ZSW)

ca wari cacha wari cacha wari baine vrar/(Dolakhä) bini wari (Sindhupalcok)

A semantic field similar to that of wari is occupied by the Thangmi kinship term daman (Dolakhä) QT jyamari (Sindhupalcok). The term daman ~ jyamari is also rarely used without a preceding qualifier. Without qualification, daman ~ jyamari occupies a domain identical to Nepali jovoi, which has a range of meanings encompassed within 'husband of female kin younger than ego'. It is possible that Thangmi daman ~ jyamari is etymologically related to Nepali jovat In combination with a prefixed qualifier, Thangmi daman ~ jyamari has a range of meanings which include 'younger sister's husband' (yZH), 'brother's daughter's husband' (BDH), 'daughter's husband' (DH), 'son's daughter's husband' (SDH), 'daughter's daughter's husband' (DDK) and 'sister's daughter's husband' (ZDH), as shown in Tables 18 and 19 below. Table 18. The range of meanings for Thangmi damari m the Dolakhä dialect younger sister's husband brother's daughter's husband daughter's husband son's daughter's husband daughter's daughter's husband sister's daughter's husband

(yZH) (BDH) (DH) (SDH) (DDH) (ZDH)

humi daman camai daman camai daman each/ daman cachi daman' bini daman

Table 19. The range of meanings for Thangmi jyamari in the Sindhupalcok dialect younger sister's husband brother's daughter's husband daughter's husband son's daughter's husband daughter's daughter's husband sister's daughter's husband

(yZH) (BDH) (DH) (SDH) (DDH) (ZDH)

jyamari camai jyamari camai jyamari cacha jyamari cacha jyamari bini jyamari

114

Mark Turin

Related to the term daman is the kinship term da/nanu', with a general meaning of 'sister of husband of female kin younger than ego' and specifically used for younger sister's husband's sister (yZHZ) and daughter's husband's sister (DHZ), as shown in Table 20. This term is used only by speakers of the Dolakhä dialect. Table 20. The two meanings for Thangmi damarnim the Dolakhä dialect younger sister's husband's sister daughter's husband's sister

7.

(yZHZ) (DHZ)

damami damarni

Sex of speaker

The kinship systems of certain Tibeto-Burman languages distinguish for sex of speaker. In certain dialects of Tibetan, for example, kinship terms distinguish between elder siblings on the basis of the speaker's gender. A sex distinction is also made in Limbu, and Benedict maintains that it "must be regarded as archaic for the group as a whole" (1941: 319, cited in Davids and van Driem 1985: 125). While sex of speaker is not generally distinctive in Thangmi, a select few kinship terms, such as those for siblings-in-law and uncles- and aunts-in-law, do reflect the Tibeto-Burman sex of speaker criterion. A number of structural features of the Thangmi kinship terms which distinguish for sex of speaker should be highlighted. In the Dolakhä dialect of Thangmi, all such terms relate to affinal rather than blood kin, people related by marriage rather than descent, and are secondary extensions of the primary meaning of a specific kinship term. For example, qji, which means first and foremost 'mother-in-law' has the secondary meaning 'wife's elder sister'. Likewise, tete has a primary meaning of 'elder sister' and a secondary meaning of 'husband's elder sister'. Kinship terms from the Dolakhä dialect which are distinctive for sex of speaker are shown in Table 21 below, and those from the Sindhupälcok dialect are presented in Table 22.

Thangmi kinship terminology

115

Table 21. Kinship terms distinctive for sex of speaker in the Dolakha dialect wife's elder sister husband's elder sister

(WeZ) (HeZ)

aji tete

wife's mother's younger brother husband's mother's younger brother

(WMyB) (HMyB)

malam chuku mou chuku

wife's mother's younger brother's wife husband's mother's younger brother's wife

(WMyBW) maJam aji (HMyBW) nini aji

wife's mother's younger sister's husband (WMyZH) husband's mother's younger sister's husband (HMyZH)

pacyu chuku ocyana chuku

Table 22. Kinship terms distinctive for sex of speaker in the Sindhupälcok dialect wife's elder sister husband's elder sister

(WeZ) (HeZ)

jhya?ama malam ca

wife's elder sister's husband husband's elder sister's husband

(WeZH) (HeZH)

bubu jhya?apa

wife's elder sister's son husband's elder sister's son

(WeZS) (HeZS)

ca baine

wife's elder sister's daughter husband's elder sister's daughter

(WeZD) (HeZD)

camai bini

wife's elder sister's son's wife husband's elder sister's son's wife

(WeZSW) (HeZSW)

ca wari bini wari

wife's elder sister's daughter's husband husband's elder sister's daughter's husband

(WeZDH) (HeZDH)

camai jyamari bini jyamari

wife's mother's younger brother husband's mother's younger brother

(WMyB) (HMyB)

mou chuku malam chuku

wife's mother's younger brother's wife husband's mother's younger brother's wife

(WMyBW) (HMyBW)

nini aji malam aji

wife's mother's younger sister's husband husband's mother's younger sister's husband

(WMyZH) (HMyZH)

ocyana chuku pacyu chuku

116

Mark Turin

It is apparent that there is a greater differentiation for sex of speaker in the Sindhupälcok dialect than in the Dolakhä dialect. In the Sindhupälcok dialect, the sex of speaker criterion is applied to offspring-in-law, specifically the children of ego's spouse's elder sister, while in the Dolakhä dialect the differentiation does not extend so far. Second, aside from the extension of the system to include children of one's spouse's elder sister, the kin for whom sex of speaker is differentiated are the same in both dialects: spouse's elder sister, spouse's mother's younger brother and his wife, and spouse's mother's younger sister's husband. Why these terms should be differentiated for sex of speaker when others are not is unclear, since none of the differentiated kin are socially or ritually more important to ego than other similarly positioned kin in Thangmi society. Finally, it is intriguing to note that the sex-differentiated terms for mother's younger siblings are reversed in the two dialects. This reversal may be attributed to linguistic attrition and language decay, processes by which rarely used and socially less important kinship terms may be the first to be confused and eventually lost. Moreover, while Benedict posited that the archaic Tibeto-Burman sex of speaker distinction applied to elder sibling kinship terminology, the above examples show that Thangmi also differentiates for spouse's elder sister and mother's younger siblings. In the Dolakhä dialect, a further terminological distinction is made between siblings-in-law related to ego via a spouse (married kin) versus relationships through ego's siblings (blood kin). For example, the terms for sister-in-law differ depending on whether the referent is related to the speaker through the speaker's sibling or through the speaker's spouse: 'elder brother's wife' (eBW) pain versus 'wife's elder sister' (WeZ) aji. The Dolakhä dialect makes this distinction for all siblings-in-law. Referring to siblings-in-law and their spouses with one of the terms for ego's own siblings, bubu (eB), tele (eZ), hu (yB) and humi (yZ), may be explained in one of two ways. On the one hand, the relationship between an ego and his or her spouse's siblings is often close, which may be reflected in the extension of kinship terminology from ego's own siblings to those of ego's spouse. Many Thangmi men and women view their spouse's siblings as de facto brothers and sisters.7 On the other hand, the use of sibling kinship terms for siblings-in-law may not carry a particular cultural meaning. After all, familial kinship terms, such as the terms denoting grandparents, parents, siblings and children, are the most commonly used forms of ad-

Thangmi kinship terminology

117

dress for non-kin in both professional and social encounters. Kinship terms for uncles and aunts are less frequently used to address non-kin, since their use invokes a notional kinship bond which reckons descent through one parent over the other.8 In both dialects of Thangmi, the kinship terms bubu 'elder brother' and tete 'elder sister' are widely used as respectful terms of address for male and female strangers around the age of the speaker.9 A terminological distinction absent in Thangmi, but present in Nepali and in many of the Tibeto-Burman languages of Nepal, is one which indicates an individual's age within generation by order of birth. In certain Tibeto-Burman languages, such as Limbu, kinship terms for age within generation are derived from ordinal numbers, while in Nepali, specifically gendered kinship forms exist for 'first-born male', 'first-born female', 'second-born male', 'second-bom female', etc.10 The only comparable Thangmi form isjekha 'large, big', which is occasionally used as a term of address to attract the attention of the eldest male child. Its use conveys a sense of anger or dissatisfaction on the part of the speaker, which may explain its limited use. The examples and tables provided above illustrate a number of the most commonly occurring kinship terms and structural issues present in both dialects of Thangmi. The kinship terms and the relationships they encode are given in full schematic form at the end of this article.

8.

The morphology of Thangmi kinship terms

A number of Thangmi kinship terms yield readily to language-internal morphological analysis. A point worth noting at the outset is the frequency of reduplicative, near reduplicative or mirrored forms, such as bubu 'elder brother', cacha 'grandson', mama 'father's sister's husband', nini 'father's sister' and tete 'elder sister'. This manner of doubling is a common and well-attested feature of kinship terminologies in many of the world's languages, including the Tibeto-Burman languages spoken in Nepal. The kinship term cacha 'grandson' is a reduplicative form of ca 'son' combined with an aspirated second element cha. Younger and imperfect Thangmi speakers frequently use an unaspirated second element, thus *caca, a form universally rejected by fluent Thangmi speakers. The morphological composition οι cacha is transparent: 'son of son'.

118

Mark Tür in

Both dialects of Thangmi have a number of kinship terms in which female gender is indicated by a final /-i/ or I-ml, such as cacha 'grandson' versus cachi 'granddaughter', thone Old man' versus thoni Old woman', daman 'son's wife' versus damarni 'daughter's husband's sister', and jarphu 'elder sister's husband' versus jarphuni 'elder sister's husband's sister'. This sex differentiation may be a remnant of an old system of grammatical gender or an indication of more recent contact with Indie languages such as Nepali, in which the ending is common in female kinship terms such as chord 'son' versus chori 'daughter', and is also used to indicate biological sex in non-human animates, such as kukur 'hound' versus kukurm 'bitch'. The suggestion that this final /-i/ may be loaned from Nepali gains credibility in light of the gendered nature of other Thangmi kinship terms. In Thangmi kinship terminology, as in the lexicon in general, biological gender can be emphasised or disambiguated with one of two gendered prefixes, (FEM) or (MALE). In the non-kinship lexicon, examples of this productive morphological process are mamct-sya (FEM-bovine) 'cow', papa-kucu (MALE-dog) 'hound' and mama-yante (FEM-quern) 'lower part of a millstone or quern'. In the kinship lexicon, these same gendered prefixes are found as fused elements of the noun, such as macyu (MyZ) 'mother's younger sister' versus pacyu (MyZH) 'mother's younger sister's husband' or malam (MyB) 'mother's younger brother' versus palam (MeB) 'mother's elder brother'. These terms function as gendered pairs, offering both a female and a male side of a respective kinship relationship. It is interesting to note, however, that the structural opposition between macyu : pacyu is semantically distinct from malam : palam. In the Dolakhä dialect, the first pair refers to an aunt and her husband, while the second pair refers to a younger uncle and an elder one. My working hypothesis is that the and prefixes are more likely to be indigenous to Thangmi, given their presence throughout the whole lexicon, while the final element /-i/ found in some female kinship terms is a later loan from Sanskrit or Nepali. A further interesting morphological feature relates to the compounding of kinship forms, such as jekhama in Dolakhä and jhya?ama in Sindhupälcok meaning 'mother's elder sister' (MeZ), and both derived fromjekha ~ jhya 'big, large, senior' and ama 'mother'. In the Dolakhä form, the /a/ elides and the form is fused, while in the Sindhupälcok dialect, the bound-

Thangmi kinship terminology

119

ary of the two elements is glottalized producing /-a?a-/. The same morphophonological patterns are at play in jekhapa (Dolakhä) and jhya?apa (Sindhupälcok) 'father's elder brother' (FeB) respectively. Other compounded kinship terms make use of the elements ocyana or ucya, both meaning 'small', as in the non-fused kinship term ocyana chuku, literally 'small father-in-law', meaning 'husband's mother's younger sister's husband' and the fused ucyapa, literally 'small father', meaning 'father's younger brother'. These forms may be caiques from Nepali, as in Thangmi jekhapa, literally 'big father', 'father's elder brother' (FeB) from Nepali thülo bä 'father's elder brother' (FeB), with the same literal meaning. Another qualifying prefix used in compounded kinship terms is thone (Dolakhä) or chyode (Sindhupälcok) Old male', and the associated female form, thoni (Dolakhä) or chyodi (Sindhupälcok) Old female'. Used as prefixal qualifiers, these lexical items add a generation to the kinship term which they prefix, such as chuku 'father-in-law' versus thone chuku 'father-in-law's father', or aji 'mother-in-law' versus thoni aji 'mother-inlaw's mother'. Although infrequent, I have heard the kinship term thorn apa, literally Old male father', used to address a grandfather, a compound created by logical extension given the absence of an indigenous Thangmi kinship term for 'grandfather'.

9.

Thangmi kinship terms and their Tibeto-Burman cognates

Here, I relate Thangmi kinship terms to reconstructed Tibeto-Burman etyma and offer a number of cognates from the extant Tibeto-Burman languages of Nepal. The Thangmi terms ama 'mother', apa 'father' and ca 'son' correspond to the Tibeto-Burman roots reconstructed by Benedict *ma 'mother' (1972: 148), *pa 'father' (1972: 19) and *tsa ~ *za 'child (offspring)' (1972: 27). The Thangmi kinship term hu 'younger brother' (Dolakhä) or 'younger sibling' (Sindhupälcok), may well be cognate with Classical Tibetan nit 'younger sibling', the latter also being present in compounds to give gender-specific younger sibling terms. Thangmi ca 'son' and camai 'daughter' are cognate with Thakali and Darjeeling Tamang ca 'son' and came 'daughter' (Vinding 1979: 208-215), and Gurung cxa 'son' (Glover, Glover, and Gurung 1977: 22) and cami or cami 'daughter' (Glover, Glover, and Gurung 1977: 17). Thangmi bubu 'elder brother' is cognate

120

Mark Turin

with Tibetan phu 'elder brother', corresponding to the Tibeto-Burman root *puw meaning 'grandfather', which has retained this meaning in some extant Tibeto-Burman languages. According to Benedict, however, *puw has undergone a "striking semantic transference" (1941: 319), and adopted the meaning 'elder brother' in certain Tibeto-Burman languages, such as Limbu phu 'elder brother' (van Driem 1987: 502) and Kulung bu 'elder brother' (Tolsma 1999: 197). Thangmi nini 'father's sister, mother's brother's wife' corresponds to Tibeto-Burman *ni(y) 'aunt' (Benedict 1972: 69), and also to Newar nini 'the husband's sister, the father's sister' (J0rgensen 1936: 101), Limbu nya? 'ego's paternal aunt, wife of ego's maternal aunt' (van Driem 1987: 483), Dumi nini 'paternal aunt' (van Driem 1993: 402), Bumthang nene 'father's sister' (van Driem 1995: 54), Yamphu niyi 'wife of mother's brother' (Rutgers 1998: 560) and Kulung ni 'paternal aunt' (Tolsma 1999: 223). Benedict posits the root *is'e 'great, old' underlying many honorific kinship terms in Tibetan, such as che-ba and chen-po 'great'. Thangmi chyode 'senior, big, large' (Sindhupälcok) and thone (Dolakhä) may be cognate with the same Tibetan or Tibeto-Burman root. Benedict (1972: 27) posits a root *tsa for 'grandchild' as well as 'child' for the 'Tibetanized languages of Nepal', related to or ultimately derived from Tibeto-Burman *tsa~*za 'child (offspring)'. Thangmi cacha 'grandchild' would fit well within such an analysis. A number of Thangmi kinship terms appear to be cognate with Classical Newar forms, and are not readily attested in other Tibeto-Burman languages spoken in Nepal. These Thangmi-Newar lexical correspondences include Thangmi cacha jyamari 'granddaughter's husband' and Classical Newar chayaflri 'grand-daughter's husband' (Malla 2000: 131), Thangmi tete 'elder sister' and Classical Newar fata 'an elder sister' (J0rgensen 1936: 75) or tatäju 'elder sister (hon.)' (Malla 2000: 170), and Thangmi pairi 'elder brother's wife' and Classical Newarpairabe 'an elder brother's wife'(Malla 2000: 296)." 10.

Conclusion

In Thangmi, kinship terms are employed to address and refer to consanguineal and affinal relatives. Kinship terms generally replace an individual's given name, both as a term of address and for reference. Moreover,

Thangmi kinship terminology

121

when kinship terms are used metaphorically as terms of address and reference for non-kin, a person's age and social position with respect to the speaker determines the choice of kinship term. For example, an elderly man may address a younger man as ca 'son' or cacha 'grandson', depending on how great he perceives the age difference to be. The metaphorical usage of kinship terms to apply to non-kin is widely observed among languages spoken in Nepal, including Nepali. Unlike Nepali, however, Thangmi has no system of ordinal terms to refer to the birth order of children within a generation. In their study of Limbu kinship, Davids and van Driem conclude that "no inferences about the social structure amongst the Limbus can be made on the basis of the fieldwork conducted with its emphasis on linguistic aspects of kinship terminology" (1985: 141). They explicitly follow Kroeber's view that "terms of kinship reflect psychology, not sociology. They are determined primarily by language and can be used for sociological inferences only with extreme caution" (1909: 84, as cited in Davids and van Driem 1985: 140). While I agree that great care must be taken when attempting to construct social meaning from the lexicon of a language, I would suggest that a detailed analysis of kinship terminology supported by in-depth ethnographic research may provide a basis for forming a number of preliminary hypotheses about the characteristic features of Thangmi culture. For example, I have noted the relative isolation of grandparents in Thangmi society and their absence from the daily social lives of their grandchildren, as well as the Thangmi preference for nuclear rather than joint family household composition. It comes as little surprise then, that the kinship terminology of Thangmi shows no indigenous terms for grandparents:12 their social absence is mirrored by a lexical absence. A further interesting socio-linguistic feature is the presence of female clan structures for women. Although practically defunct, the very existence of these clans indicates the presence of a gender distinction in Thangmi society, a feature which may be reflected in the linguistic differentiation for sex of speaker. In short, gendered clans and nuclear families are cultural characteristics of Thangmi life which have observable social and linguistic forms.

122

Mark Turin

Appendix: Thangmi kinship charts Thangmi kinship terms for offspring in Sindhup lcok

A=O gai

uma

Δ=Ο

child

ca

Δ=Ο

grandchild

cacha

cay

cacha jyamari

A-O yu

yu wari

cuy

camai

cacha

Ο

wari

cayam

Δ=Ο cuy

camai jyamari

Δ=Ο

cacha wari

great-grandchild cay

Δ=Ο

cawan

O

wari

great-great-grandchild

cuy m

Q • yuni

great-great-great-grandchild

Thangmi kinship terminology

/"

N

a, υ , Ι

I>%τ>l lε s3 «g 1 SS· 1« v y

suα «Ο

rO*» i!

J

-Οι

ICH

II

Ο

Q c

r: (38)

a. [na-kha khog-rna] khe-yu. DEM-NS play-CIT must-NPT They should play'. (Ilfaut qu 'Us joueni.) b. [a-tak-qa lu-ma-chi] khe-yu. iPOSS-friend-OBL tell-CiT-NS.u must-NPT 'My friend should tell them.' (Ilfaut que mon ami le leur dise.) c. [thuk-pi-ma-chi] khei-ka. cook-BEN-CiT-NS.U must:NPT-2 'You must cook for them.'

There is no explicit control in these quasi-impersonal constructions, and the main verb acts as a regular bivalent but formally intransitive verb. One argument is the person who has to do something, the second argument is the action that has to be done, expressed by the nonfinite clause.

Hidden syntax in Belhare

159

Formal evidence for this analysis comes from the morphological possibilities of the embedded verb. In all types of light verb control constructions, the embedded clause may not be inflected, as all morphological marking is relegated to the matrix. This includes undergoer number marking (39) and negation (40): (39)

a. lu-ma n-nu-yu. tell-CiT 3NS.S-may-NPT They may be told.' b. */? Iu-ma-chi n-nu-yu. tell-CiT-NS.u 3NS.S-must-NPT 'They may be told.'

(40)

a. ten-ma n-nui-?-ni. hit-ciT 3NS.s-may-NPT-NEG '[They may not] [be hit].' b. *min-ten-ma nu-yu. NEG-hit-ciT may-NPT '[They may] [not be hit].'

This is different in the quasi-impersonal constructions of type (38). There, the embedded verb may be inflected for the number of undergoers (although not for negation). Another piece of evidence derives from the following observation. In light verb constructions, nominals are preferably left in the lower clause. To 'move' them into the matrix clause is considered highly inelegant. This is presumably due to the fact that the matrix verb forms a tight semantic unit with the embedded verb, sharing the controlled argument outside this unit. While the predicate unit can be broken up in afterthought constructions, as illustrated in (32) above, to separate the two verbs by the controlled element results in questionable grammaticality: (41)

a. ηka lu-ma nui-?-ija. ISG tell-CiT may-NPT-EXCL Ί may be told.' b. ?lu-ma nka nu/-?-na. tell-ciT ISG may-NPT-EXCL Ί may be told.'

160

Balthazar Bickel

In quasi-impersonal construction, by contrast, nothing prevents a nominal from occurring in the main clause: (42)

[lu-ma]

rjka

khei-?-ija.

tell-CIT ISO must-NPT-EXCL Ί must tell him/her.'

From a syntactic point of view, (42) could also be interpreted without coreference of the lower A with the matrix S, e.g., as Ί must [have] somebody tell him' or Ί must be told'. However, the first reading is very rare and the second reading is pragmatically preempted by the light verb control construction in (43). (43)

[rjka

lu-ma]

khei-?-Tja.

ISG

tell-CIT

must-NPT-EXCL

Ί must be told.'

Moreover, the reading Ί must be told' for (42) is difficult to reconcile with focus structure. In Belhare, a lexical pronoun has usually a contrastive effect. In (42), the focus structure is thus something like Ί, not somebody else, must [tell]', whereas in (43), it is Ί-must [somebody tell to me, not to somebody else]. Thus, the strong bias towards the reading Ί must tell' in (42) has independent pragmatic reasons. There is no need to invoke syntactic control of the lower A in order to account for this interpretational bias. And, given the distribution of inflection in the lower clause in (38) and (39), (42) is also better accounted for without analysing it as a light verb construction with syntactic control. Type I light verbs monitor, as we have seen, an {S, U}-pivot. This is different from type II verbs, which are not constrained by a syntactic function but by a semantic role: the U-argument of a plurivalent verb. The undergoer inflection in the matrix clause registers person and number of the lower U-argument: (44)

a.

tin-ma

mai-talokt-he.

teach-ClT ISG.U-begin-PT 'He began to teach me.' (Literally, 'he began me to teach'.) b. rot/ khut-ma=cha ka-ij-na-rend-he. bread bring-ClT=ADD lNCL.U-3NS.A-stop-TEL-PT They stopped bringing us even bread.'

Hidden syntax in Belhare

161

c. lu-ma n-tog-he-ga. tell-ClT 3A-can-PT-2 'He had a chance to tell you.' The lower Α-argument is syntactically free, that is, not controlled. A sentence like the following means simply that 'they-stopped [x beating me]', where χ is unbounded: (45)

ten-ma ma-n-nar-he. beat-ciT lSG.U-3NS.A-stop-PT 'They stopped beating me.' or They stopped him/her/it/you from beating me.'

In many cases, however, the pragmatically most natural interpretation is one in which the lower Α-argument is coreferent with the higher Aargument. (It is indeed difficult to imagine how this could be different with a light verb like tokma 'can, have the opportunity or chance to', as in 44c.) Just as the reference of the lower A is syntactically free, so too, does the Α-inflection of the matrix predicate not necessarily constrain the reference of a lower argument. This is confirmed by the following. As with type I light verbs (cf. 41 above), to 'move' a nominal U-argument into the matrix is inelegant: (46)

a. [ma?j-chi ten-ma] na-rend-he-chi. human-NS beat-ατ stop-TEL-PT-3NS.u 'He stopped the people from being beaten.' b. ?[ten-ma] ma?i-chi na-rend-he-chi. beat-ciT human-NS stop-TEL-PT-3NS.u 'He stopped the people from being beaten.'

Since it is not controlled, the Α-argument, by contrast, can easily occur in the main clause. Notice that, similar to what we observed above in (42), the occurrence of an overt actor pronoun in the matrix clause induces a contrastive focus that implicates coreference with the lower actor: (47)

a. [ma?i-chi ten-ma] un-na na-rend-he-chi. human-NS beat-ciT 3-OBL stop-TEL-PT-3NS.U 'He stopped beating the people.'

162

Bahhasar Bickel

b. [rAksi alcoholic.spirit

khut-ma-na]

na-kha-ga

bring-ciT-TOP DEM-NS-OBL

ka-g-na-rend-he. INCL.U-3NS.A-StOp-TEL-PT

'They stopped bringing us'ncl raksi.'

Type II verbs do not allow control of just any U-argument, but only of U-arguments of plurivalent verbs. With monovalent verbs, syntactically controlled light verb constructions are impossible. In this case, there are two possibilities, labeled here Ha and Ilb/c. Type Ha predicates are always transitive. With embedded monovalent verbs, their U-inflection refers to the embedded clause as a whole rather than to the S-argument in it. This is shown by the following data. (48)

a. un-chik ta-ma n-tou-t-u / *tou-t-u-chi. 3-NS come-CiT 3NS.A-can-NPT-3u can-NPT-3u-NS.u 'They can come.' b. cek-ma mun-ditt-he / *mai-mun-ditt-he. say-ciT forget-ACCEL-PT ISG.U-forget-ACCEL-PT 'He forgot to say it to me.' c. ηka khik-ma nar-he / *mai-nar-he. ISO bitter-CiT stop-PT lsc.U-stop.PT 'It stopped being bitter to me.'

The matrix U-inflection must not register the number of the lower argument (48a), and must always be in the third person (48b). This also extends to unaccusative experience predicates (48c). Thus, monovalent predicates are embedded as U-arguments of the higher predicate, without explicit control marking. Type Ilb/c predicates, by contrast, are not lexically specified for transitivity. Rather, the main verb assimilates in transitivity to the embedded clause. With monovalent verbs, the embedded clause is treated as a secondary, nonmacrorole argument of an intransitive matrix clause, in the same way as in (38) above. (49)

a. ta-ma Iab-yu / *lap-t-u. come-CiT be.about-NPT be.about-NPT-3u 'He is about to come.' b. ta-ma kond-he / *mai-kond-he. come-ciT want-ρτ Iso.u-want-PT 'He wanted to come / me to come.'

Hidden syntax in Belhare

163

In its counterexpectative meaning 'already V, surprisingly V, the verb hir- combines with an embedded citation form in -ma. With the potentiality meaning 'be able to, have the power and requirements to',11 hir- requires the lower verb in a nonfinite subjunctive form in -a. This nonfinite form occurs exclusively in this environment. (50)

a. la um-ma ηη-hir-e-chi / *qi)-hir-e-ch-u. walk-CiT 3NS.S-already-PT-DL 3NS.A-already-PT-DL-3u 'The two of them already set off.' b. khar-a hi-yu / *hiu-t-u. go-SUBJ be.able-NPT be.able-NPT-3u 'She can go there.'

If the embedded clause contains a plurivalent verb, the matrix is inflected transitively and controls the lower undergoer, in the regular fashion observed above in (42) through (47): (51)

a. lu-ma n-lapt-he-ga i? / *lapt-he-ga i? tell-ClT 3A-be.about-PT-2 Q be.about-PT-2 Q 'Was he about to tell you?' b. kh bo let-ma hir-e-m-ga i? / *hir-e-i-ga i? pillar plant-ατ already-PT-2PL.A-2 Q already-PT-2PL-2 Q 'Have youpl already put in the main pillar (for the new house)?' c. hit mett-a ka-hiu-ka i? / *hiu-ka i? look CAUS-SUBJ lSG.U-be.able-2 Q be.able-2 Q 'Can you show me [the way]?'

As with type I, we need to distinguish between an obligatory and an optional version of type II control. This is the difference between type lie on the one hand, and Ha and lib on the other. Apart from control constructions as in (52), the type He predicates konma 'to want, to be about to' and mitma 'to think' also allow other constructions (53). (52)

a. kubaij-qa ten-ma ka-q-kol-yu, ani! monkey-OBL hit-ClT iNCL.U-3NS.A-want-NPT and.then 'The monkeys want to hit us, so [what do you think]!' b. un lu-ma m-mitt-he-ch-u. 3 tell-ciT 3NS.A-think-PT-DL-3u 'Theyd thought about telling him.'

164 (53)

Balthasar Bickel a. lu-ma koi-yu. tell-CIT want-NPT 'He wants to tell.' or 'He, wants somebody to tell him,·./.' b. lu-ma mitt-he-rja. tell-CIT think-PT-EXCL Ί considered telling him.'

As example (53a), repeated here from (15), illustrates, the matrix predicate can be inflected intransitively, in which case there is no control and the reference of the lower clause arguments is syntactically free. However, disjoint reference of the actors is often implausible for pragmatic reasons (53b). With embedded intransitive clauses, the type He predicates behave like lib predicates. As shown by (48b), they are inflected intransitively and treat the embedded clause as a nonmacrorole argument. The only difference between type lib and lie is that where an U-argument of an embedded plurivalent verb is available, it is obligatorily controlled by type lib verbs but optionally so by type He verbs. This is evidenced by the contrast between (51), where the lower U must be controlled, and (53), where the lower U is not registered in the matrix.

4.

Participant coding devices

Having established that Belhare syntax operates with three different types of pivots, {S, A}, {S, O}, and {S, U}, I want to explore what formal means there are to denote these pivots. I will first discuss phrase-structure and then verb and noun morphology.

4.1.

Phrase structure

As in many other languages of South Asia, the order of elements in a Belhare sentence is syntactically constrained only by the requirement that the verb be final. Thus, both orderings in (54) constitute a grammatical sentence, although they widely differ as to their information structure or functional sentence perspective. As a rule of thumb, the earlier elements tend to be topical and presupposed, while the later elements are focal and encode new information:

Hidden syntax in Belhare (54)

a. a-ni-rja

iqa

165

thukg-hett-u.

iPOSS-aunt-OBL beer cook-TEMP-3u 'Auntie is making beer.' b. iqa a-ni-ija thukg-hett-u. beer iPOSS-aunt-OBL cook-TEMP-3u ""Auntie is making beer.'

Surface word-order is obviously no indicator of sentential phrase structure, and cannot α fortiori encode pivot structure. Let us therefore look at some less obvious evidence for phrase structure or configurationality in general. In our survey of pivots, we have seen one construction, type II control (Section 3.4), that uniquely monitors the U-argument of plurivalent verbs, and other constructions that uniquely monitor an {S, A}-pivot (Section 3.1-2). Such a distribution is highly reminiscent of the subject/object asymmetries commonly taken as evidence for a verb-phrase (VP) constituent including the transitive U-argument, but excluding the A-argument. However, to invoke a W as an explanation for and exponent of subject/ object asymmetries not only presents difficulties in terms of the equally present {S, U}-pivot but is also problematic from the point of view of other aspects of Belhare phrase-structure: We do not expect idiom-formation to operate across phrases or 'constituents', and one may even take idiom-formation as evidence for phrasestructure (Marantz 1984). In Belhare, there are enough idioms involving a transitive Α-argument and a verb in order to raise serious doubts about the reality of the VP as a constituent:12 (55)

a. bo?wa him-ma ka-ta?~yu. famine-OBL iNCL.u-bring-NPT There will be a famine here.' (Literally, 'the famine will bring us'.) b. cuq-rja mai-tar-he. cold-OBL 1 SG.U-bring-PT Ί have got fever.' (Literally, 'the cold brought me') c. khalamba-ija mai-lab-he. catarrh-OBL lSG.U-catch-PT Ί have got a cold.' (Literally, 'the catarrh caught me')

The idiomaticity of these constructions is evidenced by the fact that normally transitive Α-arguments are restricted to animate entities (in line with many other languages, e.g., Lakhota, cf. Van Valin and LaPolla 1997). This constraint is systematically violated by idioms like the ones in (55)

166

Balthasar Bickel

(55) but cannot be ignored in non-idiomatic grammar. The inanimate cause argument in (56a) may not be coreferent to an Α-argument denoted by the transitively inflected verb: (56)

a. *cuij-ija sei-t-u. cold-OBL kill-NPT-3u The cold will kill him.' b. cutj-ηα si-yu. cold-OBL die-NPT 'He will die from the cold.'

The only piece of evidence that could support a VP derives from detransitivisation, a construction in which a plurivalent verb is inflected intransitively (cf. Section 3.4 above, and Bickel 1995b: 418-420). In detransitivised clauses, there appears to be a constraint against 'moving' the U-argument into a postsentential afterthought position. This position is separated from the sentence by a short intonational break, indicated here by a comma: (57)

a. i-na-qa khui?-t-u, wa. DIST-DEM-OBL steal-NPT-3u chicken 'That one steals them, the chicken.' b. wa khu?-yu, i-na. chicken steal-NPT DIST-DEM 'He steals chicken, that one.' c. * i-na khu?-yu, wa. DIST-DEM steal-NPT chicken * 'He steals, chicken.'

Whereas such an extraction is possible with a U-argument of fully transitive sentences (57a) and with an Α-argument of a detransitivised sentence (57b), it is extremely odd if not ungrammatical to extract the U-argument from the detransitivised sentence in (57c). This looks as if the U-argument and the verb taken together form a tight verb phrase ('VP'), out of which it is impossible to extract. However, this 'VP-hypothesis' might not be the best explanation of the grammaticality distribution in (57). An initial doubt concerning the VP-hypothesis arises from the fact that the U-argument need not be adjacent to the verb. Given a suitable context, where the U-argument is topical, it can be easily preposed. Such a context is given in the following exchange, which involves the verb thukma 'to cook' that is normally inflected transitively but used here in detransitivised sentences:

Hidden syntax in Belhare (58)

167

— sati iija thug-yu? who beer cook-NPT 'Who makes beer?' — irja na-kha n-thug-yu. beer DEM-NS 3NS.S-cook-NPT 'Beer is made by these [people].'

This fact is difficult but not impossible to reconcile with a VPhypothesis. However, there is evidence that we do not need the hypothesis in the first place. The grammaticality distribution in (57) can be fully accounted for in functional terms: Detransitivisation has the semantic effect of making the U-nominal nonspecific. The nominal does not refer to a specific referent but to a kind of referent. Therefore, it may not be pluralised (59a) or specified by a demonstrative (59b). (59)

a. *wa-chi khu?-yu. chicken-NS steal-NPT 'He steals chicken.' b. *i-na phak sei-7-ηα. DIST-DEM pig kill-NPT-EXCL Ί will kill this pig.'

Detransitivisation is possible only if the U-nominal allows an interpretation as a 'kind' instead of as a specific referent. For instance, cama 'food' can be understood both as a specific thing or as a generic kind. 13 (60)

a. cama cai-t-u. food eat-NPT-3u 'She will eat the food (i.e., the specific rice, maize, or millet food that we have).' b. cama ca-yu. food eat-NPT 'She will eat (whatever food).'

With an inherently specific U-argument like ucha 'his/her child', detransitivisation is impossible: (61)

a. u-cha tel-t-u. 3POSS-child beat-NPT-3u 'She beats her child.'

168

Balthasar Bickel b. *u-cha te7-yu. SPOSS-child beat-NPT 'She beats her child.'

The nonspecific meaning of a detransitivised undergoer makes it also incompatible with the article -na, which serves as a nominaliser of relative clauses. In a circumnominal construction (cf. Section 3.3), the article signals specificity of the internal head noun. If the relative clause is detransitivised, the article can therefore only refer to the actor, and only the actor can be understood as head of the construction; cf. (22), repeated here as (62a). By contrast, if the relative clause is marked by a general nominaliser rather than the article, both arguments qualify as potential heads (62b). (62)

a. tombhira wa sei?-sa-na chitt-he-m. wild.cat chicken kill-TR.PERF-N find-PT-lPL.A 'We found the cat that had killed chicken.' b. tombhira wa sei?-sa-ha chitt-he-m. wild.cat chicken kill-TR.PERF-N find-PT-lPL.A 'We found the cat that had killed chicken.' or 'We found chicken killed by a cat.' 14

The semantic constraint on undergoers in detransitivised clauses is highly reminiscent of noun-incorporation and one might want to adopt Mohanan's (1995) analysis of similar constructions in Hindi. According to this analysis, the undergoer in a detransitivised clause forms a single lexical, i.e., nonphrasal (X°) category together with the verb. Thus, the Uargument is incorporated into the verb - at least in terms of syntactic category structure. In terms of phonological and morphological structure, the undergoer and the verb are still two distinct entities in Belhare, whereas in Hindi the two elements merge into a single prosodic and morphological word (Mohanan 1995: 94^98): (63)

a. Morphological structure:

b. Phonological structure:

phak N-set-iuk pig 3NS.S-kill-NPT 'They kill pigs.' [prWd 'pharj] [PrWd 'se?yu]

Hidden syntax in Belhare

169

As shown by (63 a), the verb acts as an independent morphological unit and can host inflectional affixes. By a general rule of prosodic word ('PrWd') formation, the nasal prefix N- associates with a preceding word (63b), giving way to a phonological bracketing different from the morphological structure (see Bickel 1996: Chapter 3, for some discussion.) Nevertheless, phak 'pig' and nse?yu 'they kill' may still form a single nonphrasal category in syntax.15 While this analysis would cover many cases and would explain the semantic properties of the construction, it does not, however, fully account for all instances of detransitivisation. First, as shown in (58), the undergoer can be preposed and need not be adjacent to the verb. It can even be omitted, if the context is clear enough: (64)

un thug-yu. 3 cook-NPT 'She cooks.'

Second, the undergoer is accessible to topicalisation and can be marked as a topical constituent: (65)

iija-na thug-yu, cia-na n-thuu-t-ni. beer-TOP cook-NPT tea-TOP NEG-cook-NPT-NEG 'Beer she makes, she doesn't make tea.'

By the same token, the undergoer may serve as the head of a relative clause, a position that is inherently topical (Kuno 1987: 14-15). This is exemplified by the second reading of (62b) above. Third, focus is also a possible function of the undergoer nominal in a detransitivised clause. This is illustrated here by the 'identifying' focus clitic =ro, which marks a nominal as instantiating something that is judged inactive in the addressee's consciousness: (66)

wa=ro khu?-yu. chicken=iD steal-NPT 'It's chicken that he steals.'

170

Balthazar Bickel

Since the U-argument can form the focus of the sentence, there is nothing to prevent it from being questioned, either: (67)

i-na yeti khu?-yu? DIST-DEM what steal-NPT 'What does this one steal?'

Thus, there are some detransitivised constructions that do not fit an analysis in terms of noun incorporation, not even in a purely syntactic sense. However, semantically, these constructions are highly similar to what is known about incorporation structures. It is this that allows an account of the apparent extraction constraint in (57) without invoking a VP: There is a general constraint in Belhare that a postsentential noun phrase must have fixed and specific reference. Thus, a question word in this position can only be interpreted as an echo question, where the speaker is convinced that the addressee knows the answer because the addressee has just said it but the speaker didn't understand it. The question word refers to a specific person that the interlocutors accept in principle as an established referent, but that needs clarification because of a low-level processing error. (68)

khar-e, sati? go-PT who 'He went, as to 'who?". (So, please repeat!) Impossible reading: 'Who went?'

Given that detransitivisation induces a nonspecific reading of the Uargument, moving the U-argument to a postsentential position violates the specificity constraint attached to that position. This explains why the Uargument in (57c) cannot be postponed. It is not that its 'movement' would violate an extraction constraint involving the notion of a tight VP, but rather that the postsentential 'landing site' does not semantically fit. In summary, there does not seem to be compelling evidence for a VP, and scientific parsimony suggests that in the absence of such evidence, the VP should not be postulated. Therefore, it is impossible to uniquely identify the {S, A}-pivot as the NP outside the VP but inside the sentence, i.e. as the NP immediately dominated by the 'sentence' node. In fact, no phrase structure position at all seems to identify any of the pivots in Belhare.

Hidden syntax in Belhare

111

S3 r·^

Z

Ώ

^

cg Π3

εou

"r^* C^J CiJ SJ SJ S3 SJ

."" ."".

1;. ;:;:;.:.·: ;..;

~ ? bqj ,^i £j

ΙΛ

ntahe mintaatni ntahechi mintaatchi

S3

ΛΛ

171

-q S

1E? (^1-2 1 ^

^J

"& "q^ ^i es

ε ε ε ε ε ε ο

en ΓΟ

Q

ΓΙ

m

,S ·§I

3

3 ΐ zi "g

Ο en m

α

Ο

-j

Q

(N U

u

3

q q

q · ': .

