The Yearbook of South Asian Languages and Linguistics: 2000 9783110245257


246 55 33MB

English Pages [320] Year 2011

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

The Yearbook of South Asian Languages and Linguistics: 2000
 9783110245257

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

EDITORIAL BOARD CHIEF EDITOR: Rajendra Singh, D£pt. de Linguistique, Universite* de Montreal C.P. 6128 Succ. Centre-ville, Montroal H3C 3J7, Canada. Voice: (514) 343-2113; Fax: (514) 343-2284; e-mail: [email protected] ASSOCIATE EDITORS: Probal Dasgupta, Centre for Applied Linguistics, University of Hyderabad, Hyderabad 500046, India. Voice: (91-40)258-500; Fax: (91-40)258-120; e-mail: [email protected] K.P. Mohanan, Department of English, National University of Singapore, Singapore 119260. Voice: (65) 772-6042; Fax: (65) 773-2981; e-mail: [email protected] REGIONAL EDITORS: R.K. Agnihotri, University of Delhi, India; Tfej K. Bbatia, Syracuse University, NY, USA; W.S. Kaninatiltake, University of Kelaniya, Sri Lanka; Baber SA. Khan, UAE Air Force Academy, Al-Ain, United Arab Emirates; Rajend Mesthrie, University of Cape Tbwn, South Africa; France Mugler, University of the South Pacific, Fiji. Tsuyoshi Nara, Seisen University, Tokyo, Japan; John Peterson, Universität München, Germany; Tariq Rahman, National Institute of Pakistan Studies, Islamabad, Pakistan; Udaya Narayana Singh, University of Hyderabad, India; £. Tiffou, Universite" de Montr£al, Canada; Yogendra P. \adav, Royal Nepal Academy, Kathmandu, Nepal. EDITORIAL ADVISORS: E. Annamalai, Mysore, India. Ron E. Asher, The University of Edinburgh, UK; Bernard Comrie, Max Planck Institute, Leipzig, Germany; Wolfgang U. Dressler, Universität Wien, Vienna, Austria; Aravind Joshi, University of Pennsylvania, Philadelphia, USA; Ashok R. Kclkar, Pune, India; Paul Kiparsky, Stanford University, USA; E.F. Konrad Koerner, University of Ottawa, Canada; Bh. Krishnamurti, Hyderabad, India; Jayant K. Lele, Queen's University, Kingston, Canada; Marvin Minsky, M.I.T., Cambridge, MA, USA; Pieter Muysken, Universiteit van Amsterdam, The Netherlands; N.S. Prabhu, Bangalore, India. EDITORIAL ASSISTANT: Sylvain Neuvel, University of Chicago. South Asia is home to a large number of languages and dialects. While the number of linguists working on South Asia has grown considerably in the recent past, there is as yet no recognized international forum for the exchange of ideas among thejn. The Yearbook of South Asian Languages and Linguistics is designed to be just that forum. It will consolidate empirical and theoretical research and provide a testing ground for the articulation of new ideas and approaches grounded in a study of South Asian languages but which have universal applicability.

-f

The Yearbook of South Asian Languages and Linguistics 2000 Chief Editor

RAJENDRA SINGH

Available From: ___w».·

é i w i « V » l « W I V h II

SUBSCRIPTION AGENT · BOOK SELLERS PUBLISHERS · STOCKISTS OP BACK-VOLUMES 4, B.S.S. HALL, KHYBER PASS MARKET, CIVIL LINES DELHI-110054 (INDIA), PH.: 3811656 FAX ; 91-11-3612678, E-mail: serials Ï iatyim.n«t.in BRANCH OFFICE: 4 30/24, AN8ARI ROAD, DARYA QANJ NEW DELHI-110002 (INDIA). PH.: 3245225

Sage Publications New Delhi · Thousand Oaks · London

Copyright © Rajendra Singh, 2000 All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage or retrieval system, without permission in writing from the publisher. First published in 2000 by Sage Publications India Pvt Ltd M-32 Market, Greater Kailash, Part 1 New Delhi 110 048 Sage Publications Inc 2455 Teller Road Thousand Oaks, California 91320

/^m\ Sage Publications Ltd 1^1 6 Bonhill Street >Af London EC2A 4PU

Published by Tejeshwar Singh for Sage Publications India Pvt Ltd, typeset by Line Arts, Pondicheny and printed at Chaman Enterprises, Delhi.

ISSN: 0971-9539

UNIVERSlTAiS BIBLIOTHEK LEIPZIG

^--

--

.—

_,

;3

CM.

ISBN: 0-7619-9464-5 (US-HB) 81-7036-936-3 (India-HB)

Sage Production Team: Abantika Chatterji, N.K. Negi and Santosh Rawat

Contents Commentum Editoris A. INVITED CONTRIBUTIONS Kannada Clause Structure

R.AMRITAVALU

Functional Constraints on Word-formation Rules D.N.S. BHAT Dravidian Hindi in South Africa: An Historical Variety

7

11 31

RAJEND MESTHRIE

49

RAJENDRA SINGH and AD BACKUS

61

K.V. SUBBARAO

93

Bilingual Proficiency and Code-switching/Mixing Patterns Syntactic Typology and South Asian Languages Tesniere Indicators and Indian Languages PROBAL DASGUPTA B. OPEN SUBMISSIONS Gerundial Aspect and NP Movement

TANMOY BHATTACHARYA

The Passive and Related Constructions in Maraihi

PRASHANT PARDESHI

Language-teaching and World View in Urdu Medium Schools in Pakistan

TARIQ RAHMAN

Pragmatic Explanations for Expressing Obligations of the Agent Referred to in Hindi

GHANSHYAM SHARMA

109

123 147 173 185

C. REGIONAL REPORTS, REVIEWS, AND ABSTRACTS Regional Reports

1. Africa: Mauritius

RAJEND MESTHRIE

2. India: Dravidian

BH. KRISHNAMURTI

205 208

6 · Contents 3. North America TEJ K. BHATIA 4. South Asia: Historical HANS HENRICH HOCK 5. South-East Asia and the South Pacific FRANCE MUGLER

213 220 237

Review Article

Substantiating Blasphemous Claims for the Belabored Nativity of these other Englishes, a review of Rajendra Singh (ed.), The Native Speaker: Multilingual Perspectives ARJUNA PARAKRAMA

243

Reviews

R.K. Agnihotri et al. (eds), Social Psychological Perspectives on Second Language Learning by KATHLEEN CONNORS M. Hariprasad et al. (eds), Phases and Interfaces of Morphology by SYLVAIN NEUVEL Robert D. King, Nehru and the Language Politics of India by JAYANT K. LELE Bh. Krishnamurti, Language, Education and Society by ROBERT D. KING Annie Montaut (ed.), Les langues d'Asie du Sud by Luc V. BARONIAN R. Narasimhan, Language Behaviour, Acquisition and Evolutionary History by OTTO M. IKOME Rajendra Singh, Lectures Against Sociolinguistics by KEN C. ERICKSON

255 257 262 265 266 270 277

D. DIALOGUE On the Non-hierarchical Structure of Compounds: A Reply to Singh and Dasgupta THOMAS BECKER 283 Second Degree Morphology: A Difficulty for the One Variable Constraint?

SYLVAIN NEUVEL

In Praise of Sakatyayana: Some Remarks on Whole World Morphology RAJENDRA SINGH and ALAN FORD E. ANNOUNCEMENTS The Chatterjee-Ramanujan Prize Housekeeping About the Chief Editor Notes on Contributors

293

303

313 313 315 317

Commentum Editoris It is a pleasure to put this issue of The Yearbook of South Asian Languages and Linguistics in your hands. With the help of South Asianists from around the world, The Yearbook has clearly arrived—it has become the standard it was intended to be. Even at the risk of appearing to be immodest, we are happy to report that it has received rave reviews: The Book Review (22:12), for example, credits The Yearbook with upholding the highest standards of publication. The Invited Contributions section of this issue contains Amritavalli's indepth analysis of Kannada clause structure, Mesthrie's thoughtful and sensitively written account of a relatively unknown variety of Hindi, and Subbarao's synthesis of what is known about the typological dimensions of South Asian linguistic diversity. A rich feast indeed. It also includes contributions authored or co-authored by two of us (Dasgupta and Singh). Our explorations are invited contributions in the sense that colleagues who saw their preliminary versions convinced us that we should publish them in The Yearbook. The Open Submissions section, a selection from more than a dozen fine pieces received for consideration, includes contributions on specific aspects of the grammar of Bangla (Bhattacharya), Marathi (Paradeshi), and Hindi (Sharma) as well as a special look at language-teaching on the ground in a corner of South Asia (Rahman). As Regional Editors cannot always have things to report—research even in this publish or perish age is not that fast—we invited Professors Krishnamurti and Hock to provide us with their assessments of what has recently been accomplished in the domains of their expertise. We are grateful to them for accepting our invitation, and are delighted to include their assessments as special reports. The section also contains reports filed by Regional Editors Bhatia (North America), Mesthrie (Africa) and Mugler (Southeast Asia). Dr. Mugler has agreed to replace Dr. Saleemi, who no longer works in Southeast Asia, and we are grateful to her for accepting our invitation to take care of that region. We also take this opportunity to thank Dr. Saleemi for sending us a regional report for the inaugural issue.

8 · Contents In his review-article, Parakrama, the author of Dehegemonizing Language Standards, carefully reviews most of the individual chapters of The Native Speaker: Multilingual Perspectives and thoroughly examines the problematic that book is devoted to. His review should generate a lot of discussion, as should some of the other papers in this issue. Our Dialogue section will always be available for such discussions. As for book reviews, it is a matter of pleasure to acknowledge the cooperation received from such distinguished South Asianists as King and Lele, who took the time to do what they were requested to do. No less important is the cooperation received from the non-South Asianists who agreed to put their expertise at the service of South Asian Linguistics. It is also a matter of great satisfaction and pleasure to report that the Singh and Dasgupta piece in the Dialogue section of the 1999 issue of The Yearbook elicited two empirically and theoretically rich responses (Becker and Neuvel). They are both included in this issue. Rather than responding to them, the architects of the theory the Singh and Dasgupta piece was based on and inspired by have contributed a brief note that opens up the possibility of a discussion of a wider range of issues in morphology. We are grateful to our referees, who, unfortunately, must remain unnamed, and to Sylvain Neuvel, who gladly accepted the responsibility of helping us with technical matters. We are also grateful to the Sage team responsible for this issue for its help and cooperation in producing a flawless volume. Rajendra Singh Probal Dasgupta K.P. Mohanan

Invited Contributions

Kannada Clause Structure* • R. AMRTTAVALLI · This investigation of clause structure in Kannada (a Dravidian language of South India) shows that: (a) Finiteness cannot be equated with Tense. Kannada lacks Tense, with fmiteness residing in a Mood P whose head is one of the elements Indicative (Agr), Modal or Neg. (b) What is interpreted as Tense is an Aspect in the domain of a finite head. Perfect aspect is interpreted as past tense, imperfect aspect as non-past tense. (c) Aspect is present in nonfmite verb forms (the gerund and the infinitive). In English as in Kannada, bare infinitives and dative case-marked infinitives (or 'for-to' infinitives) differ in aspect: bare infinitives carry perfect aspect, dative case-marked infinitives carry unrealized aspect.

• 1. The Problem and an Analysis Finite clauses in Kannada fall into the following superficial subtypes, with the superficial structure indicated: I. Verbless clauses: NP NP (copular clauses) II. Clauses with verbs (including copular/existential 'to be'): (i) affirmative clauses: S O V -Tns -Agr (ii) negative clauses : S O V Inf/Ger -Neg (iii) modal clauses : S O V Inf - Modal * Parts of this paper were presented at the GLOW Colloquium, CIEFL, Hyderabad in January 1998, and at the conference on the Syntax and Semantics of Tense and Mood Selection, University of Bergamo, in July 1998. My thanks to the audiences at these two fora. The research reported here was undertaken during the year I was on sabbatical leave from the CIEFL.

12 · R. Amritavalli If we take finiteness to be what allows a clause to 'stand alone' as a root clause, it is a common current assumption (e.g. Shlonsky 1997:3,55) that finiteness is synonymous with Tense.1 Clauses of type (I) above—that is, 'nominal clauses' or copular clauses with no overt verb—are prima facie counterexamples to this assumption. Nominal clauses occur also in the present tense in Hebrew. For Hebrew, analyses differ on whether these clauses are indeed tenseless. Rapoport (1987) considers them so, offering a choice of structure of a matrix small clause and a tenseless full clause headed by Agr in Infl. Shlonsky (1997), however, assimilates Hebrew nominal clauses into regular clause structure by postulating a null present tense auxiliary with abstract tense and agreement features. Kannada copular sentences would appear to be a little more recalcitrant to a similar assimilation effort.2 But my focus in this paper is not on copular sentences. I concern myself with negative sentences in Kannada (clauses of type II [ii]) that are the counterparts of affirmative sentences with verbs fully inflected for what looks like Tense and Agreement morphology (clauses of type II [i]). In standard accounts of clause structure, affirmative and negative sentences differ primarily in the presence of the Neg P: The Neg P is simply an additional functional projection optionally introduced into the functional architecture of the non-negative clause. (Modals, too, are similarly 'added into' the functional frame of the basic clause.) In Kannada, however, the negatives of the simplest present and past tense sentences look dramatically different from the affirmatives, as a cursory comparison of the Schemas II (i) and II (ii) shows. I present here an analysis of negative sentences in Kannada which leads me to deduce three claims: (a) Finiteness cannot be equated with Tense. Kannada appears to lack Tense, with finiteness residing in an XP (perhaps a Mood P) whose head X is one of the elements Indicative (Agr), Modal or Neg. (b) Tense being absent, what is interpreted as Tense is an Aspect in the domain of the finite head X. Perfect aspect is interpreted as past tense, non-perfect aspect as non-past tense. (c) Aspect is present not merely overtly in the standard 'perfect' (-en) and 'progressive' (-ing) verb forms, but also covertly in nonfinite verb forms, e.g. in the gerund, and, crucially, in the infinitive forms. 'Bare' infinitives and 'for-to' infinitives (or dative case-marked infinitives) appear to consistently differ in their aspectual specifications in two such genetically and areally unrelated languages as Kannada, and English: 'bare' infinitives carry perfect aspect, 'for-to' (or dative case-marked) infinitives carry unrealized aspect. A consequence of the first and fundamental claim, delinking Tense from finiteness, is that it explains the simultaneous existence in Kannada (as in

Kannada Clause Structure m 13 Hebrew) of two major finite clause types—with and without a verb. For theories according Tense a pivotal role in clause structure, with Tense carried by the verb, 'nominal clauses' are an anomaly, to be explained away by assimilating their structure to that of verbal clauses. A consideration of 'nominal clauses' in their own right suggests that they are 'built around' Agr (Rapoport 1987; Amritavalli 1997). We are now arguing that in Kannada, affirmative Verbal clauses' are also 'built around' Agr, and not Tense. This argument is entirely due to properties intrinsic to 'verbal clauses'; its rationale is to explain the superficial variety of clause types exhibited in affirmative, negative and modal sentences in terms of a single clause structure. What is interesting, then, is that a claim motivated by the analysis of Kannada clauses of subtype II generalizes to and explains clauses of subtype I.3

• 2. Negation in Main Clauses Consider the affirmative sentences la and Ib. The verbs in them are standardly analyzed as having the tense and agreement markers identified in the glosses. (1) a. avanu bar he come 'He comes.' b. avanu ban he come 'He came.'