,

-_

α>

il

•

•5 "S

C^H • —,

•a ε

q

03

^3

Γ*

q 8

11

-t; 3

£ P

α δ "·*-»

3. "5

.q

Si ο 4-»

o^-S

.q j=

SP *»-ϊ

Sl •s §

||

1

c ο

i?

luitumma nlui?nimn luichuna nluichunn luitum nlui?nim luichu nluichun

ο

il

« *

oa

I

i;

uituchi inluitunc

•ΐί

i'l 'S

3a

nluitu minluitun nluichu minluichu

Z

§3 "0

:

itumchin ui?nimch

C/3

•ι ι 03 ov,

%> ^ -S 5 υ ς> 'ca

luitumga nlui?nimg luichuga nluichung

ft

3 q

··*»

itumchin ui?nimch

ί^ϊ

•il

.bO

itumchin ui?nimch

&t

•5 q ;a •5 .0

itaichi Tlintaichin

"aitina itaitinna 'aichin_a itaichinna

Balthosor Bickel

ntransitive

\72

r^

ζ

r^

3 3

-J

^

Ό

03

q

3 q

•q -q

a' 3

3 3

C3

03

υ ο

' —·

03

c\3

03

03

03 bo 03 b 03 bO 03 \ bo q -q -C •^ O^-H c a υ ο

•S? >ε

3 -J

cd ** cj> t^» t? q> CvS C>J S S q ε s ε ε ^K

· *

3 3 3 3 3 3 "οςϊ 03 03 03 03 03 •Αί Λί -Χ Αί_

;;

_J

1^1

Γ")

*-

Ο

u

g Γί

0 ΓΛ

_

εfr;

Q ro

Hidden syntax, in Belhare 4.2.

173

Verb agreement16

In this section I want to explore to what extent the verb agreement markers, which are abundant in Belhare, refer to syntactic functions, i.e., to the pivots found in Section 3. Verb inflection involves both prefixes and suffixes and registers both actors and undergoers in transitive verbs. Tables 3 and 4 illustrate the past and nonpast paradigms of the transitive verb lur- 'to tell' (infinitive citation form lumä) and the intransitive verb ta- 'come' (citation form tarna). In each cell, the form in the upper row is affirmative and the one in the lower row is negative. The past forms are marked by the suffix -he in the affirmative and -att in the negative. The nonpast paradigm is characterised by a diphthongue stem form followed by the tense marker -t. Where this marker would appear in word-final position, it is replaced by the suffix -yu, which requires a monophthongue rather than a diphthongue stem. For prosodic reasons discussed in Bickel (1996: Chapter 3), the -t suffix shows up frequently as a glottal stop or as an empty consonant position that blocks intervocalic voicing of a following obstruent. In most other environments, stops are voiced between vowels or between nasals and vowels. This is why the marker -ka, which signals a second person, frequently appears as -ga (e.g., taitiga 'youp will come'), but retains its voiceless onset in some nonpast forms (e.g., taika 'you5 will come'). Another important morphophonological rule to notice is that vowel sequences are reduced according to either of two schemes: /i/+/u/ and /a/+/u/ become /u/, but /e/ -f /u/ becomes Id. For example, sequences like -chi-u [DL-3U] regularly appear as -chu, but an -u '31T after the past marker -he is deleted and -he-u appears as -he. This results in the neutralisation of frequent transitive and intransitive forms, as illustrated by the following example. (69)

a. n-ca-he-u —> /ncahe/ 3NS.A-eat-PT-3u 'They ate it.' b. n-ca-he —> /ncahe/ 3NS.s-eat-PT 'They ate.'

174

Balthasar Bickel

G O *>-* υ v

> u >

"00 «L» C

•o es rt

U

r/1 J

O

^i AS Aj AS AS i A;

z

. Λ "VT

Ϊ

^

M

'S

AS 03

r i. ι

q !

•^c h? *? i-δ |·Λ ·Λ l o --g ll -q Λ |ι»] t«lj

cn x: J

q

;

:-5 ! υ |u!j

cc;_ ^ cv; c^; ; /lurheq/ tell-PT-3U-lSG.A 'Hold him/her.'

The accusative alignment of first person markers is also limited by the fact that with second person undergoers, it is not the 'First Person Singular A' marker (-#) that is used, but rather a special portemanteau morpheme -na 'First Person A and Second Person U' (e.g., meu-?-na Ί love you'.) Finally, there are some agreement morphemes that are general in terms of role differences. For instance, the dual marker -chi and the second person marker -kak~ -ka cover S-arguments as much as A- and U-arguments: (75)

hond-he-chi-ga appear-PT-DL-2 ' Youd showed up.'

lur-he-ch-u-ga tell-PT-DL-3u-2 ' Youd told him/her.'

n-Iur-he-chi-ga 3A-telI-PT-DL-2 'S/he/they told youd.'

178

Balthazar Bickel

Thus, even though there are a few traces of accusative morpheme alignment, the agreement system is not characterised as a whole by such a pivot. This is also corroborated by neutralisations of number marking found in the paradigm. For instance, with first person exclusive U-arguments, the third person marks dual in opposition to an undifferentiated singular/plural form: ma?i-lur-he 's/he/they told us6' vs. ma?i-lur-he-chi 'theyd told use'. With other undergoers, e.g., first person inclusive, however, there is a ternary number distinction: ka-lur-he 's/he told us" vs. ka-q-lur-he 'theyp told us" vs. ka-q-lur-he-chi 'theyd told us1.' Obviously, the paradigm structure involves rules sensitive to both Α-arguments and U-arguments simultaneously (see Bickel 1995a: 113f for further discussion). These rules are at odds with pivot-coding since they systematically crosscut both accusative and ergative alignments. Thus, the morpheme structure of the verb paradigms does not appear to encode pivots in any consistent way. From the point of view of theories which allow only 'lexical morphology' (e.g., Lexical-Functional Grammar; cf. Bresnan and Mchombo 1995; Mohanan 1995) or which insist on the autonomy of morphological organisation (e.g., Autolexical Syntax; cf. Sadock 1991), this may come as no surprise. In such theories, it is the semantic or 'functional' structure of the inflected words as a whole, rather than of the individual morphemes, that we can expect to be in a one-to-one correspondence with syntax. Let us therefore look more closely at the semantic patterns underlying the paradigm as a whole. There is only one candidate of an inflectional semantic category that neutralises semantic roles and which could take part in a pivot definition. This is the category 'S', which neutralises semantic roles in intransitive verb inflection. Intransitive inflection does not register the agentive or patientive nature of the argument in any form. This is different from transitive inflection, where the macroroles actor and undergoer trigger different verb forms. This state of affairs allows a morphological definition of both {S, A} and {S, U} pivots, but not of the {S, O} pivots operative in internal head relative clauses, since the nonmacrorole argument of a trivalent verb is not registered in verb inflection (cf. Section 2). Yet even this rudimentary pivot-sensitivity of inflectional semantics is more apparent than real. Although the S-argument in inflectional semantics seems to correspond by and large to the S-argument in syntactic pivots, the relationship is not one-to-one. There are two systematic exceptions, the 'possessive of experience' and the 'unaccusative experience verbs' (Bickel 1997). In a possessive of experience construction, the intransitively in-

Hidden syntax in Belhare

179

fleeted verb agrees with a noun (ris) denoting an experience, and the experiencer appears as a possessive prefix (Λ-) (76a). In the unaccusative construction the verb agrees with the stimulus noun (76b) (cf. Section 3.3): (76)

a. n-ris ka?-yu. 2POSS-anger come.up-NPT 'You will get angry.' b. ijka iqa su-yu. ISG beer sour-NPT To me, the beer tastes sour.'

What the verb encodes as an S-argument in these constructions does not qualify as S in the syntax. Notice that although verb inflection treats them as intransitive, both constructions in (76) involve bivalent verbs, each with both stimulus and experiencer arguments. According to the actorundergoer hierarchy in (3), the stimulus is assigned an U-role, the experiencer an Α-role. This hierarchy, rather than verb morphology, predicts the syntactic behaviour of the two constructions. Where the syntax requires an ergative-style pivot, the experiencer is excluded as the pivot. Therefore, the experiencer argument of possessive constructions cannot be controlled by nus- 'may': (77)

a. n-ris kat-ma n-nui-?-ni. 2POSS-anger come.up-ciT NEG-may-NPT-NEG 'You shouldn't get angry.' b. *(n-)ris kat-ma n-nui-?-ni-ga. (2POSS-)anger come.up-ciT NEG-may-NPT-NEG-2 'You shouldn't get angry.'

The same holds for control of unaccusative constructions, as we have seen in (35), repeated here as (78): (78)

a. yka su-ma nu-yu. ISG sour-CiT may-NPT 'To me, [the beer] may be sour.' (Ί like it sour.') b. * su-ma nui-?-ija. sour-ciT may-NPT-EXCL To me, [the beer] may be sour.' (Ί like it sour.')

With internal head relative clauses, too, the experiencer argument is excluded as pivot since it is an Α-argument (cf. (22) above in Section 3.3).

180

Balthasar Bickel

This holds, again, for both possessive of experience (79a) and unaccusative constructions (79b). (79)

a. *ma?i u-ris

kai?-qa-na

he-ne

khar-e?

human 3POSS-anger come.up-iNTR.PERF-ART which-LOC go-PT 'Which one is the person who's got angry?' b. *ma?i lim-?-na he-na cai? human delicious-NPT-ART which-ART TOP 'Which one is the person to whom [the beer] tastes delicious?'

Since they require an {S, A} pivot, embedded -sa clauses and root nominalisations, by contrast, allow the experiencer as the pivotal Aargument, rather than the stimulus which counts as an U-argument. The stimulus nominal can overtly appear in the embedded clause since it does not qualify as an S-argument in syntax. It counts as an S-argument only for the purpose of verb inflection in examples like (76) above: (80)

a. a-ppa la-har-e, inj, u-ris kas-sa? Iposs-father return-TEL-PT Q Sposs-anger come.up-ss/T 'My father went back angrily, didn't he?' b. hagliua lu-sa thaq^har-e-ija. sweat perceptible-SS/τ go.up-TEL-PT-EXCL Ί started to climb up in the heat.'

The same holds true of root nominalisation: it is the experiencer argument that qualifies as the pivot of the construction. The stimulus noun can again be overt, unlike S-arguments: (81)

a. (*u-)ns ka-ta-ba 3POSS- anger N-come.up-M 'an angry person' (One whose anger came/comes/will come up') b. irja ka-lim-ba beer N-delicious-M One to whom the beer tastes good'

These observations show that there is no one-to-one correspondence between 'S' as a morphological notion relevant for verb agreement and 'S' as a syntactic notion relevant for pivot syntax. This in fact reflects a systematic property of Belhare that extends to the morphological notions of Ά' and 'U', which are not exactly identical in syntax, either.

Hidden syntax in Belhare

181

Mismatches between morphological and syntactic A and U notions prevail in what corresponds to dative subject constructions in other South Asian languages. There is a set of idiomatic experience constructions in which the experiencer argument is encoded by an U-affix on the verb. The stimulus corresponds to an Α-affix, which is zero in the case of a third person singular. The stimulus is also realised by an Α-nominal in the oblique ('ergative') case: (82)

a. cutj-rja

mai-tar-he.

fever-OBL 1 SG.u-bring-PT Ί have got fever.' (Mal T jvaro yo.) b. iga-a mai-ser-he. beer-OBL lSG.U-kill-PT Ί got drunk from the beer.' This obviously violates the semantic role hierarchy in (3). However, the hierarchy is postulated for syntactic purposes, not for role coding in morphology. With respect to the syntax, the hierarchy holds in the same way as it did in the examples (77) through (81): the stimulus qualifies as U, the experiencer as A. As a result, the stimulus can appear as the overt argument of an embedded -sa clause (83a) or in a root nominalisation (83b) while the experiencer assumes the role of the (S, A} -pivot: (83)

a. ευη-ηα tas-sa=ro Dhankuta tas-e-η. fever-OBL bring-ss/T=io Dh. reach-pT-lSG.A Ί reached Dhankuta although I indeed had this fever.' b. cuij-rja ka-tat-pa fever-OBL N-bring-M 'the one who has got fever.'

For passivisation, which seems to straddle the borderline between syntax and morphology, the experiencer also counts as an Α-argument, the stimulus as an U-argument. Therefore, it is impossible to form a passive from the 'undergoer of experience' construction and promote the morphological U-argument, i.e., the syntactic Α-argument, to S as in (84b). (84)

a. hja-a

sei?-s-u-ha.

beer-OBL kill-TR.PERF-3U-PERF 'The beer has made him drunk.'

182

Balthasar Bickel b. *sei?-ija-ha. kill-INTR.PERF-PERF

'He has been made drunk.' (OK as 'he has been killed.')

The mismatch between macrorole assignment in syntax and morphology probably results from the competition of two functional constraints on the coding of experiencers (cf. Bickel 1997). First, experiences belong to what Bally (1926) called the 'personal sphere', which gives them a high degree of empathy (Kuno 1987). Second, experiencers are usually human participants of high value for the current purpose of verbal exchange or narration. This gives them a high degree of topicality. The first constraint requires experiencers to be encoded in an oblique or 'inverse' fashion, so as to highlight their affectedness. This results in their morphological coding as possessors, undergoers or nonmacrorole arguments. The second constraint requires that experiencers be easily accessible to topicality sensitive constructions. This is why they are treated as {S, A}-arguments in the syntax, i.e., like those arguments that have the highest degree of topic continuity in connected discourse - in Belhare (Bickel 1995a: 83-84) as much as in many other languages (Van Valin 1993: 59-62). The result of this mismatch of syntactic and morphological treatment of experiencers is that none of the semantic units of verb morphology exactly corresponds to a syntactic notion or even to a macrorole notion relevant for syntax. 4.3.

Case

As a final coding device that could in principle correspond to syntactic functions, I want to look at case marking. Of the nine cases that Belhare distinguishes, most have highly concrete meanings, corresponding to prepositions in English. The only candidates for syntax-sensitive case marking are the oblique suffix and bare nominals without case. The oblique case is realised by -Na, where TV again represents an assimilating nasal with a velar default articulation after vowels. In vowelfinal words, there is a alternative exponent in -a. The difference is minimal, -a suggesting a slightly lower degree of emphasis than -Na. An exception is the first person plural inclusive pronoun i]ke which only allows -a (i.e., qke-a 'we-OBL'). This is perhaps to avoid confusion of *gke-ija 'we-OBL' with gke-g-ija 'we-EXCL-OBL'. The first person singular pronoun (ijkä) is

Hidden syntax in Belhare

1 83

never marked by the oblique case (*qkaa, *ηkaη ), in line with universal hierarchies of case marking splits. The oblique case denotes the semantic role of effector, that is, a participant "that brought something about" with "no implication of its being volitional or the original instigator" (Van Valin 1993: 42). However, effectorcoding is subject to two conditions. First, there must be at least two participants involved. This is the case with (85a) or (85b), where the effector is an instrument and a natural cause, respectively. It is not the case, however, with the agent of a monovalent verb (85c). (85)

a. dabhek-qa n-cept-he. khukuri.knife-OBL 3NS.A-cut-PT 'They cut it with a khukuri knife.' b. cuq-tja si-yu hola. cold-OBL die-NPT probably 'He will probably die from the cold.' c. a-tak(*-qa) la us-e. iPOSS-friend(-OBL) dance-PT 'My friend danced.'

The second condition is that if the effector argument is coreferent with an argument of a plurivalent verb, this verb may not be inflected intransitively. The oblique case is excluded from a detransitivised clause: (86)

tombhira(*- ηα) wa se?-yu. wild.cat(*-OBL) chicken kill-NPT The cat kills chicken.'

Notice that the second condition does not positively require that the verb be inflected transitively. In fact, the verb may not be inflected at all for person and number. In this context, only the first condition for effectorcoding is relevant, viz. that there be at least two participants involved: (87)

a. [un-chik-qa yka min-lu] n-]iu-?-ni! 3-NS-OBL ISO NEG-tell NEG-become-NPT-NEG 'That's not OK without them telling me!' b. [a-tak(*-ga) la migg-u] n-]iu-?-ni. iPOSS-friend(-OBL) dance NEG-dance NEG-become-NPT-NEG 'That's not OK without my friend dancing.'

184

Balthasar Bickel

Thus, the distribution of the oblique case does not rely on a simple notion of inflectional transitivity, as shown by (87), nor on a simple notion of semantic valence, as shown by (85). Belhare case is neither reducible to a syntactic pivot notion nor to a macrorole notion that is relevant for pivot structure. Thus, as with verb inflection, there is no one-to-one correspondence between morphology and syntactically relevant notions. As for the 'zero' case, i.e., bare nominals, the observation is similar. Bare nominals function not only as 'absolutive' nominals which correspond to a {S, O}-pivot. They also occur as both the stimulus and the experiencer argument in intransitive experience constructions, as is evident from (76b) in the preceding section.

5.

Theoretical implications

In this essay I have adopted a language-specific onomasiological perspective on morphology and phrase-structure, that is, I have first established the pivots operative in Belhare syntax and then looked at phrase-structural and morphological devices that might identify them. The result is that none of the three syntactic functions, i.e., the accusative, the ergative and the restricted ergative pivot, is identified by either a phrase structure position or a morphological category. In fact, neither domain is sensitive to functional (relational) syntax at all, except for some relevance of transitivity notions. The relational syntax of Belhare, and possibly other Kiranti languages,17 is hidden. The question that naturally arises now is semasiological in perspective: what function do these formal devices have if it is not the indexing of syntactic functions? From what we saw in Section 4, these devices are sensitive to specific semantic participant roles such as 'experiencer' or 'effector' and their function seems to lie precisely in the detailed structuring of the argument-predicate relationship: they explicate the semantics of argument roles. To conclude, I briefly discuss in the following what this finding implies for syntactic theories. Belhare phrase structure does not seem to have any syntactic expression power, except for the fact that it restricts postsentential noun phrases to specific reference. Thus, Belhare is 'nonconfigurational' in the sense of Hale (1983). Morphology, by contrast, has much expressive content, but not of the kind that current theories lead us to expect. Most theories assume person and number inflection in verbs to mark some sort of 'agreement', and this assumption underlies terminological practice even in Himalayan linguistics. Taken at face value, 'agreement' is the

Hidden syntax in Belhare

185

phenomenon by which inflectional features are mechanically copied from one element to another so that the elements "bear the same imprint" and "are thereby stamped as somehow related" (Sapir 1921: 114). Obviously, 'agreement' presupposes that the elements have the same format of representation. It does not make sense to talk of agreement between a morpheme and a semantic unit. This is simply denotation, not agreement. Whether or not an inflected verb form is thought of as a phrase structure element or as a morphological entity, the trigger of the verb's agreement must be a configurational or a morphological entity. This condition is satisfied, for instance, in English by the fact that the trigger of verb agreement, i.e., the 'subject' or {S, A}-pivot, is identifiable by a phrase structure position (the noun phrase immediately dominated by the category 'sentence'), and in Russian by the fact that the trigger can be identified by an inflectional property (the nominative case). If there is no configurational or morphological property consistently related to the trigger, there cannot be 'agreement' in the strict sense of the term. The putative trigger is a syntactic entity sui generis and cannot in principle be represented in the same format as the category 'verb', i.e. a category defined in phrase structure and/or by morphological properties. The verb does not 'agree' with this entity, but directly 'denotes' it or 'refers to' it. If this entity is a syntactic function (a pivot), the language corresponds to what Jelinek (1989) calls a 'pronominal argument language' (such as the Australian language Warlpiri). In Belhare, there is no single noun phrase property that correlates with a pivot. Therefore, there is no 'agreement' in the strict sense. Nor does verb inflection relate to a syntactic function. Therefore, Belhare does not qualify as a pronominal argument language in Jelinek's sense, either. The language does not seem to correspond to any of the types that have been identified so far in syntactic theory. Belhare shares many properties with the nonconfigurational languages that have received attention in the literature, but it is unclear whether the language can be accommodated in existing theories of nonconfigurational grammar. In Belhare, there is a systematic and functionally motivated mismatch between the treatment of arguments in morphology and in syntax. There is, in other words, no uniqueness constraint on the projection of lexical arguments into grammatical representations. Current theories of nonconfigurational languages maintain that, just as in configurational languages, arguments are projected uniquely and uniformly into grammatical structure, either as nominals or as inflectional elements. If they are realised as both, there is either 'agreement' sensu stricto, or the inflection realises the arguments alone and the nominals stand in a 'resumptive' (Hale 1989) or 'adjunct' (Jelinek 1989) relation to the verb. Nei-

186

Balthazar Bickel

ther analysis seems suitable for Belhare since there is no uniqueness constraint in the first place. Abbreviations ACCEL ADD ADV.PASS ART BEN CIT COM DEM DIST DL EXCL FOC ID INCL INTR IPFV MED N

accelerative Aktionsart additive focus adversative passive article benefactive or malefactive citation form comitative demonstrative distal dual exclusive of addressee focus identifying focus inclusive of addressee intransitive imperfective aspect mediative ('via') case nominality marker

NPT NS OBL PERF PL PT Q REP RP SS/T SUBJ SUP SG TEL TEMP TOP TR =

nonpast nonsingular oblique case perfect plural past question reportative resultative perfect same subject and tense subjunctive mood supine ('purposive') singular telic Aktionsart temporary aspect topic transitive clitic boundary

Notes This paper was first drafted in 1996 and was accepted in 1997 for publication in a volume on Himalayan languages to be edited by George van Driem. As this volume had still not been realized as of Spring 2002,1 withdrew the paper and resubmitted it to the editor of the current volume. Given the primarily descriptive orientation of the paper I refrained from updating the theoretical discussion and limited myself to minor changes and corrections. Further theoretical discussion of the data presented can now be found in Bickel (2001) and Bickel and Nichols (2001). The research presented is based on fieldwork conducted in 1993, sponsored by the Max-PlanckGesellschaft zur Förderung der Wissenschaften, and in 1996, financed by the University of Mainz. Preliminary versions were presented to the Oriental Department of St. Petersburg State University, 18 March 1996 and to the Project Meeting of the European Cooperation Project on Himalayan Languages, Zürich, 20-21 June 1996. I wish to thank these audiences for helpful and stimulating discussions. I am also deeply indebted to Lekh Bahadur RaT, my main consultant in Belhärä, without whom 1 would never have been able

Hidden syntax in Belhare

1. 2. 3. 4.

5.

6.

7. 8. 9.

10.

187

to collect the crucial data on the not so obvious aspects of Belhare grammar. Thanks are also due to Kathrin Cooper and Robert Van Valin for helpful comments. All mistakes and misconceptions are of course my own. Nepali is transliterated according to indological tradition except for mute a which is not written. If this causes ambiguity, I introduce names in devanagarl. This notion is roughly equivalent to the Dependency Grammar concept of circonstant or freie Angabe. See Bickel (1993) for some discussion of this with respect to Belhare. In addition to this, there are two nonfinite passive participles with no known functional differentation, one in -pilat —pilau, one in -palak. On nöminapalienfis, see Section 3.2. As W. Winter suggested (personal communication, 17 June 1995), the sibilant marker in the transitive form may be interpreted historically as a reanalysed causative suffix. The different meanings and uses of the two types of 'perfect' are discussed in Bickel (1996: Chapter 9). The names of adults are teknonyms in Belhare, built on Nepalese personal names. Apart from teknonyms, the personal names are hardly ever used, for they are preempted by the ordinal names jethä 'first bom', mällä 'second born', sällä 'third born' ... känchä 'last born' borrowed from Nepali. In Bickel (1995a: 84), I also postulated an accusative pivot for zeroanaphora in topic clauses. Closer inspection has shown that the putative pivot is pragmatic rather than syntactic in nature. Given an appropriate meaning, zero anaphora can monitor any argument, including undergoers of transitive clauses: Kanchakubarj-chi ten-ma talokt-u-chi-naa last.born-OBL monkey-NS hit-ciT begin-3u-NS.u-TOP 0 pu-sa rj-khat-ca-he. run-SS/T 3NS.S-go-TEL-PT 'When Kancha began to hit the monekys, they ran off.' Lexico-syntactic units in Belhare may consist of two distinct morphological words; see note 15. First person singular is marked here as the singular value of 'exclusive', i.e., of [+speaker, -addressee]. See Section 4.2 for some discussion. The only other function of the ca-auxiliary is to encode reciprocal constructions: ten-ga-ten n-ca-he [hit-N-hit 3NS.S-AUX-PT] 'they hit each other (cf. Bickel 1997). The formative may be cognate with the verb root ca- 'eat, ingest, enjoy (a party), have as employment or problem, need (energy, of a technical device)', partially reflecting a pattern attested also in other languages of the Tibeto-Burman area; see Ebert (1994: 66). This observation obviously violates the Lexical Integrity Principle. However, since in terms of Lexical-Functional Grammar pivots are part of functional rather than categorical structure, Aa-nominalisations confirm the revised hypothesis that has been proposed by Mohanan (1995) and Bresnan

188

11. 12. 13.

14. 15.

16.

17.

Balthazar Bickel and Mchombo (1995): words are integer only with respect to categorical, not also with respect to functional structure. The conflation of ability and counterexpectancy in hir- closely parallels Nepali saknu 'be able to V; finish V, have already V-ed'. Thanks to William Foley (personal communication, December 1993) for drawing my attention to the syntactic relevance of Belhare idiom formation. The difference between (60a) and (60b) does not correlate with an Aktionsarten difference as one might expect from the point of view of Role and Reference Grammar. Both clauses behave as activities and differ from accomplishments. Inserted in both (60a) and (60b), the phrase ek chin One moment' would refer to a stretch of eating activity ('She will eat for a moment and stop.'), whereas with an accomplishment verb like kamma 'to tie up', ek chin can only refer to a post-event state, i.e., ek chin kam-t-u [tieNPT-3u] 's/he will tie it for a moment.' (The verb kamma is an accomplishment sensu stricto, i.e., it is durative and therefore compatible with the imperfective aspect: kam-yakt-he [tie-ipr-v- ] 'he was tying it (but didn't manage eventually).' See Bickel (1996: Chapter 12) for a disussion of Aktionsarten in Belhare. This sentence is translated incorrectly in Bickel (1995b: 419). It is a general feature of Belhare, briefly explored in Bickel (1996: Chapter 3), that the morphological word is not necessarily the same unit as what is treated as a word by the phonology and as what counts as a word for syntactic principles; cf. examples (9b), (20a), (50a), (85c), (87b). I retain the label 'agreement' for the sake of terminological simplicity, although there is, as we will see, nothing that the Belhare verb could 'agree' with in the strict sense of the term. This point will be taken up in the conclusions. For instance, the syntax of the Western Kiranti language Hayu also seems to be hidden, i.e., following principles entirely distinct from verb inflection and case marking. The {S, A}-pivot that needs to be postulated for the language (Michailovsky 1988: 201-203) has no analogon in morphology. In contrast to Belhare, no single construction in Hayu syntax appears to operate with ergative pivots.

References Bally, Charles 1926 L'expression des idees de sphere personelle et de solidarite dans les langues indo-europeennes. In Festschrift Louis Gauchat, F. Frankhauser and J. Jud (eds.), 68-78. Aarau: Sauerländer. Bickel, Balthasar 1993 Belhare subordination and the theory of topic. In Studies in Clause Linkage, Karen H. Ebert (ed.), 23-55. Zürich: ASAS-Verlag.

Hidden syntax in Belhare 1994

189

Mapping operations in spatial deixis and the typology of reference frames. (Working Paper No. 31, Max-Planck-Research Group in Cognitive Anthropology.) 1995a In the vestibule of meaning: Transitivity inversion as a morphological phenomenon. Studies in Language 19: 73-127. 1995b Relatives a antecedent interne, nominalisation et focalisation: syntaxe vs. morphologic en belharien. Bulletin de la Societe de Linguistique de Paris 90: 391^27. 1996 Aspect, Mood, and Time in Belhare: Studies in the Semantics - Pragmatics Interface of a Himalayan Language. Zürich: ASAS-Verlag. 1997 The possessive of experience in Belhare. In Tibeto-Burman Languages of the Himalayas, David Bradley (ed.), 135-155. Canberra: Pacific Linguistics. 2001 The syntax of experiencers in the Himalayas. In Working Papers of the International Symposium on Non-Nominative Subjects, Tokyo, December 18-21, 2001, Peri Bhaskararao (ed.), 207-237. Tokyo: Institute for the Study of Languages and Cultures of Asia and Africa. Bickel, Balthasar and Johanna Nichols 2001 Syntactic ergativity in light verb complements. Proceedings of the 27th Annual Meeting of the Berkeley Linguistics Society. Bresnan, Joan and Sam A. Mchombo 1995 The lexical integrity principle: Evidence from Bantu. Natural Language and Linguistic Theory 13: 181-254. Driem, George van 1991 Tangut verbal agreement and the patient category in Tibeto-Burman. Bulletin of the School of Oriental and Asian Studies 54: 520-534. Dixon, R. M. W. 1979 Ergativity. Language 55: 59-138. Dryer, Matthew 1986 Primary objects, secondary objects, and antidative. Language 62: 808-845. Ebert, Karen H. 1994 The Structure of Kiranti Languages: Comparative Grammar and Texts. Zürich: ASAS-Verlag. Foley, William A. and Robert D. Van Valin, Jr. 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Hale, Ken 1983 Warlpiri and the grammar of nonconfigurational languages. Natural Language and Linguistic Theory 1: 5—47. 1989 On nonconfigurational structures. In Configurationality, Laci Maracz and Pieter Muysken (eds.), 293-300. Dordrecht: Foris.

190

Balthasar Bickel

Jelinek, Eloise 1989 The case split and pronominal arguments in Choctaw. In Conßgurationality, Laci Maräcz and Pieter Muysken (eds.), 117-141. Dordrecht: Foris. Keenan, Edward 1976 Towards a universal definition of 'subject'. In Subject and Topic, Charles N. Li (ed.), 305-333. New York: Academic Press. Keenan, Edward L. and Bernard Comrie 1977 Noun phrase accessibility and universal grammar. Linguistic Inquiry 8: 63-99. Kuno, Susumu 1987 Functional Syntax: Anaphora, Discourse and Empathy. Chicago: University of Chicago Press. Lehmann, Christian 1984 Der Relativsatz. Tübingen: Narr. Marantz, Alec P. 1984 On the Nature of Grammatical Relations. Cambridge, Mass.: MIT Press. Michailovsky, Boyd 1988 La Langue hayu. Paris: Editions CNRS Mohanan, Tara 1995 Wordhood and lexicality: Noun incorporation in Hindi. Natural Language and Linguistic Theory 13: 75-134. Sadock, Jerrold M. 1991 Autolexical Syntax: A Theory of Parallel Grammatical Representation. Chicago: University of Chicago Press. Sapir, Edward 1921 Language, an Introduction to the Study of Speech. New York: Harcourt, Brace & Co. Van Valin, Robert D., Jr. 1993 A synopsis of Role and Reference Grammar. In Advances in Role and Reference Grammar, Robert D. Van Valin, Jr. (ed.), 1-164. Amsterdam: Benjamins. Van Valin, Robert D., Jr. and Randy LaPolla 1997 Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press.

On the notion of sentence in Classical Tibetan Claus Oetke

1.

The concept of 'sentence-in-Classical-Tibetan'

At the beginning of § 12 in Stephan V. Beyer's book The Classical Tibetan Language the following statement is made: A Tibetan SENTENCE consists of a proposition followed by a PERFORMANCE PARTICLE: The PERFORMANCE PARTICLE signals the way the proposition is being used. There are three basic ways of using a proposition in Tibetan - as a STATEMENT ... ,asaQUESTION ..., or as a COMMAND ... The STATEMENT PARTICLE is -o, the QUESTION PARTICLE is -Am, and the COMMAND PARTICLE is 73f/g. (Beyer 1992: 351)

In the present context the following fact is significant: If one leaves some details out of account, which call for minor modifications and some of which are explicitly acknowledged by the author, the above cited statement entails the following proposition: A sentence in Classical Tibetan is a unit which ends either in a statement particle or a question particle or a command particle.

I want to demonstrate that although the above formulated theorem is fraught with difficulties it is possible to defend this view, provided that one understands it against the background of theoretical assumptions which need to be made explicit. A difficulty for the above formulated tenet originates from the fact that Classical Tibetan possesses a number of other particles that can be employed at the end of units which could be considered as sentences. To give one example: at the beginning of the Tibetan version of the Buddhist story of king Udrayana, we find the following passage U 1,2 dehi tshe na hdzam buhi glin na gron khyer chen po Rgyalpohi khab dan / Sgra sgrog ces by a ba gnis yod de / gan gi tshe Rgyalpohi khab rgud pa dehi tshe na ni /Sgra sgrog hbyor par hgyur la /gan gi tshe Sgra sgrog rgud pa dehi tshe na ni /Rgyalpohi khab hbyor par hgyur ro //

192

Claus Oetke

'At this time there were in Jambudvlpa two big cities, namely Räjagrha and Roruka. Whenever Rajagrha declined, Roruka came to blossom; whenever Roruka declined, Rajagrha came to blossom.'

J. Nobel's German translation reads: "Zu jener Zeit gab es in Jambudvipa zwei große Städte, Räjagrha und Roruka mit Namen. Wenn Räjagrha verfiel, dann kam Roruka zur Blüte; wenn Roruka verfiel, dann kam Räjagrha zur Blüte."

In the Tibetan original a unit ending in a so-called "semifinal particle" (SP) corresponds to the first sentence of the translation(s). The units which have been separated by a semicolon in the translation(s) correspond to phrases which are linked by the particle la in the Tibetan version. It is not difficult to see that instead of the semicolon a full-stop could be used as well so that the units that are connected by la could also be considered as sentences under that aspect. There are even more extreme cases - and these are not at all rare - in which units ending in one of the initially mentioned final particles (FP) have the length of a third of a page or more and appear to correspond to paragraphs rather than to sentences. Therefore the question arises as to whether the correspondence between segments ending in final particles and sentences which has been suggested in the statement quoted at the beginning of the paper really holds true. An alternative view could be that the socalled final particles or performance particles mark the end as well as the character of larger discourse units, e.g. paragraphs, in such a way that if a series of subsequent sentences belong to the same type, say the type of declarative sentences, this character is only marked once, namely at the end of the last unit of the series. The semifinal particle which occurs in Classical Tibetan in variant forms - ste, te and de - often connects units which, as in the above-cited example, resemble sentences and which are often rendered into full sentences in translations into European languages. Accordingly, the emphasis of the present investigation lies on this particle. The results hold good, mutatis mutandis, for other "non-final" particles, too. My thesis is that the view of a correspondence between ending in a final particle and sentencehood can be vindicated for Classical Tibetan, provided that this theorem is associated with a specific explication that prevents the possibility of ruling out alternative conceptions of a sentence in the language.

On the notion of sentence in Classical Tibetan

193

The arguments - arranged according to increase of strength - are the following: 1. The semifinal particle also connects phrases which must be regarded as constituents of sentences, e.g. M I 174, 12 gan su yan run sie / de Ita buhi dnos pos hdihi srog bskyabs te / mi nams par byed nus pa yod dam / 'Is there anyone - whoever he might be - who is able to save her life and keep her free from damage?' M I 205,21 dehi tshe na kun dga bo dan hkhor man po thams cad bcom Idan hdas kyis gsuns pa la yi ran ste mnon par bstod do 'Thereupon Änanda and the many followers rejoiced in that which was spoken by the Venerable One and praised it.' 2. Punctuation. Units ending in semifinal particles are mostly treated in the same manner as clauses which are constituents of sentences. Sometimes no punctuation marks occur, e.g. U 2,4 ji tsam na rgyal po U tra ya na blon pohi tshogs kyis yons su bskor te don gyi gnas na hdug pa na fshon pa de dag skyes thogs te rgyal po U tra ya nahi gan du don ba dan / rgyal po U tra ya nas de dag la hons pa legs so / zhes bya bahi tshig gis kun tu spyad de gnasji Ita ba bzhin du hkhod do / 'Just at the time when king Udräyana, surrounded by the crowd of his ministers stayed at his ruling office, those merchants went together with gifts to the place where king Udräyana was, and king Udräyana greeted them with the words 'Welcome' and sat down at an appropriate place.'1 3. Ergative constructions. If an intransitive and a transitive verbal phrase are connected by a semifinal particle the employment of an ergative marker is possible - but not obligatory, e.g. Tib. Ill 38,24 nas den grofi dan /gron khyer dan /gron rdal de dag tu son ste rgyal po dan blon po chen po dan ...la sags pahi pho bran dan khyim dag tu bsod snoms bslari bar byaho will go now to that town, to that village, to that hamlet and ask for alms in a palace or house of a king, a great minister... [etc.]' M II 231,20 buhi ma ma yin pa des ni bu la snin rje med pas snad kyis mi dogs te mthu ci yod pas drafts so // The one who was not the mother of the child, since she had no compassion for the child, did not fear hurting [it] and pulled with all her strength.'

194

Claus Oetke

M II 229, 19 chu dehi nan nas tshur sin mkhan zhig stehu kha nas khyer te hon no 'From that river a carpenter was coming carrying an axe in his mouth.'

4. "Dependent" verb-forms. Although the future-stem is employed in Classical Tibetan in order to convey what the speaker intends to do, the socalled perfect stem is used in non-final members of a sequence of clauses expressing the speaker's intention to perform actions, if those non-final clauses are connected by a semifinal particle, e.g. Tib. Ill 301,7 den bdag gi Ins hdis yans sin rgya ehe bahi las mnon par bsgrubs nas skye sihi rgya mtshor gru gzins chen po byas te / hkhor ba yons su spans nas nes par hbyun ba thob par byaho 'Today I will perform by this body a great deed and make it into a big vessel in the ocean of the Samsara and [thus] arrange for my own salvation after completely giving up the circle [of births].'

5. Embeddings. The semifinal particles - as well as other "nonperformative particles" - can be used in order to connect phrases which constitute embedded sentences, e.g. i/wf-sentences or subordinate sentences. This is (as far as we can see) not permissible with respect to any of the "performative" particles.2 Examples of occurrences of semifinal particles within embedded clauses or subordinate sentences are not rare in Classical Tibetan, e.g. Tib. Ill 286,11 bsod nams kyi dhos po sin tu legs par bskyed de bdag cag la phan btags nas bde bar htsho bar byas pa legs so legs so 'It is good, it is good that you have in a good manner fostered the merits and having brought help to us made us live happily.' U 35, 20 lha / ji liar de gnis kyis sgyu mas hjig rten bslus te dad pas sbyin par bgyi ba chud gzan pas byi lahi skye gnas nan par skyes te / den sah nid du yan ran ran gi mchod rten na mchis pa / lhahi mnon sum du bgyiho / 'Your Majesty, we will demonstrate to your Majesty that those two, since they have cheated the world by trickery and embezzled the pious gifts have been [re]bom in the bad rebirth of cats and (that they) live even today each in his own stupa.'

On the notion of sentence in Classical Tibetan

195

M I 174, 6 stag mo zhig bu byun nas zhag du ma Ion pa bkres sin skom pas nen te / phyiryah bu za la thug pa zhig mthon nas ... 'After they saw a tigress, who had given birth to offspring a few days ago, who was hungry and plagued by thirst and moreover was ready to devour her offspring ....' M I 174, 19 lam de nid du zhugs te stag mohi tshan gan na yod par myur du son nas... 'After they entered the same way [back] and went quickly to the place where the abode of the tigress was ...' Tib. 11127,31-28, 10 gal te de bzhin gs'egs pa yan dagpar sems can rnams la byams pa dan / thugs rje chen po mna ste / thugs brtse has phan hdogs sin bde ba hthob par mdzad de pha ma ha bur gyur pas gzhan de dan mtshuns pa ma mchis pa dan .... sems can rnams la snoms par gzigs par mdzad de sgra gcan zin la ci Ita ba bzhin du thugs brste ba la mi snoms pa mi mna ba zhig lags na bcom Idan hdas kyis bdag la smon lam gcig gnan bar mdzad du gsol // 'If the Tathägata indeed possesses great compassion and pity towards the beings, and [if] no other equals him since he has become like father and mother [to them] inasmuch as [he] is of benefit [to them] and makes [them] obtain happiness on account of his compassion and if he looks upon the beings in an impartial manner and is [thus] [someone whose] affection [which is] like [that which he has for his son] Rahula possesses no partiality, then the Venerable One should please grant me one gift to choose.'

The question could arise as to why the above-mentioned phenomena constitute arguments for the thesis. To this query the following answer can be given: It appears, prima facie at least, quite improbable that a particle which is frequently employed in order to connect sentence-constituents like nounphrases or verbal phrases which are themselves constituents of verbal phrases should equally possess the function of connecting independent sentences, i.e. sentences which are not themselves constituents of more complex sentences. This bestows some weight on the observations mentioned in (1). However, this fact is not decisive. Conjunctions like 'and' might occur both as constituents of sentences and as connectives between sentences or even larger "discourse units". Unfortunately, we do not have access to intonation-patterns of Classical Tibetan since it is a dead language. However, as the rules of punctuation

196

Claus Oetke

are not explicitly and strictly regimented it appears probable that the use of punctuation signs in individual cases reflects intonational and rhythmical properties of utterances. Therefore the circumstance that the semifinal particle is often treated in a manner similar to other "non-performative" particles but dissimilar to the common final particles ho, ham and cig possesses significance. It is not always easy to draw a theoretically motivated distinction between a connection between verbal phrases of a single sentence on the one hand and a connection between sentences with some ellipsis on the other. For example, if one encounters a structure like a) NP - VPi - CON - VP2' in surface structure, one might wonder whether this should be interpreted as b) NP - (VP, - CON - VP2)vp or as c) NP, - VP, - CON - NP2 - VP2 with deletion of identical NPs. In this connection the fact is relevant that sometimes noun-phrases are syntactically marked in accordance with a distant VP. The available material suggests that in constructions in which a VP with an intransitive verb is connected by a semifinal particle with a VP containing a transitive verb, such that the agent of the transitive predicate is identical with the subject of the intransitive predicate, one can optionally use either an ergative or non ergative marking for the pertinent nounphrase. The same probably holds good if an intransitive VP follows a transitive one (See e.g. M II 229, 19 ....sin mkhan zhig stehu kha nas khyer te hon no}. In contradistinction to this, the category of a verb cannot affect in syntactical respect any noun-phrase that occurs in a clause which is separated by any final particle. A unit which is separated from another unit by a semifinal particle can also affect the latter one from the viewpoint of interpretation, as the phenomenon (4) shows. Even in this respect, the semifinal particle differs from any final one. A sequence of the form 'NPi - VPi p f- FP + NP2 - VP2fut - FP' can never be understood as saying that somebody should carry out the action expressed by VPi as well as the action expressed by VP2 or as an expression of a speaker's intention to perform both actions, even if both NPs are identical. Accordingly it appears sensible to assume a correspondence between the difference between final and semifinal particles (or other clause final particles) on the one hand and the distinction between different sentences versus one sentence on the other. The available material suggests that a clause of any complexity can be syntactically embedded, provided its constituents contain only non-final particles. One can therefore assert that any unit which ends in a non-final

On the notion of sentence in Classical Tibetan

197

particle is not merely incomplete in the sense that it requires some other unit to follow it but that it is also "syntactically open", i.e. its status with respect to a hierarchy of syntactic levels is by itself undetermined. The syntactic level to which such a unit belongs is only determined by the context. The situation is different with regard to units ending in one of the "performative" particles = final particle. Units which end in semifinal and other non-final particles are thus undetermined in two respects: 1. They are (partially) undetermined under the aspect of grammaticality, i.e. it can happen that a phrase ending in a nonfinal particle would be ungrammatical in the context of some subsequent units but grammatical in the context of others. 2. They are undetermined with regard to syntactical level in the manner described above. These facts together vindicate the "traditional view" which, as far as I can see, has not been supported by explicit arguments, that the particles ho, ham and cig constitute a separate category that is distinct from that of the semifinal and other particles. But they also corroborate the view that units ending in other than final particles are not separate sentences but at best sentential phrases. Accordingly, the contention that many texts written in Classical Tibetan contain extremely long sentences possesses a basis. There is, however, another side of the coin. If one considers the informational content of many units ending in semifinal particles - and sometimes of those units ending in other non-final particles - it appears appropriate to regard them as complete sentences. How should one resolve the conflict? To exploit the well-known distinction between langue and parole and to say that in a number of cases one deals with complete sentences from the viewpoint of parole but with sentential phrases from the viewpoint of langue might be one possibility. But one could also refer to the equally well-known type-token distinction and acknowledge that certain units occurring in speech or in texts are on the one hand complete sentencetokens and on the other hand instantiate incomplete sentence-types. Other solutions are conceivable, too. At any rate, the problem of the notion of 'sentence-in-Classical Tibetan' calls for a differentiated answer and it seems that Classical Tibetan is an interesting object of study from the aspect of explicating the concept of sentence from a linguistic point of view.

198

2.

Claus Oetke

Some peculiarities of the use of semifinal particles in Classical Tibetan

There can be no doubt that the semifinal particle is employed in Classical Tibetan in order to connect linguistic units of various levels, such that in some contexts of occurrence a semifinal particle could be replaced by a final one and in others by particles which are used in order to connect either sentential or verb- or noun-phrases without affecting the status of grammaticalness, such as nas, la, ein. On the other hand, it appears certain that some occurrences of the semifinal particles could not be replaced by any final ones while still preserving grammatical acceptability. The question arises as to whether it holds true that if a semifinal particle is replaceable by another particle without causing grammatical unacceptability it is also substitutable without affecting truth values and if a substitution salve veritate is possible, the replacement is semantically irrelevant. After the investigation of textual material amounting to roughly five hundred pages I feel confident in claiming that the supposition can be definitely rejected that substitutions of semifinal particles by other grammatically possible alternatives or vice versa are generally semantically irrelevant. On the other hand, it is more difficult to decide whether replacements affect truth value. There are apparently a number of cases in which replacements do not change the truth value and possibly do not even affect truth conditions. But the question as to what extent the differences entailed by substitutions remain irrelevant with respect to truth conditions cannot be definitively settled here. In order to avoid the danger of distortions ensuing from the peculiarities of metrical texts, above all those which are generated by metrical constraints, only non-metrical passages have been investigated. The textual corpus comprises both translations from Sanskrit and translations from Chinese sources. The special emphasis on translations is partly motivated by the fact that translations represent the overwhelming majority of texts from the early Classical period and partly by the circumstance that translated texts offer possibilities of comparison which are methodologically advantageous. The mode of employment of the particles which are relevant here is almost entirely independent of the grammatical features of the languages of the source texts and this holds good in particular for translations from Chinese. Accordingly, it can be safely assumed that the results presented below do not reflect idiosyncratic features of Tibetan translational language. On the other hand, due to the limitation of the corpus it is not yet

On the notion of sentence in Classical Tibetan

199

possible to settle in a definitive manner whether observed phenomena represent linguistic or stylistic features. However, in view of the circumstance that the data are taken from diverse sorts of texts and since the results pertain only to conspicuous facts and tendencies, they must be considered significant. Semifinal particles that connect verbal phrases can often be substituted by other particles. This is made manifest by different translations of identical textual passages. Two Tibetan translations of a passage of the Suvarnaprabhäsasütra which are based on Yi-jing's Chinese version of this Mahäyäna-sütra reveal relations of interchangeability between the semifinal particle and the particle nas. The following is a literal rendering of the Chinese original, in which the sentential phrases are separated by hyphens and the particles occurring at the end of the phrases are given in brackets. The first expression before the oblique refers to a (non-canonical) translation which has been preserved in fragments from Turkestan, whereas the expression after the oblique relates to the translation which is found in the Tibetan Buddhist canon (Tib.III). 'At that time the four heavenly kings all got up from their seats (nas / SP) put their garments on one side of their shoulder (SP / nas) - worshipped with their heads the feet [of the Buddha] (nas / SP) - placed their right knee on the ground (0 / nas) - put the palms of their hands together [in order to] pay respect (nas I SP) - [and] praised Buddha's merits with [the following] verses (f? I pa)' (T432a8-10)3

The symbol '0' expresses that no particle after a verbal stem occurs, pa refers to a particle which can be attached after verbal stems and transforms verbal phrases or sentences into noun-phrases. One can easily recognize that, with one exception, a semifinal particle in one version corresponds to nas in the other version and vice versa. Moreover, both versions "switch" between semifinal particle and nas. The above cited passage represents a "stereotype" which occurs at various places in the Sutra and this circumstance allows one to recognize that the tendency to "switch" between semifinal particle and nas is a recurrent phenomenon. Moreover, the "stereotype" sometimes exhibits minor variations, insofar as certain phrases are omitted or additional ones occur and in such cases one can observe that even within one and the same version the same verbal stems are connected with different particles, e.g. Tib. Ill 163, 29 ... lans te ... gzar nas ... btsugs te ... sbyar nas ... gsol to; Tib. Ill 156,32 ... (dga mu rans par gyur nas) ... lans te ... gzar nas ... btsugs te ... sbyar nas

200

Claus Oetke

... phyag byas te ... gsol to; Tib. Ill 4,10 ... lafis nas ... don ste ... phyag byas nas ... lan gsum bskor ba byas te ... hkhodpar gyur to\ Tib. Ill 8,19 ... lans nas ... don ste ... phyag byas nas ... bskor ba byas te ... hkhodpar gyur to l Th 508 Ia4 ... lans te ... son nas ... phyag htshal te ... bskor nas ... hkhod do. There are passages in which identical particles follow each other, e.g. Tib. Ill 13,19 ... lhags te ... phyag byas nas ... bskor ba byas nas ... hkhod do but since violations of "switch" are relatively exceptional in those contexts, they do not falsify the diagnosis that a pattern A — B - A - B ... was a preferred mode of formulation. These observations suggest that at least in contexts in which sequences of events are told, the choice between semifinal particle and nas is mainly determined by stylistic considerations like avoidance of repetitions and not - or less - by semantic facts. It would be a mistake to believe that nas and semifinal particle are the only alternatives that occur in contexts in which sequences of actions and events are mentioned. That both semifinal particle and nas possess a number of other possible substitutes is made clear by the Tibetan translations of the following passage of Yi-jing's version of the Suvarnaprabhäsasütra: Again, a literal translation is given and the equivalent particles occur in brackets: 'When [he] obtains and keeps [this] magical formula [he should] first recite it a thousand times (nas /nas) - after this [he should] in a clean room spread Gomaya on the ground (0 / SP) - make a small altar-place (la / nas) - offer with an attentive mind drinks and food according to the circumstances (ein / FP) - always burn excellent incense (nas / nas) - let the smoke never interrupt (la / FP) - recite the afore(mentioned) essential formula (nas / ein) - day and night focus the mind [on it] (na / nas) - [he should] hear [it] only with his own ears (pa tsam du /pa tsam du) - [and] let not others grasp [it] (par / na) - at this time a son of king Vaisrävana, called Sanisi, [will] manifest the form of a young man (SP / SP) - come to his place (nas /nas) - [and] ask (FP /pas): For what reason do you need to call my father? - thereupon [he] should reply (FP / FP): I need treasures in order to worship the three jewels (SP / na) - [I] want [them to be] given now.' (T431al4-20)4

It is easy to see that in this passage some occurrences of semifinal particle and nas are to be found in identical contexts, whereas in other places a semifinal particle of one version corresponds to na or does not possess any verbal particle as a correlate at all - because the original verbal-phrase 'spread Gomaya on the ground' has been rendered in the version of Th 507

On the notion of sentence in Classical Tibetan

201

by a noun-phrase meaning 'by dung of an ox'. On the other hand, the circumstance that one phrase which ends in nas in one version correlates to a phrase ending in na in the other probably reflects a difference of interpretation of the original and can therefore not be employed as an argument for a substitutability relationship. Nevertheless, the above-cited passage is also noteworthy for other reasons. On the one hand, it shows that units which end in a final particle might consist only of a verb phrase in the surface structure and indicates that there are contexts in which final particles are replaceable by verbal particles other than semifinal ones (ein, la}. On the other hand, the fact that the unit which precedes the segment referring to Sanisi's arrival is concluded by na in both versions demonstrates that verbal particles other than semifinal ones can connect units possessing different "logical subjects". One could raise the question, however, as to whether the existence of a common topic in the form of the person who recites the magical formula is an essential condition of connecting the units in this way. The following passage illustrates even more clearly that sentential units possessing different "logical subjects" can be connected in various ways: 'Then the noble goddess had spoken those words (pa dan J ma thag tu) - all beings [in] each and every one of the evil worlds of the five defilements [got] a golden colour (0 / nas) - possessed the marks of great men (SP / SP) - were neither men nor women (nas / nas) - sat on jewel-lotuses (nas I SP) - obtained immeasurable pleasure (pa ni / pas) - like [in] the heavenly palaces of the Paranirmitesvara[-gods] (SP / pa bzhin du) - there were no evil forms of existence ((0 / SP) - jewel trees were placed in arrays (0 / ein) - lotus-flowers of seven [kinds of] jewels completely filled the world (0 / FP) - moreover it rained wonderful heavenly flowers of seven [kinds of] jewels (0 / 0) - one played heavenly music (FP / SP) - the noble goddess Älokacintämani immediately changed her female body into the body of a Brahman-god (pa dan I FP) - then the god Mahabrahman asked the Bodhisattva Älokacintämani ... (FP / FP)' (T 425c5-l I) 5

We can distinguish four main segments: 1. the section in which it is said that Älokacintämani spoke the afore-mentioned words (which are not cited here), 2. the section referring to various miraculous events, 3. the segment referring to Ä's transformation of her body and 4. the section which expresses that Mahabrahman spoke to A. The second segment can be subdivided into a section relating to changes affecting all living beings and a section referring to general changes of environment. Sections 1 and 2 are connected by the conjunctive particle dan ('and') following after a nomi-

202

Claus Oetke

nalized sentence in Th 507 and a construction possessing the meaning of 'immediately after' - consisting of the compound expression ma thag tu after a sentence ending in a final verb. Sections 2 and 3 are linked by a semifinal particle in Tib. Ill, whereas in Th 507 the preceding segment ends in a final one. Sections 3 and 4 are linked in Th 507 in the same manner as sections 1 and 2, but in Tib. Ill a final particle separates both segments. Th 507 and Tib. Ill differ considerably in the manner in which the internal parts of the second main section are connected: Tib. Ill contains two units separated by a final particle - obviously triggered by the circumstance that in the original an expression with the meaning 'moreover' occurs. Although no final particles appear in the corresponding section in Th 507, the existence of units which are "zero-connected", i.e. which are placed after another without any connecting particle, is remarkable. There is no reason to suspect that those differences are due to substantial divergences regarding the analysis of the original text. Taken together, the observations suggest the following consequences: I. Units which under the aspect of semantic content correspond to sentences can be treated in Classical Tibetan in (at least) four different ways: 1. They can be concluded by a final particle. 2. They can be marked by a semifinal particle at the end. 3. They can be linked by connectives like dan ('and') occurring after a nominalizing particle. 4. They can be connected with subsequent units by various expressions either complex (like ma thag tu) or non-complex (in particular particles like na, ein) - occurring after verbal stems.

II. There are contexts in which various alternatives are admissible. It does not generally hold that the content of a sentence or a textual passage uniquely determines the way in which a sentential unit is treated. One might wonder why there are so many possibilities at all. It appears a priori improbable that there should be a natural language which possesses such a great number of divergent means for achieving one and the same goal. Since, on the other hand, the above presented observations admit such a conclusion, one should critically ask whether it is really true that no functional differences exist between the above-mentioned modes of expression. On the one hand, there can be hardly any doubt that many occurrences of the semifinal particles in Classical Tibetan admit various replacements, including "zero-replacement", such that the resulting expression is not only grammatically acceptable but also pragmatically admissible in the sense

On the notion of sentence in Classical Tibetan

203

that the achievement of important communicative goals which can be realized by using units with semifinal particles is not seriously impaired by using the alternatives. Otherwise, the facts which are revealed by comparisons of different translations of identical source texts are hard to explain. But, on the other hand, the textual material exhibits conspicuous tendencies pertaining to the employment of the semifinal particles, which suggest that neither the function nor the semantic import of such particles occurring after sentential units completely coincides with any other alternative. The following features are noteworthy: 1. The semifinal particles are used in connection with explanations and specifying remarks. LPP 90 (35) (= P 247a7) dehi don ni hdi yin te / hdi ni nes pa las gzhan thob pa ma yin gyi / hon kyan skad cig mama yin pa nid dam .... gzhan yah ruh sie / nes par byas sin lhams cad mi thob par zad do 'Its meaning is this: This is not an obtainment of anything else than [ofthat which has been] ascertained; nevertheless the non-momentariness or ... any other [property] whichever it may be, all these are not obtained [even if they] have been ascertained.' LPP 112 (46) (= P 251a2) chos mthun pahi rnam pa gnis te / snah ba dart brtags paho 'There are two kinds of similarity: A visible and a conceptual (one) // The kinds of similarity are two: [Namely] visible and conceptual.' M I 174,2 rgyal po de la sras gsum mna ste / rab ni sgra einen po zhes byaho // hbrin po ni lha chen po zhes byaho// tha chun ni sems can chen po zhes bya ste / sras tha chun de chun hu nas byams pa dan snin rjer Idan te / thams cad la bit gcig pa dan hdraho // That king had three sons: The eldest was called 'Mahänäda'. The middle one was called 'Mahädeva'. The youngest one was called 'Mahäsattva': This youngest son was from childhood endowed with love and compassion and was [accordingly] for everybody like [an own] single child.' The literature contains innumerable examples in which semifinal particles occur in connection with explanations of meaning, of elucidations of concepts and of specifying genera by species as well as in narrative texts after the introduction of new figures, which are more specifically described by subsequent phrases. Moreover, other alternatives hardly ever occur in phrases of this type, which is characterized by the fact that one could employ a colon ':' in the corresponding place in English or German translations.

204

Claus Oelke

2. Semifinal particles occur at the end of phrases which are followed by sentences in which reasons are adduced. AVS 67 don med pa. nid yin te I skad cig ma nid kyis khyab pa yod pa de bzhin du chos nan la rtogs na skad cig ma nid du yah rtogs so zhes zer na (vaiphalyam eva ksanikatvavyäptasya sattvasya tathätvena dharmini pratitau ksanikatvasyäpi pratiter iti cet) Objection: "It is indeed useless. For if existence which is pervaded by momentariness is recognized as such (i.e. as existence which is pervaded by momentariness) in some property-bearer the momentariness [of the property-bearer] is also recognized'". P 323,13 (Chap. Ill, 1) yod pa ma yin te / hdir mig ni ha has na Ita bar byedpa yin la / gzugs ni dehiyul nid yin par fie bar bstan to // (na santi/ iha hipasyatiti darsanam caksuh / tasya ca rupam visayatvenopadisyate) 'They (i.e. the instruments of seeing etc) do not exist. [For] here (i.e. in Buddhist dogmatic) it is taught that the eye, inasmuch as it sees, is that which makes see and [on the other hand] the visible form/colour is its object.' Again, there are innumerable similar instances which could be cited. In translations of Sanskrit texts the particle hi ('for') often occurs in the corresponding source passage. 3. Semifinal particles are used in main clauses if a subordinate clause follows. The standard rule is that subordinate clauses precede main clauses in Classical Tibetan but, especially in translations from Sanskrit texts, the reverse order is equally common. However, whenever a subordinate clause is postponed, a semifinal particle is used at the end of the main clause. This phenomenon is extremely common in constructions with causal clauses, e.g. PV I 36, 13 gzhan khofi du child par bya bahi don du bstan bcos kyan byed la / rjes su dpag pa yah hdor ro zhes bya ba ni mi rigs te / rjes su dpag pa de spon na hjugpa hbras bu med par hgyur bahi phyir ro 'That one on the one hand writes a scientific treatise in order to instruct other people and on the other hand rejects inference [as a means of knowledge] is unreasonable, because if one discards inference the enterprise [of writing a treatise] becomes meaningless.'6 AVS 99 gan yod pa de ni skad cig ma yin te / bum pa bzhin no (zhes ...) (yat sat tat ksanikam yathä ghata (iti...)

On the notion of sentence in Classical Tibetan

205

'That which is existent is momentary, like a pot // in the same manner as a pot [which exists is also momentary]' AVS 101 de dag kyan gan yan run bahi chos can der gzun bar bya ste / gan du gzun bar nus par ro (tau punar yatra tatra va dharmini grahitavyau yatra sakyau grahuum) 'These two are to be grasped in any property bearer where they can be grasped.' 4. There are a significant number of cases in which a semifinal particle occurs before a phrase which expresses a consequence or a resume of what has been said before: P (111, 7) 329, 7 hdir sred pahi rkyen gyis len pa zhes bya ba la sags pas len pa dan srid pa dan skye ba dan rga si la sags pa ni mam par ses pa la sags pa bzhi las skye ste / dehi phyir rnam par ses pa la sogs pa dag yod de de dag gi hbras bu yod pahi phyir ro (iha trsnäpratyayam upädänam ityädinä upädänabhavajätijarämaranädikam vijnänädicatustayäd utpadyate / tasmät santi vijnänädlni tatkäryasadbhävät) 'Here, on account of [the principle] that upadäna [arises] from craving as cause etc., upadäna, becoming, birth, old age and death etc. originate from the collection of four, [namely] cognition etc., [and] therefore cognition etc. [must] exist because its effects are existing.' Tib. Ill 290, 27 hdi hgags pahi phyir de hgag par hgyur ba hdi ha ste / ma rig pa hgags pas hdu byed hgag / hdu byed hgags pas rnam par ses pa hgag .... rga si hgags pas mya nan dan smre snags hdon pa dan /sdug bsnal ba dan hkhritg pa rnams hgag par hgyur te I hdi liar sdug bsnal gyi phun po hdi hba sig hgag par hgyur ro 'The [manner of) removing that in order to remove this is as follows: By removing ignorance the samskäras are removed, by removing the samskaras vijnana is removed, .... , by removing old age and death misery, lamentation, distress and disturbance are removed; in this manner alone the mass of distress is removed.' One might suppose that the various phenomena pertaining to uses of the semifinal particles can be accounted for by assigning to the semifinal particles the function of indicating that a subsequent clause represents an explication, with respect to the unit which ends in a semifinal particle. After all, the occurrences in which the explicative or specifying nature of a subsequent unit is obvious. Cases in which one could naturally employ expressions like 'this is to say', 'namely', 'more specifically' - 'das heißt', 'näm-

206

Claus Oetke

lieh', 'insbesondere', 'und zwar' - in English or German translations are too numerous in order to consider this as accidental. Moreover, one could assume that both the employment in the context of subsequent causal phrases and the occurrences of semifinal particles before phrases expressing consequences or summaries can be explained as an explication of the legitimacy of the linguistic act consisting in the production of the unit which ends in a semifinal particle. Whereas in the former case the explication of legitimacy relates to the truth of what is said by giving evidence for its truth, in the latter case the explication of legitimacy relates to relevance in the form of making manifest in which respect the act of saying something is important and significant. Although I see no way of disproving this account, at present I am inclined to prefer an alternative solution. Considering the fact that the expressions 'P because (of) Q' and 'Q and therefore P' or 'Q and accordingly P' are equivalent, one can explain the observed phenomena equally well by assuming that the basic function of a semifinal particle consists indeed in indicating the relation of an explication or specification but in such a way that the meaning of the particle itself is neutral with respect to the order of explicans and explicandum or specificator and specified. This means that in a similar manner as the English expression 'and' by itself does not entail any specification regarding the temporal order of events although it is commonly employed in order to connect clauses expressing events which happen in a temporal order which is analogous to the order of their formulation the semifinal particle in Classical Tibetan does not semantically entail that the subsequent clause represents an explication or specification of the content of the preceding clause or vice versa. If the sequence ' explicandum — semifinal particle - explicans' occurs more frequently than the reverse order, the reasons might be closely related to those which are responsible for certain phenomena that can be observed regarding the use of conjunctors of natural languages. There are at least two reasons for preferring the last-mentioned account: 1. It is better suited than the first mentioned alternative to account for constructions like the one which one encounters in the passage of Tib. Ill 33,23-34,24. Here it is first said that the Bodhisattvas should realize that it has to be understood in ten different ways that the Buddhas have taught the Nirvana as existent; then follows a question to the effect of what the ten ways are and a specification of the ten ways and after this follows a concluding remark to the effect that those are the ten ways in which the Nirvana has been taught as existent. Whereas the introductory

On the notion of sentence in Classical Tibetan

207

remarks ends in a final particle, the last sentence of the middle part - which relates to the tenth point - and the concluding remark are separated by a final particle. Moreover, the phrases which describe the first nine items end in final particles. This means that we have a structure of the form Ά - FP - Question - B, - FP - ... B9 - FP - BIO - SP - C'. A most natural explanation of this phenomenon is that the semifinal particle serves to indicate the "switch" which leads from the level of specification back to the ordinary level. The phenomenon that items which equally belong to a specifying level are separated by final particles is not uncommon. In the above cited example M l 74,2 two phrases belonging to the same level of specification follow a phrase ending in a semifinal particle. Here the structure obviously is: (A-SP) Le veii - (Bi - FP)Leve|2 - (B2 - FP) Le vei2 (B3 - SP)Level 2 - ((C - SP)Level 3a ~ (D - FP)Leve, 3b)Leve| 3.

2. On the basis of the suggested account the fact that semifinal particles often occur after verbal phrases which function like adverbial modifiers and could be rendered by adverbial phrases in translations into European languages does not present any problem. If semifinal particles typically occur in connection with sterns like bcas or Idan and form expressions which can be rendered by 'together with', one can explain the phenomenon as an outcome of the rule that specifying verbal phrases ending in a semifinal particle can precede verbal phrases expressing something which is specified by the former ones. The occurrence of semifinal particles in examples like MI 174, 12, where a semifinal particle is used together with gan su yan run in order to generate a parenthetically employed indefinite phrase (translatable as 'whoever it may be'), is in accordance with the hypothesis that semifinal particles indicate switches of level of specification. The circumstance that semifinal particles are often replaceable by nas ('after') and vice versa is not mysterious either. If events occurring in a sequence are mentioned, describing them as occurrences which happen one after the other or as episodes which take place under circumstances which have been created by another episode merely represent two different ways of looking at the matter. The proposed hypothesis entails some consequences which are worth mentioning. 1. Constructions of the form Ά - SP - Β - phyir - SP - C - (phyir) FP', e.g. in PV II 22,8-11; 46,18-20; 68,3-8, should be interpreted as containing second-order reasons and are thus not equivalent to constructions of the form Ά - SP - Β - dan - C - phyir - FP', e.g. in PV II 24,10-13, or Ά - SP - Β - la - C -phyir- FP', e.g. PV II 66,33-68,3.7

208

Claus Oetke

2. Since the semifinal particle possesses an "anti-coordinative" import there is a contrast not only with respect to coordinations of nominal phrases by the conjunctive particle dan, but also with regard to connections of verbal or sentential phrases by the particle la as well as to asyndetic juxtapositions of phrases. The particle la after verbal stems typically occurs in contexts in which one can employ On the one hand - on the other hand' in a translation. Accordingly la implies parity and not disparity. Apart from the examples which some of the previous citations provide (like U l ,3-5), numerous other instances can be found in other texts, including translations from Chinese, e.g. Tib. Ill 38,2 hkhor ba ni nes dmigs dan bcas pa yin la my a nan las hdas pa ni ne bar zhi ba yin ... — 'The Samsära is connected with faulty ideas, [and/but] the Nirvana [on the other hand] is complete tranquility'.8 It seems that la, although it emphasizes a contrast, differs from the (concessive) particle kyi inasmuch as does not entail any reference to the circumstance that the truth of one proposition lets one expect the falsity of another.9 Asyndetic juxtaposition appears to be another way of expressing coordination. Possibly, the employment of asyndetic constructions is associated with the function of suggesting lack of dependence between the propositions which are coordinated. If this is true, one should gather from the manner in which the occurrence of several miraculous events is formulated in the above cited passage of Th 507 III Ial4— Ib5 that those events happened independently of each other, although they (might) possess a common cause. The way in which asyndetic constructions are employed in other instances corroborates such a diagnosis. In M II 230,11 bdag nam khun nas phyir hbyun bahi tshe ni bde / slar hjug pahi tshe sdug bsnal na / cihi phyir de Itar gyur = 'Why does it happen that I am happy whenever I come out from the hole [and] I am unhappy when I enter again?' it is natural to suppose that the asyndetic construction has been employed in order to insinuate the non-existence of a connection between the mentioned phenomena and in this way underscores the bewilderment about the fact that they coexist. Since other instances of uses of asyndetic juxtapositions suggest similar consequences, the supposition that asyndetic constructions exemplify an iconic feature of Classical Tibetan is by no means eccentric. On the other hand, one has to keep in mind that it is still not certain whether the observations which one can make on the basis of the investigated textual corpus represent linguistic or stylistic features. Nevertheless, it should be safe to conclude that the existence of replaceability relations between semifinal particles and a number of alternative forms of expression

On the notion of sentence in Classical Tibetan

209

in Classical Tibetan does not seriously endanger the principle that it is uncommon for natural languages to possess a large number of different terms which completely coincide regarding meaning and use.

Abbreviations The following abbreviations are used in this paper: FP: final particle(s); SP: semifinal particle(s).

Notes 1.

2. 3.

In the Tibetan translation of the Pram naviniscaya the general rule is that units ending in a final particle are marked by '//', whereas after units ending in a semifinal particle '/' occurs - or no punctuation mark at all. On the other hand, in this and in many other texts '//' is not - or at best in exceptional cases - employed after a semifinal or other non-final particle. It is true that final particles can occur in units which are embedded by the "citation-particle" ces but this is a special case. The passages read as follows (non-nominal particles occurring after verb stems are in bold type): Th 507 V, 7al5-7b2 de nas rgyal po bzhi hdugpa las lans nas // bla gos phrag pa geig du bzar te //spyi bos zhabs la phyag htshal nas // pus mo gYas pa hi lha nas sa la btsugs /gus par thai mo sbyar nas // tshigsit bead pa hdis sans rgyas gyiyon tan la bstod do Tib III 191, 8-12 de nas rgyal po chen po bzhi po gcig car stan las lans te bla gos phrag pa gcig la gzar nas bcom Idan hdas kyi zhabs gnis la mgo bos phyag byas te pus mo gYas pahi lha ha sa la btsugs nas gus pas thai mo sbyar te tshig su bead pa gya nom pa dag gis bcom Idan hdas kyi yon tan la mnon par bstod pa

4.

Th 507 V, 5al4-5b5 snags bsgrub pa hi tshe / dan por snags hdi ston bton nas / de hi hog du khyim gtsan mar ba Ian gi lei has man dal chim nu gcig byas la dus dus su kha zas gyis sems thagpa nas mchod ein rtag par bdug pa bsregs nas // du ba myi chad par byas la snar smras pa hi snags gyi snin pos gzhan gyis ma thos par bdag gis πια bar grags pa tsam du bton nas // sems gcig du byas na / de hi tshe be sa ra ma na hi bu zhen li si zhes bya ba gzhon nu hi gzugs su snah ste / hons nas hdi skad ces hdri ho //khyod ci hiphyir na yiyab hbod//hdi la Ian hdi zhes smros sig //ha dkon mchog gsum la mchod pa byed pa ni nor rdzas hdodpa hi don te // myur du sbyin par smon to

210

Claus Oetke Tib. Ill 5-16 snags hdon pahi (she snar Ian ston tshan bar bzlas nas dehi rjes la khan pa gtsan ma dag tu ba Ian gi lei has dog sa la zhal zhal bgyis te dkyil hkror chun nu zhig bgyis nas zas ci hbyor has sems rise geig tu mchod par bgyiho // rtag par pog spos dam pa bsregs nas dud pa rgyun mi hchad par bgyiho / snags kyi snin po sna ma gdugs mtshan du sems dran pa ne bar gzhag nas bdag nid kyis thos pa tsam du zlos sin gzhan gyis rig par mi bgyid na dehi tshe rnam thos kyi bu sa ni si zhes bgyi ba gzhon nuhi gzugs ne bar bstan te drun du mchis nas hdi shad du cihi phyir bdag gi pha hbod ces hdri bar hgyur has Ian du hdi shad ces bdag dkon mchog gsum mchod pahi don du nor htshal na de bzhin du stsal bar gsol zhes smra bar bgyiho

5.

Th 507 III Ial4-lb5 de nas rigs gyi lha mos de skad ces smras zin pa dan / rnog ma rnam lhahi hjig rten gyi skye bo yod do // cog thams cad kyah gser gyi kha dog I myi chen po hi mtshan dan Idan te // skyes pa han ma yin / bud med kyah ma yin bar gyur nas // rin po ehe hi pad mo la hdug nas I Ions spyod dpag tu myed pa ni gzhan hphrul dban byed gyi gzhal myed khan dan hdra ste // nan hgro thams cad kyan myed / rin po ehe hi sin ni hphren bar dnar // rin po ehe sna bdun gyi pad mos ni hjig rten khyab par yan gan // yan mchog gi rin po ehe sna bdun gyi lha hi men tog gi char ni bob // lha hi rol mo yan byed do // de hi dus na yid bzhin rin po ehe hod hphro ba rigs gyi lha mo bud myed gyi Ins las htshans pa hi lha hi lus su gyur pa dan //de hi tshe tshans pa chen po hi rgyal pos yid bzhin rin chen po hi hod hbar ba / byan chub sems dpa la / hdi skad ces smras so Tib. Ill 157,28-158, 8 dehi tshe rigs kyi lha mos de skad ces smras ma thag tu dus nan pahi snigs ma Inahi sems can ji sned pa thams cad gser gyi kha dog tu gyur nas skyes bu chen pohi mtshan dan Idan te skyes pa yan ma yin bud med kyan ma yin par gyur nas rin po chehi pad ma la hdug ste / bde ba tshad med pa dag la Ions spyod par gyur pas gzhan hphrul dban byed kyi lhahi gzhal med khafiji Ita ba bzhin du nan hgro rnams kyan med de / rin po chehi Ijon sin phren du bsgrigs sin rin po ehe sna bdun gyi me tog gis hjig rten gyi khams kun tu gan bar gyur to /gzhan yan lhahi rin po ehe sna bdun gyi me tog gi char yan mnon par bab lhahi rol mo rnams kyah byas te / rigs kyi lha mo yid bzhin rin chen hod hphro yan bud med kyi lus yohs su bsgyitr nas tshans pahi lus su gyur to // de nas tshans pahi rgyal po chen pos byan chub sems dpa yid bzhin rin chen hod hphro la hdi skad ces smras so

On the notion of sentence in Classical Tibetan 6.

211

It is equally possible that a relative pronoun occurs in the causal clause, e.g. P (IV, 8) 335, 34 dehi de ha bu de thams cad ni Ian btab pa ma yin te / gan gi phyir tshor ba la sags pa rnams kyi yod pa yah bsgrub bya gzugs kyi yod pa dan mtshuns par rig par byaho (tad etat sarvarh tasyäparihrtarh bhavati / yasmäd vedanädlnäm api sadbhävah sädhyena rüpasadbhävena samo veditavyah) 'All this remains unanswered for him, because one has to acknowledge that the [condition of the] existence of vedanä etc. is the same as [that of] the existence of rüpa, which is to be proven.'

7.

8. 9.

One should not suppose that a "switch of level" indicated by a semifinal particle must always relate to the entire subsequent unit. In PV II 104, 23-26 a construction of the form -phyir - SP - (B - na - C)-pas - D - FP' occurs and it seems that the semifinal particle is used in order to indicate that the immediately subsequent unit, i.e. '(B - na - C)-pas' or the conditional phrase 'B - na', represents a level of lower embedding. In this way it is made clear that the logical structure corresponds to 'Since A, therefore, because (if B then C), D' and not to 'Because (since A, if B then C) therefore D' or to 'Because C (if B because of A), therefore D' The same chapter of the same text offers a number of other instances, e.g. 33,15; 35,16; 36,2; 37,17; 38,5; 40,1; 40,17. M. Hahn (1996: 134) adduces examples for non-contrastive occurrences of kyi, where the particle allegedly indicates "adverbial" and "temporal" relations. It is doubtful, however, that kyi possesses such completely divergent uses and Hahn's own examples put this thesis into question. The sentence legs par slob kyi le lo mi byed do, which Hahn translates: 'Indem (oder: Dadurch daß) man ordentlich lernt, ist man nicht faul', manifests an underlying contrast if one considers the non-negated counterpart of the proposition of the main-clause: Being lazy is a property which makes the attitude of not learning properly expectable. If one reverts the order of the clauses, the employment of 'but' or 'sondern' - e.g. 'Man ist nicht faul, sondern lernt ordentlich' - is natural. Perhaps one should consider kyi - if it occurs after verb-stems - as a "symmetric counterpart" of English 'but' or German 'sondern', i.e. kyi is a contrastive particle which is neutral with respect to the order of constituents; one might say that it merges 'P but Q' and its converse 'Q but P' or that it represents the union of the German sentenceoperator 'sondern' and its converse. Hahn's other examples could be explained in a similar manner.

212

Claus Oetke

References AVS =

Antarvyäptisamarthana (Page numbers refer to: Y. Kajiyama, The Antarvyäptisamarthana of Ratnäkarasänti, Tokyo 1999.) Beyer, Stephan V. 1992 The Classical Tibetan Language. New York: New York University Press. Hahn, Michael 1996 Lehrbuch der klassischen tibetischen Schriftsprache. SwisttalOdendorf. LPP = Laghuprämänyapariksä (Page numbers refer to: H. Krasser, Dharmottaras kurze Untersuchung der Gültigkeit einer Erkenntnis Laghuprämänyapariksä, Wien 1991.) MI = Mdzans blun, Chapter 2 (Page numbers refer to: M. Hahn, Lehrbuch der klassischen tibetischen Schriftsprache, 7th edition, SwisttalOdendorf 1996.) M II = Mdzans blun, Chapter 39 (Page numbers refer to: M. Hahn, Lehrbuch der klassischen tibetischen Schriftsprache, 7th edition, Swisttal-Odendorf 1996.) P= Prasannapadä (Page numbers refer to: J. May, Candrakirti Prasannapadä Madhyamakavrtti, Paris 1959. The numbers in brackets relate to the chapter of the work and the corresponding verse of the Mülamadhyamakakärikäs.) PV I = Pramänaviniscaya, l. Chapter (Page numbers refer to: T. Vetter, Dharmakirti's Pramanaviniscayah, l. Kapitel: Pratyaksam, Wien 1966.) PV 11 = Pramänaviniscaya, 2. Chapter (Page numbers refer to: E. Steinkellner, Dhatmakirti's Pramanaviniscayah, 2. Kapitel: Svärthänumänam, Wien 1973.) T= Taishö-edition of the Chinese Tripitaka. Th 507, 508 = Tibetan translation of Yi-jing's Chinese version of the Suvarnaprabhasasütra which is preserved in the Dun-huang fragment Th 507, Th 508 (and others). (A transcription of the relevant portions of the manuscripts is given in my dissertation Die aus dem chinesischen übersetzten tibetischen Versionen des Suvarnaprabhäsasütra, Wiesbaden 1977. The Roman numbers refer to individual fragments.) Tib. Ill = Tibetan translation of Yi-jing's Chinese version of the Suvarnaprabhäsasütra which is completely preserved in the Tibetan Tripitaka (Page numbers refer to: J. Nobel, Suvarnaprabhäsottamasütra, 2. Band, Leiden 1958.) U= Udräyanävadäna (Page numbers refer to J. Nobel, Udräyana, König von Roruka, Wiesbaden 1955.)

On discourse functions of the finite verb in Kinnauri narratives Anju Saxena

1.

Introduction

When telling a narrative, the narrator organizes the story in units, with some units larger than others. The use of optional linguistic elements such as the decision to use a finite clause instead of a nonfinite clause, or the choice between a full NP, a pronoun or a zero anaphora - may in some cases be seen as a tool which a skilled narrator employs in order to manipulate the listener's attention. Various linguistic devices are used to bring the main storyline in focus. Distribution of tense and aspect markers is one example. The aim of this paper is to examine the structure and functions of the finite verb in Kinnauri narratives.1 The question that will be raised here is: What does a finite clause encode in Kinnauri narratives? This will be investigated by examining both the distribution of finite and non-finite units and the tense and aspect variation in Kinnauri narratives. The results presented in this paper suggest that the occurrence of the finite verb in Kinnauri narratives correlates with discourse boundaries, which may perhaps be referred to as episode-boundaries. Further, the tense-aspect variation observed in the narratives, too, has discourse-related functions. The narrative past tense (V-N.PST) is the default finite ending in most Kinnauri narratives. Divergence from the default finite verb system is a linguistic resource. For instance, V-TNS-HON occurs when the main participant, in the narrator's view, deserves respect. V-PROG 5£YPRST has two discourse-related functions. It begins a new episode and it encodes peak situations. V-PROG ££-N.PST, on the other hand, occurs in situations which are preludes to the next episode. It occurs in non-primary events, providing the background information. V-1MPF BE-N.PST occurs only in background situations (restricted to "Orientation" in the Labovian model), describing states or habitual actions involving the main participant. The organization of this paper is as follows. Section 2 provides a brief background on the Kinnauri syntax and the narratives examined in this paper. Sections 3 discusses the factors crucial in the occurrence of a finite

214

Anju Saxena

clause in Kinnauri narratives and section 4 concerns the tense and aspect variation in Kinnauri narratives.