- utt -aane -non-past -3p.m.sg -d -anu -past -3p.rn.sg

Consider now the negative counterparts 2a, 2b of la, Ib. (2) a. avanu bar -uvud (u) ilia he come -gerund Neg 'He does not come.' b. avanu bar -al (u) ilia he come -inf. Neg 'He did not come.' We note that (é) the verbs in 2a, 2b are not inflected for tense and agreement; (u) they occur in a gerundive or infinitive form, apparently non-tensed forms. Yet tense cannot be absent in (2a, b). First, the choice of the nonfmite verb form reflects a (covert) tense. Gerund plus ilia negates a verb with non-past tense, while infinitive plus ilia negates the past tense. (This pattern of negation is a feature of the modern language. An earlier stage had a negative infix on the verb in main clauses [V + Neg + Agr], giving rise to the 'negative conjugation* or 'mood'. In this [now absent] conjugation, tense was not merely

14 · R. Amritavalli superficially absent; there was also no indication of tense in the interpretation, i.e. negative verb forms in the 'negative conjugation' were apparently free with respect to the interpretation of Tense.)4 Secondly (and strikingly), the Neg ilia is not licensed in genuinely nontensed (= nonfinite) clauses, i.e., in non-root gerundive and infinitive complement clauses. This is shown by 3-4: the a sentences cannot be negated by adding ilia (the b sentences). (3) a. [avanu ivattu bar -uvud(u)] aashcharya he today come -gerund surprise 'His coming today is a surprise.' b. *[avanu ivattu bar -uvud (u) ilia (du)] aashcharya he today come -gerund ilia surprise *'His not coming today is a surprise.' (4) a. avanu [PRO iij-al (u) nooDidanu he PRO swim-inf. saw (i.e., tried) 'He tried to swim.' b. * avanu [PRO iij-al (u) ilia (lu)] nooDidanu he PRO swim-inf. ilia saw (i.e., tried) *'He tried not to swim.' In other words, what we have in 2a and 2b appear to be 'matrix' infinitives or gerunds.5 The facts presented above raise a couple of fundamental questions about the functional architecture of the clause. Given that we do not expect radical differences in clause structure between affirmative and negative sentences, the questions to answer are: (a) What happens to tense (and agreement) in 2a, 2b? (b) How is the nonfinite verb form selected in 2a, 2b, depending on the tense to be negated? The intuitive answer to the first question is that tense is in some fashion 'absorbed' by the negative element ilia. The phenomenon of Tense and Neg 'going together' is attested even in English, and is widespread enough to be encoded by Laka (1990) into the pre-Minimalist s-structure condition TCC (Tense C-command). We might execute this intuition by giving ilia a strong tense feature to check, assuming very 'surfacy' structures like 5a and 5b for affirmative and negative sentences: (5) a. V]vP+/-past]TpPNG]AgrSp b. Tense does not appear on the surface in 5b (we may say) because ilia is morphologically 'defective', and hence does not manifest tense (and agreement) overtly, (ilia is in fact historically a 'defective verb' of negative existence.)

Kannada Clause Structure m 15 But the more interesting question is, how does the tense of the negated sentence get 'read off' the nonfmite morphology on the main verb? Assuming a covert tense feature located in the negative, this must be the result of a 'match' between that tense feature and the nonfmite morphology. There exist mechanisms in the theory to achieve such a 'match', such as agreement between a tense feature and the nonfmite verb, or selection of an appropriate complement type by the tense feature. Either way, there remains the central question: What makes the 'match' (of non-past tense with gerundive morphology, and past tense with infinitival morphology) non-arbitrary, and therefore learnable? I shall argue that the tense-interpretation of the Kannada negative is demonstrably non-arbitrary when the aspectual specifications of infinitives and gerunds are taken into account. I shall show, moreover, that the aspectual specifications of infinitives are more complex than currently recognized, and that careful reanalyses may well reveal such aspectual specifications to be language-independent.

• 3. Aspect in the Gerund Let us begin by considering gerunds and infinitives in their more familiar position of complement clauses. We recall here Stowell's (1982) proposal that infinitives carry 'unrealized' tense in COMP, whereas gerunds, lacking both a COMP and a tense operator, are simply transparent or 'completely malleable' to the semantics of the matrix verb.6 Now neither of these observations about infinitives and gerunds sits comfortably with our data. If gerunds have no tense operator and are 'completely malleable to the semantics of the governing verb', we expect the gerund in the Kannada negative to occur indifferently in either tense, past or non-past, with the appropriate interpretation (taking the 'governing verb' here to be Neg + Tfense, specified as [+/- past]). That is not the case, however; the gerund occurs only in the non-past tense. Worse is the case of infinitives. The Kannada negative sentence uses infinitival morphology to signal past tense, whereas the suggested tense specification of infinitives is 'unrealized', or for the future! Let us consider the gerund more closely. If we look at the paradigm for gerunds in Kannada, we see immediately that the gerund must carry aspect. The gerund form V-uvudu that occurs in the negative sentence 2a is one of a paradigm of three forms, the other members of which are the 'perfect gerund' and the 'negative gerund'. (Thus the V-uvudu gerund is the 'imperfect' gerund.) The three forms are illustrated below: (6) bar-uv-udu ban-da-ddu ba-a-r-addu come-imperf-nom come-perf-nom come-Neg-nom 'coming' 'having come' 'not coming'

16 · R.Amritavdli (7) a. [avanu ivattu bar -uvudu] aashcharya. he today come -imperf. gerund surprise 'His coming today is a surprise.' b. [avanu ivattu ban -daddu] aashcharya. he today come -perf. gerund surprise 'His having come today is a surprise.' c. [avanu ba -araddu] yaarige gott-ittul he come -Neg. gerund to whom was known 'Who knew about his not-coming?' The incidence of perfect and imperfect forms of the gerund has sometimes bed to the claim that in Dravidian, Tense is present in the gerundive and participnail forms of the verb. This is because (as we shall see) the putative tense morphemes in Kannada are in fact aspect morphemes. Thus notice the vowel u im the imperfect gerund that parallels the vowel in the putative non-past tenis£ morpheme uti in la; and the consonant d in the perfect gerund identical to the putative past tense morpheme d in Ib. Note, however, the incidence in the gerund paradigm of the negative gerund form ba-a-r-addu. This form exhibits a negative infix a in the verb bar*. The negative infix is a survivor from an erstwhile 'negative conjugation'; and the negative infixed verb currently occurs only in nonfmite contexts (cf. note 5). (As shown in 3b and 4b, the Neg ilia is barred from such contexts.) This form, moreover, is unspecified for tense, as was the erstwhile 'negatively conjugated' finite verb.7 Thus the negative gerund is not a tensed form; and by extension, nor are the other gerund forms in the paradigm 6 tensed. The three gerunds in 6 are all nonfinite, but must be specified for aspect, taking negation to be an instantiation of aspect here. We now see that in 2a, it is the non-perfect gerund that must 'match' with a putative tense feature [non-past], absorbed by ilia. It follows that in 2b, for the infinitive to similarly 'match' with the hidden tense feature [past], the infinitive must (pace Stowell) be specified for perfect aspect. There is indeed a supporting piece of language-internal evidence that this is so: namely, that the Kannada passive auxiliary aag- 'happen' takes as its complement a verb in the infinitive. (8) alii ondu mane kaTTal -aagide. there one house build-inf. aux 'There is a house being built over there.' Consider the comparable fact from English: the verb form selected by the passive auxiliary is the perfect or -en form. If this selection of a perfect is a principled fact about passives in general (owing perhaps to their 'stative' interpretation, and linking stativity with perfectivity), it is possible to draw the inference

Kannada Clause Structure m 17 that the Kannada infinitive is in this aspect parallel to the perfect form of the English verb. There is, however, a complication. Within Kannada itself, exactly as in English, there exist both purposive infinitives, and infinitival complements to verbs like try, which must have an 'unrealized' interpretation. Compare example 9a—which is a slightly modified version of example 4a given earlier—with the purposive 9b: (9) a. avanu [PROiijalu] prayatnisidanu. he to swim tried 'He tried to swim.' b. naanu [PROlarkaari taralu] horaTe. I vegetables to bring set off º set off to get vegetables.' Thus Stowell's facts, and his analysis, cannot easily be dismissed, even for Kannada. The solution to the puzzle is to recognize that there are two types of infinitives, often confused, which have opposite aspectual specifications. In Kannada, these are the dative case-marked infinitive and the 'bare' infinitive. In English, these are the 'for-to' infinitive and the 'bare' infinitive.

• 4. Case-marked Infinitives and Bare Infinitives Considering first Kannada, what we have so far cited as the infinitive form in 2b and 4a, V-alu, is indistinguishable from the 'bare' form of the infinitive. However, Kannada also has a dative case-marked infinitive. The confounding 'actor is that this dative case is often dropped; but with the benefit of hind;ight, we may sort out the data. Let us take first examples like 4a or 9: infinitival complements to verbs like ry, and purposive complements, both of which have the 'unrealized* interpreation. The form of the verb we have cited in 4a and in 9 is (to repeat) the Baseless infinitive; and this is indeed the standard, literary citation form for the 'erb in such sentences. However, the same standard dialect allows this infiniive to cany an overt dative case, especially in the spoken language. Thus 4a repeated as lOa) has the variant lOb: (10) a. avanu [PROiij-alu] nooDidanu. he swim-inf. saw b. avanu [PROiij-ali -kke] nooDidanu. he swim-inf. -dat saw 'He tried to swim.'

18 · R. Amritavalli In my own spoken language, again the standard (Bangalore-Mysore) dialect,t, there is an even more revealing difference between the bare infinitive and thee dative case-marked infinitive. The bare infinitive has the expected form V-alu.i. Whereas the case-marked infinitive is actually realized as a dative case on thee gerund: (10) c. avanu [PRO«/ -uvudi -kke} nooDidanu. he swim gerund dat. saw 'He tried to swim.' Significantly, the truly 'bare' infinitive, which (we argue) occurs in the passt tense negative sentence 2b, and as a complement to the passive auxiliary 8, cam neither be optionally case-marked, nor (in my dialect) substituted by a gerund J, case-marked or otherwise.8·9 (11) a. * avanu barali -kke ilia. (*on the relevant reading) *'He did not come.' b. *avanu baruvudi -kke ilia. (*on the relevant reading) **He did not come.' (12) *alli ondu mane kaTTali-kkel kaTTuvudi-kke aagide. 'There is a house being built over there.' (*on the relevant reading) > Coming now to English, notice that Stowell in 'The Tfense of Infinitives' actuially deals only with 'for-to' infinitives in English. Let us then investigate thte tense of the bare infinitive in English. The English bare infinitive, indistinguishable from a 'bare' or simpHy tenseless verb, occurs as a complement to perception verbs. Most of thes»e verbs also take ing or gerundive complements. Akmajian (1977) credits tto Emonds (1972) the following observation about the semantics of these two types of complements: the gerundive signifies an 'incompleted' [sic] actiom, where the infinitive signifies a completed one: (13) a. We watched the prisoners dying. (incompleted) b. We watched the prisoners die. (completed) (Akmajian 1977: example 52). Akmajian adds that the same distinction holds for certain nonperceptiom verbs which take both kinds of complements: (14) a. We had them marching into the mess hall. b. We had them march into the mess hall (ibid.: example 53). (15) a. We kept them marching into the mess hall. b. We made them march into the mess hall (ibid.: example 54).

Kannada Clause Structure m 19 He observes that 'the semantic distinction between incompleted and completed action is not restricted to perception verb contexts, but seems rather to be a function of a more general structural distinction betwefep "gerundive" and "infinitive" verb phrase complements.'10 Akmajian does not speculate on what this "more general structural distinction* might be. I suggest that it is a distinction of Aspect. What Stowell Calls the Tense of infinitives is actually the Aspect of infinitives. Locating the distinction in Aspect gives us the required distinction between 'for-to' infinitives and bare infinitives, along the values unrealized and perfect. Notice that it is very difficult to claim that bare infinitives have a tense operator: they occur in 'small clauses', which lack the functional baggage of full clauses. In the selection of the bare infinitive by the passive auxiliary in Kannada, too, it seems to be Aspect that is the operative factor. Again, this relocation allows us to group together and compare the 'tense' specifications of the gerund vis- -vis the bare infinitive in the Kannada negative. We recall that Stowell argues against Tense in gerunds: gerunds lack a COMP, and if Tense is in COMP, gerunds cannot have Tfense. But Aspect is a 'content' category, a property often intrinsic to verbs as part of their semantic specification; so it seems natural to project an Aspect Phrase whenever a Verb Phrase is projected.11

• 5. The Infinitive in English: to- and 'for-to' The distinction attempted here between the bare infinitive and the 'for-to' infinitive in English is confounded by the fact that a subsection of what we lall 'for-to' infinitives are commonly treated as to-infinitives without for. There is, however, evidence that the 'unrealized' interpretation of the so-called to-infinitive is attributable to the element for, rather than to (Jayaseelan 1987; Stowell 1982). If so, the parallel to be drawn is between dative case in Kannada and for in English, or perhaps thefor-to complex.12 A second problem is that bare infinitives in English appear to 'turn into' toinfinitives when, for example, the complement of a perception verb is passivized: (16) a. We saw Mary cross the street, b. Mary was seen to cross the street. There are two facts to note. First, the apparent convertibility holds only one vay: bare infinitives turn into ßï-infinitives, but to-infinitives do not turn into bare infinitives. Second, to typically appears in the absence of a lexical subject for the infinitive (disregarding for the moment the ecm contexts). Thus, it occurs in infinitives with PRO subjects, and in bare infinitives with an NP-trace subject 16b;

20 · R. Amritawlli but it does not occur in bare infinitives with wA-trace subjects (which we takke to be case-marked variables), see example 16c: (16) c. Who did we see cross the street (march into the hall)? These facts suggest that the presence or absence of ßï may be attributable t to case theory. Suppose infinitive verbs have a Sveak' case to assign to a subjecct, and that to is actually a reflex of an unassigned subject case, just as passiwe morphology is a reflex of object case absorption. Then in 16a, the verb assigns a Veak' case to the subject, in addition to or in conjunction with a caise received by the subject across the small clause boundary. When the bare infiinitive's subject NP-moves under passivization 16b, this case has to be reaibsorbed, and to appears. But when the same subject wA-moves 16c, there is mo case absorption, and no appearance of ßï. The case-assigning property of the bare infinitive is perhaps attributable to the incidence of perfect aspect in it (Jayaseelan [1984] notes that lexical subjects become possible in Malayalam adjuncts and in the English absolutiive construction when have or be are present). We speculate that an infinitive v^rb specified for 'unrealized' aspect must obligatorily absorb its subject case; so that its lexical subject, if any, must then find an 'external' (to the IP) c#se assigner. Thus to always appears in 'unrealized' infinitives; and the suggested link between the realization of aspect and case explains why for must occiur, andfor-to function here as a unitary 'complex'. As for the to in ecm contexts, its appearance is consistent with the account of the infinitive verb absorbing its own case, if the complement subject here gets its case not in its own clause, but in the matrix clause's AgrO projection (Lasnik and Saito 1991). The absence of the 'unrealized' reading for these clauses is due to the impossibility of for in these contexts.13·14

• 6. A Finiteness Head in Mood Ñ We have accounted for the apparent 'match' in 2a, 2b between a covert tense and the nonfinite morphology on the main verb in terms of the aspectual specifications of the bare infinitive and the gerund. Our data for this argument come from two very different languages, English and Kannada. This encourages the view that such aspectual specifications for clauses may be part of the initial state of the language faculty. This would explain the acquisition of negatives in Kannada, considering that the language-internal evidence—such as the difference of aspect between bare and case-marked infinitives—is meager at best.15 Our account began with the fiction of a covert matrix tense feature [past] or [non-past], 'absorbed' by Neg ilia, that actually 'selects' the nonfinite complement. This is a fiction that we now propose to discard. For, the aspectual

Kannada Clause Structure · 21

s specifications here identified for the nonfmite complement seem to obtain i independently of the feature content of the matrix tense. For example, in We sshall see [Mary release pigeons on Republic day], the matrix tense is non-past or f future, but the bare infinitive is interpreted as a completed action, that is, what \we have identified is a phenomenon quite different from 'anaphoric' tense, ssuch as obtains in tensed pseudorelative complements to perception verbs in IRomance (where the tense of the pseudorelative must be the same as the ttense of the matrix verb, cf. Guasti 1993:148). What seems to be required in 2a, 2b is a 'dummy' matrix tense with no subsstantive or content feature, that is, a finiteness head F, that sanctions the nonifinite clause in a matrix context, allowing its aspect to be interpreted as tense. 'This finiteness head, we suggest, is none other than the Neg ilia. A potential argument against this reduction is that the resulting structure ]] VP ]NegP, or...]VP JAspP ]NegP looks very different from what appeared to tee a reasonable structure for affirmative clauses in Kannada: ...]VP ]TP 1] AgrP. But I shall show below that affirmative sentences in Kannada actually Ihave the structure 17, quite unlike their surface appearance: 17 ]VP ]AspP jAgrP. Sentences with Modals, moreover, fall naturally into the proposed structural schema. Like negatives, they surface with neither tense or agreement (the sole exception is the Modal (I) aar-, a Modal of inability, which inflects for agreement), taking a bare infinitive complement (cf. 18a). (18) a. avanu bar· beeku\ avanu bar- bahudu he come must he come may 'He must come; he may come.' The verb form in 18a looks like a stem; but it is a (bare) infinitival, as we see when the emphatic morphemes ee or uu attach to the verb. The consonant / of the infinitive now surfaces. (18) b. avanu baral -ee beeku\ avanu baral -uu bahudu. he come inf. EMP must he come inf. EMP may 'He certainly must come; he might even come.' Modals and Neg///a are in complementary distribution in Kannada. Instead of analytic modal + ilia phrases, we have 'negative modals' such as baaradu (cf. 18c). (18) c. avanu baral -ee baaradu. he come inf. EMP must not 'He must not (under any circumstances) come.' Tius modals induce a 'matrix infinitive'.16 Let us therefore propose that Kannada has a Mood P which hosts a finiteless head F.17 Mood P takes as its complement an Aspect Phrase, which takes

22 · R. Amritavalli a VP complement. A superficial possibility is that the head F of Mood P is simply one of the elements Neg, Modal, or Agr. However, there are two facts to consider: one, that Agr on the one hand, and Neg and Modal on the other, pattern differently with respect to the type of aspectual complement they select. Agr selects verbs with standardly acknowledged progressive or perfective morphology, such as occur in non-complex clauses in the 'compound' tenses. Whereas Neg and Modal, by selecting infinitive or gerundive aspectual phrases, pattern with higher predicates that select complement clauses; they behave as if they are Outside' the IP projection, in some sense. Secondly, Agr is intuitively a very different type of element than Neg or Modal. Agr is not itself a "Mood/ but a reflex of what is traditionally labeled Indicative mood. If we take Indicative mood to be the 'absence' of mood, as is sometimes suggested, this gives us the result that there must be a null head for Mood P when Agr is present. This gives us the structure 19, abstracting away from word order: (19)

Mood P M Agr P Agr

Asp P Asp VP

It may be that Mood P is actually part of CP rather than IP. Mood P, that is, may be the lowest element of a more fully articulated 'Comp complex* (Rizzi 1997).

• 7. Tense' in Affirmative Sentences I briefly summarize in this section an independent set of arguments against a category TENSE in affirmative sentences in Kannada, that is, sentences of type II (i), which have an overt verb marked for (what is standardly analyzed a§) 'tense', and agreement. I show that derecognizing tense as a functional category in this language allows us to (a) treat as non-accidental the identicality of the tense morphemes with the progressive and the perfect aspectual morphemes; (b) explain a puzzle in the tense pattern of the 'simple' vis-ä-vis the 'compound' tenses; and (c) explain a puzzle in the pattern of negation of the verb iru 'to be'. Taking up the first point, recall the 'tense' morphemes of Kannada identified at the outset in la and Ib: -utt- for non-past, and -d- for past. The aspect morphemes are (likewise) -utt- for progressive aspect, and -d- for perfect aspect. This is seen in 'compound' tenses, where (standardly) aspect and 'tense' are both taken to occur in the verb phrase. Consider thus the composition of the past perfect form of the verb bar- 'come'. Aspect occurs on the main verb, and 'tense' on the auxiliary verb iru: note the occurrence of -d- on both verbs.

Kannada Clause Structure · 23

(20) ban -d- id -d -anu V- perf. be 'past' (= perf.) -agr '(he) had come* That is, compound tenses in Kannada appear to consist of a (serial verb-like) combination of main verb + aspect with iru + aspect. The fact that the putative tense morphemes in Kannada have the same shape as the aspect morphemes might in isolation perhaps be dismissed as a curious coincidence. Given what we have said about the structure of negative clauses (and modal clauses), however, it invites us to postulate the clause structure 17 (elaborated in 19) for affirmative clauses in place of the more conventional clause structure. TUrning to the second point, a complication in the 'tense' system of Kannada is that while verbs in general illustrate a two-valued system [perfect/ imperfect], uniquely, the existential verb iru 'to be' is instantiated in three forms, with tense interpretations corresponding to 'past', 'present', and 'future'. (21) id-d-anu ir-utt-aane idd-aane be-perf-agr be-imperf-agr be-pres.agr '(he) was' '(he) will be' '(he) is' In 21, the perfect form of iru is the regular form. But the regular imperfect morpheme here has only a future interpretation; and there is an additional, irregular form of iru9 with no clearly isolable aspect morpheme, interpreted as the present (although the stem of the 'present' form in 21 resembles the perfect stem, this is merely a superficial resemblance: there is a clear differentiation of the two in the neuter forms ittu '(it) was', ide '(it) is'). The verb iru retains this idiosyncracy when it functions as an auxiliary verb in the 'compound' tenses (Recall that the compound tense is a serial verb-like combination of the main verb and its aspect with iru and its aspect.) Thus, there appear to be three-compound tenses, although the language otherwise has a two-tense system. Below we illustrate the 'perfect tenses'. (22) a. ban-did-d-anu come-perf be-perf-Agr '(he) had come' b. ban-dir-utt-aane come-perf be-imperf-Agr '(he) will have come' c. ban-didd-aane come-perf be-pres.Agr '(he) has come' A similar paradigm obtains for the 'progressive tenses', with 'past', 'future', and 'present progressive' instantiated: bar-utt- id-d-anu, '(he) was coming'

24 · R. Amritavalli

(often used in the sense 'he used to come'); bar-utt- ir-utt-aane, '(he) will be coming' (commonly used for habitual or iterated activity, cf. the typical Indian English usage 'He will be coming now and then'); bar-utt- idd-aane, '(he) is coming'. How then should we analyze the 'tense system' of Kannada, keeping in mind, moreover, Chung and Timberlake's observation (note 15) about tense in general being a two-valued system in the world's languages? Our proposal is that the third, 'present tense' that appears only in the verb iru 'be' is simply the 'finiteness element' that resides in the Mood P. Our claim is that the existential verb iru is the only verb that can occur with no aspectual specification, that is, with a null Aspect. Hence the finiteness element is allowed to surface in its 'bare' form, with no aspectual reading added in, only in the presence of this verb: whether as a main verb or as an auxiliary. We analyze the verb iddaane (which occurs, e.g., in sentences like deevaru iddaane 'God is!' or 'God exists'), as consisting of the elements ir + ^aspect + agreement. In support of this analysis of iru, we adduce the following facts about the negation of iru. The two regular forms of iru, namely the perfect and the imperfect, are negated in the usual way, by combining a bare infinitive and a gerund form of iru (respectively) with ilia: iral (u)-illa 'was not'; iruvud(u)-itta *will not be'. Negation of the third, irregular or 'present' form of iru, however, consists in the apparent 'replacement' of iru by ilia. (This is the third point, the puzzle in the pattern of negation of iru, mentioned in the beginning of this section.) (23) a. avanu illi idd-aane he here bepres.Agr 'He is here' b. avanu illi ilia he here Neg 'He is not here' Such facts have given rise to a two-ilia theory: a Neg ilia, and a full verb ilia, a negative existential verb (this analysis, proposed in Hany Babu 1996 for Malayalam, is assumed in Amritavalli 1997 for Kannada). The plausibility of the two-ilia proposal lies in that diachronically, ilia is indeed a verb of 'negative existence': that is, a form of the verb ir—'to be' in the now obsolete negative conjugation or mood (unlike other verbs in this conjugation, ilia showed no person, number or gender agreement; it was 'defective'). This analysis is, however, not without problems. The putative verb of negative existence ilia currently occurs only in one 'tense'—the 'present' tense. (The regular forms of iru, it may be recalled, are negated in the regular way.) Unlike iru, ilia lacks any other finite forms; and it 'replaces' iru only in the 'present' tense. The two-ilia theory fails to explain either of these facts.

Kannada Clause Structure m 25 We propose that what looks like the main verb ilia in 23b is actually the Neg ilia. The main verb here is indeed iru, in its occurrence with a null aspect; and the structure of the verb in 23b is ir + ^aspect + Neg (ilia), parallel to that in 23a: ir + ^aspect + agr. This should give us in 23b a surface form *ir-illa, or *id-illa, or *U-illa (where id·, i7- are variants of the stem of iru). And indeed, a reduplicated //- does show up in emphatic structures corresponding to 23b: illav-ee ilia 'is certainly not'. But in non-emphatic contexts there is probably a 'double //-' filter, which accounts for the single ilia in 23b.

• 8. Conclusion Our attempt at an adequate descriptive account of clause structure in Kannada has unveiled some issues of broader theoretical significance. Two foci of interest to emerge have been the constitution of finiteness, and the derivation of tense interpretations. We have shown that nonfmite verbs have aspectual specifications that, while not always obvious, are of sufficient generality to be theoretically interesting, demonstrating a parallel between 'bare' and case-marked infinitives in Kannada and English, two languages that are neither genetically nor areally related. The fact that the interpretation of tense in Kannada is determined by the aspectual specification of the verb points to the need for a refinement in our understanding of the categories of Tense and Aspect, and their relation to Mood and finiteness. One understanding of this issue is that in Chung and Timberlake (1985:202): 'Tense locates the event in time. Aspect characterizes the internal temporal structure of the event. Mood describes the actuality of the event in terms such as possibility, necessity, or desirability.' A question arises in what way this broad distinction between the internal temporal structure of an event, and its location in the real world (whether by anchoring it in time, or in terms of actuality), is reflected in the functional architecture of the clause. Current syntactic analyses place on Tense the burden of actualizing the event. Our suggestion is that Tense is an epiphenomenon, and that it is the specification of Mood that serves to confer finiteness and to locate a predication in the real world. • NOTES 1. 'My starting point is that every clause, by definition contains a TP (tense phrase). I will argue... that the essential difference between a full clause and a small clause is that only the former contains a TP* (Shlonsky 1997:3);'... independent or full clauses must by definition contain a TP projection .... Clauses lacking a TP are, to adapt a familiar terminology, "small clauses" and cannot occur, for example, as root sentences' (ibid.: 55). 2. Two kinds of copular sentences are extant in Kannada, with and without the verb iru 'be*. (a) avanu oLLey-avanu he good-3p.m.sg 'He is a good man.'

26 · R. Amritavalli (b) avanu oLLey-avan- aagiidd- aane he good-3p.m.sg- aagi be.pres.-3p.m.sg 'He is a good man.' Notice the element aagi in b. I have suggested (Amritavalli 1977:48, n. 5; 1997) that aagi is a postpositional complementizer, like English 'for'. The verbless clause á is a finite clause. (It is introduced in embedded contexts by the regular complementizer anta, the counterpart of English 'that'.) The other looks like a small clause (-aagi clause) complement to a verb iru 'to be*, -aagi clauses occur (thus) in Object complement' constructions ('elect X president'), and with 'raising' verbs such as kaaN 'seem', tiLi 'know, think': (c) avan -anna presidenT-aagi elecT-maaDi. he -ace. president-aegf elect-do.imp 'Elect him president.' (d) avanu oLLey-avan- aagi kaaN- utt-aane he good-3p.m.sg- aagi seem- nonpst-3p.rn.sg 'He seems a good man.' (e) naanu avan -anna oLLey-avan-aagi tiLididde. I he -ace. good-3p.m.sg-aa£i had thought º had thought him a good man.' The Neg that appears in the verbless copular clause a, namely alia, is not the same element (namely ilia) as in b. alia normally negates noun phrases but not clauses with verbs (Amritavalli 1997). 3. The properties of Agr in nominal clauses and verbal clauses are not identical: nominal Agr lacks the person feature. 4. Hence the emphasis in Kittel (1908: 332) on how 'the modern dialect expresses clearly* or 'in a clear way' the negations of particular tenses: '... forms like iruvadilla, baruVadilla, kaaNuvadiua, aaguvadilla, in the modern dialect, take the place of the simple negative to express the present tense of the negative in a clear way, kaLeyatiUa.paDeyaUUa, keeLatUla, sigalilla are used in the modern dialect to express clearly the past tense of the simple negative...' (emphasis in the original). Kind's clarity regarding the 'match' of current negative verb forms with the tenses in the affirmative (easily verifiable by tests with time adverbials) contrasts with the confusion in this regard in some later work (for pertinent remarks, see Amritavalli 1977; Hany Babu 1986). 5. Negation in non-root gerundive and infinitive clauses is by the negative infix -a- in the verb, a morpheme which goes back to the 'negative conjugation' mentioned earlier. The problem of the 'matrix gerund/infinitive', noticed as a 'prima facie surprising construction' in Italian (Zanuttini 1991), is created by a set of assumptions. Gerundive or infinitival morphology is standardly taken to be a realization of [-finite] tense in the TP. Therefore, it must occur in complement clauses, not in matrix (finite) clauses. Typically, it needs to be licensed by lexical selection (notwithstanding Laka's [1990:197-200] evidence for selection as a compositional process where 'the inflectional elements of the matrix sentence play a role'). For English, the problem goes unrecognized if the form of the main verb that occurs with modals in matrix clauses is assumed to be a bare or stem form rather than a (bare) infinitive. However, modals like ought (standardly cited as ought to) clearly select an infinitive. Perhaps in recognition of this, Kayne (1991) (reported in Zanuttini [op. cit.]), postulates an empty modal in the Italian matrix infinitive construction described below. Italian has a 'true imperative' verb form that occurs only in the second person singular (the second person plural and the first person plural use the indicative verb form). This 'true imperative' cannot be negated by non: what occurs in the negative is an infinitive verb form. *Non telefonal Non le telefonarel 'Don't call!' (2p. sg.)