2.

Background on Kinnauri

Kinnauri is a verb-final language. A finite verb in Kinnauri contains a verb stem, tense marker and a subject agreement marker. An object agreement marker, aspect marker and evidential marker may also occur in the final verb complex (for details, see Saxena 1995). Copulas may also occur in noncopula constructions, where they function as auxiliaries. The tense markers in copula constructions as well as in noncopula constructions involving the copulas are the same, i.e., -tal-lil-lel-tol-il-o (future tense), -kel-ge (simple past tense) and -gyy (narrative past tense).2 The present tense does not have any overt marking. The narrative past tense (-gyo) occurs only with third person subjects in traditional narratives. Its primary function seems to be to encode the end of a discourse unit.3 (1)

06/073 nane ki-n Pad mes-pa $au aunt you(HON)-POSS Son.in.law last.night food

am zaza/ little eat/NF

prat-o-ls pi cay luq-o du// serving.plate-LOc-ABL starch drink-pRoo BE/PRST '(The smart man said to the old woman:) "Having eaten very little last night, your son-in-law is (now) drinking starch (leftover liquid from rice) from the serving plate.".'

Kinnauri makes frequent use of the converb construction.4 A finite unit in such cases is composed of a combination of converb(s) along with one or more embeddings. For example, (2)

08/143 kam a work

lan-nu her ay / make-NOM occasion

basry-s small.hatchet-lNST tshuishu

du-k/t

tie/PERF

BE-lSG

k eke/ give/NF

'(The girl said:) "While working, (the finger) got cut with a small hatchet, (therefore) (I) have tied (it).".'

Discourse functions of verbs in Kinnauri

215

The data for the present study come from the Kinnauri narrative corpus which I have collected and compiled.5 For the present purposes, I have examined narratives 06, 07, 08, 12 and 13 in the Kinnauri corpus. The motivation for selecting these narratives is partly that I have a better grasp of these narratives and partly the fact that informants of these narratives come from the same region in Kinnaur (i.e., the Sangla county). These narratives are traditional folktales. Traditional folktales in Kinnauri are usually about one specific person and that person's journey through a series of "episodes", where other participants of varying importance show up in one or more episodes before they vanish. The moral of these narratives is often the victory of good over evil. Narrative 06 has 284 clauses (this includes finite as well as non-finite clauses), 07 has 250 clauses, 08 has 422 clauses, 12 has 256 clauses and 13 has 356 clauses. Because of the small size of the material examined for the present purposes, the results presented here should be considered preliminary until they are verified against the larger Kinnauri corpus. The presence of the narrative past tense, direct speech, repetition and gestures are some characteristic features of the traditional narratives in Kinnauri. Direct speech seems to have a special role in traditional narratives. Since most of these direct quotations are arguments of cognition-utterance verbs, technically they function as direct objects of the verb, but direct quotations in traditional Kinnauri narratives are usually long and have a complex internal structure. Direct speech is outside the scope of this paper.

3.

Finite verbs in Kinnauri narratives

A narrative may be seen as a chain of related and not so related events. Structural clues are, at times, used to draw the listener's attention to the way the narrator would like to portray the relationship between a set of events which, in part, is also determined by the interpretive conventions of the community which speaks the language. A distinction is sometimes made between the main story-line and the non-primary story-line (Longacre 1990). The former is the core of a narrative ("bone" Forster 1977) which moves the story ahead and the latter gives any additional information needed to understand the main story ("flesh" Forster 1977). Kinnauri, as mentioned above, makes frequent use of the con verb construction, where one or more converb segments are followed by a base segment (Johanson 1995).6 The latter carries the finite verb morphology.

216

Anju Saxena

Syntactically the converb segments are dependent on the base segment, although semantically the converb segment and the base segment have the same status. Converb is a semantically neutral clause-linkage device. Its default interpretation is to mark sequentiality. Converbs are propulsive and move the storyline ahead (see Johanson 1995). There are other syntactic devices available in the language which provide more specific information about the type of clausal relationship. (3)

13/029 do raza (s)he king

baba/ come/NF

puja lanlan / prayer make/NF hod that

macks fish

lanlan / make/NF

naij-o-nu mafä/j-o tata du-gyo ma // plate-PL-POSS middle-LOC keep/PERF BE-N.PST DISC.PRTL 'That king came, (he) prayed, (he) prepared the fish, (and he) put (it) on the plate.' An examination of the distribution of the converb segments and the finite verbs in narratives suggests that the finite verb in Kinnauri narratives has a discourse-related function (Converb segments may or may not end with the base segments. In the latter case they occur as part of another complex "sentence"). This distribution marks the end of a discourse unit, which may perhaps be referred to as a thematic unit or an episode. A distinction is made here between narrative information units, which move the story line ahead, and background information units, which provide background information. Some of the discourse factors which trigger the narrative information units are: change in discourse topic, shift in point of view, change in time and place and world-shift (shift between direct speech and non-direct speech).

Discourse functions of verbs in Kinnauri (4)

21 7

Change in discourse topic7 08/018-019 pfipfi / take.away/NF i-parj zaijdgal-o id kim-ts kztkat-u a-place forest-LOC a house-DlMlNU cow.dung.cakes-POSS km-ts lala/ house-DiMiNU make/NF hxl.'y-r kayag that-LOC dance

so-n-nu / dance-NOM-NOM

se-kyo// send-N.PST ruza-ts an-sya Iao=tob9ij bof-o sex/ o.man-DiMlNU self-NOM dried.cucurbitaceous. fruit tree-LOC send/NF 'Having taken (them) away, in one place in the forest (he) built a small house with cow dung cakes, there (he) sent (them) to dance. The old man himself sent [=hanged] a dried cucurbitaceous fruit on a tree, ...' (5)

Change in time 07/068-069 then

skin/NF

tsei kh£tsikhetsi / all separate/NF

khur-ai}-o se -gyo// stable-at-LOC send-N.PST s.*n k^-ma is// morning see-NOM EMP 'Then (she) skinned (the horse), (she) separated all (body parts), (she) put (them) in the stable. In the morning (when she) saw, ...'

Two factors which are crucial for invoking the background information unit are the general background information essential to understanding the narrative ("Orientation" in the Labovian model) and the narrator's comment, both of which provide additional information about the narrative or

218

Anju Saxena

its participant(s) ("External evaluation" in the Labovian model) (see examples 7-8 below). (6)

Shift between direct speech and non-direct speech 08/092-94 csste-nu lentil-POSS

den on

ao

si-k //

ONOMO

die-ISO

brsttsi-gya // fell-N.PST

lo-kyo // say-N.PST phal&-yo

me

sko-gyo //

stove-LOC fire move-N.PST '(He) fell down on the lentils. "Oh! (I) am dead." (He) said. (He) poked in the fire (to see when the girls ran away).' (7)

General background information 08/174-175 kuion-u id t^etsats du-gysl/ demon(F)-POSS a girl BE-N.PST dam good

gas-ο trana clothe-PL jewelry

kuton-K

an-u

tarftaij/ see/NF

cimed-u

cudus

rar a/

demon(F)-ERG self-POSS girl-ACC

pinch

give/NF

kram-mu / cry-NOM sed-o

du //

send-pROG

BE/PRST

'The female demon had a girl. Having seen expensive clothes and jewelry, the female demon herself pinched her daughter, (and) sent (her) away crying.' The narrator's comments are addressed to the listeners directly. They are not part of the main story-line. The background information unit ("Orientation") may be of global or local significance. It occurs at the beginning of the narrative and is also interspersed at various points in the narrative.

Discourse Junctions of verbs in Kinnauri (8)

219

Narrator's comment 08/069-070 koni kul-ic// rice. shaft crush-IMF (s)he-day

dok=suy-u those.two-ACC

nis-mi two-man

za-mu/ eat-NOM

du-gyo// be-N.PST id kag bd-kyz// a crow come-N.PST '[(The demon) took out a rice shaft, (and) gave (it).] (The demon said:) "Crush the rice shaft!" That day those two were to be eaten. A crow came.' Apart from the explicit background information units (such as the narrator's comment, as in example 8, for instance), languages also make use of more implicit ways of providing the background information given within the temporal framework of the story-line. Linguistic strategies, such as the choice of the lexical item, intensifiers, direct speech and tense-switching are some linguistic mechanisms reported in the literature (Fleischman 1990; Schiffrin 1981; Silva-Corvalän 1983). Converb segments within one finite unit encode one thematic unit in Kinnauri narratives. Each convert) segment represents one sub-unit which contributes to the making of the thematic unit. (Slobin 1995 makes a similar suggestion concerning the converb - in Turkish) The base segment (i.e., the finite clause in the converb construction) links this thematic unit to the next thematic unit. An examination of these narratives suggests that the occurrence of a finite verb in Kinnauri narratives correlates closely with discourse boundaries, which, in turn, are triggered by one of the aforementioned factors. But what is subsumed within one discourse unit is a decision that the narrator makes. It seems that when the narrator wants to emphasize each of the smaller events, or each part of a sequence, she uses the finite verb form to present each unit as a discourse unit by itself (example 9). This diverges from situations where a sequence of events is encoded by means of a converb construction, which subsumes them as part of one discourse unit and also as part of one finite unit (example 10).

220

Anju Saxena

(9)

08/098-100 dele

dele

quickly quickly koden walnut

sd-e

bi-gyo//

terrace-DiR

go-N.PST

brec-B / slip-NF

tfatluj-o da-kyo// courtyard-LOC stop(v)-N.PST tfatliq-o yag-o brime-s k9-kyo// courtyard-LOC yak-PL mother.yak-ERG bring-N.PST 'Quickly, quickly (he) went to the terrace. (He) slipped on the walnut, (he could only) stop (himself) in the courtyard. In the courtyard the yaks took (him).' (10)

08/206 raza-s king-ERG

pyuk=tsh3r-u wood-ACC

tots/ take.out/NF

kutonu cimed-u dok-c frana demon(F)-POSS girl-POSS then-ABL jewelry kuton-u

cimz-psi) ι-ρΐη-ο

gas-o Jfaylfdy/ clothe-PL grab/NF

cub-ospa laglag/

demon(F)-POSS girl-ACC a-face-LOC coal-ash smear/NF t ot 3

se-kyo //

take.OUt/NF

SEND-N.PST

'The king took out the wooden piece, (he) took the jewelry and clothes from the female demon's girl, (he) put charcoal on the female demon's girl, (he) took (her) out [=threw her out].' The presence of a finite verb in Kinnauri might thus be seen as a linguistic cue to the listener that the present episode is now finished, and that what follows is a separate discourse unit. There might also be a more narratororiented explanation for the use of the finite verb morphology to mark the end of an episode (Clancey 1980), i.e., the narrator retrieves one episode at a time from her memory (the assumption being that a narrative is composed of a string of smaller episodes).

Discourse functions of verbs in Kinnauri

4.

Tense-aspect variation

4.1.

Introduction

221

The tense-aspect system in narratives tends to differ from the tense-aspect system in non-narratives (Dahl 1985). Dahl (1985) (as well as Labov 1972: 337) states that the tense-aspect system is richer hi non-narratives than in narratives. The two basic assumptions (Fleischman 1990) concerning temporal orientation in the narratives are: 1. Every narrative has a speaker. 2. There are two temporals involved in a narrative: the time of narration ("speaker-now") and the temporal framework reflected in the narrative ("story-now").

There may be differences as to which tense and aspect occurs within "story-now" and "speaker-now" situations. The narrative past tense (N.PST in the examples) in Kinnauri occurs only in traditional narratives and there, too, only in "story-now" situations. When "speaker-now" tenses do occur in "story-now" situations (for example, the present tense marker), they can either refer to the "speaker-now" situation (where it primarily encodes its referential interpretation, which at the same time contributes to the discourse of the narrative), or to the "story-now" situation, in which case they primarily have a discourse-related function. The example below illustrates the occurrence of the narrative past tense. It further shows that while the temporal layout and sequencing of the various events is chronologically ordered, the narrative as a whole is presented as a recapitulation of the events which occurred earlier on, real or imagined (therefore retrospective-presentation) from the narrator's perspective. In this way, retracing forward is a hallmark of traditional narratives. The iconic-sequencing in narratives involves the iconic relationship between the sequence of events in the narrative and the sequence of clauses in which those events are reported. (See, for instance, example 11 below.) Linguistic description of narratives takes the iconic-sequence as the default clausal relationship. A narrative is described by Labov as "a method of recapitulating past experience by matching a verbal sequence of clauses to the sequence of events (it is inferred) actually took place" (1972: 359).8

222 (11)

Anju Saxena 08/107-112 sd-e bi-gyo// terrace-DIR go-N.PST sol-e-c

tfya-gyo //

terrace-DlR-ABL see-N.PST nappay-o-c nes danic-is bi-ts/ from.that.side-LOC-ABL over.there hill.between.two.ditches-iNST go-NF taq-gyo// see-N.PST fured-o / run-NF bi-gyo// gO-N.PST

lfab-mu catch-NOM

berarj / occassion

ti phikya-gyy// water throw-N.PST '(He) went to the terrace. From the terrace (he) saw (X). From that side over there, on a hill between two ditches, (the girls) go (are going), (he) saw. Running, (he) went (there). At the time of (demon's) catching (them), (the girls) threw water (at him).' However, the linear-ordering is, at times, not maintained. Instead, morphological, lexical and prosodical patterns describe the clausal relationship. Further, narratives can recreate past events as if the narrator is presenting an eyewitness account. The progressive aspect and the present tense are often used for this purpose (see example 7 above). The ocurrence of the present tense in "past tense contexts" has drawn some attention in the literature. Jespersen (1924: 258), for instance, suggests that the use of the historical present is a stylistic device to present a past event in a dramatic way, making it more vivid, by presenting the past event as if it is happening right in front of the listeners. Wolfson (1979), too, argues that historical present and past tense taken together have a discourse function. According to her, the alternation of these two tenses in narratives is "to structure the experience from the point of the view of the speaker and to dramatize it" (1979: 216). The direction of the switch (from historical present to past or from past to historical present) is irrelevant.

Discourse functions of verbs in Kinnauri

223

Both types of switches, according to Wolfson, organize the narrative into chronological pieces, emphasizing the event which is viewed by the narrator as the most significant. Much of the work on tense-aspect variation in narrative discourse has been done within the Labovian framework, with some notable exceptions (for example, Fleischman 1990; Levinsohn 1992; Longacre 1976a, 1976b, 1990). In the following section we will examine the tense and aspect variation in traditional Kinnauri narratives. It will be suggested below that the shift away from the default finite system (simple narrative past tense marker (N.PST) in most cases) to other tense/aspect markers is a grammatical device used to encode discourse-related functions.

4.2.

Tense-aspect variation in Kinnauri narratives

The time reference is usually specified at the beginning of Kinnauri narratives. However, the tense and aspect markers occur regularly in the narrative at the "sentence-level" as in many other languages with grammaticalized tense and aspect. The primary function of tense/aspect markers in such contexts is no longer their referential interpretation. Rather, it is their nonreferential interpretation which becomes their primary function. The finite verb systems which occur in declaratives in Kinnauri narratives are: V-N.PST, V-PST(-HON), V-PROG BE-(N.PST/ HON),9 V-IMPF BEN.PST, V-PERF BE-N.PST10 and V-PERF AUX-N.PST.11 As mentioned earlier, the default finite verb system in the narratives examined here is the narrative past tense (-gyo)· The present tense does, however, occur at times in these narratives. Future tense, on the other hand, never occurs in nondirect speech in Kinnauri narratives.12 Here we will examine the distribution of some "diverging" tense and aspect markers in Kinnauri narratives. 4.2.1. Distribution of V-PST-HON Kinnauri has three copulas: to, du and ni. Two of these (to and du) inflect for tense and agreement (including honorificity) in all tenses. The third copula (ni) inflects only in the future tense. In all instances of the copula

224

Anju Saxena

construction with the honorific marker, it is the copula to which occurs (see example 12). In the direct-elicited data the distribution of to and du with the honorific marker (-s) is related to the animacy of the subject: to occurs with animate subjects and du with inanimate subjects. This generalization holds even in these narratives. All instances of to-ke-s 'be-PST-HON' in the narratives have animate subjects. (12)

07/090

nal-κ to-ke-s// enjoy-iNST be-PST-3H 'All were joyful.' Copulas to and du also function as auxiliaries in noncopula constructions. oe-PST-HON seems to occur in noncopula constructions, when the main participant (subject/agent in the finite unit), in the narrator's view, deserves respect. Which participants might potentially receive respect also depends on the socio-cultural norms of the society. Participants such as god, cow, saintly man or the hero or the heroine of a narrative with god-like qualities receive this finite verb form in the narratives. (13)

08/128

εηιη-u ysfaq udder-POSS below

byo-e go-ADV

raij/ time

pa-gi-s // hit.w.feet-PST-3H 'When (the demon) went under the cow (examining her udder), (the cow) hit (the demon).'

4.2.2. V-PROG

BE(-TNS/HON)

In finite constructions, the progressive aspect marker is followed by a copula auxiliary which carries the finite verb morphology (see, for example, 1 above).13 It occurs both in direct speech (both as part of the main clause and inside the direct quotation) as well as in other constructions. The shift from the simple narrative past tense to the present progressive aspect seems to be a linguistic cue to its listeners that the finite unit encodes a change in orientation. The three permissible structures with the progressive aspect are: V-PROG 5£/PRST (V-o du), V-PROG 5£-N.PST (V-o du-gyo) and V-PROG 5£-HON (V-o du-s). V-PROG 5£/PRST occurs predominantly in

Discourse junctions of verbs in Kinnauri

225

two contexts, where it either begins an episode (example 14) or it encodes peak situations (for example, 18). V-PROG 5£-N.PST, on the other hand, occurs in situations which are either preludes to the next episode, or where the finite unit represents a non-primary event which provides the background information spanning over time. This information is relevant for the narrative, but it does not move the storyline forward. The distribution of V-PROG BE/PRST and V-PROG fi£-N.PST is described below.

4.2.3. V-PROG BE/PRST The historical present in Kinnauri narratives occurs only with the progressive aspect. The reference time of the finite units with the present tense is prior to the time of speaking. The temporal ordering is made explicit by discourse. The present tense occurs in certain critical junctures in Kinnauri narratives, described below.

4.2.3.1. In non-peak situations V-PROG BE/PRST occurs primarily at the beginning of a new episode. In such contexts it does not occur in more than one finite unit at a time. (14)

07/024-025 byo-o du // gO-PROG

BE/PRST

nyum-s=ko-c rakses from.behind-ABS demon

k^erya-o / follow-PROG

byo-gyo// gO-N.PST

'(She) is going. From behind, the demon went following (her).' The repetition of the last unit of the preceding finite unit as the first clause in the following finite unit is a frequent phenomenon in Kinnauri narratives (see, for example, 15). The finite verb with V-PROG BE/PRST does not usually repeat the contents of the finite unit preceding it as its first clause. This hints at the possibility that V-PROGfi£/PRSTin non-consecutive

226

Anju Saxena

finite units encodes detachment from the discourse unit preceding it (see Section 4.2.3.2 for the occurrence of V-PROG BE/PRST in consecutive finite units).14 (15)

07/005-006 dok do-mya then (s)he-day tsb€tsats girl

id BMP

tshetsats hact-gys// girl become-KPST

hac-s / become-NF

id solas a fox come-N.PST 'Then, that day (a) girl was born. The girl having been born, a fox came.' Further, there is a strong tendency that V-PROG BE/PRST in such contexts also marks a change in the focussed participant (see example 16 below). The use of the present tense in traditional narratives carries special connotation because of its potentially ambiguous nature: it can either have a present tense interpretation (i.e., simultaenous with "speaker-now") or have a past tense interpretation (i.e., simultaneous with "story-now"). Because of its present tense interpretation, V-PROG BE/PRST provides, in some contexts, the sense of an eyewitness account. In all these cases V-PROG BE/PRST is thematically neither a part of its preceding or following unit. An example of this phenomenon is provided below. (16)

06/079-081 byu// gO/IMP

IPs shit(N)

phikyakya/ throw/NF

Jd-ra// come/iMP-iMP bi-yo

du //

gO-PROG BE

Discourse functions of verbs in Kinnauri

dok then

227

lo-kyo// say-N.PST

nane ki-n

Pad

mes=pa

khau

am

zaza/

aunt you(H)-POSS son.in.lawyesterday.night starch little eat/NF '(The youngest prince said:) "Go! Throw the shit and come!" (The priest) is going. Then (the prince) said: "Aunt, your son-in-law ate little last night...".'

4.2.3.2. In peak situations Several linguistic devices are used to demarcate the peaks from the rest of the narrative. Sudden bursts of direct speech, repetition, an amassing of details and the narrator's evaluative comment just before or after the peak are some linguistic devices, as mentioned in the literature, used to draw the listener's or the reader's attention. Longacre (1976b), among others, suggests that a shift to present tense is often used to mark the peaks and to bring out the details more vividly. V-PROG ÄE/PRST is also a linguistic device used to highlight the importance of that which is being narrated or, in other words, to increase the tempo. In such cases V-PROG BE/PEST occurs consecutively in a sequence of finite units. The same prominent participant(s) continues in all the finite units which comprise a peak. (17)

12/021-027 Iay-u hadd-i COW-ACC time-EMP supaks evening pyltete retum/NF

za-gya// eat-N.PST

kim-o byo-o house-LOC go-PROG

du// BE/PRST

beraij / occassion

do rats ai n is far cai}-ts-D (s)he calf another two lion child-DIMINU-PL kim-o

bsd-o

du //

house-LOC come-pROG BE/PRST kim-o

bd-n-na

house-LOC come-NOM-NOM

fa /

BMP

228

Anju Saxena do far tar ifakaij-o swig polats-rarj (s)he lion mouth-LOC red blood-with tos-is sit-PERF do far (s)he lion galya

du BE/PRST

darkat) // content

& aq-ts-3-s ta child-DMiNU-PL-ERG BMP ran-o

ne-o du// know-PROG BE/PRST

du //

scold(N) give-PROG BE/PRST '(She) ate the cow at once. In the evening (she) is going home. When (she) returned, the calf and the two cubs are coming home. When (they) come, that lioness is sitting content, with blood in (her) mouth. That lioness* cubs are knowing. They are giving (=throwing) abuses. (The cubs said to their mother, the lioness: "Where have you sent the calf s mother?")'

In peak situations the first unit of a peak does not repeat the contents of its preceding finite clause. Within a peak situation, however, the recapitulation of the preceding clause occurs rather frequently (see, for example, 17 above). Two structural patterns are found within peak situations (with V-PROG BE/PRST). The first pattern involves a peak situation which comprises predominantly simple finite units or shorter convert» constructions. The peak situation concerns more than one participant ('lioness' and 'her cubs' in 17 above). The other pattern involves highly detailed finite units. The entire excerpt concerns the same focussed participant(s) ('king' in example 18 below). Dense coding is a rhetorical device for emphasizing the importance ofthat which is being conveyed. Language uses different means to distinguish main events from the rest of the narrative. "If all parts of a discourse are equally prominent, total unintelligiblity results ... Prominence is achieved in many ways in a discourse" (Longacre 1976b: 10). Here we have seen that V-PROG BE/PRST occurs frequently in Kinnauri narratives, which draws the listeners' attention to important events in the narrative. This finite verb system also highlights the detachment of this unit from the unit preceding it.15 This characteristic of V-PROG BE/PRST reminds one of what Turner (1963: 62) describes as "a fresh paragraph in the narrative" (quoted in Levinsohn 1992: 455). Further, the occurrence of V-PROG BE/PRST in consecutive finite units is one linguistic strategy to mark peak situations in Kinnauri. But this is, however, not the occurrence of V-PROG BE/PRST which encodes the peaks. Sometimes it is just the reverse, i.e, the whole narrative is in V-PROG BE/PRST and it is the lack of this finite ending which encodes peaks. This

Discourse functions of verbs in Kinnauri

229

seems to the case in the narrative 17 in the Kinnauri corpus. This suggests that it is not V-PROG 5£/PRST per se which is the linguisitc device marking main events. Rather, it is one (among other such) linguistic device which marks a "change in pace" crucial to the highlighting of important information units. (18)

13/129-130 teyari dok dag-c then there-ABL preparation wapss return

lan-o du // make-PROG BE/PRST

muluk-o bd-n-nu/ an-u self-ACC country-LOC come-NOM-NOM

dok kim-o then house-LOC

bs-n-na-ta/ come-NOM-NOM-EMP

wal ksmai lan-tsya/ earning make-NOM big wal lan-tsya-go/ kam 3η work make-NOM-PL big do-go lo-n-njg tsl (s)he-PL say-NOM-NOM EMP

completely=ECHO=completely do nis (s)he two mala=matra property

ta banbant EMP plenty leks-is/ ?bum-NF

hac-ΰ / become-NF ruai}-sya /// shape(N)-NOM

bad-o/// come-NF banbant plenty

ruay-sya /// shape(N)-NOM raij-o-nu den zaij-o mul-o kar-o du// horse-PL-POSS on gold-PL silver-PL bring-PROG BE/PRST 'Then they started making preparations to return to his own country. Then in order to return home, (the one) earning a lot, (the ones) doing much work, they saying, having become a lot, those two with (god's) face [god's blessing], after having accumulated a lot of property, (they are) coming, (those two) with (god's) face [=blessing] (they) are bringing lot of horses, jewelry, gold and silver.'

230

Anju Saxena

4.2.4. V-PROGBE-N.PST V-PROG 5£-N.PST describes a state or an event which has relevance for the narrative. It provides an extra dimension, but it does not move the storyline ahead. It describes either non-primary events such as general background situations (example 19) or situations which are preludes to the next episode (for example, 20). V-PROG 5£-N.PST does not continue in more than one unit. (19)

08/161-162 if an za-mu beraij / food eat-NOM occasion gud jisi-mu / hand hand.wash-NOM yaqze -c-is o.woman-DiMiNU-ERG

ran-o du-gyo// give-FROG BE-D.PST

raza -s lo-ki-s// king-ERG say-PST-3H ki-s gud-is-nn / you(H)-ERG hand-hand.wash-NOM &u ksr-o du-n// why give-PROG BE-2H 'At the time of eating, for washing hands, the old woman is giving (water). The king said: "Why are (you) giving (water) for handwashing?".' (20)

12/015-016 sub fas-is / saliva fell(from.one's.hands)-NF

ti-o water-LOC go/NF yen far over.there.on.right.side lion

ti water

tuy-o drink-PROG

du-gyoll BE-N.PST

d.? sub far-u k?ak-o byz-gyvll (s)he saliva lion-ACC mouth-LOC go-N.PST 'The saliva fell, (and) went in the water, over there on the right side, the lioness was drinking water. That saliva went in the lioness' mouth.'

Discourse functions of verbs in Kinnauri

231

V-PROG B£-N.PST is not sequential in relation to the following unit. Events described with V-PROG &E-N.PST occur over time. In some cases, the temporal span overlaps with the events described in the following unit (for example, 19-20) or describes one specific event as a long-lasting event (for example, 21). There is a gradient difference. Further, the focussed participant in V-PROG 5£-N.PST and its following unit need not be the same. (21)

06/060-062 dok bi-gyy//

then

go-N.PST

yms=ko-s

rote-nu

txl-o /

after-lNST

chapatis-ACC

take.out-PROG

zo-u

dv-gya/l

eat-PROG

BE-N.PST

bi-yo

bi-yo I

gO-PROG

gO-PROG

'Then (they) went. After that, taking out the bread, (he) was eating. (And) going (and) going,...'

5.

Discussion

Language is one part of communication. All aspects of communication are influenced by its aim and the situation in which it is undertaken. When narrating, the narrator decides on what he or she is going to tell and from which perspective. The narrator also evaluates the addressee's knowledge in order to decide how much of the background information needs to be provided, in what detail and how much persuasion is needed. In Kinnauri the aim of narrating, in many instances, is simply to narrate something, to tell others the story or inform others of an event. Very often the purpose of narrating a story is to teach others (especially children) the morals of the society and/or help them understand nature. Traditional Kinnauri narratives have a well-structured form, where finite verbs occur regularly. The corpus-based analysis, presented here, suggests that finite units in Kinnauri correlate with discourse boundaries, referred to here as episode boundaries. The default finite verb system in many Kinnauri narratives has the simple narrative past tense. It seems to be the neutral linguistic device which encodes the end of a discourse unit. The narratives are presented as distant and chronologically ordered propulsive events which happened long ago.

232

Anju Saxena

One of the linguistic devices to mark digression from this default system is the use of other tense and aspect markers. The distribution of V-IMPF V-N.PST, V-PROG 5£/PRST and V-PROG 5E-N.PST, as well as other possible finite verb systems in Kinnauri narratives, encodes the change of pace in the story. It is governed by the needs of the discourse. V-PROG BE-N.PST presents situations "analytically" (Fleischman 1990) and not as bounded events (as V-N.PST or V-PERF 5£-N.PST do). The situations presented with V-PROG 5E-N.PST do not move the storyline ahead, however relevant they are to the understanding of the narrative. Together with the default tense, it presents events as background information or as a prelude to the next event within the default temporal frame of the narrative. This differs from the finite verb system V-PROG 5£/PRST, where the choice of both tense and aspect diverges from the default system. The choice of this finite verb system, too, is governed by the discourse function of narratives. It suggests that there is a change in orientation. In these narratives the progressive aspect together with the present tense occurs in situations where it is either introducing a new episode or describing a peak situation. But this does not necessarily mean that the progressive aspect together with the present tense has a foregrounding function. Rather, it is the change in pace (i.e., digression from the default system) which contributes to this interpretation. Divergence from the default finite verb system is a linguistic resource which the narrator uses in Kinnauri narratives. It has a rhetorical function. It is, however, important to point out that even though we can see some general tendencies concerning the discourse organization and the interplay of linguistic units and discourse boundaries in Kinnauri narratives, this should not automatically lead to the conclusion that there are some general "principles" of narratives. These general tendencies leave scope for individual speakers to present the same narrative in the same language in different ways, for example, by adapting a more listener-attuned approach or a more speaker-attuned approach. It is ultimately the speaker who makes a choice (consciously or unconsciously) concerning the perspective to take in telling the narrative and also on how to present the narrative in its entirety and the chain of sub-events. Notes 1.

Kinnauri is a West Himalayish language spoken in the Himachal Pradesh region in India. I would like to express my thanks and gratitude to my Kinnauri consultants the late Smt. Jwala Sukhi Negi and Santosh Negi.

Discourse functions of verbs in Kinnauri 2.

3.

4. 5.

6.

7.

8.

233

The abbreviations used in this study are as follows. 1: first person, 2: second person, 3: third person, ABL: ablative, ACC: accusative, ADV: adverbial, AGR: agreement, AUX: auxiliary, COMP: complementizer, DAT: dative, DDVflNU: diminutive, D.PST: narrative past, DISC.PRTL: discourse particle, EMP: emphasis, ERG: ergative, F: feminine, PUT: future, H/HON: honorific, IMP: imperative, IMPF: imperfective, INST: instrumental, LOC: locative, M: masculine, NEG: negative, NF: nonfinal verb, NOM: nominalizer, NONFUT: nonfuture, NP: noun phrase, PERF: perfective, PL: plural, POSS: possessive, PROG: progressive, PRST: present, PST: past, SG: singular, V: verb. It seems that the narrative past tense marker (-gy^), in modem Kinnauri has primarily discourse-related functions. My informants have never provided this inflectional ending on their own in direct elicitation. Further, -gyz does not occur inside direct speech within narratives either. The distant past tense interpretation of the inflectional ending -gyz derives from the fact that it occurs only in traditional stories, which are usually about events which happened long ago. It is plausible that -gy represents an older verb-inflectional system (or, plausibly the tense system), which has lost its referential interpretation for the most part, and that it has taken over discourse-related functions in modem Kinnauri. The more regular verb inflectional system in modern Kinnauri would then represent a secondary development (see also Saxena 1997a). Among other West Himalayish languages, in Tinani -kyal-gya also occurs as a verbal inflectional ending with third person singular subjects (Shanna 1989). The distribution of -kyal-gyä and other tense markers which are permissible with third person subjects in Tinani seems to be semantically conditioned. Unlike other tense markers, in Tinani and in Kinnauri (except in the speech of one informant (Arjun Negi) from the Kalpa county in Kinnaur) gyo does not take any inflectional ending. The verb form in the converb segment is homophonous with the perfective verb form in Kinnauri (Saxena 1997b). Kinnauri data collection was supported in part by NSF grant BNS-8711370, an Olof Gjerdman travel grant from Uppsala University, and by the Bank of Sweden Tercentenary Foundation as part of the research program Translation and interpreting as a meeting of languages and cultures. The last mentioned has also in part financed the compilation of the corpus of Kinnauri narratives. Kinnauri exhibits all four levels of converbs (i.e., levels 1, 2, 3, 4 in Johanson 1995). Here we will concentrate our attention on the level 1 converb constructions. Evidence in favor of the change in the discourse topic is the occurrence of the progressive aspect (see section 4.2 below) and the presence of the emphatic pronoun (an-sya) in the following finite unit. Comrie provides a similar description: "A narrative is by definition an account of a sequence of chronologically ordered events (real or imaginary),

234

9. 10. 11. 12.

13. 14. 15.

Anju Saxena and for a narrative to be well formed it must be possible to work out the chronological order of events from the structure of the narrative with minimal difficulty; this constraint of minimal difficulty means the easiest way to present these events is with their chronological order directly reflected in the order of presentation" (Comrie 1985: 28). There are very few instances of V-PROG BE-HON in the narratives examined. There is only one instance of the perfective aspect together with the present tense in the material examined here. The auxiliary here is a noncopula auxiliary. The following constructions usually do not take the regular finite endings in Kinnauri: Equational copula constructions often do not end with a copula, the obligative construction usually ends with a nominalized verb, and the reciprocals. Apart from these, there are also instances in Kinnauri narratives where there is no finite morphology, even though the following unit comprises linguistic elements which show that it is a separate unit. This is indicated in the examples with a ///. The progressive marker also occurs in adverbial clauses, where it indicates simultaneity (Saxena 1997b). The repeated clauses together with V-PROG 5£/PRST occur, however, as the non-initial unit within peak situations. It seems that an exchange between two participants encoded by means of direct speech, where the head verb has the form V-PROG du, also has a similar function of highlighting the importance of the contents for the narrative as a whole. More work needs to be done in order to examine the discourse functions of direct speech in Kinnauri narratives.

References Clancey, Patricia M. 1980 Referential choice in English and Japanese narrative discourse. In The Pear Stories, vol. Ill, R. O. Freedle (ed.). Norwood, NJ: Ablex. Comrie, Bernard 1985 Tense. Cambridge: Cambridge University Press. Dahl, Osten 1985 Tense and Aspect Systems. Oxford: B lackwell. Fleischman, Suzanne. 1990 Tense and Narrativity. From Medieval Performance to Modern Fiction. London: Routledge.

Discourse functions of verbs in Kinnauri

235

Forster, Keith 1977 The narrative folklore discourse in border Cuna. In Discourse Grammar. Studies in Indigenous Languages of Colombia, Panama and Ecuador, Robert E. Longacre (ed.), 1-24. SIL. University of Texas, Arlington. Jespersen, Otto 1924 The Philosophy of Grammar. London: George Allen & Unwin. Johanson, Lars 1995 On Turkic converb clauses, hi Converbs in Cross-Linguistic Perspective. Structure and Meaning of Adverbial Verb Forms - Adverbial Participles, Gerunds, Martin Haspelmath and Ekkehard König (eds.), 313-347. Berlin: Mouton. Labov, William 1972 The Transformation of Experience in Narrative Syntax. Language in the Inner City. Philadelphia: University of Pennsylvania Press. Levinsohn, Stephen H. 1992 The historical present and speech margins in Mathew. In Language in Context: Essays for Robert E. Longacre, Shin Ja. J. Hwang and William R. Merrifield (eds). SIL, University of Texas, Arlington. Longacre, Robert E. 1976a An Anatomy of Speech Notions. Lisse, Belgium: Peter de Ridder Press. 1976b Discourse Grammar. Studies in Indigenous Languages of Colombia, Panama and Ecuador, part 1, ed. by Irvine Davis. (SIL publications in linguistics and related fields 52.) Dallas: Summer Institute of Linguistics. 1983 The Grammar of Discourse. New York: Plenum Press. 1990 Introduction. In Indian Textlinguistic Sketches, Robert E. Longacre and Dorothy J. Shaler (eds.), 1-17. Occasional papers in translation and textlinguistics 4 (1-2). Dallas: Summer Institute of Linguistics. Saxena, Anju 1995 Finite verb morphology in Kinnauri. Cahiers de Linguistique - Asie Orientale 24(1): 257-282. 1997a Towards a reconstruction of the proto-West Himalayish agreement system. In Tibeto-Burman Languages of the Himalayas, David Bradley (ed.), 73-94. (Papers in Southeast Asian linguistics, no. 14. Pacific Linguistics.) 1997b Internal and External Factors in Language Change. Aspect in Tibeto-Kinnauri. Reports from Uppsala University Linguistics 32. Department of Linguistics, Uppsala University. Schiffrin, Deborah 1981 Tense variation in narrative. Language 57(1): 45-62. Sharma, D. D. 1989 Tribal Languages ofHimachal Pradesh, vol. 2. Delhi: Mittal Publications.

236

Anju Saxena

Silva-Corvalan, Carmen 1983 Tense and aspect in oral Spanish narrative. Language 59: 760-780. Slobin, Dan I. 1995 Converbs in Turkish child language: The grammaticalization of event coherence. In Converbs in Cross-Linguistic Perspective. Structure and Meaning of Adverbial Verb Forms - Adverbial Participles, Gerunds, Martin Haspelmath and Ekkehard König (eds.), 349-371. Berlin: Mouton. Turner, Nigel 1963 Syntax. A Grammar of New Testament Greek 3. Edinburgh: T. and T. Clark. Wolfson, Nessa 1979 The conversational historical present alternation. Language 55: 168-182.

Language change

Preverbal modifiers in Sunwar* Werner Winter

1. In the following, I propose to discuss a set of preverbal modifiers that show a high degree of internal variation as well as a consistent use of this variation. With the help of these modifiers, a speaker of Sunwar is able to convey important information in addition to that contained in the verbal forms themselves. Such a specification of verbal content is found in other Kiranti languages, too. What Novel Kishore Rai and I myself adduced from Bantawa in a paper that first came out, in a badly garbled form, in 1990 and was then republished in 1997, resembles to some extent what Marlene Schulze and Dora Bieri identify as onomatopoetic modifiers in Sunwar. Here, too, reduplication or (in Bantawa) triplication is found, but in Sunwar, gradation is found, if at all, in the use of vowel change (ablaut), while in Bantawa the number of repetitions of an element appears to be significant. However, a presentation and analysis of the rich "onomatopoetic" material unearthed by Schulze and Bieri in their work on a Sunwar-English-Nepali dictionary will have to be postponed for the time being. The concern of the present paper involves what Schulze and Bieri call "intensifiers". 2. In their simplest forms, these "intensifiers" are monosyllabic items characterized by simple (C-) or complex (Cr-, Cy-) onsets, a short vowel under low tone, and a coda consisting of zero, -r, -I, -w, -n (with prepalatalization), -η, or (rarely) a voiceless stop. The entities so characterized are in their simplest form most adequately identified as preverbal modifiers and not as prefixes. They do not show a stability on a par with that of normal lexical entries, but are variable both in vowel and consonant structure and hence are best described as templates which are converted into actual forms following principles of iconicity. 3. Duration of an activity is indicated by a short (low-tone) vowel of the preverbal modifier for action without significant extension in time (referred to by Schulze and Bieri by a translation 'quickly' for the respective preverbal modifier form) and a lengthened vowel (with high-falling tone) for protracted action.