Kannada Clause Structure m 27 According to Kayne (1991), the negative licenses an empty modal that licenses the infinitive. According to Zanuttini, non itself needs to be licensed by a TP, absent in true imperatives, but perhaps present in infinitives. (Though not a finite form, the infinitive has 'some inflectional morphology', namely, the suffix re—which may be tense, or may be the head of an InfP—which suffices to license non.) What militates against similar ad hoc solutions for Kannada (for example, ilia could be said to be a matrix element that selects nonfinite complements) is the fundamental, non-peripheral nature of the phenomenon with respect to Kannada clause structure. 6. Stowell presents the contrast (i, ii): i. Jim tried [PRO to lock the door] ii. Jim tried [PRO locking the door] There is an inference in i, but not in iit that Jim did not lock the door. This would follow if infinitives and gerunds differed in their tense specification. Infinitives might have a tense operator with the specification unrealized (with respect to the matrix tense). The familiar purposive interpretation of infinitives would thus be explained, as also the inference in Hi, where the bringing of the wine is unrealized with respect to the remembering: iii. Jim remembered [PRO to bring the wine] Stowell argues that gerunds, on the other hand, have no tense operator, the 'understood tense of the gerund is completely malleable to the semantics of the governing verb*. Thus since remembering refers to the past, the wine-bringing is in the past in iv below, whereas the locking of the door in ü is simultaneous with the trying, or even unrealized with respect to it: iv. Jim remembered [PRO bringing the wine] 7. Temporal aspect can be 'added* into 7c if a dummy carrier iru 'to be* is inserted for this morpheme: bar-a~d-ee ir-uvu-du 'be without coming.' ( = 'to not come') bar-a-d-ee id-da-ddu 'having not come.' 8. Note that 'the truly "bare" infinitive' cannot be substituted by a gerund in my dialect. The significance of this point emerges in comparison with dative case-marked contexts, where the dialectal facts appear to argue for a neutralization of gerunds and infinitives in favor of a single 'nonfinite' category in Kannada. (I thank Hans Kamp for drawing my attention to the last point.) The point that in non-dative-marked contexts, gerunds and infinitives have different privileges of occurrence irrespective of dialect, can be illustrated both ways. While only infinitives are permitted in past negative and in passive sentences, only gerunds are permitted in the subject position of verbless copular sentences, and in conditional clauses: i. avanu baruvudu (*avanu baralu) aashcharya he come-gerund (he come-inf.) surprise 'His coming is a surprise.' ii. avanu baruvud-aagi (*baral-) iddare barali he come-ger aagi (*come inf) be.cond let him come 'If he is coming/ is to come, let him come.' 9. Negation of a dative case-marked nonfinite verb (examples [11] in the text) yields a modal of prohibition: maaDali-kk-illa (maaDuvudi-kk-illa), 'cannot or should not do.' Notice that these readings are in the expected direction: what is prohibited is an unrealized event, and the verb form that occurs is the dative case-marked infinitive. (Compare the English be + to inf. construction, which functions as a modal of obligation.)

28 · R. Amntavalli Like example 11, example 12 in the text also has a legitimate reading "It has been possible to build a house there/ with the nonfmite clause serving as a clausal complement to the verb aagu 'happen* ('it has happened \pro to build a house there]'). A 'dative subject* can occur in 12: nonage aUi ondu mane kaTTalikke aagide, 'It has been possible for me to build a house there.' 10. Guasti (1993:150) reiterates the facts about English bare infinitives. She, however, notes that the accusative-infinitive complement to perception verbs in Romance may have an imperfective aspect; it is in this respect like the English ace-ing construction. 11. Martin (1992b), reported in Boskovic (1997:11 and 179 n. 9) appears to have made the opposite move, generalizing the label 'tense' to aspect. Starting with the observation that eventive predicates contain a temporal argument that must be bound by (any one of) Tense, aspectual have and be, or adverbs of quantification, Martin reportedly treats have and be as specified for [+ tense] in the 'raising' complements John believed Peter to have brought the beer/to be bringing the beer. It is not apparent (however) how this meshes with his proposal (built on Stowell 1982) that [+ tense] infmitivals are control structures (since only [-1- tense, - finite] I checks null case, permitting PRO), while [- tense] infmitivals are raising structures. 12. The idea that for is 'semantically active' is due to Bresnan (1972). Jayaseelan (1987) argues from a different perspective to conclude that 'a to infinitival interpreted as describing an "unrealized" event invariably contains/or underlyingly'. He regroups fry-type verbs with want- type verbs (rather than believe- or seem- type verbs) with respect to complement selection, adducing evidence from Malayalam, as also from ing- 'infmitivals' in English: pointing out that the latter include the (little noticed) obligatorily controlled complements to succeed in, fail at, tried. In his analysis, the lexical subject of a nonfmite clause is a reflex of (/) the 'case-absorbing' properties of the matrix element (e.g. try always absorbs the case offor), and (//) the ability of ing or ßï to 'pass on' a case from a matrix element (adapting Reuland's 1983 analysis). Jayaseelan also points out that infinitive adjuncts and infinitival relatives with/or have the unrealized interpretation: ...went to market for his wife to buyapig\ the man (for you) to cultivate. ... We may add that purposive infinitives in nursery rhymes show afor. forto catch a whale (Simple Simon), for to buy a pig (To market, to market). Stowell's (1982) analysis, which locates the 'tense' of infinitives in COMP, may be taken as indirect support for our claim, since/or is located in COMP. Stowell also notes (1982:569) the following data: To kill animals ... does not have an unrealized interpretation (being aspectually equivalent to Killing animals...); but 'the unrealized tense suddenly reappears' in For John to kill his gold fish ...»where there is 'a lexical complementizer/or in COMP.' 13. The idea that perfect or progressive aspect assigns a *weak' case to the verb's subject receives support from Jayaseelan, who notes (1984: 627-28) that a have or a be in the English 'absolute' construction allows a lexical subject where only PRO is otherwise possible: i. (John) having finished dinner, Mary decided to wash the dishes. (Her children) being ill, Mary decided not to go to work, ii. (*John) munching an apple, Mary sat on the doorstep. (*Her children) covered with bruises, Mary looked pretty miserable. Jayaseelan generalizes this observation to Malayalam, where lexical subjects are permittee only in adjuncts with perfect or progressive aspect markers. We term this a 'weak' case because it sanctions a lexical subject only in nonfmite or dependent clauses. 14. Chomsky (1995:119-120) proposes that PRO receives a null Case from to or ing. The motivation for this is the argument-like behavior of PRO (which moves from non-Case positions and is barred from moving from Case positions).

Kannada Clause Structure m 29 However, Boskovic (1997:178, n. 6) credits to Lasnik the generalization of this problem (in i) to ii, and to Martin (1992a) the proposal that follows: which suggests that an alternative account of the problem is possible. i. *John tried PRO; to seem to t, that the problem is unsolvable. ii. *Hei seems to t, that Mary is ill. '...the construction is ruled out because the Case feature of to remains unchecked.' 15. There is yet another piece of interesting evidence from Kannada. The grammarian Spenser, speaking of the tense interpretations of negative sentences, tells us in a footnote (1914: 51) that in the Southern Maratha variety of Kannada, 'maaDuvudilla is present and maaDlikkilla is future*. Notice that the latter is a dative case-marked infinitive, which (as predicted) has the 'unrealized' interpretation. Spenser does not speak about whether this variety retains the bare infinitive form maaDlilla for past. If it did, it would indicate a three-way distinction past, present, future, using respectively the bare infinitive, the gerund, and the dative infinitive. However, a more likely possibility is that (in line with the tendency of languages to have a two-valued tense system) this variety distinguishes future and non-future rather than pay/ and non-past, (cf. Chung and Timberlake 1985:204: 'The direct encoding of three tenses is not particularly common. It is more usual to find only a two-way distinction in tense, either future vs. non-future or past vs. non-past.') 16. The modal of prohibition beeDa can take either an infinitive complement or a gerundive complement. The infinitive permits only a 2nd person subject, and has a prohibitive reading: i. niinu (*naavu, *avanu) noog-al-uu beeDa you (*we, *he) go inf. EMPnot 'Don't *wefyou/*him go' The gerund has a negative permissive reading: it has an impersonal rather than an imperative interpretation even with 2nd person subjects. ii. naavu (niivu,avaru) hoog-uvudu beeDa we you.pl they go gerund not 'Let us/you/them not go' These facts are parallel to the English facts in (iii): iii. Don't you (*we, *them) go Let's (let them) not go; Don't let us (him) go The stage of the language which had a negative infix in the finite clause also had modal infixes. Whereas the infixal negative is now found only in nonfmite contexts, an archaic modal construction still occurs in the matrix: niinuyeiladru bidd-ii, '(I fear that) you may fall.' The verb in the modal conjugation inflects for agreement, but tense is not present: avanu bidd-aanu, naanu bidd-e, etc. 17. Pollock (1994) proposes that 'the seldom recognized functional category of mood ... should be the head of a Mood P, which... is the highest functional projection in French and Romance as well as Old, Middle and Modern English clauses.'

• REFERENCES Akmajian, Adrian. 1977. The complement structure of perception verbs in an autonomous syntax framework. Formal Syntax, ed. by Peter Culicover, Thomas Wasow, and Adrian Akmajian, 42760. New York: Academic Press.

30 · R. Arnntovalli Amritavalli, R. 1977. Negation in Kannada. Simon Fräser University, Burnaby: Master's thesis. . 1997. Copular sentences in Kannada. Paper presented at the Seminar on Null elements, Delhi University, January 1997. Boskovk, Zeyko. 1997. The syntax ofnonfinite complementation: an economy approach. Linguistic Inquiry Monograph 32. Cambridge, MA: MIT Press. * Bresnan, Joan. 1972. Theory of complementation in English syntax. Cambridge, MA: MIT Doctoral dissertation. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press. Chung» Sandra and Alan Tlmberlake. 1985. Tense, aspect and mood. Language typology and syntactic description, volume III: Grammatical categories and the lexicon, ed. by Timothy Shopen, 202-58. Cambridge: Cambridge University Press. Emonds, Joseph. 1972. A reformulation of certain syntactic transformations. Goals of linguistic theory, ed. by S. Peters. Englewood Cliffs, NJ: Prentice-Hall. Guasti, Maria Iferesa. 1993. Causative and perception verbs: a comparative study. Torino: Rosenberg and Sellier. Hany Babu, M.T. 1986. The structure of MalayaJam sentential negation. International Journal of Dravidian Linguistics 25 (2). 1-15. Jayaseelan, KA. 1984. Control in some sentential adjuncts ofMalayalam. BLS10,623-33. Berkeley, CA: Berkeley Linguistic Society. . 1987. Remarks onfor-to complements. CIEFL Working Papers in Linguistics 4 (1). 19-35. Hyderabad: CIEFL· Kayne, Richard S. 1991. Italian negative imperatives and clitic climbing. New York: CUNY, ms. Kittel, F. 1908 [1982]. A Grammar of the Kannada Language. New Delhi: Asian Educational Services. Laka, LM. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. Cambridge, MA: MIT Doctoral dissertation. Lasnik, Howard and Mamoru Saito. 1991. On the subjects of infinitives. CLS 27, Parti: The general session, cd. by LM. Dobrin et al. Chicago Linguistic Society: University of Chicago. Martin, R. 1992a.*Case theory, -chains, and expletive replacement. Storrs: University of Connecticut, ms. . 1992b. On the distribution and case features of PRO. Storrs: University of Connecticut, ms. Pollock, Jean-Yves. 1994. Notes on clause structure. Amiens: Universite de Picardie, ms. Rapoport, T. 1987. Copular, nominal and small clauses: A study of Israeli Hebrew. Cambridge, . MA: MIT Doctoral dissertation. Reuland, E. 1983. Governing -ing. Linguistic Inquiry 14.101-36. Rizzi, Luigi. 1997. The fine structure of the left periphery. Elements of Grammar, ed. by Liliane Haegeman, 281-337. Shlonsky, Ur. 1997. Clause Structure and Word Order in Hebrew and Arabic: An Essay in Comparative Semitic Syntax. Oxford: Qxford University Press. Spenser, H. 1914. A Kanarese Grammar. Mysore. Stowell, Tim. 1982. The tense of infinitives. Linguistic Inquiry 13 (3). 561-70. Zanuttini, Raffaella. 1991. Syntactic properties of sentential negation: A comparative study of Romance languages. Pennsylvania: University of Pennsylvania: Doctoral dissertation.

Functional Constraints on Word-formation Rules* ι D.N.S. BHAT ι This paper examines a neglected, though important, aspect of word formation rules: the constraints imposed upon them by the phrasal or sentential functions for which they supply words. Our examination leads us to the hypothesis that word-formation rules manifest systematic functional differences, and that these differences are correlated with the (syntactic) classes for which these rules furnish words.

• 1. Introduction I wish to examine in this paper an evident but generally neglected aspect of word-formation rules, namely the constraints that are placed upon them by the phrasal or sentential functions for which they are to supply words. We can think of word-formation rules as having the function of producing words for different word-classes like nouns, verbs, and adjectives. Words have been grouped into these distinct word-classes on the basis of the fact that they have to perform distinct phrasal and sentential functions like identification of referents for nouns, predication of events for verbs, and modification of nouns for adjectives. Since these functions have different requirements that the words have to fulfil, and since, depending upon those requirements, words of different word-classes need to differ in their characteristics, we can expect word-formation rules also to manifest differences depending upon the type of word-classes for which they have to supply words. * An earlier version of this paper was presented at the 5th Himalayan Languages Symposium, (August 1999) at Kathmandu, Nepal. Capital letters are used for retroflex consonants throughout.

32 · D.N.S. Bhat For example, the function of identifying persons, objects or entities requires either compact names (proper names or common names) or more complex descriptions. Languages generally use words (nouns) for the former purpose and more complex constructions (noun phrases containing modifiers like adjectives or other types of complements) for the latter purpose. Wordformation rules that provide words for the nominal category must, therefore, produce compact words that can function as proper or common names, in the use of which the association of the referent with the word is determined primarily by convention. This is in contrast to the use of descriptions (noun phrases) in which the association with the referent is established on the basis of the meaning that one can derive from their constituent elements. Thus, in the case of a noun (common name) like blackboard, the association of the referent with the expression is through convention, whereas in that of a noun phrase (description) like black board it is through the meaning of its constituent elements. The former can refer to green or white objects that may or may not be made of boards, but the latter cannot. The latter can only refer to a 'board' that is 'black' in color, and further, it can refer to any board that is black in color. There is no comparable distinction in the case of verbs and adjectives. Verbs have to indicate a specific event or state in order to carry out their function of predication, whereas adjectives have to indicate a single property with the help of which the meaning of the noun that they are to modify can be constrained. Neither of these two functions involves the identification of an entity as such, and hence neither of them involves a distinction that is comparable to the distinction between names and descriptions. As a result, word-formation rules that are to provide words for these two latter word-classes do not have to produce compact words as in the case of the nominal category. We can expect this difference in the requirements of word-formation rules for nouns on the one hand, and for verbs and adjectives on the other, to get reflected in the structuring of those rules. Another interesting point that is relevant in the present context is that languages differ from one another in the number and type of word-classes that they possess and also in the type of word-formation rules that they possess. We can expect the former differences to account for at least some of the latter differences, in view of the point mentioned earlier. For example, languages that do not have a word-class of nouns would not require a set of word-formation rules that are meant for producing compact words for functioning as names. This has, in fact, been reported to be true of certain languages like Mundari (see Bhat 1985; Hoffmann 1903). However, the non-occurrence of certain types of word-formation rules may depend upon other factors as well. For example, the non-occurrence of prefixes in a language could be correlated, in some languages at least, with the occurrence of verbs in the sentence-final position.

Functional Constraints on Word-formation Rules · 33

In view of the above-mentioned points, it is possible to establish the occurrence of functional constraints on word-formation rules by examining both language-internal as well as cross-linguistic differences. In the former case, a given word-formation rule would turn out to be different for different wordclasses, or it would be restricted to only one of them. In Hindi, for example, the rules that underlie the formation of nominal compounds are quite different from the rules that underlie the formation of verbal compounds, whereas in Manipuri (Tibeto-Burman), compound formation is restricted to the nominal category. In the case of cross-linguistic differences, on the other hand, the presence or absence of a given word-formation rule would be correlatable with the presence or absence of the corresponding word-class, i.e., the sentential function for which the word-class is needed. This is exemplified by the case of Mundari mentioned earlier. There is an on-going dispute concerning the relationship between wordformation rules and syntactic rules. According to a relatively strong variant of the Lexicalist Hypothesis, syntax neither manipulates nor has access to the internal form of words (Anderson 1992:265). This claim has been questioned by other linguists who consider it necessary for the internal structure of words to be visible to syntactic rules (see Baker 1985; Bauer 1983). I would like to suggest that this dispute derives partly from a failure to differentiate between word-formation rules that belong to different word-classes like nouns, verbs and adjectives. We could expect the internal structure of names to be mostly invisible to syntax, whereas that of descriptions or of verbs and adjectives to be visible to syntax to a greater extent.

• 2. Names and Descriptions A distinction needs to be made between two different types of expressions in the case of the nominal category, namely between names (proper names or common names) and descriptions (see Bhat 1979; Bauer 1983:142; Downing 1977). I have put forth in Bhat (1979) a number of semantic as well as formal characteristics that can be used in support this distinction. Simple and derived nouns and also compounds generally function as names. That is, most of the processes of word-formation have the function of providing names in the case of the nominal category. On the other hand, noun phrases that are derived by the addition of modifiers or complements of different types to a head noun function as descriptions. Consider, for example, the following pair of sentences: (1) a. There is a blackberry on the table. b. There is a black berry on the table. Notice that la identifies an object that is on the table with the help of its name, namely blackberry, whereas Ib does so through a description. In the former

34 · D.N.S. Bhat

case, the object might or might not be black in color. It might, for example, be an unripe blackberry, in which case, it would be green or red in color. That is, the identification does not depend upon the meaning that can be derived from the constituent elements of the name blackberry, but rather on the basis of the object that is conventionally associated with it, namely a particular species of berry. In the latter case (Ib), on the other hand, the object on the table must necessarily be a berry that is black in color because Ib identifies the object through the use of a description, namely black berry. Names identify objects on the basis of certain conventions that establish relationships between those names and the objects that they identify. There is apparently a convention in English that connects the name blackberry with a particular species of berry. A speaker would be using that word with a berry of that species in mind, and a hearer can understand the speaker and identify the relevant object only on the basis of his knowledge ofthat convention. The fact that the name blackberry is made up of the elements black and berry might help him in remembering that convention, but those constituent elements cannot by themselves form the basis of his identification. On the other hand, a speaker's use of the description black berry does not directly involve any such convention. He would expect his hearer to identify the object in this latter case merely on the basis of the meaning of its constituent elements. There would be a need to make use of the notion of convention in this latter case only for establishing the meaning of those constituent elements. Linguists generally consider this particular aspect of nouns to be part of a gradation called 'productivity'. In the case of compounds, for example, nonceformation, institutionalization, and legalization are considered to be three different stages of a word-formation process (Bauer 1983:45). A compound is considered to be potentially very ambiguous at the first stage, namely nonceformation. It is considered to lose this ambiguity 'gradually' and to get restricted to a particular meaning at the stage of institutionalization. Whereas in the final stage of lexicalization, it is considered to change further through sound changes such that in some cases it may even become opaque and unanalyzable. In Kannada, for example, eNNe Oil' derives from an earlier compound form that had eL 'sesame seed' and ney 'clarified butter' as its constituents, but lexicalization has made the word un-analyzable in the present stage. I have argued in Bhat (1979), on the other hand, that the semantic transparency or non-transparency of a name does not depend upon its 'newness' and further that 'institutionalization' need not necessarily be gradual. Many compounds are institutionalized as soon as they are created as names for a particular referent. Zimmer (1971) points out, for example, that a name like bedbug was probably never used to refer to any and all bugs that might be found in beds, but was rather coined as a name for climex lectularius. Similarly, in Kannada, the namejalajanaka ('water-producer') was coined for referring to hydrogen. It was never used ambiguously for referring to oxygen or other such

Functional Constraints on Word-formation Rules · 35

entities that can also be associated with that word on the basis of the meaning of its constituent elements. It is therefore necessary to claim, I think, that a compound, if it is used ambiguously for referring to any of the possible objects that can be associated with it on the basis of the meaning of its constituent elements, is a description rather than a name. It gets established as a name in a given language only when it is institutionalized as referring to a particular object or set of objects. That is, the distinction between names and descriptions is crucially dependent upon this distinction in the connotation of forms. In English, for example, names may be written solidly as in bedroom, hyphenated as in push-button, or left open as in eating apple. They may or may not be differentiated from descriptions by stress placement as shown by blackbird with primary stress on the first element and headmaster with the primary stress on the second element (Bauer 1983). The ultimate test for the distinction between a name and a description, as pointed out by Gray (1939) (who uses, however, the terms compound and non-compound for denoting this distinction), is semantic: If it has acquired a special and distinct connotation, it is a name, otherwise it is a description. Languages may allow even complex noun phrases to be used as names and may contrast them with other similar noun phrases that occur as descriptions. In Kannada, for example, this contrast appears in the case of its genitive constructions, with structures of one particular type showing the tendency to be used as names rather than as descriptions. Consider, for example, the following pairs of constructions: Naming expression (2) hu:vi-nagiDa flower-Gen plant 'flower plant' (3) ba:la-dahuLa tail-Gen worm 'tail-worm' (4) ha:li-nakoDa milk-Gen pot 'milk-pot' (5) HaNa-da muduka money-Gen old.man 'the rich old man'

Describing expression giDa-dahu:vu plant-Gen flower 'flower of the plant' huLa-daba:la worm-Gen tail 'worm's tail' koDa-daha:lu pot-Gen milk 'pot's milk' muduka-na haNa old.man-Gen money 'the old man's money'

TTie genitive constructions occurring in the first column are all used as names vhereas the ones occurring in the second column are all used as descriptions, iannada has a tendency to use located-location, possessed-possessor, whole)art, quality-qualified and feeling-feeler structures as names and the ones

36 · D.N.S. Bhat with the reverse order (namely location-located, possessor-possessed, partwhole, qualified-quality and feeler-feeling) as descriptions. There are clearly exceptions to this rule (as this is only a tendency), which apparently derive from the fact that a given construction needs to get established (institutionalized) in a particular usage in order to be regarded as a name. Kannada also shows a distinction between names and descriptions in the use of its 'derivational' affixes. For example, we can contrast the use of two different suffixes that occur in this language, namely ige (with four different shapes, namely, ike, ige, ke andge) and ike with only a single shape. Both are used for deriving nominals from verbs, but the former is used to derive common nouns (names) whereas the latter to derive descriptions. There are several interesting differences between the two, as pointed out below: (a) The allomorphy of ige is rather irregular. One can only establish certain general tendencies. For example, it is ige oige after disyllabic bases ending in u with a short first vowel, and ike after other u-ending bases (and also after any base ending in su). It isge after bases ending in i and ke after bases ending in e (the base-final vowel changes to α in the latter case). There are, however, exceptions to all these 'rules'. The use of ike, on the other hand, involves two very regular sandhi changes, namely the addition of ν after bases ending in u and ofyuv after bases ending/ or e. (b) The occurrence of ige is also irregular in the sense that it can be attached only to a small set of verbs such as he:Lu 'to say', koDu 'to give', bayasu 'to desire', suli 'to peel', moLe 'to sprout', etc., but not to several others like kuNi 'to dance', ma:Du 'to do', bare 'to write', tinnu 'to eat', etc. (i.e. its use requires institutionalization). The suffix ike, on the other hand, can be attached to any given verbal base. It can even be attached to bases that have been derived from Sanskrit borrowings through the addition of the suffix isu. (c) The meaning that gets associated with the nouns that are derived through the use of ige is also very irregular. It may refer to an action, fl object that has been derived from an action, an instrument used for carrying out an action, and so on. The suffix ike, on the other hand, has the function of referring to the action (or event) that the verb denotes in all its usages. Verbal base anju 'to fear' koDu 'to give' teru 'to pay tax' kettu 'to chip' ha:su 'to spread' moLe 'to sprout' hiDi 'to catch' holi 'to stitch'

Nominal with ike anjuv-ike 'fearing' koDuv-ike 'giving' teruv-ike 'paying tax' kettuv-ike 'chipping' ha:suv-ike 'spreading' moLeyuv-ike 'sprouting' hiDiyuv-ike 'catching' holiyuv-ike 'stitching'

Noun with ige anj-ike 'fear' koDu-ge 'gift' ter-ige 'tax' kett-ige 'sculpture' ha:su-ge 'mattress' moLa-ke 'sprout' hiDi-ke 'handle' holi-ge 'stitch'

Functional Constraints on Word-formation Rules · 37

As I have pointed out in Bhat (1979), there are several types of differences tttiat occur between names and descriptions, in addition to the above-mentioned dlistinction in their regularity vs. irregularity. We may note the following in this cconnection: (a) While coining a name, speakers of a language generally select a permanent, striking and classificatory characteristic as its basis. The formation of descriptions, on the other hand, does not involve any such constraints. One can produce a description for referring to an object with the help of any of the characteristics that are noticeable in it. (b) Names that occur in a language are finite in number (and can be listed in a dictionary) whereas descriptions are infinite. There is no need to list the latter in a dictionary as their meaning can be derived directly from the meaning of their constituent elements. (c) It is possible, in the case of names to specify the date on which it was introduced into the language and the individual who introduced it, whereas in the case of descriptions, no such specification is possible. The former is a historical event in the language whereas the latter is not. (d) The formation of a name is constrained by the 'name-worthiness' of the entity or category that is being associated with it rather than by the derivational process that underlies its formation (Downing 1977:841). There is no such constraint on the formation of descriptions. (e) The structuring of a name may involve 'creativity', the native speaker's ability to extend the language system in a motivated, but unpredictable (non-rule-governed) way (Bauer 1983) whereas that of a description involves rule-governed productivity. The former may even involve incompetent formations or 'mistakes'. (/) The meaning of a name is to be determined on the basis of the speaker's understanding of the referent that is conventionally associated with the name, whereas the meaning of a descriptions is to be derived directly from the meaning of its constituent elements (see Bhat 1979: 23; Stern 1931: 35). There is clearly a need to differentiate between the formal distinction between compounds and noun phrases on the one hand, and the functional distinction between names and descriptions on the other. Downing (1977) points out that several authors like Jespersen have relied on semantic criteria for distinguishing between compounds and noun phrases. They have emphasized the unitary meaning that is typically associated with a compound, or the fact that the meaning of the compound as a whole typically cannot be deduced from the meaning of its constituents. That is, these scholars have identified the term 'compound' with the functional category of 'name'. On the other hand, there are other linguists who use the term compound in order to indicate

38 · D.N.S. Bhat

constructions that differ from the corresponding noun phrases only in their compactness, but not necessarily in their meaning. The former have been troubled by the fact that even less compact expressions can function as names, and hence the question as to whether such expressions are to be regarded as compounds or 'words' rather than as noun phrases has been problematic to them. On the other hand, the latter have failed to make a proper use of the distinction between names and descriptions.

• 3. Compositionality Condition The occurrence of irregularity has been one of the most troublesome aspects of word-formation for linguistic theorists. Bauer (1983:294) refers to several cases 'which seem to suggest that word-formation, at least in some areas, may not be rule-governed at all'. For example, in his discussion of -nik suffixation of English, he suggests that a residue of unexplained forms might be due to 'incompetent formation'. In his examination of neo-classical compounds also, he finds several instances that do not appear to have been formed 'according to the usual rules, and for which there was no apparent generalization to be made'. He finds it necessary to regard all such cases as 'exceptions' to the general principle of regularity in word-formation or as part of the 'creative' process of word-formation rather than as part of its productive process. He wonders, however, whether 'too much is being swept under the carpet in this way'. That is, should we regard the process of word-formation as rule-governed or as in principle irregular? In spite of these doubts, Bauer, along with several other linguists, concedes Compositionality Condition, according to which the morphological representation of a complex word is to be considered as providing the formal structure on the basis of which the semantic interpretation may be composed in a simple manner. Botha (1984) gives two different arguments in defense of this condition. The first one is based upon the notion of economy. 'Suppose one were to give up the Compositionality Condition. It would then be a puzzling accident that, in general, the formal structure required by the rules composing the semantic interpretation of a complex word is the same as the one already embodied in the independently generated morphological representation of the word.' The second argument is based upon the claim that there is a need to postulate 'the fundamental principle of Compositionality' in order to account for the occurrence of infinitely many complex constructions. It appears to me that this problem of irregularity can be reduced to a great extent by separating names from descriptions, especially because irregularity occurs most prominently in the structure and meaning of nouns. Since only institutionalized formations are to be regarded as names, their number would be finite, and hence the second argument of Botha mentioned above would be ineffective in the case of names. Regarding the first argument, it can be

Functional Constraints on Word-formation Rules · 39

claimed, as I have suggested earlier, that in the case of names identification might involve two different strategies. Conventional meaning would be the primary basis for identification, but in addition to this, it would also be possible to use the meaning that one can obtain from the meaning of constituent elements as an aid in remembering the conventional meaning. Notice, however, that this latter device would be available only in the case of 'productive' names and not in the case 'non-productive' names or names that do not involve any internal structure as such.