240

Werner Winter

4. The degree of vigor required for an action, largely determined by properties of the patient, is reflected by the placement of the vowel and the initial consonant of the preverbal modifier in two different strength hierarchies. 5. For vowels, the hierarchy has the shape a>o>e>u>i( with ">" denoting 'stronger than'). Examples, presented here in Schulze and Bieri's practical orthography, are: ba 'laaktbo laakt-

'beat with a big stick from further away' 'beat with a small stick from nearby'

bra 'sapbro 'sap-

'roof with large slates' 'roof with normal-sized slates'

ba kyembe 'kyem-

'cut large-sized vegetables' 'cut small potatoes'

shaw phrayktshew phraykt-

'scatter a lot of chickenfeed' 'scatter a little chickenfeed'

shap shupship shup-

'keep something big for oneself 'keep something small for oneself

sor 'shepser 'shep-

'sharpen a big knife' 'sharpen a small knife'

so bruyk-t su bniyk-t

'it broke loose (big animal)' 'it broke loose (small animal)'

bre 'tooksbri 'tooks-

'spit a lot' 'spit some'

shu 'greshi 'gre-

'move oneself aside a great distance' 'move oneself aside a small distance'

6. As for consonants, one expects to find, on the strength of evidence from other languages, a hierarchy aspirated > plain stop; this is indeed amply attested. Examples are:

Preverbal modifiers in Sunwar gha yaaktga yaakt-

'shift a big object' 'shift a medium-sized object'

tho 'guptto 'gupt-

'put on a small cover' 'put on a very small cover'

pher 'gaak-t per 'gaak-t

'it tore (big hole)' 'it tore (small hole)'

dhe 'dzuuktde 'dzuukt-

'land (heavy object)' 'land (middle-sized object)'

lam ihul dzuyk-t Jcun tul dzuyk-t

'smoke appeared (of large fire)' 'smoke appeared (of small fire)'

phril yap-l pril 'yap-l

241

'it leaked a little' 'it leaked a tiny bit'

7. The equally expected hierarchy fortis > lenis is, however, inverted to lenis > fortis so that the iconicity observed up to this point appears to be no longer present. Examples are numerous: ba 'riikkpa 'riikk-

'cut off a big piece' 'cut off a small piece'

bal liirdpal lurd-

'twist a rope' 'twist a thread'

braw daangpraw daang-

'light a big light' 'light a small light'

bo 'tharpo 'thar-

'set up a long thick object' 'set up a long thin object'

bro "thudpro 'thud-

'take down a big object' 'take down a small object'

be laakkpe laakk-

'disappear (big person)' 'disappear (small person)'

bei kyimkkpel kyuukk-

'lift a heavy object' 'lift a light object'

242

Werner Winter bre khrumpre khrum-

'plant big plants' 'plant small plants'

bu 'tsopu 'tso-

'burn big pieces' 'burn small pieces'

bul 'sherdpul 'sherd-

'boil much water' 'boil a little water'

bru iulpru 'tul-

'pull out a big object' 'pull out a small object'

biyuppiyup-

'suck much' 'suck a little'

bil 'sirdpil 'sird-

'cut off a big tree close to the root' 'cut off a small tree close to the root'

bri 'khriingpri 'khriing-

'thread a thick needle' 'thread a small needle'

bril tsimpril tsim-

'squeeze out a large amount' 'squeeze out a small amount'

8. Iconicity in the fortis-lenis domain can, however, easily be recovered once one considers Sunwar b-, d-, g- to have derived from pre-Sunwar *p-, *t-, *k-, and Sunwar p-, /-, k- from pre-Sunwar *b-, *d-, *g-, which would mean that the distribution of plain fortis and lenis non-affricates in the onset of Central Kiranti (Bantawa, Puma, Chamling, Kulung) forms should be taken to reflect the system of Proto-Kiranti, a system that was affected, between Proto-Kiranti and Sunwar, by a fortis-lenis reversal, just as it can be assumed to have occurred in the prehistory of, e.g., Bahing, Khaling, and Dumi. Full iconicity then would have existed, as far as the unaffricated stops are concerned, at the pre-Sunwar level, which means that the preverbal modifiers of Sunwar are phenomena of considerable antiquity. 9. Up to this point in the present paper, only pairs of preverbal modifiers have been cited in which either a vowel or a consonant was subject to modulation. Alongside such simple pairs there are found numerous sets which show both ablaut and consonant gradation and which frequently

Preverbal modifiers in Sunwar

243

have more than just two members. A few examples of such fairly elaborate chains are given here (for more data, the Appendix should be consulted): bharpaakk-

'cut wide open'

barpaakkpharpaakkberpaakk-

'slit open (large cut)' 'slit open (medium-sized cut)' 'slit open (small cut)'

bra deptbro dept bre deptpre dept-

'plaster with much mud' 'plaster with a fair amount of mud' 'plaster with a small amount of mud' 'plaster with just a little mud'

bra tubru tubri tupri tu-

'stitch loosely 'stitch loosely 'stitch loosely 'stitch loosely

bral shayk-t pral shayk-t prel shayk-t bril shayk-l pril shayk-t

'it spilled on the floor (much grain)' 'it spilled on the floor (some grain)' 'it spilled on the floor (a little grain)' 'it spilled on the floor (a little water)' 'it spilled on the ground (very little water)'

bral hem-t pral hem-t brol hem-t brel hem-t

'it fell 'it fell 'it fell 'it fell

brel 'yap-t bril 'yap-t phril 'yap-t pril'yap-t

'it leaked badly' 'it leaked moderately' 'it leaked a little' 'it leaked a tiny bit'

dar 'greekktar 'greekkthor 'greekkthir 'greekk-

'slip (large person)' 'slip (adult)' 'slip (light person)' 'slip (small person)'

do 'khruyktto 'khruyktdu 'khruykttu 'khruykt-

'hide (large object)' 'hide (medium-sized object)' 'hide (small object)' 'hide (very small object)'

with very long stitches' with very big stitches' with big stitches' with very small stitches'

on the ground (grapefruit)' on the ground (apple)' on the ground (plum)' on the ground (nut)'

244

Werner Winter dor tsimtor tsimtur tsimtir tsim-

'milk cow or buffalo, getting much milk' 'milk cow or buffalo, getting little milk' 'milk goat, getting much milk' 'milk goat, getting little milk'

ghayaaktga yaaktkhiyaakt-

'shift a large object' 'shift a medium-sized object' 'shift a small object'

ghar khraakkkhar khraakkkhir khraakk-

'churn much buttermilk' 'churn some buttermilk' 'chum a little buttermilk'

ghoyn khraaktghuyn khraaktkhuyn khraakt-

'extract much oil' 'extract a moderate amount of oil' 'extract a little oil'

gro timkro timgre tint-

'fell a very big tree' 'fell a big tree' 'fell a small tree'

gro theentgm theentkru theent-

'move a large object around' 'move a medium-sized object around' 'move a small object around'

gyal kyal gyul kyul

'become full 'become full 'become full 'become full

briingbriingbriingbriing-

gyol 'lekyol 'legyel 'lekyel 'le-

gyol dem-t kyol dem-t gyul dem-t kyul dem-t

'transfer 'transfer sel' 'transfer 'transfer vessel'

(very large vessel)' (large vessel)' (small vessel)' (very small vessel)'

large amount of grain/water to another vessel' moderate amount of grain/water to another vessmall amount of grain/water to another vessel' very small amount of grain/water to another

'it ran together (large amount of water)' 'it ran together (fair amount of water)' ' it ran together (modest amount of water)' ' it ran together (small amount of water)'

Preverbal modifiers in Sunwar

245

10. Ablaut and consonant gradation also affect the set of affricates. Examples of paired forms are: dzer 'khaakk-

'produce a big tear'

tser 'khaakk-

'produce a small tear'

dzel 'uukk-

'split (bamboo) fairly thick'

tsel 'uukk-

'split (bamboo) very fine'

dzew 'risstsew 'riss-

'laugh heartily' 'smile'

dzek rettsek ret-

'trap (a rat)' 'trap (a small bird)'

dzo 'kutttso 'kutt-

'take a big object upstairs' 'take a small object upstairs'

dzor napttsor napt-

'warm up very much (in the sun)' 'warm up a little (in the sun)'

dzol 'hoykttsol 'hoykt-

'heat up large amount of liquid' 'heat up small amount of liquid'

dzu neekttsu neekt-

'fill big hole with soft material' 'fill small hole with soft material'

dzulpookttsulpookt-

'make big person get up' 'make small person get up'

dzir 'nattsir 'nat-

'pinch a large area' 'pinch lightly'

dziw 'rom-ti tsiw 'rom-ti

Ί got many wrinkles' Ί got few wrinkles'

246

Werner Winter

11. What is absent in the case of affricates is the consonant gradation aspirated > plain stop. The sole example in Schulze and Bieri's data seems to involve a mis-hearing of plain voiced affricate as an aspirated voiced one: dzhey 'waltsey 'wal-

'deep-fry a small quantity' 'deep-fry a very small quantity'

This state of affairs agrees very well with the fact that there are no native lexemes with initial and in Sunwar (in contrast to what is found in loanwords from Nepali) - obviously a gap that resulted in some asymmetry in the consonant system of Sunwar. (In "onomatopoetic" forms, the gaps have been filled in part - a phenomenon which can also be observed in Bantawa.) When one inspects chains with more than two members that involve initial affricates, one comes to note a somewhat surprising situation: the place of and is taken in Sunwar by sh- and s-, respectively. Examples are: dzel 'haakisei 'haakttsel 'haakt-

'uncover completely' 'uncover partly' 'uncover a little'

dzew 'haak-iu sew 'haak-tu tsew 'haak-tu

'(the sky) cleared completely' '(the sky) cleared in part' '(the sky) cleared a little'

shew hamddzew hamdsew hamdtsew hamd-

'spread out for drying a large amount of cloth' 'spread out for drying a big piece of cloth' 'spread out for drying a medium-sized piece of cloth' 'spread out for drying a small piece of cloth'

Examples of paired forms provide additional evidence: dzaw bar-t saw bar-t

'it sprouted and grew (of plants sown far apart)' 'it sprouted and grew (of plants sown close together)'

shot 'shopdzat 'shop-

'clean the entire floor' 'clean things in a room'

Preverbal modifiers in Sunwar

247

The simplest way to explain these chains is to propose that pre-Sunwar *lsh- developed into Sunwar sh- and pre-Sunwar *dzh- into Sunwar s-. That what Schulze and Bieri transcribe as sh- is indeed functionally the voiced counterpart of Sunwar s- is amply confirmed by sets and pairs of preverbal modifiers: sha 'saassa 'saasse 'saas-

'get many things ready for a journey' 'get quite a few things ready for a journey' 'get a few things ready for a journey'

show 'phiikksaw 'phiikk-

'sweep the full length of the floor' 'sweep the floor nearby'

show 'khaldsow 'khald-

'mix a large quantity' 'mix a small quantity'

shong 'himdseng 'himd-

'shake (large branches)' 'shake (medium-sized branches)'

shit tilsn til-

'remove something to a far-away location' 'remove something to a nearby location'

Without consulting informants, it is hard to determine whether the pairs show 'phiikk- : saw 'phiikk- and shu til- : su til- are complete in themselves or whether they constitute the remnants of (at least) tripartite chains containing the preverbal modifiers shaw : dzaw* : saw and shu : dzu* : su, respectively. If the preverbal modifiers shaw in shaw 'phiikk- and shot in shot 'shop- 'clean the floor' are related (the evidence for variation in the coda of preverbal modifiers is too disparate at this point to permit a systematization), then dzat 'shop- 'clean things in the room' may point to the existence (former or present) of shaw : dzaw : saw. 12. Reconstructed early Sunwar *dzh- has no counterpart in items from the core vocabulary; neither does *tsh-. Instead, we have clear indications that pre-Sunwar *ch- is reflected by Sunwar is-; examples are: Sunwar 'tsats(a), Bahing caca, Khaling 'cos, Dumi tsatsu 'grandchild' : Bantawa, Puma, Chamling, Kulung cha 'child' Sunwar tsar-, Khaling cer- : Bantawa cheys-, Puma chemma, Chamling chArs-, Kulung ches- 'urinate' Sunwar 'tsard-, Bahing cart- : Bantawa chett-, Puma chetd- 'urinate on something'

248

Werner Winter

These developments have precise parallels in the replacement of preSunwar *kh- by Sunwar k-\ cf: Sunwar kur-, Bahing kur-, Khaling kur-, Dumi k'ir-: Bantawa khuy-, Chamling khur-, Kulung khur- 'carry' Sunwar 'kamd- 'carry between teeth' (Khaling kam- 'chew'): Kulung khamm-, Limbu khamd- 'chew' Sunwar kyeekt-, Khaling 'keht- 'kill lice, nits' : Bantawa khek- 'crack (nits)'

Sunwar ph- th- kh- are found in derived (i.e., causative) verb stems and, as secondary developments, also in some noun forms (cf. Sunwar 'khiin, Bahing khyim, Khaling kam, Dumi ki:m : Bantawa, Puma, Chamling, Kulung khim 'house'). 13. To explain underlying *dzh- and *tsh- in Sunwar preverbal modifiers, it seems reasonable to introduce a time factor. Supposing that the development in the core vocabulary that led to Sunwar is- antedated the development of aspirated affricates in preverbal modifiers, different results could indeed have occurred. Such an assumption of a later origin of *dzh- and *tsh- in preverbal modifiers is made attractive by the fact that it allows us to view the question of the origin of Sunwar bh- dh- gh- in the onset of preverbal modifiers in a parallel manner. These onsets, which do not recur in the core vocabulary of Sunwar, also could be claimed to have developed from a secondary aspiration of plain stops. This secondary aspiration provided a means to achieve an even finer calibration in consonantal gradation than that available through the alternation of voiced and voiceless stops in preverbal modifiers. That the aspiration was inserted in positions that agreed with the "natural" strength hierarchy, that is, that aspirated stop preceded plain stop, may or may not be more than mere coincidence. As the partial destruction of iconicity by the voicing reversal in early Sunwar was not subjected to subsequent revision within the preverbal modifiers, this makes the case for naturalness of the distribution of aspirates vs. plain stops rather weak. 14. A contrast Sunwar sh- vs. Sunwar s- occurs in a fair number of pairs of verbal preverbal modifiers. While of course, as pointed out earlier, such pairs may be remnants of complex chains involving affricates, at least some

Preverbal modifiers in Sunwar

249

of these pairs may contain original sibilants. This suggestion is supported by the fact that a number of lexemes with onsets Sunwar sh- and Sunwar shave well-attested counterparts in other Kiranti languages with initial s-. Examples are: Sunwar 'shii: Bahing sir), Khaling sang, Dumi s'i: Bantawa sürj, Chamling surf, Kulung sirj 'wood' Sunwar 'shep-, Bahing syab-: Bantawa sept-, Puma sepd- 'sharpen' Sunwar shye, Bahing sye : Khaling so, Dumi su : Bantawa sa, Chamling sa, Kulung sa 'meat' vs.

Sunwar 'saan, Bahing saw, Khaling suk-pu, Dumi suk-li: Bantawa sum-ka, Chamling sum-ra, Kulung suk 'three' Sunwar 'sat-, Bahing sat-, Khaling 'set-, Dumi sid-: Bantawa ser-, Puma setma, Chamling set-, Kulung set- 'kill' Sunwar 'sit-t 'it bore fruit', Bahing sid-1 sit-: Dumi si:- 'bear fruit' Sunwar 'surbu, Bahing syura, Khaling siir : Bantawasunya(p), Chamling sirama, Kulung si4rwey 'bee'

Of the two sibilants, Sunwar sh- (/z/) and Sunwar 5- (/s/), the former clearly is to be aligned in preverbal modifiers with Sunwar b- d- g-, which are derivable from pre-Sunwar (Proto-Kiranti) *p- */- *k. In preverbal modifiers, Sunwar s- in sets not involving affricates may then be taken to point to pre-Sunwar *z-. If so, this would make one raise the question as to whether this *z- (> Sunwar s-) originated in preverbal modifiers as a form contrasting with pre-Sunwar *s- (> Sunwar sh- /z/) or whether it existed outside the preverbal modifiers as part of the phonemic inventory of the core. At first glance, Sunwar 'saan 'three' and Sunwar 'sat- 'kill' seem to provide strong evidence in favor of a Sunwar core phoneme s- < preSunwar *2-. This *z- might itself be of secondary origin - both 'three' and 'kill' show initial clusters in Tibetan. But Sunwar 'saan and Sunwar 'satmay owe their s- to Sunwar-internal developments. The latter may have been influenced by Sunwar 'saykt- which, with its underlying *-//- rather

250

Werner Winter

than the expected -d-/-l-, is part of the system of marked causatives derived from intransitives with a voiced onset, and hence have voiceless onset as, e.g., in the case of affricates: dzeekk- 'wear out'

:

tseekk- 'tear (trs.)'

dzir- 'drip out'

:

tsir- 'drain'

dzut- 'run out'

:

isut- 'extract'

Sunwar sit-t 'it bore fruit' may show the effects of the same secondary alignment. Sunwar 'saan 'three' may owe - apart from its secondarily developed high tone - its voiceless onset to a compensatory process accompanying the replacement of a rhyme *-Vm by one showing a (lengthened) nasalized vowel. A fairly similar development is found, e.g., in Sunwar 'khiin 'house' : Bahing khyim, Bantawa khim or Sunwar 'loan 'path' : Bahing lam, Bantawa lam. For Sunwar 'surbu 'bee' no explanation in line with the suggestions just made seems to be available. On the other hand, both Bahing syura and Khaling stir can be cited in support of a reconstruction pre-Sunwar *syurwhich, however, one would expect to have yielded a Sunwar form with an initial sh-. The data available at this point do not permit a fully satisfactory solution to the problems alluded to. One may have to settle for a somewhat weak tentative proposal: Sunwar sh- may have resulted from a merger of preSunwar *sy- and pre-Sunwar *z-. 15. What seems to unfold as a possible scenario for the development of consonantal gradation in Sunwar is the following: (a) Preverbal modifiers (which ultimately may be derivable from verb stems) were made part of a fairly simple system including only unaspirated stops which made use of consonantal strength as an indicator of the amount of effort involved in an action. (b) Along with the core system, the incipient system of preverbal modifiers was affected by the change *voiceless > voiced and *voiced > voiceless (shared by Sunwar with Bahing, Khaling, Dumi, etc.). (c) *s- was changed to /z-/, and subsequently a new /s/ was introduced to match secondary/?-1- k- is-.

Preverbal modifiers in Sunwar

251

(d) The system of preverbal modifiers was further expanded by introducing aspirates outranking their unaspirated counterparts in the strength hierarchy. These aspirates were created independent of secondarily developed aspirates in the core system. (e) The doubly marked aspirated affricates were replaced by the sibilants IzJ and /s/.

16. The system of consonant gradation in Sunwar preverbal modifiers is thus both very rich and extremely transparent. Combined with the modulation attained by making use of vowel ablaut and the iconically highly plausible recourse to vowel length, the degree of flexibility found in the Sunwar verb phrase is immensely impressive. The richness is so great that only a small portion of the potentially available forms is put to actual use. Of the twenty variants of a labial, dental, or velar onset that might be called upon to express degrees of vigor in regards to an action, only a handful attested to in the material is accessible in any concrete instance (see the data in the Appendix). Further systematic searches may fill in existing gaps - but even then there are likely to be numerous actual lacunae in an ideally possible system. Future research might lead to an understanding of the reasons as to why particular variants were chosen in preference to other possible ones. To that end, the relationship between the templates that form the skeleton of the preverbal modifiers and actual lexemes - especially verb stems - of Sunwar and of related languages will have to be studied. The analysis offered here can thus offer only preliminary results and should be considered the beginning of what promises to be a lengthy and complex investigation.

Appendix: Sunwar preverbal modifiers - inventory (September 2001) Note: The arrangement in the following list is in accordance with the Latin alphabet: bh is found before hi, gh before gi, kh before £/', etc. 'bre > 'pre thept-

'throw away (nasal mucus)'

ba 'do

'dig'

ba kroptba > be 'kyemba > bo 'laaklba > bra > pa > phe 'kyorba > pa 'lamsba > pa 'laptba > pa 'riikk-

'rearrange slates' 'cut vegetables' 'beat' 'cut up' 'carry over one's shoulder' 'hang over a beam' 'saw off

252

Werner Winter

bel > pel kyuukkber > bir 'taykt-

* 'choke on food' 'choke on food' 'cuto.s.' 'step across' 'hang over one's shoulder' 'put on s.o.'shoulder' 'finish offend (rope)' 'twist o.s.' 'twist' 'twist' 'walk with legs spread' 'bend down' 'take a step' 'disappear' * (?) 'embrace' 'tie together' 'have an eye infection' * (?) 'rot' 'hang on a swing' 'lift a cover' 'lift up' 'throw away'

bha > bho kaptbha > bo 'leengbhang > bha kyoongbhar > bar > phar > berpaakkbhaw > bhur > bur > phur krolbhe 'samdbherpanbher > her > per piisbher > pher 'khraakkbho > bo> pho > po grooktbhrang 'prefbhrek retbhu phooktbhu > bu> pit 'dookkbhup 'dombhur > phur ber-

'extinguish' 'sell' 'clear woodlands' 'cut open a round object' 'set on fire and burn to ashes' 'put in storage' 'spin (e.g., wool)' 'fart' 'chum' 'fill hole in the ground' 'tear down' * (?) 'trap' * 'husk' 'fall down' 'put on weight' 'fly'

bi grayktbi > piyitpbikyupbil > phil > pil looks-

* 'fall off 'suck' 'suck' 'turn over (intr.)'

ba> pa> bo 'kardba > pa > bo 'karsba > phe 'kyorsba > wa khraaktbak 'lamsbak 'laptbal 'peekkbal romsbal > pal 'kirdbal > pal roptbe 'harsha gaakkbe 'shumdbe > pe khraaktbe > pe laakkbe > pe mupbe > pe 'phriptbe > pe raakk- in 'miikts(i) .. be > pe raaktbel gyuukk-

bel 'htiukk-

Preverbal modifiers in Sunwar bil > pil 'sirdbil > pil > phi! (sic!) looktbir > phir hirsbir > phir letbir > pir grittbir > pir khritbir > pir siingbiw ruuktbo barbo 'durbo 'goongbo 'huptbo khrabo 'kramdbo kroptbo 'muuktbo sumbo 'thutbo> bu> pu hornbo > pho > po 'iuptbo > po 'apbo > po dziikkbo > po 'griingbo > po 'humdbo > po 'huptbo > po katbo > po 'khriingbo > po khruukkbo > po 'kromdbo > po mardbo > po mooktbo > po 'sayktbo > po 'simdbo > po 'snbo > po 'suptbo > po 'tettbo > po '(harbor > bur > pur grinbra bwaakkbra dambra 'phiindbra saangbra sheektbra 'sheld-

'cut near root; kill' * (?) 'turn over (trs.) 'turn around' * 'turn around (intr.)' * 'perforate' 'pierce' 'pierce and insert thread' 'wrap around head' 'increase' * 'come off 'turn over fire' 'feel hot' 'hit' 'add wood to fire' 'cover roof with slates' 'fold cloth' 'cover with lid' 'borrow' 'swell' * 'hit' 'throw at; shoot' * 'break' 'slide through' 'throw into fire' 'put in warm place' 'bite off a piece' 'make slide down' 'put upright (stone)' 'throw at s.o.' * Overfill' 'throw at s.o. in anger' 'kill' 'kill' 'repair leak' 'season' 'hit' 'put upright' * 'perforate' 'talk' 'build bridge' * (??) 'take down (trap)' * 'be healed' 'rinse' 'teach s.o.'

253

254

Werner Winter

bra shooksbra 'shumdbra 'tharbrayoyktbra > bre > pre 'temdbra > bre > pre 'thitbra > bro 'apbra > bro 'sapbra > bro > bre > pre deptbra > bro > bre > pre 'detbra > bro > bre > pre thaaktbra > bro > bru > pru thiinsbra > bro > pro > bre > pre deektbra > bro > pro > bre > pre 'giikkbra > bro > pro > bre tayktbra > bro > pro > pre 'durbra > bru > bri > pri 'tubra > pro 'seensbra > pro 'simdbra > pro fhepbral > pral > bre] > bril > pril shayktbral > pral > brol > brel hembra! > prel > pril tembrang 'kherbraw hookkbraw > bro 'huyktbraw > praw daangbre > bri > pri 'tooksbre > phre raaktbre > pre dzeekkbre > pre khreptbre > pre khrumbre > pre lumsbre > pre thayktbre > pre thibre > pre 'thuyndbre > pre 'tuptbre > pre wotbrel > bril > phril > pril 'yapbrel > bril > pril 'yabrew haar 'paykt- (sic!) brew 'kyamdbrew 'sild- in 'miikts(i) .. brew 'nnkk-

'cross water' 'bend down' 'put upright'

* 'find unexpectedly' 'add' 'find after search' 'throw at; shoot' 'roof with slates' 'plaster' * 'be overly soft' 'remove bark' 'fall over' 'throw at' * 'give birth (animal)' 'trample on' * 'come off 'stitch' 'wake up' 'kill' 'beat rice from ears' * (?) 'spill on ground' 'fall off 'let water run together' 'chase away' 'bark' 'scold' 'light' 'spit out' * (?) 'rot' * 'tear' 'press together' 'plant (plants)' 'get o.s. dirty' 'strike' * 'give birth' 'disturb surface of a wound' * 'hit with hand' (?) 'put somewhere' * 'leak' 'melt; dissolve' 'widen' 'erase' 'deceive s.o.' 'split bamboo'

Preverbal modifiers in Sunwar brew > briw 'sildbrew > prew 'khreektbri > pri 'khriingbril > phril 'tsardbril > pril dzut- in 'hush .. bril > pril tsimbril > pril tsutbril > pril tsut- in 'hush .. briw > brew 'sildbriw > priw tsutbro dzittbro lubro 'nooktbro sumbro thepbro > bre > pre neptbro > bre > pre thimbro > bre > pre theekibro > bre > pre tsiptbro > pre 'thibro > pro dettbro > pro 'nibro > pro 'thudbro > pro 'thurbro > pro > pre 'dimdbro > pro > pre 'diptbro > pro > pre 'suptbro > pro > pre 'taldbro > pro > pre tsetbro > pru 'dulbru > phru 'toynsbru > phru 'tuudbru > pru > pri 'liiktbru > pru ngaakkbru > pru shuitkkbni > pru siingbru > pru 'tookkbru > pru 'tulbrul ngaakkbu > bru > pni > bri > pri 'hoi- (!) bu > hu 'khrubu > pu Ίορbu > pu 'suykt- in 'mii.. bu > pu 'tookk-

255

'smear floor with mud and dung' 'rub in; smear' 'thread needle' 'urinate on sth.' 'bleed from pierced hole' 'squeeze out' 'extract (liquid)' 'draw blood' 'smear floor with mixture of mud and dung' 'extract (solids)' * 'get wet' 'have sexual intercourse' 'massage' 'cover ground' 'beat fruit from tree' * 'stick on' 'bury' 'walk over' 'crush' 'touch' 'pound ground' 'knead' 'take from hook' 'scratch off 'step on' 'press from above' 'add ingredients' 'cut off 'split wood' * 'fall down' 'come down' 'stab' 'string up' 'weep' (?) 'point at' 'pierce and insert' 'let fall' 'pull out' 'weep' 'make hole' 'hit with horns' 'boil in water' 'set on fire' 'drop'

256

Werner Winter

btt > pu 'tsobu > pa nisbu > pu nissbu biikkbul 'dzibul 'hoyktbul 'sherdbul > pul 'etbung > pung tolbur > phur groyktbut > put nisbut > put nissbwal tebwar 'kaaktbwer 'khaakk-

'burnsth.' 'sit' .'sit' 'fall' 'get tired' 'heat up' 'boil' 'fall over' 'roll up' * 'be burned' 'sit' 'sit' 'spit out' 'widen a ring' 'tear off

da > de > te 'khookkda> de> te kyeptdal > tal 'kamddar phriinddar > tar > dor > thor > thir ? > tir 'greekkdaw > taw 'kilddaw > taw 'kilsdaw > taw 'shordaw > taw soongdaw > taw 'sordaw > thaw 'botdayn giingdayn > tayn kiingde > te 'berde > te kyemsde > te hiptde > te 'naynsde > te neptde > te sede > te tiiktde > the deektde > the 'samdde > the > te 'tsemde > the > te wotdew > dyuw > tyuw preekk-

* 'stick on' 'stick on' 'carry between teeth' 'cause to fall' 'slip' * 'get blisters' 'get blisters' (?)' pour off 'become dry (river)' * (?) 'evaporate' * (?) 'bloom' 'reach out' 'drag; pull' 'stick on' 'stick on o.s.' 'stain' 'rest' * 'stick on' 'pluck off 'support' 'throw at' 'put in store' 'forge' (?) 'put somewhere' 'jump'

Preverbal modifiers in Sunwar dha > te tseekkdhak > thak 'thooktdhe > te khrumdhe > te 'theldhe > de> ie 'dzuuktdhi > thi > ti tsudho dettdho > dheproongdho > du> tu liiksdho > du> lu liiktdho > te tsu-

'cut meat' 'drive in with wooden hammer' 'plant (seeds)' 'level ground' 'land; fall down' 'jab with elbow' 'pound ground' 'build (wall)' 'put on' 'put on' 'husk'

di > //' khrotdil 'dzetdil > ihil > til 'bordil > ihil > til homdil > til 'goldil > til 'Molding > ting 'golding > ting 'kholdir > tir khrotdo biikkdo 'dzarddo 'pipldo sumdo > de > the dayl 'pado > dhe > de khruukkdo > tepeitdo > to 'dzirddo > to 'gnimdo > to 'khnimdo > to kookkdo > to muyksdo > to muyktdo > to 'nuptdo > to 'rapdo > to 'relddo > to seekkdo > to > du > tu 'khrnyktdo > to > te shedo > to > tu 'pelddok > tok 'rapdol (?) > dal > fal 'ngoynsdol > te! gyuukk-

'cut through' 'light' * (??) 'swell' 'swell' * 'fall down' 'lay down' * 'fall down' 'lay down' 'cut through' 'fall into' 'cover with blood' 'insert in' 'cover' 'stumble over' (!!) 'put upright (pole)' 'cut off (ears of grain)' 'print cloth' * 'sink in water' 'immerse' * 'peck; bite' 'clothe o.s.' 'clothe s.o.' 'dye for weaving' 'stand upright' 'put on necklace' 'pick flowers/fruit' 'hide sth.' 'hold by handle' 'remove by twisting' 'stand upright' 'climb up' 'climb rope'

257

258

Werner Winter

dol > toi > tel kyatdor > tor > tur tsirdor > tor > tur > tir tsimdu > tu 'beldu > tu 'khruynsdul > thulu(i) > tul 'bruyktdur 'piptdur 'tayktdur > tur 'shuykt-

'liftup' * (?) 'bear fruit' 'hang sth. on hook' 'hang o.s.' 'hang on hook' * 'drain' 'milk' * 'break off 'hide o.s.' 'boil (water)' 'insert in slit' 'throw' 'insert inbetween'

dzaw > dzo 'bordzaw > saw bardze khreptdze > se 'bruyktdze > se 'yuukkdzek > tsek retdzel > sei tselddzel > sei > tsel 'haaktdzel > tsel 'huukkdzel > tsel 'thuungdzel > tsel 'uukkdzer > tser 'gaakkdzer > tser 'khaakkdzew 'rissdzew 'waangdzew > gyew > sew 'khordzew > sew > tsew 'haakt-

* (??) 'rise (dough)' 'spread; grow' 'press together' 'put on (whitewash)' 'whip sideways' * (?) 'trap' 'arrange bamboo strips' 'clear (area)' 'lift cover' 'uncover' 'split bamboo' * 'tear' 'tear off 'laugh' 'sprout; grow' 'scratch' * 'clear up (sky)'

dzhap > shop 'yuukkdzhey > tsey 'wal-

'whip sideways' (?) 'deep-fry'

dzir > tsir 'natdziw > tsiw 'romddziw > tsiw 'rom- * in 'kushul. dzo 'bettdzo 'budzo proongdzo teptdzo 'tsayktdzo > so naakt-

' pinch' * (?) 'get wrinkles' 'get wrinkles' 'lift up' 'carry on one's back1 'stack up' 'dip in (rice)' 'annex land' 'weigh (trs.)'

dol > tel kyuukkdol > tel 'sitdol > toi 'heekt-

dol > toi > tel gyayns-

Preverbal modifiers in Sunwar

dzo > so > dzo 'nidzo > tso 'bradzo > tso gupdzo > tso kraaksdzo > tso kurdzo > tso kur 'padzo > tso 'kuttdzol > tsol 'hodzol > tsol hoyktdzor > sor gaykfdzor > tsor naptdzor > tsor naptdzow > dzo 'palddzu > tsu neektdzul > tsulpookkdzuw 'guukt-

mix in 'spread out mat' 'pick up' 'get stuck' 'carry' 'put in container' 'bring up' 'get warm' 'heat up' 'get angry' 'shine (sun)' 'warm up (sun)' 'knead out liquid' 'fill hole with soft material' * (?) 'get s.o. up' 'shield from above '

ga 'urdgal 'tsiiktgal > kal tsuunggar > khar 'yookkgar tsittgat raaktgaw 'shorgaw > kaw tsittgaw > khaw breektgew > kew waakt-

* 'shiver from cold' 'bathe' 'wrap up' 'divide for inheritance' 'divide' * (?) 'go bad (liquid)' (?) 'pour off 'plough' 'write' 'plough with oxen'

gha > ga > khiyaaktghar shaakkghar > khar > khir khraakkgho 'tsayktgho > go 'rooktghoyn > ghityn > khuyn khraakkghra > gra > khra 'tsookkghrak > khrak diikkghya > ghe > ge > ke 'po-

'shift sth.' * (?) 'stir rice' 'churn' 'annex land' 'cultivate' 'press out (oil)' 'close' 'stop' 'damage'

gi > khi protgi > ki roptgi > ki royktgitprotgo 'buptgo 'harsha tuunggo 'herd-

'tear off 'twist' 'rob of 'tear off 'handle hookah' 'drink without touching cup with lips' 'make detour'

259

260

Werner Winter

go 'herdgo 'hoptgopegophaptgo 'wego > gu 'baptgo > gu > kit 'baptgo > kephreekkgo > ko 'retgo > ko tapgo > ko 'umgo > ko umgol > khol > kol 'priindgol > kol legor bwamsgor bwaptgor > khor waaktgor > kor breektgor > kor 'hildgor > kor 'mugoyn > khoyn 'hildgra 'delgra 'harsha tuunggraphrimdgra > gri > khri duuktgra > gro 'telgra > khra diikkgra > khra tiikkgra > kra 'deekkgra > kra 'teekkgra > kra > kri niinggrak tiiktgrak > khrak 'deektgraw 'uukkgraw > griw olgraw > grow 'kyeekk- in 'miikts(i) gre 'detgre > kre kyeektgre > kre regrew > griw neldgri > kri 'sholgri > kri shotgriw romsgriw tol-

'tilt' 'eat' 'damage (blade)' 'bring s.o. over to one's side' 'be bent low' 'lie in ambush' 'bow' 'damage' 'scratch' 'scoop out' 'feed s.o. solid food' 'eat solid food' 'fill with liquid' 'pour from one container into another' 'scratch o.s.' 'scratch' 'plough with oxen' 'write' 'grind' 'carve' 'grind' * 'gather' 'drink without touching cup with lips' 'remove' * 'move body parts' 'gather' 'stop' 'interrupt' * 'be interrupted' 'stop s.o.' 'choke on' 'support' * 'interrupt' 'split bamboo' 'bend; roll up' Open eyes' * 'become soft' 'kill lice / nits' 'make cracking sound' 'stir porridge' (?) 'take out' 'pull out from inbetween' 'twist around' 'roll up'

Preverbal modifiers in Sunwar griw > kriw olgriw > kriw olsgro megro mettgro 'moyklgro 'thayndgro tsaakkgro 'yookkgro > gru > km 'tsookkgro > gru > km theendgro > khro dimgro > kro deektgro > kro > gre timgrow > krow shooksgru > kru daakkgru > kru neektgru > kru oonggru > kru Oyndgu > khu > ku rookkgu> ku 'bregul 'lookkgur > khur tuunggur > kur 'wegya 'bripgya 'gruuktgya khrepfgya 'naakkgyaphooktgya shaakkgya > gyi 'kratgya > gyo > kyo bwitgya > ke mapgya > kya bruptgya > kya dettgya > kya 'dogya > kya khruukkgya > kya 'pholdgya > kya 'phreendgya > kya 'phriptgya > kya 'poyktgya > kya riptgya > kya > gi > ki 'nipgyal > gyol > kyol > kyew (!) 'luuktgyal > kyal > gyul > kyul briing-

'stir porridge' 'writhe' 'vomit' 'vomit on sth.' 'pour out' lift and support (with stone)' 'bring' 'divide up pots and pans' 'close' (?) 'move around' 'fall (tree, plant)' 'throw at' 'fell' 'cross (body of water)' 'swallow' 'fill hole with hard material' 'enter' 'take inside' Open' 'crawl' 'turn over soil' 'drink' 'be bent low' 'be bond friend' 'put in ground' 'press together' 'chew' * 'husk' * 'stir vegetables' 'bite' 'tear off 'cling to' 'grab' 'pound ground' 'dig' 'put upright (stick)' * 'stir rice' 'tie together' 'tie together' 'spoil' 'bundle sticks' 'test by touching' 'pour liquid' * 'become full (container)'

261

262

Werner Winter

gyar 'retgyaw > gyo bwirdgyaw > kyaw groyktgyaw > kyaw 'retgye 'nigyer 'dayktgyo buuktgyo 'krimdgyo 'yookkgyo > kyo 'hookkgyo > kyo 'hutgyo > kyopvnkkgyo > kyo > ge thuuktgyol 'brupgyol 'moyktgyol 'phrupgyol > kyol thookkgyol > kyol > gyel > kyel legyol > kyol > gyul > kyul demgyol > kyol > kyul 'priindgyol > shol shayktgyom > kyom 'walgyow hamgyoyn > kyoyn 'diinsgyu tolgynl > kyul waykt-

' remove by scratching' * 'twist' * 'be burned' 'scratch' 'knead' 'set trap (birds)' * (?) 'gather' 'crunch' 'share money' 'remove from container' 'pick up and fill hands' 'heap up in front of 'push wood closer to fire' 'descend' 'pour out' 'move downhill' 'pour from one container into another' 'pour from one container into another' * 'run together (water)' 'fill with liquid' * (?) 'spill on ground' (?) 'deep-fry' 'spread (hands, wings)' 'stretch' 'roll up' 'water down'

haw > ho 'dzihaw brahaw duplho 'kyuptho sumho > so 'gramdhop sumhu 'prethu > phti bretthup 'domhur > bhur > phur muykthur > bhur > phur muykt- in 'mii hur bra-

npen 'bum' * 'get burned' 'spill' 'cover with lid' 'threaten to punish' 'become overcast' 'tear down' * (?) 'collapse' 'put on weight' 'blow' 'kindle a fire' 'burn'

kaw > ko shiikt-

'cling to wall'

kha > ka tseekkkhal braakk-

4

cut meat' * 'become untied'

Preverbal modifiers in Sunwar khal 'braangkhal 'braaskhal 'phraangkhal 'praakkkher tsittkhiyaakskhiyaaktkhir breektkho bwamskho bwaptkho > khupriiktkhol > khul 'tuyndkhor 'dayktkhorpankhuyn 'apkhyew 'preekkkhyi boykl-

'separate s.o.' ' separate o.s.' 'separate1 Open (bag)' 'divide' 'dislocate joint' 'move house' 'write' 'scratch o.s.' 'scratch' 'remove pot' * 'give s.o. to drink' 'set trap (box-like)' 'spin (e.g., wool)' 'throw at; shoot' 'slice off (piece of meat)' 'tie two ends together'

kor malkor mekor mettkor 'miikskor 'miiktkraw krumskraw krupkya 'kokya 'kotkyal 'kyaldkyaw dtiptkyi 'kyaakkkyol mekyol mettkyul 'hin-

'search' 'vomit' 'vomit on sth.' 'comb hair' 'comb hair' 'cuddle up' 'cuddle up' 'sow single seeds' 'sow single seeds' 'mix liquid by pouring' * 'get burned' 'move supplies' 'vomit' 'vomit on sth.' (?) 'be afraid'

la komsla koptlaw 'dzilaw 'gorlaw 'giiukklaw > lew 'khele 'dayktle 'muuktle 'raptle taakk-

cover o.s. 'cover in order to protect' 'ripen' * 'rub off * 'bend down' 'peel' 'receive with open hands' 'fold cloth' 'smash on floor' * 'receive'

263

264

Werner Winter

lew 'kyemlew 'preekklew 'thaadlew 'tuukkli khriiktli liptlo bayndlo 'brettlo 'bulo dilo 'dzalo dzaakklo gaakklogelo 'guptlo 'guuktlo 'gyaplo kookklo 'kuttlolalo 'moyktlopulo 'rollo 'sheendlo 'sheenslo 'shoonglo 'soyktlo 'tettlo tsoo-diltlo tsoonglo 'tsoyndlo 'tuptlu biikklu oong-

'cut vegetables' 'cut off (chunk of meat)' 'make fit vertically' * 'lick' * 'rape' ' scrape off thorns' 'use things' 'call s.o.' 'carry on back' 'go and get' 'eat' 'arrive here' 'walk' 'give' 'put on from above' 'shield from above' 'buy' * 'bite' 'bring up' 'go to' 'trap' 'put over fireplace for drying' * (?) 'swing' 'teach' 'learn' 'talk' 'send away' 'smack' * 'take s.o. over there' 'return sth. borrowed' 'give away' * 'smack' 'fall into' 'enter'

mal 'thamol 'hims-

'faint while resting' 'sway'

ngol 'guukkngol 'himdngol 'khwukt-

* 'bend down' 'sway; shake' 'bend down'

Preverbal modifiers in Sunwar

pet ret-

* (?) r? 'trap'

pha tsetpha > pa 'tuptpha > phe 'brolpha > phe 'kyookkpher dzeekkpher > per 'gaakkphi 'gyayktphit grayktpho 'gruyndpho phettpho > phot gruungphra letphra > phre phrooktphra > phre 'yookkphra > phro > phre 'raptphre 'brolphrek > prekphrooktphrel 'yookkphru 'ihooktphul 'dzet-

' split wood in two parts' (?) 'break to pieces' * 'hit with stick' * 'break to pieces' 'shave wood' * 'tear' * 'tear' 'catch' * 'fall off 'send outside' 'remove' 'appear' * 'turn around (intr.)' 'slap' 'share food' 'slap' * 'break to pieces' 'slap in face' 'share food' 'prick' 'light'

po 'beekkpo 'thayndpo > be > pe khroptpor 'mupot kayktpra koongpra > bro > pro 'thaptpre teptpre thamprew 'thaadprew > priw thampro > phre > pre 'buukkpro > pre 'bwophu 'papung gil-

'die' 'lift and support (with thin material)' 'pile up' 'carve' * 'feel very cold' 'look at; watch' 'pay' 'dip' 'test thickness' 'make fit vertically' 'test liquid condition' * (?) 'sprout' 'lay egg'(NB!) 'wrap'

ra 'gipra 'tha rak > ra 'etraw kiikkraw 'pumd-

'bend' 'faint while walking' 'lean over' * 'burn (food)' * 'surround'

pha 'prol-

265

266

Werner Winter

ro 'peldro phettrow > rew 'kaakt-

'remove by twisting' 'remove' 'widen a ring'

sak > sok 'saakssal ruuklsar > ser 'bruyktsaw 'komdsaw soongsaw Isopsaw > so 'goongse 'riikkse gar 'patse 'tuptsel 'himdsel 'himssep 'preekk-

'get choked' 'wrap around head' 'sprinkle' * 'heal' 'dry up (land)' * 'go back (swelling)' 'turn over fire' 'cut off 'arrange' * 'hit' 'sway; shake' 'sway' 'shave head; shave off

sha 'peekksha riptsha yoongsha > sa > so (?)> se 'saassha > shot (!!) 'rasha > she 'leengshakphrimdshal 'peekkshal 'shildshal 'tsiiktshal > shil 'shamdshal > shil 'sishal > shol dzeldshap 'brnyktshap 'dzoshap > ship shupshatprookkshat > dzat 'shapshaw 'shookkshaw 'tsiiktshaw 'yoldshaw > saw 'brashaw > saw bwarshaw > saw 'puldshaw > saw > dzo > so > tso 'brashaw > saw > so 'phiikk-

Tmish off edge (cloth)' 'bundle up grass' 'select' 'be ready to leave' 'wind up thread' 'sell' 'remove' 'finish off edge (basket)' 'clean intestines' 'wash' 'abort' 'polish metal vessel' 'arrange bamboo strips' 'splash' 'graze' (?) 'keep sth. nice for o.s.' 'run away' 'clean' 'sow' 'wash' 'carry away' 'scatter grain' 'sow' * 'come in large numbers' 'spread' 'wipe; sweep'

Preverbal modifiers in Stinwar show > shew 'phrashaw > shew 'phrayktshaw > sho 'moldshow > so 'buldshaw > so heektshaw > so 'phreendshayn 'khershelpronshel sheeklshep tsumsshep tsuptshel 'tsuynssherphaaktshew khuyktshew > dzew > sew > tsew hamdshi phatshik utshik uynsshil 'sheldsho 'geendsho khuptsho lasho mursho 'payndsho 'paynssho 'thutsho tilsho > shu 'khrettsho > shu 'shusho > so bwayktsho > so ntpsho > so wayktshol 'buynsshol 'moyktshol 'yoldshom 'shumdshong > seng 'himdshop 'khuyktshop shoptshor 'moyktshot > shut 'wardshow > sow 'khaldshu khreengshu > shi 'gre-

' scatter' 'scatter chickenfeed' * 'mix in yeast' 'mix in' 'sift' 'tie together' 'chase away' 'loosen' 'rinse' 'shrink' 'reduce in number' 'do up one's hair' 'get out of the way' 'wipe clean' 'spread out for drying' (?)'exchange' * 'recover' 'recover; be relieved' 'irrigate; disentangle' 'clean (rice)' 'gather' 'go to' * 'wash sth.' 'adorn s.o.' 'adorn o.s.' 'borrow' 'remove stones from grain' 'serve food' 'adjust a rope' 'cause animals to mate' * 'gather and take along' 'mix' 'go together' 'trap' 'carry away' 'bend to the ground' 'sway; shake' 'wipe clean' * (?) 'strain beer; clean' 'pour out' 'throw, smashing to pieces' 'mix' 'align in one direction' 'move aside (intr.)'

267

268

Werner Winter

shu > su diJshu > su tilshuy 'hildshuyn prookkshyaw > ship 'krap-

'part company' 'remove' 'grind' 'run away' 'clean grain'

so brayktso bweso 'dzomssogaso 'garso gilso 'gipso gyaynsso 'gyomdso 'gyomsso kiingso 'kimsso 'kiptso kriptso kyuukkso 'leengso 'lemdso 'miiksso 'miiktso pettso phamsso 'phardso 'pheektso 'piptso 'praakkso proongso ramdso ramsso rapso retso riikk-* in 'miikts(i) so riptso 'sapso somsso taso 'thayndso thuptso 'tsimd-

'spread out for s.o. (mat, grain)' 'separate' 'be ready' 'become insane' 'agree' 'wrap' 'bend' 'hang o.s.' 'dress s.o. up' 'dress o.s.' 'drag; pull' 'gird o.s.' 'gird s.o.' 'shut (eyes)' 'lift up' 'sell' 'comfort s.o.' 'comb hair' 'comb hair' 'cut off (ears of grain)' 'seek shelter' 'block view' 'dress s.o.' 'insert in slit' 'undo (seam)' 'knit' 'cause to get sick' 'warm o.s. all over' * (?) 'singe (meat)' * (?) 'trap' 'blink one's eyes' 'bundle material' 'roof with grass' 'be venturesome' * 'see' 'lift and support (with wood or cloth)' 'call together' 'sew a hem'

Preverbal modifiers in Sunwar soyuukk- in 'miikts(i).. so > se 'goso > sepherso > se 'rupso > se 'thaptso > su bruyktso brayktso bweso gaso gilso gyaynsso kiingso kriptso kyuukkso penso phamsso proongso ramdso ramsso rapso retso riikk-* in 'miikts(i).. so riptso somsso taso thuptsoyuukk- in 'miikts(i) .. sol 'beekksom 'riikksor hamdsor 'phrasor 'preekksor pusor > ser 'shepsor > sir dzittsot 'giikksot 'samdsot 'tayktsoyn 'khuyktsoyn hiikksuprutsuk namsuk namssul dil-

'blink one's eyes' ' start sth.' 'sew'

'understand' 'measure' * (?) 'become loose' 'spread out (appl.) (mat, grain)' 'separate' 'become insane' 'wrap' 'hang o.s.' 'drag; pull' 'shut (eyes)' 'lift up' 'cut off (ears of grain)' 'seek shelter' 'knit' 'cause to get sick' 'warm o.s. all over' * (?) 'singe (meat)' * (?) 'trap' 'blink one's eyes' 'bundle material' 'be venturesome' * ? 'see' 'call together' 'blink one's eyes' 'be sick' 'cut off 'spread out for drying' 'scatter' 'slice off (meat)' 'put over fireplace for drying' 'sharpen' * 'get wet' *'give birth (animal)' 'put in store' 'throw away' 'wipe clean' 'count (coins)' 'undress' 'smell' 'smell' 'part company'

269

270

Werner Winter

sur barsut 'soyktsuyn 'apsuyn 'yoldsuyn suyn lasuyn suyn laakk-

grow 'send away' 'throw at; shoot' 'carry away' 'go without looking left or right' 'disappear without looking back'

te 'buukk- in 'miikts(i) .. tegupte kroptte nisste pete 'pherste 'prolte 'puukk-(?) in 'miikts(i) te 'shumdte 'sute tayktte 'thite 'toynste 'tupttel 'peldtep 'yap-

'be destroyed (eyes)' 'pick up' 'cover roof with slates' 'sit' 'damage (blade)' 'be stuck together' (?) 'break to pieces' 'destroy eyes' 'bend down' 'repair leak' 'kick' 'touch' 'come down' * 'hit' 'remove by twisting' * 'leak'

thak 'garthaw 'dzithe 'buukkthe bwaakkthe dzutthe 'garthe gar 'patthe khrathe thamthe tsutthe (?) > to dziikkthek 'garthil dzutthil 'tsuykttho gar 'pattho 'guukstho 'guukitho 'kromdtho > the > te 'thookttho > to 'gupt-

agree 'ripen' * (?) 'sprout' 'talk' * 'send up sparks' 'agree' 'make arrangement' 'hit' 'test the taste' 'light a fire' * 'break' 'agree' * 'send up sparks' 'hurry up' 'arrange' 'shield o.s. from above' 'shield from above' 'throw at s.o.' 'hammer in' 'put on from above'

Preverbal modifiers in Sunwar thu bronthu pronthtil > tul dzuyk-l

'be released' 'send away' 'appear'

// mimttiphruukktir 'yapto 'dayktto 'shettok setttop 'kald-

'reflect; think' 'remember' * 'leak' 'make trap' 'put in s.o.'s hands' 'address unknown person' 'keep'

tsak 'naakktsap 'naakktse 'samdtset 'moykttsil dzuttso suukttsor 'khattsul bookktsut 'ward-

'chew' 'chew' 'put in store' 'trap' * 'send up sparks' 'mend a thatched roof * 'feel pain' 'get up' 'throw, smashing sth. to pieces'

tyot 'rol-

(?) 'swing'

\val 'khewa] 'kya-

'peeF 'mix'

ya kaptya yooktyal yalyal > yol (?) 'dayo 'brimdyo dzamyo gamyo 'gipyo graptyo 'gyamyo 'khotyo 'khrimdyo khrooktyo kiptyo 'kyamdyo 'kyumdyo pettyo 'phrimd-

'extinguish' * 'become stiff 'be embarrassed' 'be tired' * (?) 'move downwards' * 'get lost' * 'become inactive' 1 bend' 'lose appetite' * 'fade' 'follow s.o.' 'reduce fire' 'damage edge' 'put wood over fireplace' 'erase' 'give less' Overtake; pass' 'adjust (beam)'

271

272

Werner Winter

yosamyo 'soongyo yapiyol 'beekkyol bremyol greengyu but-

'losesth.' (?) 'get rid of sth.' 'be alert' 'be sick' * 'become flattened' ' fall asleep' * (??) 'fall out'

Note The paper presented here would have been impossible to write without the generous help of Marlene Schulze and Dora Bieri, who gave me full access to the data included in their forthcoming dictionary, as well as answers to questions that arose in the course of my work and kindly granted me the use of their materials, this to include the presentation of my findings at the Uppsala conference.

Directional prefixes in Kathmandu Newar David Hargreaves

1.

Introduction

The inflectional morphology of the verb in Kathmandu, Dolakha and other languages of the Newar family has generated a great deal of interest over the last ten years or so, with the discussion focusing primarily on the historical status of the Newar verb suffixes in relation to Kiranti languages, and the semantics/pragmatics of intentional action and evidentiality, the socalled "conjunct/disjunct" morphology (van Driem 1993; Genetti 1994; Hargreaves 1991, 2003). Since all of the relevant inflectional morphology is suffixal, the Newar system of verbal prefixes and preverbal collocations has been a neglected area for systematic investigation. With the exception of Hargreaves and Tamot (1985), and the brief grammatical sketches by Malla (1985) and Hale (1986), there is little in the published literature focusing on the morphological elements preceding the verb stem. This paper, devoted primarily to the system of directionals, outlines some of the parameters of investigation for this neglected area of Newar language studies.

1.1.

Proto-Tibeto-Burman *s-

Proto-Tibeto-Burman causative/intensive *s- is well attested in Newar although the alternation is no longer productive. There appear to be three categories of reflexes. The most commonly attested reflex is the alternation between plain voiced and voiceless aspirated stops. Comprehensive lists are found in Malla (1985: 99) and Shresthacharya (1981: 172). gu(t)dengya(t)ba(t)dya(t)-

'tear' (intransitive) 'lie down, recline' 'be afraid' 'to be separate' lose money'

> > > > >

khit(t) thenkhya(t)pha(t)thya(t)-

'tear' (transitive) 'lay someone down' 'frighten' 'separate' 'defeat, win'

274

DavidHargreaves

dundonga(l)-

'submerge, collapse' 'stand up' "hang'

> thun> thon> kha(l)-

'immerse something' 'stand something up' "hang something'

Significantly, the initial alternations occur with the prefixed directionals, the focus of this paper. The prefixes //'- 'back, behind' and kwo' down/below' can be prefixed to both the simple and causative roots, suggesting that directional prefixation represents a historical process subsequent to causative prefixation. li-jya(t)- 'be delayed' kwo-ji(l)- 'be decided

> li-chya(t)> kwo-chi(l)-

'delay someone' 'settle, decide'

In addition to the alternation with initial stops, Hargreaves and Tamot (1985) proposed that the non-aspirated glides alternating with aspirated glides be included as additional reflexes of the Proto-Tibeto-Burman *s-. ye(l)wa(t)-

'be pleasing to' 'be spilled'

> he(l)> hwa(t)> hwa(l)

'be allured, persuaded' 'cast dice* 'scatter, sew seeds'

Based on Benedict's (1943) analysis of medial -y- in Lepcha, Hargreaves and Tamot also tentatively suggested that some cases of Newar medial -y- be considered reflexes of *s-. For example, Modem Newar ta(t)'cut' has a counterpart in tya(t)- 'mince, chop', no(l)- 'eat' has a counterpart in nhyo (l)-/nhe(l) 'chew, gnaw', 'eat up' and kon- Open eyes' (perhaps also khon- 'see') has a counterpart in kyon-iken- 'show'. Finally, there is evidence for an earlier prefixation process for nominalization. Verbal forms with plain initials correspond to nouns with aspirated initials. kontwontultwo(l)-

'tell' 'drink' 'roll' 'shine'

kho thw5 thu -thwo ihwokola

pu(t)-

'sweep'

tu-phi

'talk, discourse, topics' 'fermented rice drink, beer' 'classifier, rolled objects/bundles' 'suffix indicating the bright half of the lunar month' 'bright half of the lunar month' (Kölverand Shresthacarya 1994: 156) 'broom'

Directional prefixes in Kathmandu Newar

2.

275

Preverbal collocations

While the alternations above suggest reflexes of Proto-Tibeto-Burman prefixation processes that are no longer productive, modern Kathmandu Newar also has an innovative and heterogeneous collection of preverbal elements. Although a full account is beyond the scope of this study, it is important to note that the preverbal elements exhibit a continuum from syntactically productive preverbal positioning of nominal and adverbial elements to high frequency collocations, compoundings, and prefixation. Although we will not focus on the syntactic patterns in this paper, focusing instead on the directional system, the discourse-functional motivations and frequency distributions that lead from preverbal positioning and collocation to compounding and prefixation are apparent in the murky boundaries between Newar discourse patterns and the lexicon. Thus, in addition to syntactically productive preverbal positioning, the Newar lexicon has a large number of collocations with synchronically identifiable preverbal elements. For example, kh5 lha(t)- 'to speak' is a high frequency collocation of the nominal kh5 'talk, discourse' plus the verb lha(t)- 'speak', Similarly, the verb pujya(t)- 'worship, perform ritual' is transparently a collocation ofpuja 'worship' andya(t) 'do'. Also included here would be verb-verb collocations, such as nyasi won- 'walk' from nyasi, a nominalized verbal root 'walk, move', plus won- 'go', functioning as a deictic motion verb. Verb-verb collocations that do not involve deictic elements include ye-cal- 'be in heat (of cows)/ feel aroused, horny' fromyefl) 'be pleasing to' plus cal- 'feel/sense'. From these patterns of collocation we can also recognize older preverbal elements. For example, modern Newar ca-hil- 'stroll around, wander' is derived from the Indie noun cakra 'wheel' and the verb hil 'rotate, change, exchange'. Finally, there is a sizeable number of lexical items with prefix-verb combinations for which the prefix meaning is opaque. Thus the simple/causative pair two-du(l)- '(for a stick) to break', two-thu(l)- 'to break (a stick)' contains the prefix two-, for which there is no distinct semantic value, although the simple/causative alternation establishes the prefix as formally distinct from the root.

276

3.

David Hargreaves

The directional system

Although the preverbal elements as a whole make up a rather heterogeneous set, there is a subset that behaves systematically. This paradigm, which I will refer to as the directional system, consists of a set of locative/directional morphemes that appear in four distinct constructions: (a) as independent stative-locative adverbs or post-positions, (b) in collocation with intransitive motion verbs, (c) in collocation with caused motion verbs, and (d) as verbal prefixes. The morphemes in this directional paradigm are: du 'in', pi Out', // 'behind', nhyo 'front', thz 'up' and kwo 'down, below'. Another marginal member of the set is de- 'upper, on top of, which takes the locative form dyone and occurs as a prefixed form, e.g., de-chu(l)- 'put on stove' (Kölver and Shresthacarya 1994: 169). It does not seem to occur in the motion verb constructions. The paradigmatic set appears as follows: Table 1. Kathmandu Newar directional system

du

'in' Out' /"' //' 'behind' nhyo 'front' kwo 'down' 'up' tho de On top'

Adverb post-position du-ne pi-ne li-u-ne nhyo-ne kwo -ne/ku-ne (tho:-ne)7 dyo-ne

Motion (intransitive) du-M: pj-M: Ii-h3: nhya: kwo-h5: tho-ha: —

Motion (caused) du-t3 pi-to li-to nhe-to kwo-to tho-to

—

Verb prefix

dupilinhya:kwothz-

de-

The constructions are illustrated below with the morpheme, du 'in, inside'.

3.1.

Locative adverb or postposition

The directional morphemes may occur as adverbial or post-positional elements with the suffix -ne as in example (1) below: (1)

khotha du-ne lat-o room inside-LOC remain-PRF.DJ '(S/he) remained in the room.'

Directional prefixes in Kathmandu Newar

3.2.

277

Intransitive motion verbs

The directional morphemes occur with the suffix -ha:, appearing as preverbal directionals with verbs of intransitive motion, typically in collocation with -won- 'go' or wo(l)- 'come'. The source of the intransitive suffix -ha: is unclear, and significantly does not occur outside of the directional construction. It will be glossed as "intransitive motion" to contrast it with the "transitive/caused" motion construction below: (2)

3.3.

manoj du-hS: Manoj inside-lNTR 'Manoj went inside.'

go-PRF.DJ

Caused motion verbs

Directional morphemes with the suffix -to occur only with caused motion verbs, in preverbal collocation with ditransitive verbs, typically chwo(t)'send', ho(l)- 'bring', or bi(l)- 'give'. Although the form -to is homophonous with the dative suffix -to, neither the semantics nor syntax support positing any relationship. As with the intransitive -ha: above, transitive/causative -to does not occur outside the directional construction. (3)

3.4.

wo-yato du-to hol-o ri 3 -DAT inside-TRANS bring-PRF.DJ 'S/he was brought inside.'

Directional prefixes

Finally, there are numerous lexicalized prefix-verb constructions. A quick tally of the verbs with the prefix du- listed in Kolver and Shresthacarya's (1994) dictionary shows roughly thirty-five forms. The other directional morphemes exhibit about the same degree of productivity and lexicalization. Some of the lexical items are semantically transparent: du-swo(l)'look in on, check up on' < swo(l)- 'look, watch', while others exhibit varying degrees of semantic extension. For example, du-ka(l) < ka(l)- 'take' is transparently 'take in', but is more accurately glossed as "accept", occur-

278

David Hargreaves

ring more typically in the context of accepting the bride into the groom's family and household, or granting someone admission into the guthi, an influential and important Newar social institution. Similarly, there is dulwa(i)- 'to fight with oneself, struggle internally' from the root verb lwa(t)'fight'. More abstract usages include the verb du-bi(t)~ 'become possessed by, transformed into deity/spirit', from the root bi(t)- 'to elapse (of time), to set (of the sun)'. The morphemes may also be prefixed to nominal forms such as du-syo: 'essence, inner part, marrow' < syo: 'marrow', or du-gha: 'internal injuries'

Ο . 00

22 ^J

•4—1

^ 5 .S o ^ oo

Z

^H

S3 a.*

J. 2 S

^00-

s «

"o

bo

"^ O O

ο, υ oo

i ll ill ·§ΐ1 ·« f -§ ill 5 z«

u. O

S 5

^

-4—»

ϊ I

u i >c A Zu

to Z ~^

^ a,^ 4 4 ? ?i? 0.

n O

^ υ « u ± oo ι 2 α o.

Ιο ω cn

"V^ iZ 'S bO

OH

"Q OO'.""

2

•Λ Ο OT

r-

O

o o -s ^ Z O

bO

ι 00

f

v 2 g, J-s;~ς u

8.

M

1-,

0» 0,

o.

00 cn

° ^

,·

0,

φ^

•Si 5

O

Σ 00

1 1 iV u-^s-° Iι-s: s to 1

ι>

rt

r·

2

» fH

1

"Do " s subordinator in Tshangla

315

U.

Ζ

ώ

•s:G,

u.

s 1

υ -

"""Ϊ*

O

a S

•s: i -s: D U C/5

s.«

ι w *; Q z O R ι C Τ £ -

·*-»

·*-* OT

C3

Π

li

εο

u.

3

Τ "«> ύ 0

ί

Σ cd 0 Λ

1 έ

β ο °° •s: 8 ϋ ί>

I έ.1 3

s

ΐ?

Ό

00

k α>

ζ^ *-· 0

C

Τ *i

•4-*

D

£

? 0

^ά

'οο ο "οβ 00

ά ο ω S Ζ

ω •4—*

8^ i u £ ·*-*

3 ^ -Si ι

Ο G 00 -Ξ

• ^-» 4

ά g

i?

„

ο-

B.

i 6. i c 4> ΙΛ

u *e H O

α.

| 2 "^ 1

α

•^

|

CJ 5

JA

W

Ζ ^ 00 '

C

„ υ 'ο . 00 ^ ο Ζ ο κι ω 1

Q

ς;

Μ

S.S

4~»

t

C/5

ο ο~^ α

^

&

l

T3

Ι

Ο 00

Q

1

>->

&,

o U G

X) tH

υ

-s; w -υ

U

Ο

§

-ί"ί

·§

t/) ε

υ

D .S

l

C \

'l

1

l

*·

bO

"cd

*2

*^H

O 3 co

CH-H

O

b cd | t/3

II tJ-4

U

T3 Arabic qabäla 'tribute' > Old French gabelle 'salt tax'. The same Semitic root > Hebrew qabbäla 'received doctrine' > 'esoteric doctrine' (English cabala, kabbalah).

348

James A. Matisoff

The following example illustrates both (1) and (2): BLIND/BLEMISH/EQUITABLE COMPENSATION FOR DAMAGE/INTERMEDIATE

Proto-West-Semitic *'-w-r 'be(come) blind' > Arabic 'awar 'blemish', 'awäriya 'damaged goods' > Old Italian ccvaria 'customs duty' > Old French avarie > Old English averay 'charge above the cost of freight' > Modem English average (Legal) 'incurrence of damage or loss of a ship or cargo at sea; equitable distribution of such a loss among concerned parties' —> (Math.) 'intermediate'.

3. The co-constituents in a compound each have an association with the meaning of the compound as a whole. See Figure 1.

EYE

WATER

I

I

EYE

FOOT

TEARS

ANKLEBONE

Lahu m£?-yt; Thai naam-taa

Malay/Indonesian mata-kaki; Lahu kht=me?-si

EYE

BAG

EYEBAGS

English bags under eyes Mandarin yon-dai BfU i§

Figure 1. Association via compounding

1.1.

Heuristic distributional classification of semantic associations

Naturally any classes we might propose are idealized, provisional, and porous; usually we lack the necessary information to decide difficult cases. A perfect classificatory scheme would imply that we had recorded all semantic associations in all the world's languages. At first I thought I would operate with a fourfold classification, or tetrachotomy - but that sounds ominously like a tracheotomy - so a five-way classification, or pentachotomy, seems preferable. See Figure 2. "Parallel but independent" associations can verge on the "universal", if they are extremely widespread, though as Greenberg (1963) has taught us, universality is a statistical notion.

Areal semantics - Is there such a thing?

349

Semantic associations

Idiosyncratic

Paralleled elsewhere

Independent

Genetic inheritance

Areal influence

Trans-areal influence

Figure 2. Types of semantic associations

1.2.

Idiosyncratic semantic associations

In favorable cases, we can show that a particular metaphorical or metonymic association can only be understood within the context of the language/culture which produced it, or has so far only been attested in a single language. Of course any claim to idiosyncrasy can be instantly demolished when somebody finds a parallel example elsewhere.4 Good candidates for idiosyncratic status include the following assortment of examples: English toast

GRILLED BREAD -> LIBATION

The association here is via the Elizabethan custom of floating bits of grilled bread in their wine on especially festive occasions. Jingpho my it ksbt^ dum 'vacillate, be indecisive' < my it 'mind' + kdbuq dum 'sound the 3 gongs for the death-dance; perform the death-dance' (done every day in the house of a recently deceased person, until his soul is sent off to the land of the ancestors). The metaphor is based on the fact that this dance involves much swaying from side to side - an exteriorization of the notion of mental vacillation. (SeeMatisoffl986:47.)

Lahu phi-qhe-lä?-nD

DOGSHIT + FINGER -> INDEX FINGER

< pht-qhe 'dogshit' + la?-no 'finger'. The association is via the size and shape of the finger. Oiang SALT + STEAL(er) -> BAT (Longxi dial.) tsht-qii-mir, (Taoping dial.) tshjssx/oJJm9JJ [tshi I tsh]5i < Proto-Tibeto-Burman *tsa 'salt'; qii I xka33 < Proto-Tibeto-Burman *r-kaw 'steal'; mil I ms33 'noun suffix1] (See Evans 1999.) Evidently the Qiang believe that bats are especially fond of salt, and will go to any lengths to get it.

350

James A. Matisoff Manange5 NOSE + WATER -> DOG 2 ΏΑ 'nose' + kyu 'water' —> 'nAkyu 'dog'. Healthy dogs have wet noses.

Kokborok6

PENIS + FRUIT -» BANANA

tha-U 'banana' < bathay 'fruit' + bahy 'penis'. Although I have so far not encountered this association elsewhere, it seems so natural that new data will probably cause it to be moved into the "parallel but independent" class (below 1.3.1.). Gnau7 basvilape

GRASSHOPPER/CRICKET -> FONTANELLE

The association is via the cricket-like jumping pulse of the fontanelle's membrane when the infant coughs or sneezes. Japanese itoguchi THREAD+MOUTH -> CLUE TO A SOLUTION < ito 'thread' + kuchi 'mouth'. The metaphorical image is the unravelling of a tangled ball of thread. Cf. expressions like shussei no itoguchi to naru 'be the first steps towards one's success in life'. Japanese tatsu no otoshigo DRAGON + LOVE-CHILD —> SEAHORSE < tatsu 'dragon', otoshigo 'child born to a nobleman's mistress'. The anomalous, non-fishlike appearance of this creature makes it look as if it were engendered by mistake, perhaps as the result of an extracurricular sexual escapade by a dragon. Japanese baa-koodo no ojiisan BARCODE + OLD GENT —> BALDING GUY This brilliant new coinage is applied to a man who attempts to hide his baldness by a bad "comb-over". Japanese kingyo no fun GOLDFISH + SHIT —» GROUPIE Another inspired coinage, referring to a servile devotee or "groupie", who trails behind his/her idol the way droppings trail behind a goldfish in its bowl. Lahu ye-SS-ma

MISTRESS OF THE HOUSE -> HOUSE-LIZARD

See Figure 3, and Matisoff (199la: 391-392, 1991b: 331).

Areal semantics - Is there such a thing?

FEMALE BODY y5 d=se-ma da? ja 'She has a nice figure.'

(FEMALE) PROPRIETARY SPIRIT i-ka?=se-ma 'water spirit

351

HOUSE LIZARD ye=se-ma CRITTER; BUG pu=se-ma 'silkworm'

FEMININE REFLEXIVE y5 5=se-ma 'she herself

GERM; MICROBE na=se-ma FEMININE AGENTIVE NOMINALIZER

FEMALE OWNER

-qä d=se-ma ya ga ga po pi se-ma 'buffalo's owner (fern.)' 'midwife ("she who must help give birth")'

LADY OF THE HOUSE

Figure 3. Female owner house lizard

1.3.

Semantic associations paralleled elsewhere

1.3.1. Parallel but independent associations Associations of this type are so "natural" to human thought processes that they have been made independently in genetically and geographically unrelated language families. A few examples: FIRE + TONGUE8 -> FLAME

This metaphor is found widely in Southeast Asian language families (e.g. in Khmer, Malagasy, Yao/Mien, Jingpho, Lahu), but also crops up elsewhere, in fact throughout the world: Swahili ndimi za moto (p.c. A. Y. Lodhi), English tongues of flame, French les flammes lechent ("flames lick"), etc. BAG + CARRIER -» TOADY/"GO-FER"

Italian portabursa (carry + bag) / Japanese kaban-mochi (bag + carry).

352

James A. Matisoff FIRE + MOUNTAIN -> VOLCANO

Chinese huöshän and Japanese kazan, both written with the same characters (A ill), with the latter borrowed from the former. But the

same formation occurs on the other side of the world, e.g. Icelandic eldßall. HORN -» CORNER/ANGLE

Both in Indo-European (Proto-Indo-European *ker-no underlies both English horn (< Proto-Germanic *hornaz) and English corner (< Old French, ult. < Latin comu), and in the East/Southeast Asian linguistic area, e.g. Chinese A [Mandarin jiao] 'horn; comer, angle'; the same character is used in Japanese both for tsuno 'horn' and kado 'comer'; Proto-Tibeto-Burman *g-rug/k 'horn; corner, angle'.

A number of these widespread associations are grounded in what one might call "botanical universale": CABBAGE + FLOWER -> CAULIFLOWER

English cauli- (» kale » cole (slaw) < Latin caulis 'stalk of plant; cabbage'; French chou 'cabbage', chou-fleur 'cauliflower'. Lahu gS-cä 'cabbage', ve/'flower', go-cä-ve?'cauliflower'. GROUND + BEAN/NUT -» PEANUT

English groundnut (chiefly British and South Atlantic U.S.) = peanut; Lahu / ?i-n3?'peanut' (< mi 'earth' + /?A"bean'). SUN + X -> SUNFLOWER9

(Indo-European) German Sonnenblume (Sonne 'sun' + Blume 'flower'); Russianpodsolnecnik (< solnce 'sun'); French toumesol (tourne 'turn' + soleil 'sun'); Yiddish zunroiz (zun 'sun' + roiz 'rose'). (Asian) Burmese ne-kra (ne 'sun' + kra 'lily'); Lahu mCi-ni-ha-pa-ku-ve? (mii-ni 'sun', ha-pa 'moon', ve?'flower'); Thai dook-thaan-tawan (dook 'flower', thaan 'withstand', tawan 'sun').

Areal semantics - Is there such a thing?

353

Similarly widespread associations may be found with verbal roots: HANG (LOOSE) -> DEPEND

English -pend is a xenomorph from Latin pendere 'hang' (cf. also suspend, perpendicular, impend, append, etc.). Other European languages have similar abstract developments from the notion of hanging, e.g. German hängen 'hang', abhängig 'dependent', Unabhängigkeit 'independence'; Russian vesat'lpovesit' 'hang', viset' 'be suspended', zaviset' 'depend', nezavisimost' 'independence'. The Lahu verb cu (< Proto-Tibeto-Burman *m-dyu-k)™ in the physical sense means 'hang loosely'; figuratively it means 'depend on (as a fish depends on water, or a Christian on Jesus)'.

CUT -4 DECIDE

Lahu

'cut/sever' ehe?

Jingpho

phrat; töij

'decide' d5-che?

d5 'think' my it 'mind'

Burmese phrat

my it phrat; my it to dt püiij-phrat

Chinese

duan iff

duan -diij; jue-duan

il Hmong

tat txiav

tat-sw-caj txiav siab

diq £ 'settle', jue tfc 'resolve' caj "breath; heart' siab 'liver'

omoikiru

omoi- 'think'

Japanese kiru Latin caedere

decidere

English

decide

incise

cit 'mind'

GET/OBTAIN HAVE TO MANAGE TO RED -> EARTH

Cf. Hebrew ?ädäm 'human being' » dam 'blood' x ?ädöm 'red'11 ?ädämä 'earth'. Semitic *g-b-I MOUNTAIN BORDER Cf. Arabic jabal'mountain' / Hebrew gvul'border'.

Areal semantics - Is there such a thing?

355

rossc» WAIT

TUil DOTATION

BAT

JULNESS ?AIN

METAKE

Figure 5. Semantic ramifications of the copula

An association EGGWHITE *-> MOON has been reported for Hmongic by Martha Ratliff (p.c., ca. 1997). WHITE -> MOON is attested elsewhere (e.g. Heb. hvana 'moon', lit. "the white one (fern.)", but the ovoid association may well be unique to Hmongic. TibetO-Burman

MIND + DIE -» SETTLED/SERENE

A candidate for unique Tibeto-Burman status is the metaphorical extension of 'die' to mean 'settled/stabilized', or even 'satisfied/serene', in the combination MIND + DIE: Lahu s t 'die', ni-qhä 'viewpoint', ni-qha s t 'come to a meeting of minds, be satisfied'; Burmese se 'die', cit 'mind', cit se-wap, cit se-sat (sat 'kill') 'be quietly settled in mind; steady in purpose and conduct'; Jingpho si 'die', fly/V'mind', my it s/'be satisfied; content to let things be as they are'; Lai Chin thi 'die', kdluq 'my heart', kalutj sthi 'my mind is made up'. Lahu st, Burmese se, Jingpho 57, Lai Chin thi are all cognate, < Proto-Tibeto-Burman *ssy.

356

James A. Matisoff

Yet further investigation reveals that this association is actually areal. In Hmong, as in many other Southeast Asian languages, the locus for psychological phenomena is deemed to be the liver (White Hmong siab), not the heart or mind, and the collocation tuag siab (DIE 4- LIVER) means 'be totally satisfied with' (Jaisser 1990: 171). On the other hand, this same combination of morphemes in Chinese means something quite different.12 Tibeto-Burman

PIG + INTESTINES -> SCREW

I was at first delighted when I ran across Jingpho we?-wu and Pa-O Karen wek-?u 'screw', since their first syllables looked like an excellent match with Lahu 3-K£/"id.', for which I then had no etymology. Could this be a precious example of the rare Proto-Tibeto-Burman rhyme *-ek> But the screw is hardly an artifact of any great antiquity, and it would be prima facie implausible that a root with such a meaning could have existed in Proto-Tibeto-Burman. The truth quickly became apparent. The modern Burmese form for 'screw', we?-?u (Written Burmese wak-?u), the obvious source from which Jingpho, Pa-O, and Lahu borrowed these words, means literally "pig-intestine". The semantic association is the squiggly corkscrew-like appearance of a pig's small intestine. This etymology is also interesting from the viewpoint of distinguishing native vs. borrowed coallofams. The usual, native words for 'pig' in Jingpho and Lahu are wä? and vä?, respectively, but the doublets borrowed from Burmese have front vowels, as in spoken Burmese. Unless a native speaker of Jingpho knows Burmese, s/he is unlikely to realize that the first syllable of we?-wu means 'pig', especially since this syllable is in the high-stopped tone, while 'pig' is low-stopped. The native Lahu speaker is even less likely to recognize the source of 5- ?, since the morpheme for 'intestine' has been completely dropped from the original Burmese compound,13 rather like the way our word camera (< Latin 'room; chamber; vaulted enclosure') is a shortening of the old compound camera obscura ("dark chamber").14 Two more semantic associations which seem peculiar to Tibeto-Burman are discussed below (2.3) in the context of cross-linguistic similarities and differences: EARTHWORM TESTICLE and BIRD + WATER -> EGG.

Areal semantics - Is there such a thing? 1. door: mouth of house 2. bark: skin/back of tree 3. knee: head of leg 4. wrist: neck of hand 5. calf: excrement/belly of leg 6. eye: fruif seed/bean of face 7. bile: bitter 8. finger: child of hand 9. boa constrictor: deer snake 10. moon: grandmother 11. ring: coyol palm hand 12. witch: owl, sleep(er) 13. cramp: (associated in some way with) deer 14. fiesta: ceremony; (big) day 15. root: hair of tree 16. twenty: man 17. lime: (stone-)ash 18. egg: stone/bone of bird 19. wife: intimately possessed 'woman' 20. porcupine: thorn-opossum, thorn-lion thorn-peccary, thorn-pig 21. cougar: red jaguar 22. anteater: honey sucker, suck-honey 23. to kiss: to suck 24. to smoke: to suck 25. branch: arm (of tree) 26. to marry: to join, to find 27. gold/silver: excrement of sun/god 28. eclipse: eat the sun/moon; the sun/ moon dies; sun/moon to rot 29. coral snake: mother of driver ant

357

30. bladder: house (of) urine 31. vein: road (of blood) 32. canine tooth: dog-tooth, snake-tooth 33. molar: grindstone (metate) 34. edge: mouth 35. thumb: mother of hand 36. mano (of metate): hand/child of metate 37. poor: orphan, widow 38. rainbow: snake, cougar, turtle, squirrel, or weasel 39. otter: water-dog, water-fox 40. cedar: god tree 41. medicine: liquor, poison 42. to cure: to suck (to smoke) 43. pataxte (non-domesticated cacao): tiger-cacao (jaguar-cacao) 44. town: water-mountain 45. soot: nose/mucus of fire 46. to write: to paint, to stripe 47. to read: to look, to count, to shout 48. alive: awake 49. son and daughter: man's are distinguished, but a single term for woman's 50. head: bottle gourd (tecomate) 51. thirst: water-die 52. need: want, be wanted 53. enter: house-enter 54. cockroach: contains the root for 'house', often compounded with 'in' or something equivalent 55. feather: fur

Figure 6. Meso-American areal semantic features (Campbell, Kaufman and Smith-Stark 1986)

1.3.3. Parallel semantic associations due to areal influence In their study (1986) of Meso-American areal features, Campbell, Kaufman, and Smith-Stark listed 55 semantic associations (compounds, metaphors, polysemies) that are characteristic of the "Meso-American linguistic area", and that in fact help to define that area in opposition to outside languages (e.g. those of the South American area). See Figure 6. Nevertheless, at least a dozen of these typically Meso-American features are found in East/Southeast Asia or other areas of the world:

358

James A. Matisoff 4. wrist: neck of hand 5. calf: belly of leg

6. eye: fruit/bean of face

7. bile fur Typically Tibeto-Burman languages have the same morpheme for 'body-hair', 'fur' (of mammal), and 'feather (of bird)', with the latter meaning usually made more specific by compounds meaning "birdhair": e.g. Lahu mu 'body-hair, fur', ηα?-ιηιι 'feather' (ga?'bird').