• 4. Nominal and Verbal Compounding As I have mentioned earlier, the distinction between names and descriptions is restricted to the nominal category—a restriction that apparently derives from the fact that the distinction is concerned with the categorial function of nouns, namely reference (identification of persons, objects, etc.). As pointed out by DuBois (1980: 220), nouns denote entities that have continuous identity over time, whereas verbs denote ephemeral and unique entities. Reference to a new entity by a noun involves the opening of a 'cognitive file* for that entity; subsequent references to it add information to update that file. Verbs, on the other hand, do not establish any such entities on their own; rather, they provide information about an entity that has been established earlier by a noun. Successive verbs ordinarily do not refer back repeatedly to a single event, and are very different from nouns on this point. This functional difference between nouns and verbs is responsible, I think, for the fact that the above-mentioned distinction between two different types of expressions, called names and descriptions, is restricted to the nominal category. And it is this restriction that is responsible for the fact that we do not find word-formation rules that are meant for producing names to be occurring in the case of the verbal or adjectival category. Notice that the distinction between common names and proper names is also restricted to the nominal category. It does not occur in the case of verbs or adjectives. We do not have any 'proper verbs' or 'proper adjectives'. We can exemplify the absence of word-formation rules that are meant for deriving names in the case of the verbal and adjectival category by examining the notion of compounding that is in use in the grammars of different languages. Linguists generally use terms like derivation and compounding while describing the internal structure of words that belong to the category of nouns as well as the ones that belong to other categories like verbs and adjectives. No distinction is generally made between words of these different categories while using those terms. However, the actual processes that occur in the case of these two sets of words are quite different from one another, and the use of the same set of terminology (like compounding) in the case of both of them is rather misleading. For example, linguists talk of compound nouns as well as

40 · D.N.S. Bhat

compound verbs, but there are important differences in the processes that underlie these two types of structures. Some linguists prefer the term 'complex predicates' in the latter case, which apparently reflects the recognition of this difference. The point that needs to be emphasized here is that this difference derives from the functional distinction that occurs between the two (reference versus predication), and also from the fact that only nouns involve the notion of names. As I have mentioned earlier, meaning particularization (or institutionalization) is the hallmark of nominal compounds that are used as proper or common names (see also Meys 1975: xiii). When the speakers of a language desire to have a name for a given object or concept, they produce a compound (or some other type of derived word) by choosing one of the characteristics of that object as its basis. There would be other objects or concepts that would also be manifesting that characteristic, but the activity of naming has the effect of restricting its reference to the object or entity under consideration. In the case of scientific terminology, for example, we generally 'define' a word (a technical term) as having a specific connotation, whereas in our normal use of language, the connotation gets established as a result of its acceptance (in that particular connotation) by the community. This acceptance of a coinage by the society, or what is called its institutionalization, is of crucial importance in our use of words that belong to the nominal category. We do not find any comparable situation existing in the case of the verbal category. For example, compounding, in the case of verbs, is either indirect, or represents a process that is quite different from the one occurring in the case of nouns. In English, for example, verbal compounds are introduced into the language indirectly through nominal compounds. Most of them can be shown to derive from nominal compounds through backformation (Adams 1973; Bauer 1983): Nominal Compounds globe-trotter air-conditioner double-header ghost-writer house-hunting sleep-walking

Verbal Compounds (derived through back-formation) globe-trot air-condition double-head ghost-write house-hunt sleep-walk

The so-called compound verbs of Hindi, on the other hand, exemplify the second type of situation mentioned above. These are derived by joining a conjunctive participle (or a verbal root) with a vector verb. This vector verb, also called a light verb, belongs to a small class of verbs, such as/a: 'to go', le 'take', de 'give' Da:l 'put', ma:r 'hit', paR 'fall', etc. Only this vector verb loses its semantic content when used in this construction; the conjunctive participle

Functional Constraints on Word-formation Rules · 41

(or verbal root), also called host verb, retains all its original meanings. These compound verbs function like simple verbs, but their argument structure is more complex than that of either of their constituent verbs. Both the constituents appear to contribute to the argument structure of the compound verb (Bashir 1993; Butt 1997). (6) a. vo ro paR-a: he weep fall-Perf 'He burst into tears' b. us-ne ro Da:l-a: he-Agt weep put-Perf 'He wept copiously on purpose' Notice that the light verb Da:l 'put' occurring in 6b has lost much of its semantic content. It provides only the meaning of volitionality to the compound verb. However, it also affects the case marking of the agent. The crucial point to be noted here is that the use of these compound verbs of Hindi does not involve any institutionalization as such. In other Indo-Aryan languages like Marathi (Berntsen and Nimbkar 1982) compound verbs of the above type form part of a more productive and regular construction in which a conjunctive participle is joined with a finite verb in order to indicate two different events that are related in various ways. In this latter usage, both the verbs retain their meaning fully, but the deictic tense is controlled by the second verb. There are also certain (»referential constraints regarding the arguments of the two verbs that affect this participial structure. We may regard compound verbs as resulting from the grammaticalization of these periphrastic constructions in these languages. (7) mi tyala tikDe jaun bheTto I him there go (Part) meet Ί will go there and meet him' (8) to gh^ri yeun basla he home come (Part) sat 'He came home and sat down' (9) mi te sdgLz: phekun dih: I they all threw (Part) gave Ί threw it all away' (10) he kam hwkzr szmpwun Taka this work quickly finish (Part) put 'Finish up this work quickly' Notice that the combinations of a participle and a finite verb, occurring in 7 and 8, indicate two different events in each case, whereas in 9 and 10 they

42 · D.N.S. Bhat

denote a single event each. We have a complex sentence in the former case and a compound verb (or conjunct verb) in the latter case. The latter has apparently resulted from the grammaticalization of the former, but this clearly did not involve any change that is comparable to the institutionalization (or a 'coining' of a word) that forms the basis of nominal compounds. Similar developments have taken place in almost all the languages of India. In the case of Dravidian languages, for example, Steever (1993) points out that the compound verbs have resulted from a gradual process of grammaticalization of analytic constructions involving participles or infinitives. The bases that occur as light verbs in these constructions show a gradation concerning the amount of semantic bleaching that has affected them. The change was also not sudden as in the case of nominal compounds, and further, it has affected systematically the second constituent of the compound verb by making it a light verb. Both the constituents continue to have syntactic relevance. In Tamil, for example, the past participle of a host verb is joined with the finite form of one of a set of light verbs like viTu 'to leave', muTi 'to finish', a:ku 'to become', koL 'to take', etc., to form compound verb constructions. (11) viSam kuTittu cettam poison drink (past) died 'He took poison and died' (12) puttakattai paTittu muTitta:n book (Ace) read (Past) finished 'He finished reading the book' (13) puttakattai paTittu pa:rtta:n book (Ace) read (Past) saw 'He tried reading the book' Notice that the second verb in 11 clearly functions as a main verb, whereas in 12, and more clearly in 13, it functions as a light verb. Annamalai (1985) points out that the gradation that occurs among these light verbs concerning the amount of semantic bleaching that has been undergone by them is both vertical as well as horizontal. That is, the light verbs can be arranged into a gradation of main verbs, semi-verbs and auxiliaries, and further, the same verb may function as a main verb in some contexts and as a semi-verb or light verb in other contexts. In addition to this, there is the possibility of having an alternative interpretation for these complex structures in all their usages in which the light verb functions as an independent (main) verb. These facts of Tamil 'compound verbs' indicate clearly that the process that underlies their formation is very different from the one that underlies nominal compounds.

Functional Constraints on Word-formation Rules · 43

Another important difference between compound nouns and compound verbs is that grammaticalization in the latter case can lead to the change of a light verb into an inflectional affix. Several languages have been reported to show the development of aspect and mood affixes, transitive and causative affixes, reflexive, reciprocal and beneficiary affixes, and so on from periphrastic constructions of the above type. We do not generally find comparable developments taking place in the case nominal compounds apparently because institutionalization affects individual compounds (names) in the case of the nominal category whereas grammaticalization affects the whole paradigm (in which a light verb occurs) in the case of the verbal category. For example, Old Kannada had a periphrastic construction made up of a past participle of the main verb followed by the future forms of the verb a:gu 'to become' (tindapem Ί eat', ke:Ldapem Ί hear', bandapem Ί come'). In the modern non-coastal dialects of Kannada, this form (or rather the paradigm) has coalesced into a simple form that has the future subjunctive meaning. A peculiarity of this form is that it contains the past stem of the verb, even though it does not involve any past meaning. This peculiarity is clearly a remnant of its historical antecedent. tinnu baru ho:gu aLu

'to eat' 'to come' 'to go' 'to weep'

tind-e:nu band-e:nu ho:d-e:nu att-e:nu

Ί Ί Ί Ί

may eat' may come' may go' may cry'

Another change of the same type has taken place in Tblu. It has a reflexive form (paradigm) derived by adding the reflexive suffix oNu to the verb. In the case of two of the four classes of verbs, this suffix is attached to the past stem of the verb. koy su: bare

'to pluck' 'to see' 'to write'

kaDpu 'to cut' kuLLu 'to sit' tin 'to eat'

koyy-oNu su:v-oNu barey-oNu

'to pluck oneself 'to see oneself 'to write oneself

kaD~t-oDu 'to cut oneself ku-d-oDu 'to sit oneself tin-d-oNu 'to eat oneself

Diachronically, this form also appears to have developed from an earlier periphrastic construction made up of a past participle followed by the finite forms of the verb koL 'to take'. This diachronic antecedent can be exemplified with the help of the reflexive verbal forms of Kannada (no:Du 'to see': no:DikoLLu 'to see oneself, hoDe 'to beat': hoDedu-koLLu 'to beat oneself) which is clearly a periphrastic construction.

44 · D.N.S. Bhat

Some linguists also use the term 'compound verb' in order to denote certain noun + verb constructions. These are also different from nominal compounds in exactly the same way in which the verb + verb compounds mentioned above are different from them. For example, Mohanan (1997) describes the noun + verb constructions of Hindi, called 'complex predicates', as consisting of the noun as the host and the verb as a light verb. The noun and the verb jointly determine the number, meaning, and case of the arguments and the construction is mono-clausal. The nominal host, though contributing to the argument structure, can itself be an argument of the light verb; it may passivize, and the light verb may agree with it. However, the nominal host allows no modification. It also does not allow conjoining, or replacement by the gap of a relative clause. (14) a. ila: ~ne mohan-ka: Ila -Erg Mohan-Gen 'Ila insulted Mohan' b. ila: -ne Mohan-Id: Ila -Erg Mohan-Gen 'Ila praised Mohan'

apma:n kiya: insult (M) did (M) prasamsa: Id: praise (F) did(F)

(15) ni:na: -ne ra:m-ld: kaha:ni: par dhya:n diya: Nina -Erg Ram-Gen story Loc attention gave 'Nina paid attention to Ram's story' This kind of difference between nominal and verbal word-formation rules can be seen in the case of Tibeto-Burman languages as well. In Manipuri (Bhat and Ningomba 1997), for example, the formation of nouns involves primarily the process of compounding, whereas that of verbs involves either affixation (as in the case of causative, reflexive and beneficiary forms, or various types of deictic and spatial forms) or noun incorporation. Some grammarians have described this latter construction as a compound verb. The rules that underlie this latter construction, however, are quite different from the ones that occur in the case of nominal compounds. The incorporated noun, for example, can function as one of the arguments of the verb in which it has been incorporated. (16) synz ka-dz way sitli I-Nom room-Loc rubbish swept Ί swept rubbish in the room' The main purpose of noun incorporation in Manipuri is to disambiguate certain homophonous verbs. hat pa lam pok

'to kill' 'to read' 'to reach' 'to beget'

ssmhat mssa pa caklam mgypok

'to comb' (s&ncet 'comb') 'to be thin* (mosa 'body*) 'to be hungiy* (cak 'food') 'to burn' (may 'fire')

Functional Constraints on Word-formation Rules · 45 Notice, further, that the incorporated noun, even while functioning as an argument of the verb, is unlike other arguments in that its specification is obligatory, and further, it cannot take any modifiers on its own. (17) thaij QsimQ ya knife this tooth blunt 'This knife is blunt' (18) layrik aw mz mal tarjrji book this price costly 'This book is costly' The foregoing comparison of compound nouns and compound verbs indicates clearly that there is a basic difference in the word-formation rules that are employed in their formation, and further that this difference is derivable from the sentential functions for which the words so derived are to be employed.

• 5. Indian and Western Approaches to Word I have pointed out in Bhat (1984) that there is an important difference between the Indian and the Western approaches to the notion of word. The former deals primarily with two distinct entities, namely the nominal base and the verbal root, whereas the latter deals with a single, unified entity called Vord' (which is sub-divided into several distinct parts of speech). For example, the Indian philosophers and grammarians postulate two different sets of theories for dealing with the meaning of words, of which one is concerned with the meaning of nominal bases and the other one with that of verbal roots. Western theorists, on the other hand, have only a single theory of word meaning, which applies to nouns, verbs, adjectives, etc. In the case of nominal bases, for example, the dispute among the Indian philosophers and grammarians concerns the distinction between the 'individual' (yyakti) and the 'universal' (ja:ti). Some philosophers (like sa:mkhyas) regard nouns to be denoting an individual while others (like mi:ma:msakas) regard them to be denoting the universal. A third view of compromise is held by the logicians as well as grammarians, namely that the individual and the universal jointly form the meaning of nouns. In the case of verbal roots, on the other hand, the dispute concerns the distinction between the 'activity' (vya:pa:ra) and 'result' (phala}. The logicians of the old school consider the activity to be the meaning of the verbal root, whereas Mandana Mishra argues that verbal roots denote the result of an action; he considers the inflectional affixes to be denoting the activity. The grammarians, on the other hand, are of the opinion that both the activity as well as its result forms the meaning of verbal roots.

46 · D.N.S. Bhat This dichotomous view of word and its meaning also gets reflected in the Indian grammarians' treatment of word-formation rules. The three major types of word-formation processes, namely the use of primary suffixes (kRt\ secondary suffixes (taddhita) and compounding (sama:sa), are considered to be relevant for the derivation of nominal bases only but not that of verbs. Western grammarians of Sanskrit do describe certain verbal forms like sva:ha: karoti 'sacrifices by uttering sva:ha:\ svi:karoti 'accepts (makes one's own)', etc. as Verbal compounds' (see Whitney 1879) but Indian grammarians do not appear to regard them as compounds. Notice that Sanskrit does not have a distinct category of adjectives. It brings adjectival words under the category of nouns. Hence its grammarians need to recognize only two major word classes, namely nouns and verbs. In contrast to this, Western grammarians have apparently started with a complex classification of words into several distinct word classes (parts-ofspeech), and apparently because of this, they have not been able to recognize the crucial distinction between nouns and other words while establishing the characteristics of words.