Areal semantics — Is there such a thing?

359

Such extra-areal parallels by no means invalidate these semantic associations as Meso-American areal features. Rather they indicate that these features must have developed independently in widely separated linguistic areas which had never been in contact. While we can hardly assume that this makes them "universal", at least they are "areal-plus", and seem to indicate thought-processes that are characteristic of large segments of humanity. Areal semantic features imply calquing or "loan translation",16 a process which can occur between related as well as non-related languages. In the past geographical contiguity was usually necessary for such calquing to occur, and it was a gradual process.17 Nowadays geographic contiguity is no longer a factor, and loan-translation can be instantaneous and global, so that the distinction between "areal" and "universal" is breaking down (see Section 5 below).

2.

Cross-linguistic similarities and differences in the semantics of compounds and collocations

2.1.

Different constituents but same overall meaning

Sometimes there is no areal or even familial unanimity in the morphemic composition of collocations that designate the same object: UVULA

This is an organ the functions of which are totally obscure, leaving the way open for great diversity in compound formation possibilities.18 The English word is from Latin uvula "little grape"; Tibeto-Burman languages typically form compounds with other bodyparts, but these range all over the anatomy, e.g.: Burmese hlya-khaq "tongue-knoll"; Tangkhul Naga kharok-dmathin "throat-liver"; Written Tibetan Ice-chug "little tongue"; Lushai day-man "palate clitoris"; Lahu ha-cuni "tongue clitoris".19 Japanese, apparently via the phenomenon that Benedict (1979) dubbed "genital flipflop", has the expression nodo-chinko "throat-penis".

360

James A. Matisoff ZOO European languages tend to romanticize the displaying of captive

animals, with expressions meaning ANIMAL + GARDEN: English zoo < zoological garden, French jardin des animaux, German Tiergarten, Dutch dierentuin, etc. The Lahu view the matter differently, expressing this with their compounds tönütösä?-th5 or 3?- 5- 3?- 5 "animal + jail". BELLOWS

Chinese and Japanese describe this object by totally different compounds, Chinese by the more matter-of-fact jengxiäng Ä iff (WIND + BOX), while Japanese has the more fanciful jabara IS (SNAKE + BELLY). PINE + APPLE -» PINE-CONE

The original meaning of English pineapple, attested since the late 14th century,20 was what is now called pine-cone, i.e. the fruit of the pine tree. French pomme de pin 'pine-cone' retains this original meaning. When the tropical fruit we now call pineapple reached Europe in the 17th century, its resemblance to a pine-cone caused the name to be transferred from the fruit of the pine to the fruit of the genus Ananassa. Other languages (e.g. French, German) took the path of adopting the botanical name (ananas) for the exotic fruit.21 PLEASURE AT ANOTHER'S PAIN

This complex concept has been lexicalized to a compound in German Schadenfreude (lit. DAMAGE + JOY), such an apt coinage that it has been borrowed by many English speakers. Chinese expresses the same idea by a four-syllable literary expression:22 xing-zai-le-huo ^JA!^||O| "rejoicedisaster-enjoy-calamity".

2.1.1. Comparative proverbs The last example leads us to the question of how folk wisdom is encoded into set expressions or proverbs. Sometimes these proverbs correspond exactly, in which case one can assume that contact has led to calquing. Thus English Don't look a gift horse in the mouth is so similar to French Un cheval donne on ne regardepas a la dent and German Einem geschenkten Gaul sieht man nicht ins Maul that calquing is a reasonable explanation. On the other hand, it is hard to determine whether the similarity between English Time flies like an arrow and Japanese Tsukihi no tatsu no wa ya no

Areal semantics - Is there such a thing?

361

gotoktt ("The passage of the days and months is like an arrow") is due to calquing or is simply based on universal human experience. More often perhaps the same proverbial message is conveyed by somewhat different images, as in the rough equivalents to be found for English A burnt child dreads the fire or Once bitten, twice shy: French

Chat echaude craint Veaufroide. scalded cat is afraid of cold water.'

Russian

Obzegsis' na moloke, budes' dut' i na vodu. 'Having been burnt on milk, you'll even blow on water.'

Swahili

Aliyeumwa na nyoka, huogopa gamba lake. One who has been bitten by a snake is afraid of snake skin.'23

Chinese

y zhäo bei she yäo, shi nian pa Jingsheng. 'One morning bitten by a snake, for ten years fear a well-rope.'

Lahu

pf ehe? jo go, -ma la ka? ko?. 'If you've once been stung by a bee, you're afraid even when a fly comes.'

Some proverbial expressions are so original that even rough equivalents are difficult to find in other languages. The vivid Chinese saying tä yinggäi tuö küziföngpl ('He has to take off his pants to fart'), applied to someone who puts himself to unnecessary effort, has no close English analogue.

2.2.

Same constituents but different overall meaning

Conversely, sometimes morphemically identical collocations may mean totally different things in different languages: MIND + DIE

We have seen (above 1.3.2.) how this combination of morphemes means 'be settled/serene' in many Asian languages. In Mandarin Chinese, however, this same combination sixw $£ ^ (si 'die'24 + 'heart') means rather 'give up on something or someone; drop an idea; have no further illusions'; when yaw 'eye' is added, the trimorphemic sixinyanr ("die in heart and eye") acquires the meaning 'be stubborn, obstinate as a mule'. None of these expressions have the unfavorable Western associations of "brain-death".

362

James A. Matisoff

DAY + EYE Old English dceges-ege 'day's eye' > Modern English daisy. With the opposite order of constituents, eye + day, the same combination means 'sun' in Indonesian/Malay mata-hari and Thai tswan (< taa 'eye' + wan 'day'). NIGHT + CRAWL

In English, a night-crawler is a kind of worm used for fishbait. The superficially similar Japanese word yobai (< yo 'night' + hau 'crawl') occurs in the expression yobai ni yuku $L ill ^ ^ < 'visit ladies at night for amorous purposes'. (A somewhat similar image of dragging oneself around for non-virtuous purposes is conveyed by English pubcrawler 'someone who visits one bar after another'.) HAND + PAPER

A famous example of the same ideographic compound with vastly different meanings in Japanese and Chinese is ^f- £K (HAND + PAPER), read tegami in Japanese, with the meaning 'letter, epistle', but pronounced shduzhl in Chinese, with the meaning 'toilet paper'. GRAB/GRASP + TESTICLE

In Japanese, the idiom hito no kintama wo nigiru (literally "grasp a person's balls")25 means 'curry favor with someone; try to ingratiate oneself with someone'. The English expression have somebody by the balls is quite different, meaning 'have someone at a decisive disadvantage'. TAKE + TWO English dilemma 'situation that requires a choice between unattractive alternatives; problem for which there is no good solution' derives from Greek di- 'twice' + lemma 'proposition; something taken' < lambanein 'take'. Similar morphemically, though vastly different semantically, is American English double-take, a show-business term meaning 'a delayed reaction to an unusual remark or circumstance, often used as a comic device'. STAR + DATE

Chinese xingqi IÜ $JJ, literally "star-date", is the ordinary Mandarin word for 'week'. Quite independently, the creators of the classic TV science fiction program Star Trek coined the English compound star-date to refer to the time-keeping system of the far future.

Areal semantics - Is there such a thing?

363

2.2.7. Incongruities in the literal translation of collocations Sometimes collocations with similar morphemic structure also mean more or less the same thing in two languages, but differ in their stylistic level, so that literal translation is highly incongruous.26 HANG + HIPS -> SIT DOWN

Japanese koshi wo kakeru 'sit down' (literally "hang one's hips") is an honorific expression much more polite than the simple verb suwaru. Yet it would be quite rude or highly jocular in English to invite someone to "hang his/her hips" ("Hang your hips right down on that chair, honey!"), or to poser sesfesses in French. NOT + STOP -»INTENSIFIER

A more recherche example of Japanese/English stylistic incongruity may be illustrated from the first sentence of Lady Murasaki's Tale of Genji (ca. A.D. 1000): ... itodo yamu goto naki kiwa ni wa aran ga ... 'although her rank was not so very distinguished', lit. "although she did not have an endless(ly high) rank" An awful English literal translation might be "it wasn't that she had a rank that wouldn't quit", rendering the Heian Japanese intensifier yamu goto naki (lit. "there is no stopping") by the slangy American superlative that won't/wouldn't quit, as in She has a body that won't quit.

2.3.

Allocomps and alternative paths of semantic association

A consideration of words for 'testicle' illustrates many of the associational categories discussed so far, ranging from universal to language-family specific to idiosyncratic. See Figure 7a. TESTICLE ROUND OBJECT ^-""^"^-^

ball/ stone/ egg/nut/jewel

ORCHID

kidney/ heel

Figure 7a.Testicular associations (general)

EARTHWORM

TESTIMONY/ WITNESS

364

James A. Matisoff

Most of the associations under ROUND OBJECT are "natural", and not localizable in any linguistic area. Less widespread are metaphorical associations with other paired organs, e.g. the kidneys?1 or pairs in general (e.g. twins).29 The connection with heels is attested in Tibeto-Burman, although it is quite indirect: Mikir ti:is glossed as 'egg, testicle; heel' (Walker 1925), via the compound EGG + FOOT —» HEEL. The concept EGG itself in many KukiChin languages has been identified as a compound of BIRD + WATER (Benedict 1939),29 e.g. Lushai ke-ar-tui 'heel' ("foot-egg": ke 'foot', ar 'bird', tui 'water'), so that we may diagram this complex of associations as in Figure 7b: BIRD

WATER

EGG

TESTICLE

FOOT

\ HEEL

Figure 7b. Testicular associations (Tibeto-Burman)

English orchid derives from Greek or%is 'testicle; orchid', apparently so called because of the shape of the roots of most species. This association seems confined to Indo-European, perhaps only to Greek. But see also English dogstones (Testiculus canis\ a name for various British species of orchids (OED).30 English avocado ultimately derives from Nahuatl ahuacatl, which meant both 'fruit of the avocado tree' and 'testicle', but with the pronunciation contaminated by Spanish speakers by analogy with Spanish avocado 'lawyer'. The association with earthworm has so far only been unearthed in Tibeto-Burman. The Proto-Tibeto-Burman root *zril means 'worm' in most Tibeto-Burman languages, including Written Tibetan sril » srin-bu, Written Burmese tit Thado til); but in Lushai, Lai, and several other Chin languages til means 'testicle'. The original meaning seems definitely to have been WORM (Thado also belongs in the Chin group, but retains the meaning 'worm'), and one can only speculate on the exact path of semantic development in this case.

Areal semantics - Is there such a thing?

365

The striking homophony between Latin testis 'testicle' and testis 'witness' has nonplussed dictionary-makers. Some propose deriving the latter term from *tri-sto- 'standing as a third (party)'. The OED, however, says "etym. uncertain". Buck (1949: 257) guesses that the morphemic identity is probably in imitation of Greek parastotal 'testicles', a word which was "wrongly associated with the legal sense of paras tates 'supporter, defender', and so with 'witness'." I personally feel that there is a clear semantic connection between the two concepts because of certain ancient Near Eastern oath-taking practices. In the Old Testament there are numerous references to swearing an oath to a man by placing a hand under his loins (Hebrew svü'ä taxat matnäv)^ By swearing on his interlocutor's most precious possessions the oath-taker signified his willingness to be exposed to the utmost wrath of the other man should the oath be broken. The question arises as to whether we need a special term to refer collectively to all the points in semantic space that are associated with the same concept in one language or another, e.g. do we ever need to refer to the relationship holding among BALL, ORCHID, WITNESS, and EARTHWORM? In the case of actual alternative compound formations where one constituent is the same but the other is different, is it useful to speak of allocompsl Just because BRAIN is sometimes expressed as HEAD + MARROW, but also as HEAD + SNOT, should we recognize a semantic connection between MARROW and SNOT? See Figure 8. BRAIN

SNOT

HEAD

MARROW

Figure 8. Ways of conceiving brain

3.

Southeast Asian lexico-semantic areal features

On the simplest level, Southeast Asian areal semantic features include the presence of words referring to aspects of the physical environment of Southeast Asia, e.g. regional flora and fauna terms.32 It is not surprising that all Amazonian languages have a word for tapir, but that no Sino-Tibetan or

366

James A. Matisoff

African language does; or that the whole world has borrowed kangaroo from an Australian aboriginal language.33 Historical linguists have taken advantage of such facts when attempting to establish the Urheimat of a language family.34 Within East and Southeast Asia, names of animal species that were domesticated at any early date (e.g. DOG, HORSE, CHICKEN),35 as well as a number of feared, spectacular, or ubiquitous species (e.g. BIRD OF PREY, GER, ELEPHANT, CROCODILE, RABBIT) have diffused widely through the area.36 Technological terms constitute another important category of areally diffused vocabulary, e.g. a root for CROSSBOW of the shape *s-na that is found in Chinese ^, Karen, Hmong-Mien, Tai, and Mon-Khmer.37 Important clues to the semantic preoccupations of a linguistic area are found in the parts of the vocabulary where fine lexical distinctions are made. While the Eskimos might not really have as many words for 'snow' as we used to think, at least it is safe to say that they must make more subtle conceptual distinctions38 in that area than Southeast Asians do. Southeast Asian languages are rich in verbs of manipulation (e.g. CARRY; CUT; PLUCK) reflecting the physical lifestyle of the region. While English also has a huge number of verbs of cutting, many of them are highly technical or not in common use (e.g. excise, amputate, julienne), whereas in Southeast Asia verbs of cutting are simple in phonological shape (mono- or sesquisyllabic) and used by everybody. In general, the Mon-Khmer languages do not make much use of compounding, making up for this by a high degree of lexical specificity.39 Morphology plays a key role in the lexicalization of complex concepts, even in the monosyllabic Sinospheric languages (Chinese, Tibeto-Burman, Tai-Kadai, Hmong-Mien, Vietnamese), which typically have hundreds of four-syllable elaborate expressions, some literary but many in common use. These often involve repeated syllables (either the 1st and 3rd, or the 2nd and 4th), with the other two syllables being synonyms, antonyms or otherwise correlative in meaning.40 Of particular interest are compounds which have been calqued throughout Southeast Asia, but which have so far not been attested elsewhere,41 e.g.: PIG + CRAZY/ILLNESS -» EPILEPSY

(Khmer, Mon, Thai, Indonesian/Malay, Burmese) FLY + SHIT -> FRECKLE/MOLE

(Khmer, Mon, Thai, Indonesian)

Areal semantics - Is there such a thing?

367

EYE + X -> ANKLEBONE: EYE + FOOT (Indonesian, Burmese, Lahu) EYE + COW (Khmer); EYE + ELEPHANT (Mon); EYE + FISH (Vietnamese) TOOTH + INSECT -» DENTAL CARIES

(Khmer, Vietnamese, White Hmong, Thai, Jingpho, Burmese, Chinese) STAR + SHIT -> METEOR

(Hmong, Lahu) The metaphorical extension of the opposition MOTHER/CHILD to AUGMENTATTVE/DIMINUTTVE is clearly to be considered a Southeast Asian areal semantic feature, though it is certainly to be found elsewhere.42 While MOTHER/FEMALE + HAND/FINGER -> THUMB is found in other linguistic areas, so far MOTHER —» LOCK vs. CHILD -> KEY has not been observed outside of Southeast Asia. Cf. the following data: Indonesian ibu 'mother' / anak 'child' ibu kota 'capital city', ibu rod 'yeast' (kota 'city', roti 'bread'); ibu djari ~ ibu tangan 'thumb', ibu kaki 'big toe' (djari 'finger', tangan 'hand', kaki 'foot'); ibu panah 'bow' / anakpanah 'arrow'; ibu kuntji 'skeleton key; lock' / anak kuntji 'key' Cham inö' tatjin 'thumb', inö' /äfc?/'big toe' (/no"mother', taqin 'hand', takai 'foot') White Hmong niam 'mother; woman' » niag 'augmentative' (-m and -g are tonemarks) Mien (Chiengrai) puad-do'q-nei 'thumb', zauj-do'q-nei 'big toe' (puad 'hand', zaf//'foot', nei 'mother') Vietnamese cat- 'mother' / con- 'child' ngon tay cai'thumb\ngon chan cai 'big toe' (ngon 'digit', tay 'hand', chan 'foot')

368

James A. Matisoff Thai mee 'mother' / JQuk 'child'43 hua-mee-myy 'thumb', hua-mee-tiin 'big toe' (hüa 'head', myy 'hand', tiin 'foot') Hlai /w/J-2/77-(Baoding dial.), mei3-zi:g2 (Zhongsha dial.) 'thumb' (pai3! mei3 'mother', zi:q2 'finger') Chinese44 (da)-mu-zhl 'thumb' ^ n Ja ( umm-al-nujum 'Milky Way' ("mother of stars"; umm-al-qiira 'Mecca' ("mother of cities"); umm-qam 'rhinoceros' ("mother of horn")

On the level of morphosyntax, there have been strikingly parallel developments of root-verbs to abstract auxiliaries that reflect massive grammatizational calquing,46 e.g.: DWELL

OBTAIN

GIVE

'progressive'

'manage to; must; able to'

'causative; benefactive'

Chinese

zai&

Thai

JUU

gei haj

Hmong (Miao)

nyob yhm

de; dei ff daj tan

Mien (Yao Samsao) Vietnamese Khmer Burmese Lahu

d

duoc

— pun cho

— ne ehe

baan

?aoy

ra

pe

ga

P'

til?

Certain collocations exemplify concepts characteristic of a broad regional worldview, what one might call the Southeast Asian sociocultural world of ideas. The Jingpho expression my it ?3wäm ~ my it ?agam 'to be deterred by feelings of respect, embarrassment, fear of offending; be generally restrained in one's interpersonal behaviour by the knowledge that selfassertiveness is not socially approved' (Maran 1979) is paralleled in many other Asian languages (e.g. Thai kreeq-caj, Burmese ?a-na, Japanese enryö sum), reflecting a mind-set more typical of the region than the more aggressive interpersonal ideal in Western competitive societies. Greeting behavior falls into the sociocultural category as well. Greetings like "Where are you going?" and "Have you eaten yet?" are common throughout the region, but formulas like "Good morning" or "God be with you" are rare. Some collocations involve metaphors reflecting the material rather than the social dimensions of life in Southeast Asia - images that spring naturally to the minds of people living closer to nature than we do:

370

James A. Matisoff Jingpho my it makün ~ my it dakun 'be resolute; make a resolution to do something' < kün 'prepare to spring or charge, as a wild beast; crouch, as a tiger ready to spring'. Jingpho my it daguk 'be boorish; have a klutz-like mentality' (Maran's gloss!) < daguk 'curve-horned' < Proto-Tibeto-Bunnan *guk « *kuk 'crooked, bent, curved'; the metaphor seems to refer to the sluggish behavior of Southeast Asian bovines, rather than to the inertness of hunks of wood (Yiddish klofs, pi. kletser).

The verbalization of psychological phenomena also has characteristic features in Southeast Asia: -

Sentence-final particles with the exclusive function of expressing emotional attitude (particularly characteristic of omnisyllabic tone languages, where intonation alone seems inadequate to the task). - Large repertoires of "expressives" that typically function as vivid adverbiale encoding visual or auditory aspects of experience. These are especially highly developed in Mon-Khmer (see e.g. Diffloth 1976), as well as in Korean and Japanese (giongo and gitaigo). - "Psycho-collocations", i.e. expressions referring to mental processes, emotional states, or qualities of character which contain a "psychonoun" that explicitly mentions the locus of the activity or state (heart, liver, gall, spirit, etc.).47 Finally one may perhaps speak of "negative areal semantic features", i.e. associations that are well exemplified elsewhere, but never seem to turn up in Southeast Asian languages. These are of course potentially unlimited in number. A couple of random examples: -

The association POOR UNFORTUNATE has not been attested in Southeast Asia, though it is common in European languages (e.g. English poor guy, French pauvre homme, Spanish pobrecifo, German armer Kerf). - The conceptual opposition between HEART and REASON (cf. Pascal's aphorism, Le coeur a ses raisons, que la raison ne connait point) seems quite foreign to East and Southeast Asia. These negative facts are important when one is evaluating an etymology that involves a semantic leap that has never been attested anywhere in the language family or linguistic area in question (see below Section 6).

Areal semantics - Is there such a thing?

4.

371

Vicissitudes of cultural prestige and the directionality of semantic influence

At the heart of diffusional phenomena are the dynamics of relative prestige among the languages in cultural and/or geographical contact. The following are brief characterizations of some of these relationships, in East/Southeast Asia and elsewhere.

4.1.

Sanskrit/Pali and East/Southeast Asian languages

Even semantically obscure terms connected with Buddhism were dutifully borrowed phonologically or calqued from Sanskrit into Tibetan and Chinese, and from the latter into Japanese: Sanskrit dhyäna > Chinese W (Mandarin chan) > Japanese zen 'Zen Buddhism'. Sanskrit Tathägata (lit. "went thus"; an appellation of the Buddha and certain Buddhist saints) > Chinese Rulai #0 3fc (> Japanese Nyorai), Tibetan de-bzin gsegs-pa ("ace. to the explanation now generally accepted: he that walks in the same ways (as his predecessors), a very frq. epithet of the Buddhas": Jäschke [1881] 1958: 565-566). A website I visited explained it as "one/those who have gone to suchness ... (having just) walked onto the scene of the indescribable essence of reality".

This is similar to grammatical terminology inappropriately calqued into Tibetan from Sanskrit. Cf. Written Tibetan mam-dbye (lit. "piece-section"), "Grammatical cases ... of which the Tibetan grammarians, from an excessive regard of the Sanskrit language and in fond imitation of its peculiarities, have also adopted seven in number" (Jäschke 1958: 314-315). Some words have filtered all the way down from Sanskrit/Pali to humble minority languages like Shan, Lahu, or Phunoi, via the great regional literary languages Mon, Burmese, Khmer, and Thai (e.g. Sanskrit/Pali > Mon > Burmese > Shan > Lahu; or Sanskrit/Pali > Khmer > Thai > Phunoi): Sanskrit äcärya 'teacher, sage1

> Written Burmese chsya > Lahu sälä; > Khmer ?aacaa(r) > Thai aacaan

372

4.2.

James A. Matisoff

Classical European languages

4.2.1. Greek > Latin Latin is full of words calqued from Greek, e.g.: hypothesis > supposition hupo- 'under' + tithenai 'to place' > sub- 'under' +ponere 'to place' metamorphose > transform meta- 'beside, beyond' + morphe 'form' > trans- 'across, beyond' + forma 'form' entomo- > insect en- 'in' + temnein 'to cut', because insect bodies are divided into segments > in- 'in' + secäre 'to cut' sympathy > compassion sun- 'with' + pathos 'emotion' > com- 'with' +pati 'to suffer' melancholy > atrabilious melon- 'black' + khole 'bile' > ätra- 'black' + mis 'bile' metaphor > transfer meta- 'beside, beyond' +pherein 'carry' > trans- 'across, beyond' +ferre 'carry'

4.2.2. Greek/Latin > German, Russian, Celtic Similarly, the classical languages have furnished models for the calquing of abstract vocabulary in modem European languages: -

WITH + FEEL: sympathy I compassion > German Mitleid, Russian socuvstvie IN/OUT + PRESS: Eindruck I impression : Ausdruck I expression FROM + HANG: depend > German abhängen, Russian zaviset' (see above 1.3.1.) Names of the grammatical cases in Russian, e.g. vinitel'nyj padez 'accusative case' (cf. vinovat 'guilty', izvinit' 'exculpate, excuse') Irish magairle 'testicle' magairlin Orchid'; caique < Greek? (see above 2.3.)

Areal semantics - Is there such a thing?

4.3.

373

Chinese and other East Asian languages

Prehistoric Cisyangtzeana (China south of the Yangtze) was a "mothersoup" (Ballard 1984); a "corridor of nationalities" (Sun Hongkai 1990, speaking of the Gansu/Tibet/Sichuan/Yunnan border area), where Han Chinese were not a majority until the Han Dynasty. Loanwords and caiques must have gone in all directions (see Benedict 1975). Later the overwhelming cultural prestige of Chinese ensured that the flow of linguistic influence went from Chinese to the other languages of the area, especially Vietnamese, Korean, and Japanese. In the case of Japanese, this directionality changed decisively in the 19th century. Chinese had been exporting its morphemes (characters) to Japan since around the 7th c. A.D.; but in the 19th century Meiji Japan was influenced by Western culture and technology well before Qing Dynasty China, and combined these morphemes into new combinations, many of which were subsequently adopted into Chinese. This oscillation of mutual influence may be diagrammed as Chinese > Japanese > Chinese. 4.4.

French and English

Ever since the Norman Conquest (A.D. 1066), French has had an enormous lexical influence on English, but beginning in the mid-20th century the tide has been turning, and by now the reversal of influence has reached truly terrifying proportions.48 Desperate attempts are being made to popularize new "pure French" coinages in technological vocabulary like computer jargon, e.g. logiciel 'software'; ordinateur 'computer'; courner electronique > courriel'e-mail'\pourriel 'junkmail' (< pourrir 'to rot'); fureteur 'browser'; escargotique 'snailmail';49 bavardoir (< bavarder 'to gossip'), or clavardoir (< clavier 'keyboard') 'chatroom'. Some of these defensive coinages are considered ridiculous by the French themselves, e.g. la vacancelle (instead of le weekend). Some recent loans from English refer to unpleasant aspects of hurried modern life to which the French seem to prefer giving an Anglo-Saxon twist: le stress, le crash (of a plane). Others have been adopted because of their pop culture glamour, e.g. a poster advertising golf lessons seen in Paris Metro stations (fall 2001) inviting people to swinguer comme Tiger Woods. Newspapers routinely use caiques from American English to express modern geopolitical concepts, e.g. une re-

374

James A. Matisoff

ponse ciblee < English targeted response; cibler (< la cible 'target') did not use to be a verb in French. Many former faux amis ("false friends", i.e. English words originally borrowed from French but having slightly different meanings),50 are now converging under English influence: -

le trafic used to mean only 'illegal commerce' (as in le trqfic des drogues) - now it is replacing la circulation in the sense of 'vehicular traffic'; — formel used to mean only 'strict; categorical', as in une interdiction formelle 'strict prohibition'; now it has come to mean 'formal' in the sense of'ceremonious'; - le papier can now be used in the sense of a scholarly communication (le papier que j'ai presente a la derniere sino-tibetaine), a usage formerly frowned upon.

But perhaps the situation is not as hopeless as it seems, since English is still borrowing the occasional lexical item from French (e.g. disco; microfiche; tampon; crampon), and even caiques a collocation once in a while, e.g. this (having been) said< ceci dit.5] We may diagram this historical dynamic as French > English > French.

4.5.

Spanish and English

Latin American Spanish has been calquing American English vocabulary for almost 200 years (e.g. perro caliente < hot dog), and there are areas of the U.S. where a virtual Spanish/English Creole is developing, e.g. the "Tex-Mex" tongue with which President Bush is so familiar. The interweaving of the two languages will only become more intimate as the "Hispanic" population of the U.S. increases.

4.6.

German and English

German has traditionally preferred calquing over borrowing when coining technical vocabulary (e.g. Fernsprecher 'telephone',52 Fernseher 'television', Sauerstoff Oxygen', Wasserstoff'hydrogen'*), but has recently shown much greater receptivity toward loans from English, both in technical and everyday vocabulary.

Areal semantics — Is there such a thing? 4.7.

375

Icelandic

Languages differ greatly in their receptivity to foreign influence, i.e. they have widely varying "acquisition tolerance quotients",53 with English and Japanese ranking near the top, and a language like Icelandic near the bottom. Geographically isolated from the rest of Europe, and with a small homogeneous population and an unbroken literary tradition of over 1000 years, Icelandic has resisted both phonetic borrowings and caiques, and has successfully been purged of loans from the previous occupying power, Denmark. The whole country debates the suitability of new additions to the vocabulary. Among recent distinctive coinages are: skjar tölva simi bota myndband eyoni

'TV screen' (originally 'sheep's placenta', once used by farmers for windowpanes 'computer' < tola [pi. tölitr] 'number' + völva 'an ancient prophetess' 'telephone' < an old word for 'thread'54 'jet' RAILROAD

French chemin defer > Vietnamese difctng sat. Japanese tetsudoo Hie ill (> Chinese, e.g. Mandarin tiedao) is probably from German Eisenbahn (cf. also Swedish järnväg, Russian zeleznaja doroga).

376

James A. Matisoff FIRE + CAR -> TRAIN

Mandarin huöche yk.$; Bunnese mi-yatha; Lahu a-mT-lz, Indonesian kereta api; Thai rotfaj, Vietnamese xe Ida. These latter compounds may actually be indigenous Asian creations, since the compound "fire-car" does not seem to exist in European languages.

5.1.

English and Chinese

The enormous size of the Chinese lexicon, along with the sporting possibilities provided by its logographic writing system, has made possible a wide variety of creative meldmgs of caiques, phonetic borrowings, and puns in the adaptation of English vocabulary to the strict monosyllabic structure of Chinese:55

5.7.7. Complete phonetic borrowings with stab at the semantics

m & 3P

ni

m mm

julebü

keköukele

mätini

baituöqüqi

'club'

'coca-cola'

'Martini' (Y.R. Chao)

'butter cookie'

"horse-kick-you"

"white-removedcookie"

"complete"can-tasteenjoy-place" can-enjoy"

fä

0 aa iiti ^

5.7.2. Compounds with one real Chinese constituent and one straight phonetic borrowing 'apple pie'

m^mpingguöpäi(pingguö 'apple';/?ai < English)

'bar (for alcohol)' M HE jiübä (jiü 'liquor'; bä < English) 'mango'

^ mangguo (mang- < English or Thai?); guo 'fruit'

Areal semantics - Is there such a thing?

377

5.1.3. Compounds -with one calquedand one real Chinese constituent which clarifies the semantic category 'bridge' u) W qiaopai The etymology of the card game bridge is actually quite obscure. Random House suggests it is "apparently an alternation of earlier biritch < ?". It is called. "Etym. uncert." in Shorter OED Vol. I: 221 (1964), but an earlier (1933) edition of the unabridged OED (Vol. I: 1098) implies it is related to a verb referring to a card-cheating practice.56 But Chinese does not hesitate to identify it by folk-etymology with the usual word for 'bridge'. After all, it makes equal lack of sense in both languages! 'cocktail' ü M Mß-wei-jiü Here the first constituent is already a caique of 'cock + tail' (actually "fowl-tail"), buty/ 'liquor' is added to make the semantic category clear. (By coincidence one of the slang Chinese words for 'penis' is jl-ba ÜG, where the first element is identified with 'fowl', reminding one of the slang sense of English cock.)

5.1.4. Collocations with both constituents calqued 'make love' ffc *Jt zuo äi A recent coinage, characterized by Mary Erbaugh (p.c.) as "the ugliest loan translation I've ever heard in my life!"

5.7.5. Graphic puns 'ping-pong' jr- -^ An incredibly clever coinage based on the phonologically and graphically similar word A bmg 'soldier'.

5.1.6. Senseless borrowing The following is an exact transcription of the writing on a T-shirt worn by a little boy in Taiwan (observed December 1995), under a picture of two cuddly bears:

378

James A. Matisoff Bear Spirit Club

sunny day is SEED TIME for HOURSEBEA Tra ra ra sings jen Found HORSEBEANS

and in the

Anything at all written in Roman letters looks glamorous and exciting to a six-year-old who doesn't understand a single syllable.

5.1.7. Chinese influence on English In the opposite direction, we may note a few recent caiques from Chinese into English: χϊηάο ί&ΙΙι > brainwash', zhih &K I/u > paper tiger. A somewhat earlier caique is Long time no see, probably < ^pAv^M, h ojlu mei ji n. A handful of phonological borrowings have been around for quite a while (catsup/ketchup, tycoon, typhoon, kowtow, lichee, gung ho). It is a safe assumption that the influence of Chinese on English will greatly increase in the course of the next couple of centuries. 5.2.

English and Japanese

The influence of Western languages, especially English, on Japanese has been growing since the mid-19th century. The earliest wave of influence was provided by the "returned students" who brought back modern technical concepts from their studies abroad, e.g. from foreign medical schools. Thus, the names of the bones of the middle ear: Latin/ Technical English

Everyday English

Japanese

malleus stapes incus

hammer stirrup anvil

tsuchi-kotsu (tsuchi 'hammer'; kotsu 'bone') abumi-hone (abumi 'stirrup'; hone 'bone') kinuta-hone (kinuta 'fulling block')

The first two of these are direct caiques, but the Japanese equivalent of anvil is a new creation. Instead of using kanatoko 'anvil' (lit. "metal-bed") for the bone of the ear, Japanese uses kinuta 'fulling block' < kinu 'silk' + uta- 'to pound'), a wooden block on which silk cloth was spread and

Areal semantics - Is there such a thing?

379

pounded in order to straighten the fibers and increase their gloss. Perhaps this metaphor was preferred in order to avoid the jarring associations of metal banging in one's ears. DIABETES Similarly creative is the Japanese coinage for 'diabetes', loonyoobyoo $tl$c$j ("sugar-urine-disease"),57 now also used in Chinese (Mandarin tangniaobing), a compound that is much more transparent than the English word, which has an obscure Greek etymology: < Greek diabainein < dia- 'across, apart' + bainein 'walk, stand' Si diabeinein 'stride or stand with legs asunder', hence diabetes One that straddles; a compass, siphon'. The latter sense led to a referential shift to this disease that involves the discharge of excessive amounts of urine. FONTANELLE

Another apparent caique on Western bodypart terminology is Japanese senmon I Mandarin quanmen & PI (lit. "fountain-gate") 'fontanel(le)' < English < French 'little fountain'.

Since Japanese has by now borrowed virtually the entire lexicon of English, it need not confine itself to calquing pre-existing English collocations,58 but can instead freely amalgamate English morphemes into new combinations, truncating many of them in the process,59 e.g.: aruhara

< alcohol harassment, i.e. 'peer pressure to drink'

baajin roodo

< virgin road, i.e. 'the aisle down which the bride marches during her wedding'

baikingu

< Viking, i.e. 'a buffet' (cf. smorgasbord)

chaifii

< Chinese food (there was already a perfectly good Sino-Japanese word chuukaryoori)

chiekaratcho

< check it out, Joe (a casual greeting in koo-gyaru-go 'high-school girl talk')

heaa nuudo

< hair nude, i.e. 'a photo that shows pubic hair'

natsumero

< Japanese natsukashii 'nostalgic' + English mero[jii] 'melody', i.e. 'a golden oldie'

380

James A. Matisoff pokeberu

DAISY, below 2.2. The sunlike appearance of the flower's center is usually the basis for the association; in some languages (e.g. French, Thai) this appearance seems to have suggested that the flower is heliotropic. See Matisoff (1989). Sino-Tibetan seems to have the BLOOD RED association (cf. Written Tibetan khrag 'blood'/Chinese Ufa Old Chinese Fjak 'red'), but not the others. See below 2.2, "Same constituents but different overall meaning". The Lahu cognate to Written Burmese ?u 'intestine' is γύ (usually in the compound a-yu-fe/). There is a difference in detail between the two cases, however: the deleted 'intestine' is the head of the compound "pig-intestine", but the deleted obscura is the modifier in the collocation "dark-chamber". Interestingly the same semantic association is to be found with another Tibeto-Burman root for 'intestine', *ri:l (> e.g. Lushai riil 'intestine', Darang Deng ha:-ri 'screw'). Cf. the vulgar English slang expression suck face. Caique is a loanword (< French) and loanword is a caique (< German Lehnwort). Cf. the interesting studies of areal features in the African-Semitic-IndoIranian-Turkic culture area by A. Y. Lodhi (2000,2001). See Matisoff (1978: 67-68). This latter formation also apparently occurs in Carrier, an Amerindian language of British Columbia (p.c. William Poser, April 2002). See the Compact Edition of the OED (1973: 879). While we are on this subject, the word pineapple is used in U.S. slang for hand grenade (usually called grenade for short). The word grenade itself is from the French word for 'pomegranate'. See below Section 3. Another Swahili proverb (of Indian origin) in this vein is more similar to the Russian version: Aliyeungua maziwa, hupuliza mtindi One who has burnt his tongue when drinking hot milk, blows air on sour milk'. Data courtesy of A. Y. Lodhi.

Areal semantics - Is there such a thing? 24. 25. 26.

27.

28. 29.

30. 31.

32. 33.

34.

35. 36.

37. 38. 39.

387

This word is cognate to the forms for 'die' in the Tibeto-Burman languages cited above. The familiar Japanese word for 'testicle' means literally "golden ball" (kin 'gold', tama 'ball, sphere'); see below 2.3. As a cynical Frenchman (perhaps Sacha Guitry) once put it: "Les traductions sont comme les femmes - lorsqu'elles sont belles, elles ne sont pas fideles; lorsqu'elles sont fideles, elles ne sont pas belles." Cf. dialectal Latin nefrones » nebrundines » nefrundines 'kidneys; testes', as well as Chinese H (Mandarin shen), glossed 'testicles; kidneys' in Karlgren(1923: 130). See Matisoff (1978: 221-222, 281). Cf. Greek didumoi 'twins; testicles' More recent data have shown that the Chin roots for 'water' differ in tone and/or vowel from the element in the compounds for 'egg', though it is not yet clear whether separate roots are involved or merely allofams of the same etymon. Unless otherwise specified, references to the Oxford English Dictionary are to the 1973 printing of the 2-vol. Compact Edition. See Genesis 24:2, Genesis 47:29, Ezekiel 17:18, etc. Whatever the etymological truth may be, classically trained French schoolboys often make the association explicit, referring to their testicles as mes temoins (Buck 1949: 257). Many of the items on the Meso-American list (Fig. 5) are in this category, e.g. #'s 9,11,20,21,22, 29, 39,43. The 200-word CALMSEA list (Culturally Appropriate Lexicostatistical Model for Southeast Asia) offered in Matisoff (1978: 283-296) was an attempt to rectify some of the more glaring defects of the standard Swadesh list by eliminating items like SNOW and substituting appropriate areal concepts like MONKEY, BANANA, BAMBOO. Caution must be exercised even here, however, since the referent of a flora/fauna term may change when speakers migrate to another area. The phytonym HEMLOCK known to Socrates (French cigue) was later applied to a totally unrelated New World evergreen (A.-G. Haudricourt's example). See also Friedrich (1970). There are even a few zoonyms which apparently reflect Indo-European contact with East Asia (HONEY/MEAD; DOG; HORSE). In several cases it can be demonstrated that Mon-Khmer is the source of animal-name loans into Tibeto-Burman. See Shorto (1973); Matisoff (1973). The best study of technological and cultural Wanderwörter in Southeast Asia remains Benedict (1975). The same is undoubtedly true of Scandinavians. See Hoeg (1993). Diffloth (p.c.) reports an Aslian language with a monosyllabic verb that means 'to stack up flat round objects (e.g. pancakes or dried cowflaps)'.

388

James A. Matisoff

40.

See, e.g. Matisoff (1986). Sometimes a morpheme which has become obsolete in other contexts is preserved in an elaborate expression, where its meaning can be deduced from that of its "couplet-partner".

4L

See Purtle (1971); Matisoff (1978: 70-71).

42. 43.

See Matisoff (199Ib). There are interesting similarities and differences between Vietnamese and Thai in the usage of these morphemes in particular compounds. Sometimes they correspond exactly (e.g. 'top (spinning toy)' Vietnamese con quay I Thai luuk-khaaij (luuk is also used in compounds referring to small round objects), but sometimes they do not (e.g. 'jack (lifting tool)' Vietnamese con do// Thai mee-reen ("mother of strength"). Two Chinese morphemes for 'child', 5iL and "?, appear as noun-suffixes in Mandarin, sometimes with diminutive force. The latter morpheme is used neologistically in the names of subatomic particles. Cf. Hmong niam-ntawv'alphabet' (niam 'mother', ntawv'writing'). See Matisoff (1991a: 414-440). See Viberg (2000, 2002) for comparative studies of the grammaticalization of GO and GET in English and Swedish. English has something similar, at least for qualities of character, which can be expressed by adjectives plus a suffixed form of a psycho-noun (e.g. good-hearted, bad-tempered, ill-humored, narrow-minded). But mental operations are not expressed this way, and emotions only rarely so (e.g. heartsick). See Matisoff (1986). The most frenetic resistance to this trend has been hi Francophone Canada, where French speakers fear being drowned in a sea of English. Resistance has proved futile, however, and calquing from English to Quebecois has been proceeding apace. See, e.g. Heller (1979). This of course is still a caique from English. A couple of years ago I had independently suggested escargourrier. The culturally important snail makes its reappearance hi the French word for the symbol "@" used in email addresses (le petit escargot), for which no universal term seems yet to have evolved. It is called Strudel in Israel, again a tribute to a beloved food, but in Japanese it is called simply atto-maaku ("at-mark"), a phonetic loan. By the way, computer terminology has diverged quite a bit between Mainland and Taiwan Mandarin, e. g. the word for computer itself: Taiwan diann o ί& U^ ("electric brain"), Mainland jisuanjJ πί"$ϊίΠ. ("calculating machine"). This is either because the sense shifted in French itself while the older meaning was retained in English, or because of later semantic developments within English itself. The directionality of some older caiques is far from clear, e.g. spitting image (as in He's the spitting image of Pierre I C'est du Pierre tout crache). According to Roland Bielmeier, this word was propagated by the Nazis, though it never really caught on as a substitute for Telefon.

44. 45. 46. 47.

48.

49.

50. 51. 52.

Areal semantics — Is there such a thing?

389

53.

This concept was developed at length in Matisoff (1978), although it is there labelled "replacement tolerant quotient" or RTQ.

54.

A. Y. Lodhi notes the Swahili terms simu 'telephone', barua ya simu 'telegram' < Persian sim 'metal wire, thread' (cognate to the Icelandic word?).

55.

For an entertaining account of the many outside influences acting on the creation of Taiwan Chinese vocabulary, see Hansell (1990). OED offers this quotation in support: "By slightly bending a card - termed bridging - he could force, as it were, his opponent in the game to 'cut' the cards wherever he wished (1879)". It is probable, however, that this coinage was influenced by German Zuckerkrankheit or Dutch suikerziekte. Although such caiques still occur in abundance, e.g. osoi sutaato 'a slow start (as in a baseball season)' < Japanese osoi 'slow' + English. While most of these sound quite bizarre or even incomprehensible to a native English speaker, it must be admitted that even some compounds invented by native English speakers seem quite strange on first hearing, e.g. trophy wife and arm candy (roughly synonymous). A few Japanese inventions have in fact caught on in the English-speaking world, e.g. oiru shokku Oil shock'; Walkman; salary-man. Japanese morphemes are subtly being introduced into model names of Japanese cars, e.g. Camry (< kammuri 'crown'), so called because of the success of previous Toyota models with crown-like names (Corona and Corolla). By the way, French and Japanese usually treat English -ing quite differently. French seems to like it: le meeting, le building, le smoking [jacket], le parking [lot]. On the other hand, Japanese usually drops it: happi endo 'happy ending'; biru 'building' (vs. biiru 'beer'). In Thai such mixed-race individuals are called luuk-khryq 'half-children', and used to be the objects of popular prejudice. Now, however, they are acquiring a certain chic, and figure prominently in popular culture. Japanese also uses dabburu in an imaginative pun: Many trendy coffee or ice cream shops have two size categories marked with the roman letters S and W, standing for shingaru (< single) and dabburu (< double-u), respectively. Loans from English are in boldface. What MARROW is related to conceptually (both within and without the Southeast Asian linguistic area) is BRAIN. See Matisoff (1978: 181-184), and Section 1 above. It has been objected that these Indo-European roots are really quite distinct, since they are reconstructed with two different laryngeals, but surely they are close enough to be considered co-allofams, or at least likely candidates for intercontamination. The degree of phonological alternation between Proto-Indo-European *pla- and *ple- seems no more extreme than that between other variants that are accepted members of the same Indo-European word-family, e.g. *wrg- » *~wrf- χ *-wrk- 'work; turn; twist'.

56.

57. 58. 59.

60.

61.

62. 63.

64.

390

James A. Matisoff

References Ballard, William L. 1984. Mother soup: A south Chinese recipe for tonometamorphogenesis. Computational Analyses of Asian and African Languages (Tokyo) 22: 43-64. Benedict, Paul K. 1939 Semantic differentiation in Indo-Chinese. Harvard Journal of Asiatic Studies 4 (3-4): 213-229. 1972 Sino-Tibetan: A Conspectus. Contributing Editor, James A. Matisoff. Cambridge: Cambridge University Press. 1975 Austro-Thai Language and Culture, -with a Glossary of Roots. New Haven: Human Relations Area Files Press. 1979 Karen genital flipflop. Linguistics of the Tibeto-Burman Area 5(1): 1-35. 1990 Japanese. Austro-Tai. (Linguistica Extranea, Studia 20.) Ann Arbor: Karoma Publishers, Inc. Buck, Carl Darling 1949 A Dictionary of Selected Synonyms in the Principal Indo-European Languages: A Contribution to the History of Ideas. Chicago: University of Chicago Press. Campbell, Lyle, Terrence S. Kaufman, and T. Smith-Stark 1986 Meso-America as a linguistic area. Language 62: 530-570. Diffloth, Gerard 1976 Expressives in Semai. In Austroasiatic Studies, part I, P. N. Jenner, L. C. Thompson, and S. Starosta (eds.), 249-264. Honolulu: University Press of Hawai'i. Evans, Jonathan P. 1999 An introduction to Qiang diachronic phonology: Synchrony and diachrony. Ph.D. dissertation, University of California, Berkeley. Friedrich, Paul 1969 On the meaning of the Tarascan suffixes of space. International Journal of American Linguistics 35 (4, part ii): 5—48. 1970 Proto-Indo-European Trees: The Arboreal System of a Prehistoric People. Chicago: University of Chicago Press. Greenberg, Joseph H. 1963 Some universale of grammar, with particular reference to the order of meaningful elements. In Universals of Language, Joseph H. Greenberg, (ed.), Cambridge, MA: MIT Press. Hansell, Mark 1990 Lexical borrowing in Taiwan Chinese. Ph.D. dissertation, University of California, Berkeley.

Areal semantics - Is there such a thing?

391

Heller, Monica 1979 What do ice and opera glasses have in common?: Struggling with car part terminology in Quebecois. U.C. Berkeley term paper. 29 pp. ms. Hoeg, Peter 1993 Smilla 's Sense of Snow. New York: Farrar, Straus, and Giroux. Jaisser, Annie 1990 DeLIVERing an introduction to psycho-collocations with SIAB in Hmong. Linguistics of the Tibeto-Burman Area 13 (1): 159-178. Jäschke, Heinrich August 1958 A Tibetan-English Dictionary. London: Routledge and Kegan Paul Ltd. Original edition 1881. Karlgren, Bernhard 1923 Analytic Dictionary of Chinese and Sino-Japanese. Paris. 1957 Grammata Serica Recensa. Bulletin of the Museum of Far Eastern Antiquities #29. Stockholm. Kurath, Hans 1921 The semantic sources of the words for the emotions in Sanskrit, Greek, Latin, and the Germanic languages. Ph.D. dissertation, University of Chicago. Published by Collegiate Press, Menasha, WI. Lewis, Gilbert 1974 Gnau anatomy and vocabulary for illnesses. Oceania 45 (1): 50-78. Lodhi, Abdulaziz Y. 2000 Oriental Influences in Swahili: A Study in Language and Cultural Contacts. Gothenburg: Acta Universitatis Gothoburgensis. 2001 Areas of Iranian-Semitic-Turkic convergence: Linguistic contact in Western and Central Asia in past and present. Uppsala: 18 pp. ms. Maran, LaRaw 1979 A dictionary of modern Jinghpaw. Unpublished ms. Matisoff, James A. 1973 The Mon-Khmer substratum in Tibeto-Burman. Paper presented at First International Conference on Austroasiatic Linguistics, Honolulu. 1977 The lexicon of Loloish and Loloish lexicography. Paper presented at ICSTLL #10, Georgetown University. 1978 Variational Semantics in Tibeto-Burman: The Organic' Approach to Linguistic Comparison. Philadelphia: Institute for the Study of Human Issues. 1980 Stars, moon, and spirits: Bright beings of the night in Sino-Tibetan. Gengo Kenkyu 77: 1—45. 1985 God and the Sino-Tibetan copula, with some good news concerning selected Tibeto-Burman rhymes. Journal of Asian and African Studies 29: 1-81.

392

James A. Matisoff

1986

1988a 1988b

1989

1991a

1991b

2001

OED 1973

Hearts and minds in Southeast Asian languages and English: an essay in the comparative lexical semantics of psycho-collocations. Cahiers de Linguistique - Asie Orientale 15 (1): 5-57. The Dictionary of Lahu. Berkeley: The University of California Press. Universal semantics and allofamic identification - two Sino-Tibetan case Studies: STRAIGHT/FLAT/FULL and PROPERTY/LIVESTOCK/TALENT. In Languages and History in East Asia: Festschrift for Tatsuo Nishida on the Occasion of his 60th Birthday, A. Sato (ed.), 3-14. Kyoto: Shokado. A new Sino-Tibetan root *dyu-k BELONG/TRUST/DEPEND/ACCEPT/TAKE, and a note of caution to megalo-reconstructionists. In Prosodic Analyis and Asian Linguistics: To honour R. K. Sprigg, David Bradley, E. J. A. H. Henderson, and Martine Mazaudon (eds.), 265-269. Canberra: Pacific Linguistics C-104. Areal and universal dimensions of grammatization in Lahu. In Approaches to Grammaticalization, vol. II, Elizabeth C. Traugott and Bemd Heine (eds.), 383-453. Amsterdam: Benjamins. The mother of all morphemes: Augmentatives and diminutives in areal and universal perspective. In Papers from the First Annual Meeting of the Southeast Asian Linguistic Society, M. Ratliff and E. Schiller (eds.), 293-349. Tempe: Arizona State University, Program for Southeast Asian Studies. Phonosemantic problems in Sino-Tibetan/Tibeto-Burman bodypart etymologies. Paper presented at International Conference on Linguistic Data Structures, University of Tübingen. Oxford English Dictionary. Compact Edition, 2 volumes. Oxford: Oxford University Press.

Purtle, Dale 1971 Some w(v)- roots in the languages of East and Southeast Asia. Paper presented at ICSTLL #4, Bloomington, IN. Sagart, Laurent 1990 Chinese and Austronesian are genetically related! Paper presented at ICSTLL #23, University of Texas, Arlington. Shorto, Harry L. 1973 Mon-Khmer contact words in Sino-Tibetan. Paper presented at First International Conference on Austroasiatic Linguistics, Honolulu. Sun Hongkai 1990 Languages of the "ethnic corridor" in Western Sichuan. Linguistics of the Tibeto-Burman Area 13 (1): 1-31.

Areal semantics - Is there such a thing?

393

Thomason, Sarah G. and Terrence Kaufman 1988 Language Contact, Creolization, and Genetic Linguistics. Berkeley, Los Angeles, London: University of California Press. Viberg, Ake 1999 The polysemous cognates Swedish go and English go: Universal and language-specific characteristics. Languages in Contrast 2 (1): 87115. 2002 Polysemy and disambiguation cues across languages: The case of Swedishes and English get. In Lexis in Contrast: Corpus-Based Approaches, Sylviane Granger and Bengt Altenberg (eds.), 119-150. Amsterdam: John Benjamins. Vovin, Alexander 1993 Is Japanese related to Austronesian? Paper presented at the Conference on Austronesain-Mainland Asian Connections, University of Hawai'i. Walker, G. D. 1925

A Dictionary of the Mikir Language. Shillong.

Wang Lida i^U 1958 Japanese loanwords in modem Chinese, [in Chinese] Zhongguo Yuwen 2: 90-94. Watkins, Calvert (ed.) 1985. The American Heritage Dictionary of Indo-European Roots. Boston: Houghton Mifflin Co.

Shafer's proto-West Bodish hypothesis and the formation of the Tibetan verb paradigms Roland Bielmeier

1.

Introduction

To understand Robert Shafer's notion of "West Bodish" it is necessary to look into his classification of Sino-Tibetan languages. In his monumental Introduction to Sino-Tibetan (Shafer 1974) he gives de facto two different classifications of the "Bodish Section". Both are mainly to be found in part I (Shafer 1974: 1-120), published 1966, only the last 7 pages (Shafer 1974: 121-127) were published 1967 in part II.1 The first classification of what he calls "Bodic Division" (Shafer 1974: 2-3) simply repeats the classification given in Shafer 1955 with a few slight changes. The second classification (Shafer 1974: 78-127), however, differs considerably. In his first classification the "Bodic Division" contains among others a "Bodish Section", and this "Bodish Section" among others a "Bodish Branch", containing four units: 1. West Bodish Unit: Balti, Purik, Ladakhi 2. Central Bodish Unit: all varieties known to Shafer, from Southern Tibet, spilling over into India, Nepal, and Bhutan, and including also Amdo and Khams 3. South Bodish Unit: Tromowa, Sikkimese, Dandzongka 4. East Bodish Unit: Takpa

These four units are taken up in his second classification in chapters 6 to 9 (Shafer 1974: 78-116), and the three other Branches of the Bodish Section are treated in chapter 10 (Shafer 1974: 117-127): chapter chapter chapter chapter chapter

6: Archaic West Bodish dialects 7: Southern Bodish 8: Central Bodish 9: Eastern dialects 10: Other Bodish languages

396

Roland Bielmeier

Shafer's concept of the "Bodish Branch" is closest to what we call "Tibetan dialects" and contains the internal classification of Tibetan. But we must be cautious to identify the "Bodish Units" with groups of Tibetan dialects determined by their closer genetic relationship and to trace them back uniformly to Old Bodish (Classical Tibetan). As we will see, the East Bodish Unit "Takpa" of his first classification and the "Other Bodish languages" of chapter 10 in his second classification cannot be considered Tibetan dialects. And as far as "West Bodish" in Shafer's concept is concerned, it consists of the "Archaic West Bodish dialects": Balti, Purik and Ladakhi. In BSOAS 13(4) (1951: 1020) Shafer expressed the idea that "West Bodish diverges sharply both phonetically and morphologically from East and South Bodish, which are probably true descendants of Old Bodish." And in the chapter on the "Archaic West Bodish Dialects" in Shafer's Introduction to Sino-Tibetan (1974: 78) he repeats: "In my article on the Bodic verb, I noted that Central Bodish is descended from Old Bodish, but that West Bodish is descended from proto-West Bodish." And again he explicitly states this hypothesis in the introductory remarks to the chapter on "Southern Bodish" in Shafer 1974: 87: "Except for West Bodish ... and Dwags, the modem Bodish dialects that we know appear to be direct descendants of Old Bodish ... All of these dialects more or less resemble Gtsang and Dbus dialects, long known as Central dialects, in their phonetic degeneration; hence I have grouped them under the general designation of Central Bodish."2 For his hypothesis Shafer (1951: 1020) gives two reasons: a) "If one assumes that the West Bodish simple verb roots which resemble the Old Bodish perfect, except that they lack the b- prefix and the -s suffix, arose through degeneration of an ancient complex morphology such as we have in Old Bodish, we are faced with difficulties. For that would imply that only the transitive perfect with its generally surd initial was more used than the present and the future combined, which have sonant initials in certain types of verbs. This is improbable." Or in other words: We would have to assume that only the perfect stem of Old Bodish has survived in Western Archaic Tibetan (WAT)3 without leaving any trace of the Old Bodish present and future stem, b) "Derivation of West Bodish roots from the Old Bodish perfect also leaves unexplained the 'dropping' of b- prefix and -s suffix in Sbalti when it preserves the former and traces of

Shafer 's proto- West Bodish hypothesis

397

the latter in non-verbal categories." Therefore, Shafer (1951: 1020) offered a different approach: "An alternative to a theory of West Bodish verb roots being derived by degeneration from the Old Bodish perfect requires the assumption that West Bodish preserves a root form of the verbs which in prehistoric times had no prefix or suffix and no consonantal alternation for tense or the imperative; that to this simple root the Eastern dialect, which later became the literary language, developed prefixes and consonantal alternation which had no effect on the popular speech of Western Tibet, except in borrowed words." This hypothesis, however, we have to modify in the following way: the -5 suffix in final and postfinal position, well known from Classical and Old Tibetan, is kept in most parts of West Bodish (i.e. Balti, Purik, Lower and Central Ladakhi) in all parts of speech, and it can be traced at least in Shigatse Tibetan as well as in conservative Amdo dialects. Therefore, the (post)final -s suffix is a feature of Common Tibetan and not significant of a separate proto-West Bodish. But as far as the absence of the b- prefix, as well as the lack of the change between surd and sonant initials in certain types of verbs, connected with the lack of certain "ablaut" phenomena with verbs in Balti, Purik and Ladakhi is concerned, Shafer's conclusion is completey correct and allows to posit a proto-West Bodish differing significantly from the usual model and resembling strongly E. H. Srurtevant's (1933) "Indo-Hittite hypothesis":

Proto-Tibetan Old and Classical Tibetan

Western Western Archaic Innovative Tibetan Tibetan

Central Tibetan

Southern Tibetan

Eastern Kham Tibetan

Eastern Ando Tibetan

etc.

Figure 1. Usual classification of Tibetan based on geography, native conception, history of the problem, language change, esp. sound change

398

RolandBielmeier Bodish (proper)

Old Bodish

proto-West Bodish

Central Bodish

West Bodish

Southern Bodish Eastern dialects

Figure 2. R. Shafer's proto-West Bodish hypothesis of 1951 and 1966 based on Tibetan verbal morphology (cf. Shafer 1951, 1974)

Indo-Hittite

Hittite

Indo-European

Sanskrit

Greek

Latin

etc.

Figure 3. E. H. Sturtevant's Indo-Hittite hypothesis (cf. Sturtevant 1933)

2.

Excursus: R. Shafer's proto-East Bodish hypothesis of 1954 and 1966 (cf. Shafer 1954 and Shafer 1974)

In classifying "Dwags", the "Täkpa" of Hodgson (1853), Shafer tried to establish also a proto-East Bodish in his article of 1954 on "The linguistic position of Dwags". In his note 2 on page 351 Shafer quotes Hodgson (1853) and George Campbell (1874) as sources.4 On page 142-147 Campbell gives, besides Tibetan and the English translations, material of three Tibeto-Burman languages. One of it, "Bhootea of Twang" can be identified with Shafer's Dwags, Twang = district of rTa-wang. Comparing Hodgson's Takpa material, Shafer (1954: 349-350 = Shafer 1974: 112-113) found "certain archaic phonetic features not preserved in Old Bodish" and concluded "that Dwags must be descended from a proto-East Bodish dialect" and that "these features that are more archaic than Old Bodish and that are shared by Rgyarong are the reason for considering Dwags descended from proto-East Bodish. But Rgyarong is considered a language and not a Bodish dialect, because of its divergent morphology ... and vocabulary." Thus he came to the following diagram, which suggests a kind of "balance" between proto-West Bodish and proto-East Bodish (Shafer 1974: 113):

Shafer 's pro to- West Bodish hypothesis "Bodish I Bodish proper proto-W. B.

Old Bodish

I

I

Sbalti, Burig, etc. Gtsang, Dbus, etc. Tsangla

399

1 proto-E. B.

I

Dwags Rgyarong Gurung Branch

Figure 4. R. Shafer's proto-East Bodish hypothesis Shafer's error to consider Dwags a Tibetan dialect5 or an "East Tibetan Unit" of the "Bodish Branch" in his terminology, has clearly been recognized by Michael Aris (1980: XV-XVI): The term Dag-pa (or 'Dap' in abbreviated form) is applied by the Bhutanese to a small group of pastoralists on the eastern border who are related to the Mon-pa people of the rTa-wang region where a group of villages are classified in the Tibetan administrative system as the Dag-pa Tsho-lnga ('The Five Hosts of the Dag-pa'). Hodgson's informant in the mid-19th century was undoubtedly a native of the area: 'Takpa, the Towang Raj of the English, is a dependency of Lhasa. Its civil administrator is the Chonajung peun [mTsho-na rDzong-dpon]; its ecclesiastical head, the Tamba [?] Lama, whence our Towang [rTa-wang]'. (Hodgson 1853: 124). The confusion arose when Shafer accepted Hodgson's mistaken identification of Dag-pa with the well-known Tibetan district of Dwags-po which lies south of the gTsang-po river, west of Kong-po. We can instead be positive now that the language studied by Hodgson and Shafer belongs to the Bhutanese heartland in Bum-thang and to one segment of the 'Mon-yul Corridor'. Therefore, Shafer's proto-East Bodish hypothesis is wrong, but the diagram, in which he derives all varieties with the exception of Balti, Punk and Ladakhi (West Bodish) and Dwags (East Bodish) from Old Bodish (Classical Tibetan), explains why he had problems to posit "Southern Bodish" and the "Eastern dialects". In his first classification he distinguished a "South Bodish Unit", corresponding to "Southern Bodish" in the second classification, but it does not appear in his diagram, where it is subsumed under "Gtsang, Dbus, etc." There is no other possibility. The same holds true for the "Eastern dialects" of Kham and Amdo. In the first classification

400

Roland Bielmeier

he indeed lists them under the Central Bodish Unit, but in the second classification he separates them from Central Bodish and groups them in a separate chapter "Eastern dialects". By using "Eastern dialects", he is forced to abandon his own terminology, as "East Bodish" is reserved for Dwags on the same level as "West Bodish".

3.

Basic remarks on the formation of Tibetan verb paradigms

If we follow Shafer's proto-West Bodish hypothesis we have to abandon the idea of the Common Tibetan origin of the well-known Tibetan verb paradigms, consisting of four stems at the most. These paradigms are taken from indigenous grammatical treatises. We do not yet know who first put up these paradigms. There is no trace of them in the first known treatises of the Sum cu pa or the Rtags kyi 'jug pa, which are traditionally ascribed to the 7th century. They can be found, however, more than one thousand years later in the work of the native grammarian Situ (1699-1774), cf. the edition of Das (1915).6 And as the Rtags kyi 'jug pa, according to the native tradition the second oldest grammatical treatise written by Thonmi Sambhota, mentions in verse 12 the "three tenses" or better the "three times" (dus gsuni) da ha for present, 'das for past, and ma Ons for future, it seems possible that later grammarians who usually refer in their writings to these two first "canonical" treatises have styled the paradigms according to the three times mentioned, adding the imperative. And behind this system we may suppose the Indian philosophical and not linguistic concept of the "three times" or "three periods" (Skt. trikola: afita 'past', adya 'present', anägata 'not yet come'). The paradigms, having been taken over from the indigenous tradition, then show up in the first grammars written by western authors such as Csoma de Körös (1834) or Isaac Jacob Schmidt (1839). One of the first attempts at a linguistic analysis may be attributed to August Conrady in his pioneering work (Conrady 1896). And one of the most recent major attempts was the one of South Coblin (1976), whereas, e.g., Stephan Beyer takes a different approach in his book (Beyer 1992: 164-185). Most of these analytic attempts have certain common features of approach. They take the verb forms from dictionaries, of indigenous or western origin, rather than from texts. The danger with such material is that it reflects to a large extent the philosophical and often pseudo-linguistic ideas of the native grammarians and lexicographers. Usually the attempts at least implic-

Shafer 's proto- West Bodish hypothesis

401

itly consider the forms as belonging to a more or less homogeneous layer of language, i.e. to a kind of "Old Tibetan" or "Common Tibetan". Therefore, those attempts usually try to (re)construct a comprehensive synchronic grammatical system, where all the changes to be met with in the paradigms of all Tibetan verbs can be explained by linguistic functions within one descriptive system minimizing the exceptions. In my opinion, such an approach is a typically synchronic and descriptive approach and not a historical one. And this seems to me to be one of the reasons why it has not worked out so far. My argumentation on the other hand follows Shafer's concept and is based on the evidence of the modern spoken dialects. If we consider the dialectal evidence, we find that in Eastern Amdo Tibetan, especially in the conservative northern Amdo nomad varieties, we still find verb paradigms similar to those which we know from Old and Classical Tibetan, or simply Written Tibetan (WT), showing stem alternation especially with controllable (c) verbs, expressing an action controllable by an animated agent, whereas with the non-controllable (nc) verbs, expressing an event happening by itself or a state, there is only little or no stem alternation. One reason for this is that nc-verbs usually do not have imperative forms. A major difference between the Written Tibetan paradigms and those occurring in the conservative Amdo varieties is that there is no separate future stem in any of the Tibetan dialects. The imperfective stem of the dialectal variety either corresponds with the present stem or with the future stem of Written Tibetan. Therefore, we find three stems in Amdo Tibetan at the most.7 There are a few traces of such stem alternation in the varieties of Eastern Kham Tibetan. The variation found in Southern Tibetan, Central Tibetan and Western Innovative Tibetan, however, is usually of a different nature, mainly going back to the loss of postvocalic final -s. This coincides very well with Shafer's statement (Shafer 1951: 1020), quoted above, that "the Eastern dialect, which later became the literary language, developed prefixes and consonantal alternation which had no effect on the popular speech of Western Tibet, except in borrowed words". This could imply that some of the varieties spoken nowadays in Central and Southern Tibet once belonged to West Bodish in the sense of Shafer's concept. To illustrate the building up of the fourfold paradigms in Old and Classical Tibetan, I will show that in those Amdo varieties which have different verb stems, the formation of these different verb stems is governed by rules which are still partly productive. In this paper I will demonstrate only one of these rules, the suppletive formation, especially of imperfective stems.

402

Roland Bielmeier

The perfective stem is the basic one also in the Amdo varieties, as in Westem Archaic Tibetan and in Written Tibetan. Other rules governing the formation of the imperative or the formation of the controllable past by prefixing b- or/and suffixing -s, etc., will be treated elsewhere. The Amdo material I use, is from the highly conservative northern Amdo nomad dialect of Themchen (The.), spoken northwest of the Blue Lake. All the material has been collected and analysed by Felix Haller, who has finished a comprehensive grammatical description of this variety, which will go to press very soon.8

4.

Suppletion in Tibetan verb paradigms

The basic principle is that certain verb paradigms with stem alternation in Written Tibetan and in Themchen consist of originally two separate but semantically close verbs, which still exist as separate verbs in other varieties of Tibetan. Sometimes the verbs are etymologically related, and sometimes the suppletive present or imperfective stem is based on the etymologically corresponding nc-verb. Themchen and other Amdo varieties have three verbs, whose imperfective stems have been taken over from different, etymologically unrelated, but semantically close verbs. In Written Tibetan and in other spoken varieties these verbs are still separate and used independently of each other: The. sler, fasn, (φ)ς3η The. ser, bzi, bzi The. ndax, bdal, rdol

'to give' 'to say' 'to sit'

WT sbyin, byin, sbyin, byin9 WT zla, bzlas, bzia, zlos WT sdod, bsdad, bsdad, sdod

The. ipfv. ster The. pfv. imp. faan The. ipfv. ser The. pfv. bzi The. ipfv. ndax The. pfv. bdal, imp. rdol

'to give' 'to give' 'to say' 'to say' 'to sit, to stay' 'to sit'

< WT ster < WT pf. byin < WT zer < WT pf. bzlas < WT 'dug < WT pf. bsdad, imp. sdod

The imperative variant esn and the imp. bzi are innovations within Themchen and not from WT byin or sbyin or from WT zlos. For the imperative variant can cf. The. φ&ιη 'to get' < WT byun, The. rdzug, bdzurj, rdzuq 'to learn' < WT sbyon, sbyans, sbyan, sbyons as well as The. rdzor, bd?ar, rdzor 'to paste' < WT sbyor, sbyar, sbyar, sbyor. For the imp. bzi cf. The. rdza 'moon' < WT zla.

Shafer 's proto- West Bodish hypothesis

403

Apparently the suppletive imperfective stems in the following paradigms are quite old, as they also occur in Written Tibetan: The. len, blag, luy The. γηοη, mnan, non The. mphen, hag, hug

'to take' 'to press' 'to throw'

WT len, blans, blan, Ions WT gnon, mnan, mnan, non WT 'phen, 'phans, 'phan, 'phons The. ndzsn, bzun, zuq 'to seize' WT 'dzin, bzun, gzun, zun(s) The. nfaen, [§aq, [suq 'to pour in (hon)' WT 'dren, dram, dran, drons

Again, these paradigms are formed on the basis of two verbs, which in some of the spoken varieties still function as separate verbs independently of each other. And even in Written Tibetan, traces of the older separate usage can be found in some cases. The present stem is to be considered suppletive and the future stem may represent the old present stem. In the spoken varieties, the correspondences of WT len, blans are taken together to one paradigm only in the varieties of Eastern Amdo Tibetan and in some varieties of Eastern Kham Tibetan as, e.g., in the variety of Nangchen. In Western Archaic Tibetan, however, only correspondences of WT len are used for all formations. This is also the case in some varieties of Central Tibetan as, e.g., in Southern Mustang, Kyirong and Shigatse. In the variety of the Western Drokpas we find len as imperfective and lay as perfective. But in Dingri len is used as imperfective and lay as perfective and imperfective as well. In the spoken varieties, the correspondences of WT gnon, mnan are taken together to one paradigm only in the varieties of Eastern Amdo Tibetan. In Western Archaic Tibetan, nc-correspondences to WT non 'to catch hold of (by chance)' are used as well as c-correspondences to WT mnan 'to press'. In the variety of the Western Drokpas, however, we do not only find the nc-verb ngn 'to catch hold of (by chance)' < WT non and the c-verb ηεη 'to press down, to subdue' in perfective contexts < WT mnan, but also the c-verb n&n 'to catch, to seize (power)' in imperfective contexts < WT gnon. Thus we see how the mechanism works. The nc-verb non 'to catch hold of (by chance)' has been transformed into a c-verb gnon 'to catch, to seize (power)' including a semantic modification by prefixing g-. Due to the resulting semantic proximity to mnan 'to press (down), to subdue' gnon became the imperfective counterpart of pfv. mnan in certain Tibetan varieties. I do not want to exclude the possibility of an etymological relation between non and *nan > mnan. On the contrary, such a supposed etymological relation could account for the semantic closeness which

404

Roland Bielmeier

makes it possible to unite the different verbs in one paradigm. This is often the strategy of the Tibetan varieties with different stem forms. They integrate etymologically related and semantically close "new" present or imperfective stems into the verb paradigm. It may be mentioned in this context that in some varieties we find correspondences to the nc-verb Ion 'to (be able to) get, to reach' which is a variant of the nc-verb non, not of the c-verb len 'to take',10 which has again a variant nen 'to take' known in Western Archaic Tibetan, Western Innovative Tibetan and Southern Mustang. Up to now no correspondences to WT 'phen, 'phans have been found in Central Tibetan. In the varieties of Eastern Amdo Tibetan and in some varieties of Eastern Kham Tibetan, WT 'phen, 'phans are taken together to one paradigm similar to Written Tibetan. In Western Archaic Tibetan and Western Innovative Tibetan only correspondences to WT 'phans with the basic meaning 'to throw' are used in all formations. I have not come across a correspondence to WT 'phen outside Eastern Kham Tibetan and Eastern Amdo Tibetan so far. In the case of WT 'dzin, bzun, gzun, zun already Jäschke (1881 s.v.) has pointed to variants used in all tenses. A paradigm corresponding to Written Tibetan is found only in Eastern Amdo Tibetan. All other dialects use either forms etymologically related to WT 'dzin or forms etymologically related to bzun, gzun, zun. Most Central Tibetan dialects use forms related to WT 'dzin. In Western Archaic Tibetan, Western Innovative Tibetan, Southern Tibetan and Eastern Kham Tibetan, however, usually forms related to bzun, gzun, zun are found. With the exception of Jirel, all dialects using forms related to WT 'dzin show initially a simple fricative which clearly points to a protoform *zin rather than to 'dzin. And if we look more closely we find that WT zin does not only mean 'to be finished', but has a homonym with the meaning 'to seize', also with the mind, i.e. 'to understand, to know' (like German ergreifen and begreifen), cf. the following phrase from mdzans-blun (Hahn 1996: 144): bu ... gnis ... nor bdaggis zin nas 'after the two sons were seized by a rich man'

Therefore, apparently two verbs are involved. One is *zuri, pf. bzun and the other 'dzin, pf. zin, which were mixed up in Written Tibetan and Eastern Amdo Tibetan due to their semantic proximity and perhaps also due to their similarity on the sound level. Finally it may be interesting to note that also in Balti both etyma are present but without being paradigmatically

Shafer 's proto- West Bodish hypothesis

405

connected, cf. Balti zunma 'to catch', related to *zun ba, and Balti tjozin 'recognition' < *no zin, cf. WT no 'dzin. In the paradigm WT 'dren, drans also two originally separate verbs are involved. The various meanings listed for WT 'dren, drans may be summarized as 'to draw, to pull, to lead' and also as 'to serve (food), to invite', a meaning which may be derived from the first one. With the meaning of 'to invite' the Written Tibetan verb 'dren, in honorific contexts spy an 'dren, still works as a full verb in Old Tibetan, occurring in imperfective as well as in perfective contexts. In the li-yul chos-gyi los-rgyus, the Religious Annals of Khotan, preserved in a Dunhuang manuscript and edited by R. E. Emmerick (Emmerick 1967), we find on p. 85-86, §§ 65-67 the following two phrases (translation by myself): phyogs bzir dge-'dun spy an 'dren-pa'ipho-na bkye-ste 'sending a messenger in the four directions to invite the Sanghas' dge-'dun spy an 'drend-pa'i pho-nas ... rgyal-po-lagsold-pa 'the messenger, who had invited the Sanghas, said to the king'

In the first phrase we clearly find an imperfective usage for 'dren and in the second clearly a perfective one for 'drend. It shows that (spyan) 'dren 'to invite (hon)' was used as a full verb in Old Tibetan. The originally independent verb WT (')dran, drans 'to draw, to lead, to conduct (water); with spyan to invite (hon)' may be derived from a noun *dran, cf. WT dranpo 'straight'. Its imperfective as well as its perfective usages can be found in Old Tibetan, as, e.g., in line 56 and 59 of the south face inscription of the Zhol inscriptions (dated after 763), cf. Li and South Coblin (1987: 145, 160). In the first phrase we see a clearly imperfective usage, in the second a clearly perfective one, marked by final -s: ken sir dran ba'i dtnag dpon chen po(r) 'the great army commanders to attack Keng-shi' lit. 'the great army commander leading (the army) to Keng-shi' ken sir drans n as 'after attacking Keng-shi' lit. 'after having led (the army) to Keng-shi'

With the meaning 'to pour (hon)' the paradigm is used in Themchen. But in Nangchen the correspondence of WT 'dren is found in perfective contexts with the meaning 'to offer (chang, etc.)'. In Central Tibetan the

406

Roland Bielmeier

meanings of the records are also mainly 'to serve (food), to invite', but the different forms do not constitute a paradigm as in Written Tibetan and Themchen. In certain Central Tibetan varieties the correspondences of WT 'dren and of WT dran(s) are still used as separate verbs, usually with ten (WT gdan) in the meaning 'to invite', cf: WDro. [en (ipfv., pfv.)

'to serve (food), to water (plants); to invite' Dingri [hen (ipfv., pfv.) 'to serve (food); to invite' Dingri [haq (ipfv., pfv.) 'to serve (food)'; (with ten) 'to invite' Kyir. [ghg (ipfv., pfv.) (with 4gmbo) 'to serve up'''

< WT 'dren < WT dren < WT dran(s) < WT dran(s)

The general analysis must take WT 'dron, 'drons, 'dron 'to find, to come, to arrive' into account, which occurs in Shigatse t§oq 'to make straight' and in Dingri ten [hoij 'to invite', both in imperfective and perfective contexts.12 A different etymon, however, is present in Southern Mustang thak c[en 'to set up the warps in weaving' and in the Tabo variety of Spiti tha: fen 'to prepare the loom, to spread', despite a certain closeness in pronunciation and meaning. Exactly the same meaning is found in Western Archaic Tibetan with several correspondences to an undocumented WT (thags) *sbran, which also accounts for the forms in Southern Mustang and Spiti. The Purik variety of Kargil has zbran 'to fasten, to tie (e.g. threads or ropes)' and the Lower Ladakhi variety of Nurla ran 'to lead (water to the fields)' and thaks ran 'to set up the warps in weaving', etc. The sound change from WT sbr- > r- in some parts of Lower and Central Ladakh is completely regular, and this leads to Jäschke's (1881) remark under the entry 'dren: "ran 3. col. for 'dren to lead (water)". We now see that 'dren has to be replaced by *sbran here. The paradigm for 'to go' is suppletive in Written Tibetan as well as in many spoken Tibetan varieties: WT 'gro, son, 'gro, son

'to go'

The. pdzo, shy, s^rj Balti go, so/7, soy Leh tfha, sog, soij

'to go' 'to go' 'to go'

The. ipfv. fldzo < WT 'gro Balti ipfv. go is a loan Leh ipfv. tfha < WT cha(s) 'to go'

Western Archaic Tibetan shows that again the imperfective stem is the suppletive one. Balti has go as imperfective stem, a loan from neighbouring Shina, and all the other varieties of Western Archaic Tibetan have tfha as

Shafer 's proto- West Bodish hypothesis

407

imperfective stem < WT cha(s). As WT son (pf.) and WT son (imp.) are extremely stable and their correspondences are used in most spoken varieties, they may safely be taken as the basic pattern for 'to go'. Only in certain Central Tibetan varieties, correspondences of WT phyin are found as perfective stem, sometimes along with correspondences of WT son. A suppletive paradigm for 'to come' is also known of many varieties in Western Innovative Tibetan, Central Tibetan, Eastern Kham Tibetan, Eastern Amdo Tibetan and even occurs in Written Tibetan, cf. WT yon, yon, yon/sog and The. jug, jug, ςοχ. This paradigm is exceptional as it is the imperative which is suppletive here. We can see that it is a younger phenomenon, because we find both in Written Tibetan, the old form yon and the suppletive imp. sog belonging etymologically to WT gsegs, used for 'to go, to come' in honorific context. In the above-mentioned spoken varieties, however, the imperative has lost its honorific character. Only in Balti the verb is still used with its honorific meaning as it is to be supposed for Common Tibetan, cf. Balti ipfv.pfv. faxs, imp. foxs 'to go, to come, to sit (hon)'. In BTC (1985) we find the Written Tibetan paradigm 'ded, bdas, bda', ded 'to expel, to pursue, etc.' The Fifth Gser tog, Bio bzan tshul khrims rgya mthso (1845-1915) gives two paradigms 'ded, ded, ded, ded (Gser tog 1995: 284) and 'ded, bdas, bda', dos (Gser tog 1995: 251). Other lexicographers, however, list two separate verbs, like e.g. J schke (1881), who gives: 'ded, ded, -, ded 'to drive, to pursue, to chase, to run after, etc.' and bda', pf. bdas = 'ded 'to drive, to chase, etc.' Both verbs are widely spread among the spoken Tibetan varieties. But they are always clearly separated. An important reason for mixing these verbs in Written Tibetan is probably the semantic proximity. But in addition to that it may be pointed out that at least for the forms used in Kham today it is difficult to establish whether they are etymological correspondences of WT 'ded or of WT bda'. A further suppletive paradigm in Written Tibetan is 'byin, phyun, dbyun, phyun 'to cause to go out or to come forth'. Correspondences of V/Ύ phyun in imperfective and perfective contexts with the basic meaning 'to take out' are widely spread in Western Archaic Tibetan, but we have no records in any other Tibetan dialect. Therefore, it is not astonishing that we have not come across the paradigm in any spoken variety of Tibetan so far. In perfective contexts, the dialectal records always take (post)final -s, cf. e.g. Balti ptyug 'to take out' andptyugs 'took out'. Already J schke (1881 s.v.) has pointed out that WT 'byin, phyun, dbyun, phyun is the transitive verb (or c-verb in our terminology) to WT 'byun, byun, 'byun 'to come out, to

408

Roland Bielmeier

become visible, to appear'. And the correspondences of WT byun are used with the same meaning as full verbs throughout the varieties of Western Archaic Tibetan in imperfective and perfective contexts, cf. e.g. Balti bjuij 'to come out' and 'came out'.13 Especially in Central Tibetan it is often used as an auxiliary. And in Themchen the correspondence of WT byun is used as a full verb in the phrase nal