• 6. Conclusion There are important differences between nouns, verbs and also adjectives in the kind of meanings that they provide, and also in the kind of word-formation rules that they get associated with. These differences derive from the fact that the three word classes have to carry out three distinct functions in language, and in order to do this they have to manifest distinct characteristics. The approach of the Western linguistic tradition to word as a unified entity has apparently had a masking effect upon this interesting aspect of the nature of words. • REFERENCES Adams, V. 1973. An introduction to Modem English word-formation. London: Longman. Annamalai, E. 1985. Dynamics of verbal extension in Tamil. IHvandrum: Dravidian Linguistics Association. Anderson, S.R. \992.A-morphous morphology. Cambridge: Cambridge University Press, Baker, M. 1985. The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16. 373-415. Basnir, E. 1993. Causal chains and compound verbs. Complex predicates in South Asian languages, ed. by M.K. Verma, 1-30. Delhi: Manohar. Bauer, L. 1983. English word·formation. Cambridge: Cambridge University Press. Beratsen, M. and J. Nimbkar. 1982. Marathi structural patterns. Book One. Pune: GranthaJi and Marathi Abhyas Parishad. Bhat, D.N.S. 1979. The referents of noun phrases. Pune: Deccan College. . 1984. Certain developmental tendencies of the Dravidian verb. International Journal of Dravidian Linguistics 5.249-58.

Functional Constraints on Word-formation Rules · 47 Bhat, D.N.S. 1985. Word and its meaning in the Indian linguistic tradition. International]ournai of Dravidian Linguistics 13 (1). 47-59. Bhat, D.N.S. and M.S. Ningomba. 1997. Manipuri grammar. Munich: Lincom Europa. Botha, R.P. 1984. Morphological mechanisms. Oxford: Perganon Press. Butt, M. 1997. Complex predicates in Urdu. Complex predicates, ed. by A. Aisina, J. Bresnan and P. Sells, 107-49. Stanford: Center for the Study of Language and Information. Downing, P. 1977. On the creation and use of English compound nouns. Language 53.810-42. DuBois, J.W. 1980. Beyond definiteness: the trace of identity in discourse. The Pear stories: cognitive, cultural and linguistic aspects of narrative production, ed. by W.L. Chafe, 203-74. New Jersey: Ablex Publishing Corporation. Gray, L.H. 1939. Foundations of language. New York: Macmillan. Hoffmann, F. 1903. Mundari grammar. Calcutta: Bengal Secretariat Press. Meys, WJ. 1975. Compound adjectives in English and the ideal speaker-listener. Amsterdam: NorthHolland. Mohanan, T. 1997. Multidimensionality of representation: NV complex predicates in Hindi. Complex predicates, ed. by A. Alsina, J. Bresnan and P. Sells, 431-71. Stanford: Center for the Study of Language and Information. Stecver, S.B. 1993. Analysis to synthesis. Oxford: Oxford University Press. Stern, G. 1931. Meaning and change of meaning. Bloomington: Indiana University Press. Whitney, W.D. 1879. Sanskrit grammar. Cambridge: Harvard University. Zimmer, K.E. 1971. Some general observations about nominal compounds. Working Papers on Language Universals 6.

Dravidian Hindi in South Africa: An Historical Variety ι RAJEND MESTHRIE ι This paper aims to characterize a particular variety of Hindi in Kwazulu—Natal, South Africa, once used quite commonly by people of south Indian descent when conversing with South Africans of north Indian origin.

• 1. Introduction This paper aims to characterize a particular variety of Hindi in Kwazulu— Natal, South Africa, once used quite commonly by people of south Indian descent when conversing with South Africans of north Indian origin.1 My approach is largely a socio-historical one. Why should one want to write about a second language variety spoken by a small community in a remote part of the universe as far as the Hindi-speaking world is concerned? The answers are manyfold. First, the sociolinguistics of second languages (L2s) in India and elsewhere tends to be neglected, though the social histories of L2s are often as compelling as those of first language (LI) varieties. Second, the intriguing morphology of a second language such as South African Dravidian Hindi raises many questions about relationships between the different varieties of Hindi on the Indian subcontinent, its bazaar forms as well as the general Dakkhini variety of the south. Finally, I suggest that the humble and little known Dravidian Hindi of South Africa had a significant bearing in formulations of Indian nationality and an Indian national language in preIndependence India. This paper thus deals with what I call 'the first diaspora': that is, rather than dealing with the movements of people from the South Asian subcontinent

50 · Rajend Mesthrie

which has brought large numbers of speakers of Punjabi, Gujarati, Bengali, Hindi, and others to contemporary UK and US, this paper focuses on the (largely) nineteenth century, semi-forced migrations of Indian workers to the European colonies (Natal, Trinidad, Guyana, Mauritius, Fiji, etc.). This may well not be the first Indian diaspora by any means since Indians had travelled to Indonesia, Malaysia, and the east coast of central and north Africa centuries earlier (see e.g. Chattopadhyaya 1970; Tinker 1974). Nor should the fairly large numbers of slaves taken from India to other parts of the world (e.g. to Mauritius in the eighteenth century [by the French] and to Cape Town in the seventeenth and eighteenth centuries [by the Dutch]) be forgotten. But from a sociolinguistic perspective, settled communities that still speak extraterritorial modern Indian languages (to varying degrees) date from the nineteenth century. It is worth stressing that for a long time one of the largest Indian communities outside south Asia was to be found in South Africa. This population totals about a million people today. 'First diaspora' communities tended to come froni a more circumscribed area—from the territory that is India today, rather than from Pakistan, Bangladesh or Sri Lanka. There is a moderately sized sociolinguistic literature of this 'first diaspora', chiefly dealing with processes of koineization (dialect adaptation and convergence) of the many forms of Hindi that were taken to Mauritius, Guyana, Suriname, Fiji, and South Africa. A good collection of essays reflecting this research tradition was edited by Barz and Siegel (1988). This literature also deals with external changes due to contact with the dominant colonial languages and the localized languages of the territories involved. For Natal, the territory with which this paper is concerned, my study of Bhojpuri-Hindi (Mesthrie 1985,1992a) was the impetus for further studies of Telugu (Prabhakaran 1991) and Gujarati (Desai 1992). In Mesthrie (1985,1992a) I focused on the linguistic practices of people from the Bhojpuri-Hindi community. With hindsight, other users of Hindi in Natal might have provided an interesting counterpoint to that research. Sadly—from the point of view of at least historical research—this mixed bag of Other tongue' speakers of Hindi in South Africa is an ever-shrinking one: English people with colonial experience in India, Gujarati traders, Dravidian speakers, and a few Zulu speakers.

• 2. Dravidian Hindi in South Africa Although Hindi was more prominent as an L2 than LI in Africa (outside Natal and Mauritius) little has been written about the features of this lingua franca. Desai (1992: 85-86) refers in passing to the Hindi of Gujarati speakers in Durban, Natal as a pidgin, though this view remains to be substantiated. This claim has also been made for Bombay Bazaar Hindi (Apte 1974). However, my impression is that the two varieties (the Hindi of Gujarati traders in South Africa and Bazaar Hindi in Bombay) are interlanguages rather than pidgins.

Dravidian Hindi in South Africa m 51

TThis paper presents a brief analysis of the characteristics of the L2 variety of Hindi spoken by some older Tamil and Telugu speakers in Kwazulu-Natal. Thereafter it compares the features of this variety with the one genuine Hindi pidgin known to exist outside India—Fijian Dravidian Pidgin Hindi. The variety that I will for convenience call South African Dravidian Hindi ((SADH), has a long history—as suggested by M.K. Gandhi's frequent allusion tto the learning of Hindi by south Indians in nineteenth-century Natal. For example, in Satyagraha in South Africa (1928:147) he characterizes the competence of one of his ardent followers, Thambi Naidoo as follows: Thambi Naidoo was a Tamilian born in Mauritius where his parents had migrated from Madras state. He was an ordinary trader. He had practically received no scholastic education whatever. But a wide experience had been his school-master. He spoke and wrote English very well, although his grammar was perhaps not free from faults. In the same way he had acquired a knowledge of Tamil. He understood and spoke Hindustani fairly well and he had some knowledge of Telugu too, though he did not know the alphabets of these languages. Again, he had a very good knowledge of the Creole dialect current in Mauritius which is a sort of corrupt French, and he knew of course the language of the Negroes. A working knowledge of so many languages was not a rare accomplishment among the Indians of South Africa, hundreds of whom could claim a general acquaintance with all these languages.2 In his later attempts at formulating a language policy for India, Gandhi remembered his South African experiences. When proposing Hindi as a language of unity he specifically cited the South African case: little does anyone know that almost all Tamils and the Telugus in South Africa can carry on an intelligent conversation in Hindi' (YoungIndia 16 June 1920). I have suggested elsewhere (Mesthrie 1993) that Gandhi's thinking on a variety of linguistic topics, no less than his politics, was, in fact, shaped in South Africa. The roots of SADH appear to be multiple, if one is to judge from its presentday characteristics as well as the contexts in which it was used in the nineteenth and early twentieth centuries. There was a large measure of multilingualism in the plantation period when indentured workers were housed in barracks in many parts of Natal. Some Bhojpuri speakers learned Tamil; but more Tkmil and Telugu speakers learnt a variety of Hindi, which proved more viable as a language known outside the Bhojpuri community. It enabled communication with Gujarati traders, with Urdu-speaking Muslims, and with some Europeans. The latter included some Franco-Mauritian planters who were among the first to establish plantations in Natal, missionaries with experience in north India, and British soldiers previously stationed in India.3 This Hindustani lingua franca appears to have been different from vernacular South African Bhojpuri; today's SADH might well be a confluence of the two varieties. However, SADH never rose to great prominence because a rival lingua franca,

52 · Rajend Mesthrie

which had developed prior to the arrival of Indians, proved more useful for communicating not only with the Zulu majority of the colony but in many instances with plantation bosses and other European settlers (see Mesthrie 1989; 1992b: 23-24). This pidgin, Fanakalo, based on English, Zulu and some Afrikaans was sometimes pressed into service amongst some Indians when they had no other common language. English, which was initially rare among Indians, soon developed into a prestige link language amongst educated young Indians (see further Mesthrie 1992b: 26). For much of the twentieth century, there was accordingly a pecking order of lingua francas amongst Indian South Africans: English amongst the educated; failing that, Hindi or— less commonly—Tamil; failing that, Fanakalo pidgin. Today some Tamil and Telugu speakers still speak a fluent variety of SADH. Typically, these are old female speakers who did not have access to English in their youth (but who may have much later acquired some English, on account of the influence of their grandchildren). My characterization of SADH is built on four interviews (one in 1982, two in 1993, and one in 1998). Brief information concerning the four speakers follows: Mrs F.R. was 57 years old at the time of the interview in 1982. She is a Tamilspeaking Catholic living in a northern Natal city, married to a Bhojpurispeaking Hindu. She had learnt Hindi from her neighbors and friends from a young age, in addition to her Tamil, English, Fanakalo and Afrikaans. Whilst her speech is morphosyntactically close to that of the Uplands form of South African Bhojpuri (see Mesthrie 1992a: 40-55), her phonology is a distinctly Dravidian-influenced one. Mrs T.M. was about 80 years old when interviewed in 1993, living in a rural part of the south coast of Natal. She too learnt Hindi from a young age and is fluent in Tamil (her first language) and Fanakalo; she knew no English until the recent arrival of several great-grandchildren. Her speech is a focused, fluent form of SADH that differs in many respects from that of her Bhojpuri neighbors. Mrs N.R. was 76 when interviewed in 1993. She grew up with Telugu as her first language, but speaks more Tamil than Telugu today (having married into a Tamil family). She learnt Hindi as a child; she also speaks Fanakalo and has recently acquired a little speaking knowledge of English. While not struggling at all during the interview, she seemed on the whole slightly less proficient in Hindi than Mrs T.N. Mrs R.N. was about 90 when interviewed in 1998. She is a Tamil speaker who learnt Hindi as a child whilst playing with her Bhojpuri-Hindi neighbors. Though she had suffered from a stroke, her knowledge of Tamil, Hindi and Fanakalo remained, though she lost her erstwhile rudimentary command of English. The four speakers were thus all born between 1910 and 1925, a period that saw the end of the era of indentured immigration (in 1911) and of the earl)

Dravidian Hindi in South Africa · 53

phase of plantation life in South Africa. They were thus placed in a position where they must have been permeable to the influence of varieties of Hindi from south India and Bombay, as well as Bhojpuri and closely related forms of north Indian Hindi in South Africa. Presumably, distinctions of region and Language were still important during this period, so that whilst a form of Hindi was being learnt as a lingua franca by south Indians, it had to be distinct from the norms of the Bhojpuri community in South Africa.

I 3. Characteristics of SADH l 3.1. Salient Features Shared with South African Bhojpuri (SABh) • Honä as desiderative verb: Ilög ke sädl nei honä they GEN marriage not want 'They don't want to get married' (N.R.) M

**S£

·**·

t>t*M«

rM*»

rftVTM«

Honä thus differs greatly from its north Indian counterpart hona 'to be*. Its likely origins are Dakkhini Hindi-Urdu and Bazaar Hindi of Bombay (Mesthrie 1992a). Whether it was adopted independently by SADH and Bhojpuri speakers via contact with speakers from the south of India and with Gujarati traders, or whether it first entered Bhojpuri and then SADH or vice versa can only be speculated upon. Working out the algebra of this borrowing would cast crucial light on the history of SADH, in particular, the issue of its Asian or African origins. • Lexical items mundl/mür 'head', bagal 'side, direction', etc. As a Dakkhini and Bombay Bazaar Hindi form mundi presents the same challenges as honä described above. The next four features shared by the two varieties are less problematic, insofar as they can be confidently assigned to north Indian antecedents of South African Bhojpuri. • Log as periphrastic plural for [+ human] nouns: I log 'these people, they'; ü log 'those people, they'. • Ham as 1st person, sg. pronoun T; hamär as genitive sg. possessive pronoun 'my'. • Ke (not kär) for conjunctive constructions: Barä business rakh ke rätä wähä. big keep CONJ stay.pres prog, there 'They've established a big business and remained there' (N.R) • Phono-lexical forms du 'two'; sob 'all' which are typical of eastern Bhojpuri of India.

54 · Rajend Mesthrie

• Absence of grammatical gender for nouns and adjectives; and of number for verbs. Such simplification is common in all the transplanted varieties of Hindi, but not in Indian varieties of Bhojpuri or closely related varieties. It is not possible to prove whether SADH follows the simplificatory trend on account of its Dakkhini antecedents or because of the influence of South African Bhojpuri. • Loanwords slk 'sick', rum 'room', etc. These are likely to be South African neologisms. • 3.2 Differences between SABh and SADH Despite the above similarities it is doubtful if SADH can be considered a form of Bhojpuri, since differences between the two varieties are quite striking. The chief differences in verb morphology, pronouns, and phonology are outlined next. 3.3.1. Verb Morphology Table 1 describes the differences between the morphology of the two SADH speakers and their Bhojpuri neighbors.4 Ibble 1: A Comparison of Some Copular Verbs and Verb Endings in SABh and SADH Verbform Copula Present

Past

Present habitual

Present progressive

Future

Past

Coastal SABh

SADH

StdHn

1. hailhe 2. hawe 3. he 1. rahati 2. rahale 3. rahäl 1. dekhi-la 2. dekhehe 3. dekhehe 1. dekh-alhai 2. dekh-athawe 3. dekahat he 1. dekh-ab 2. dekh-be 3. dekhi 1. dekh-ll 2. dekh-le 3. dekh-lasllak

rätä rätä rätä rätä/reiyä rätälreiyä rätä/reiyä dekh-tä dekh-tä dekh-tä dekh-tä dekh-tä dekh-tä dekh-egä dekh-egä dekh-egä dekh-ä dekh-ä dekh-ä

hü hai hai thä the the dekh-tä hü dekhtähe dekhtähe dekh rahä hü dekh rahä hai dekh rahä hai dekhügä dekh-egä dekh-egä dekh-ä dekh-ä dekh-ä

SADH thus shows tense syncretism in relation to Bhojpuri paradigms. The present habitual and progressive is denoted by the invariant -tä ending in

Dravidian Hindi in South Africa m 55

SADH, which may even be used for the present perfect. Use of the copula hai is rare in SADH; it may occasionally occur in response to an LI SABh speaker's utterance with hai. The usual form is the distinct ratä, which derives from Hindi rahatä. It may alternate with zero copula for SADH speakers, at a significantly higher level than zero copulae in SABh speech. For one speaker (Mrs N.R.) ratä alternates in the past with reiyä, which I presume to be based on rähä. Only Mrs R.N. used the form rahä as past copula.

3.3.2. Pronominal Differences Again the characteristic forms of SADH are of a general 'Hindustani' lingua franca type, rather than the Bhojpurian type. Table 2 gives some of the similarities and differences in pronoun usage among SABh, SADH and Std Hn. Tfcble 2: A Comparison of Pronominal Forms in SABh, SADH, and Std Hn5 SABH

SADH

StdHn

neun

ham

tu

mai

tu/tum ham log

ham

ham log

1 ü Hog

hamke toke

eke oke

ekar okar ilögke erne

ome kekar

i ü

lldg hamke toke iske uske iske uske ilogke isme usme kiske

tu/tum yah/ye vah/ve

ve

mujhe tujhe iselisko uselusko iskA uskA

mkA

isme usme kiskA

Gloss 'you' (neutral) 'we' 'he/she/it' (proximal) 'he/she/it' (distal) 'they' (proximal) 'me' (dative) 'you' (dative) 'hinVher/it' (proximal) 'him/her/it' (distal dative) 'his/her/its' (proximal) 'his/her/its' (distal) 'their' (proximal) 'in this' 'in that' *whose'

3.3.3. Phonology The following influences from Dravidian languages (more so, Tamil) occur in the data: •

-dropping: Forms like am for ham T, usiar for husiar 'clever' occur quite frequently, together with occasional pronunciations with voiced (rather than breathy voiced) h in ham, hamke, hona, etc. Note that the copula form rätä has loss of the entire medial syllable containing [h]. • Variable aspiration and murmur: Forms like gar for ghar 'house* and cokrä for chokrä 'boy* and pin forphin 'also' are quite frequent in the data. • Loss of final nasals: Occasionally forms like mälö for malum 'know', äpö for 'self (reflexive) occur.

56 · Rajend Mesthrie

• Initial devoicing of stops: The forms poltä for boltä 'says'; tu for du 'two* occurred.6 The process isnot widespread however. There were occasional 'one off' hypercorrective forms like bhustak forpustak 'book' andphäni forpani 'water'. • Voicing of medial stops: Occasionally there was the opposite process the voicing of/k/ especially in phrases like ayi ke 'having come'. This was also rare. • Changes to initial /c/: This was changed occasionally to [J] or [s], as in saltä for caltä 'moves' and sot for cot. • Initial euphonic lei: On a few occasions there were pronunciations like ye/: for ek One'. (As occurrences of initial /o/ were few, it could not be ascertained whether there is a similar tendency for euphonic [wo].)

• 4. A Comparison with Fijian Pidgin Hindi It should be stressed that apart from -dropping and de-aspiration, the phenomena described here were not widespread; nor should they be taken as denigration of what is a clear form of L2 communication. That we are dealing with a variety that falls into the 'second language' rather than 'pidgin' end of a language learning continuum can be seen from comparisons with the one form of pidginized overseas Hindi that developed in Fiji, as described by Siegel (1987: 173-83). Siegel distinguishes between several varieties of pidgin Hindi in Fiji: that of native Fijians, former European overseers and south Indians in former times. Note that it is not the case that all south Indians in Fiji speak a pidgin form of Hindi. According to Siegel the general form of Hindi on the island (Fiji Hindi) has been adopted as a lingua franca amongst almost all people of Indian descent. The linguistic situation in Fiji is, therefore, quite unlike that of the other territories to which indentured Indians immigrated, where the lingua francas adopted by Indians were either the colonial language or a Creole form thereof. Fijian Dravidian Pidgin Hindi (FDPH) was used by older south Indians who were not born on the island, and who consequently did not have much exposure to Fiji Hindi. This pidgin will form a useful counterpoint to my description of SADH. The broad phonological characteristics of both varieties, FDPH and SADH, are quite similar, though both Siegel's and my descriptions would need to be more finely-grained and quantitative before detailed comparisons can be made. I shall instead list the salient morpho-syntactic features of FDPH given by Siegel (1987: 178-79) and consider their relation to SADH forms. • The generalized use of originally imperative endings in for indicative tenses reported for FDPH does not occur in SADH. • The use ofbaitho 'to sit' as a copula in FDPH does not occur in SADH. Although the copula rätä (literally 'to stay') of SADH might seem a clear

Dravidian Hindi in South Africa · 57

analog, it has a greater basis in the Bhojpuri input—the (suppletive) past tense form of the copula in SABh being rah-. The use oikhalas as a completive marker in FDPH is partially parallel in SADH (and SABh). Khalas, which originates from Calcutta Bazaar Hindi and Bombay Bazaar Hindi, is used as a phrasal verb in SABh and SADH ('to finish'), rather than the all-purpose perfective element of FDPH. The SVO order, common in FDPH, is not common in SADH. SADH, like SABh is verb-final. The use ofsako as a full verb in FDPH, as in Ham sako kam can do the work', has no parallel in SADH, where sak is a post-verbal modal. The use of wälä to mark adjectives in FDPH has no counterpart in SADH, where all tokens of wälä were attached to a noun, with agentive, descriptive or relative function.

• 5. On the Origins of SADH The origins of SADH can only be conjectured, pending further descriptions of Dravidian Hindi in India and elsewhere. Is it a variety with direct links with nineteenth century India? If so, does it have roots in varieties from India of the sort described by Tinker (1974:52): 'The Biharis were steady and patient and accustomed to hard toil ('docile' in the words of the planters). Also, they mostly spoke one of the dialects of Hindi; and Hindustani was becoming the lingua franca of the emigration traffic.' Was such a lingua franca of the north Indian 'traffic' passed on to south Indians in Kwazulu-Natal? Another line of investigation might disregard the north Indian connection and enquire instead whether SADH has more significant ties with Dakhini Hindi-Urdu of Hyderabad and Bombay? Despite the similarities suggested by the three very salient lexical items, khaläs 'finished', mundl 'head' and honä 'to want', SADH of today does not seem close to the three descriptions of Dakhini available to me—Schmidt (1981), Mohiddin (1980), Grierson (1916: 188-213). On the other hand, SADH is definitely closer to what might be considered Bhojpuri-Urdu in South Africa, rather than Bhojpuri-Hindi.7 Briefly, almost all the features characterizing SADH in Tables 1 and 2 can be found in South African Bhojpuri-Urdu. A fair number of Urdu-speaking Muslims of nineteenth-century South Africa originated from Hyderabad, and would have been proficient in Dakhini as well as a Dravidian language like Tamil or Telugu. Was the Urdu of Muslims emanating from Hyderabad a more ready form of input to early SADH speakers? In contrast to such a continental Asian theory of origins, one could equally ask whether the variety described in this paper is a distinctly South African creation, with reasonably close links to the nineteenth-century lingua franca that Gandhi commented on and so admired in south Indians of Natal? (Two of my informants were born on or before 1913, the year Gandhi left South

58 · Rajend Mesthrie

Africa.) Finally, one needs to investigate whether there is any relation between the more salient features of SADH and the foreigner-talk registers of Bhojpuri speakers. In my interviews, the Bhojpuri speakers used the salient morpho-syntactic forms of SADH (except rata), even before the SADH speakers did.8 Clearly, they had a notion of an appropriate register for talk with Outsiders' and/or *L2 speakers'. Is their foreigner-talk a distinctly South African one formed in response to SADH speech, or does it, too, have its roots in the Hindustani lingua franca of the 'emigration traffic'? Much archival work will be needed to uncover further data and historical information needed to decide the extent to which each of these influences applied. • NOTES 1. I would like to dedicate this paper to Yamuna Kachru, on the occasion of her recent retirement. Regarding terminology in this paper: Hindi is used as a cover term for a number of varieties in South Africa, India and elsewhere. The term 'Bhojpuri' is used to draw a special distinction between one community within South Africa (who actually use the term 'Hindi' or 'Kalkatya bat' for their language) and speakers of other languages or other varieties of (L2) Hindi. The territory known as Natal in colonial times is now known as Kwazulu-Natal province of South Africa. 2. These astute and sociolinguistically enlightened remarks need to be modified in just one respect: it is unlikely that Thambi Naidoo could have learned Zulu. The reference to the 'language of the Negroes' must be taken to refer to FanakaJo pidgin. 3. Some missionaries with experience in south India had a knowledge of Tamil. 4. There is by no means a perfect fit between the SABh and the Standard Hindi categories in the respective columns. Whereas the primary meaning of the second row is 'present habitual* in SABh, the Std Hn paradigm has the primary semantics of 'imperfect*. The comparison is further clouded in the past tense by the ergative morpho-syntax of Std Hn as against nominativeaccusative of Bhojpuri. I hope I will be forgiven for choosing the masculine forms of the Hindi verbs as exemplars of the paradigm, simply for reasons of space. 5. The symbol Ά* in the table denotes that the form is inflected for gender and number. 6. Note that as the [t] was dental and not alveolar, the speaker is using the Bhojpuii-Hindi numeral, and not the English form. 7. South African Urdu has not been described structurally, though there is a good macrosociolinguistic treatment of its status as a minority language in South Africa. 8. I had enlisted the help of several Bhojpuri speakers in making contact with the elderly SADH speakers and had encouraged them to initiate the conversations and participate as much as possible. • REFERENCES Apte, M.L. 1974. Pidginization of a lingua franca: A linguistic analysis of Hindi-Urdu spoken in Bombay. International Journal ofDravidian Linguistics 3.21-41. Ban, R.K. and J. Siegel. 1988 (eds.) Language Transported—The growth of Overseas Hindi. Wiesbaden: Otto Harrassowitz. Chattopadhyaya, H. 1970. Indians in Africa—A socioeconomic survey. Calcutta: Bookland. Desai, U.K. 1992. The Gujarati language amongst Gujarati-speaking Hindus in Natal. University of Durban-Westville: Master's thesis.

Dravidian Hindi in South Africa m 59 Gandhi, MLK. 1928. Satyagraha In South Africa. Ahmedabad: Nayjivan. Grierson, GA. 1916. Indo-Aryan family, central group, Parti: Western Hindi and Panjabi (Vol. IX of Linguistic Survey of India). Calcutta: Government of India. (Reprint: Delhi: Motilal Banarsidass, 1961.) Mesthrie, R. 1985. A history of the Bhojpuri (or 'Hindi') language in South Africa. University of Cape Town: Ph.D. thesis. (Revised as Mesthrie 1992a.) . 1989. The origins of Fanakalo. Journal of Pidgin and Creole Languages 4 (2). 211-40. . 1992a. Language in indenture—A socioUnguistic history of Bhojpuri-Hindi in South Africa. London: Routledge. . 1992b English in language shift—The history, structure and sociolinguistics of South African Indian English. Cambridge: Cambridge University Press. . 1993. Gandhi and language politics. Bua 8 (4). 4-7. Mohiddin, S.K. 1980. Dakhani Urdu. Annamalainagan Annamalai University. Prabhakaran, V. 1991. The Telugu language and its influence on the cultural lives of the Hindu 'Pravasandhras' in South Africa. University of Durban-Westville: Ph.D. thesis. Schmidt, R.L· 1981. Dakhini Urdu—History and structure. New Delhi: Bahri Publications. Siegel, J. 1987. Language contact in a plantation environment: A socioUnguistic history of Fiji. Cambridge: Cambridge University Press. Tinker, H. 1974. A New System of Slavery—The export of Indian labour overseas, 1830-1920. Oxford: Oxford University Press.

Bilingual Proficiency and Code-switching/Mixing Patterns* • RAJENDRA SINGH and AD BACKUS · Purism sees foreign words more clearly than does a lax defense of them. — Theodor Adorno (1992:287)

The purpose of this paper is to explore the possibility that despite some empirical and conceptual difficulties associated with the determination of bilingual norms (cf. Backus 1993), codeswitching/mixing patterns can and should be considered as indices of bilingual proficiency.

• 1. Introduction In a paper read at the Max Planck Institute in 1992, Singh attempted to make sense of his code-switching/mixing findings in north India and presented his informal observations about the facts of the matter in Montreal against the findings presented at the European Science Foundation Symposia on the subject (cf. Singh 1995c). These latter findings seemed to him to create a set of contradictions or paradoxes that had to be resolved. Although we shall not summarize that paper here, it may be worthwhile to restate some of its conclusions, as they set the agenda for this article: (1) a. Perhaps it is an error to assume bilingualism and then study patterns of code-switching/mixing. We should study these patterns to find out if we can, in fact, speak of bilingualism. * This paper is related to and builds on Singh (1995c). We are grateful to R.K. Agnihotri, Rik Boeschoten, and Pieter Muysken for listening, and to Stig Eliasson for saving us from overstating our doubts about Grasshopper violations of the Free Morpheme Constraint.

62 · Rajendra Singh and Ad Backus

b. Whereas perfect bilinguals switch when they want to, other 'bilinguals' switch when they have to. c. Patterns of code-switching/mixing constitute an excellent window on the degree of "bilingual competence'. d. Some cases, like that of English in north India, require making a distinction between indiginizational and non-indiginizational contexts. e. It is possible to correlate patterns of switching/mixing with degrees of bilingual competence. f. The correlation Singh proposed was not entirely consistent with the one proposed or implied by Poplack (1980) or Nortier (1990), though it found support in Backus (1993) and Bentahila and Davies (1991; 1992). In recent years, a number of studies have appeared in which differences in code-switching (CS) styles between various groups within the same community were reported. In all of these, the differences correlated well with variation in bilingual proficiency, which in turn correlated well with the generation to which a particular informant belonged. These studies include Jacobson (1989) and Torres (1989) on Spanish-English CS in the US, Bentahila and Davies (1992,1995) on CS between Moroccan Arabic and French in Morocco, Li Wei and Milroy (1995) on Chinese-English CS in Newcastle-upon-Tyne, UK, and Nishimura (1995) on Canadian Japanese. Similar patterns have been mentioned in passing by earlier researchers, for example Di Pietro (1978: 276), or Dabfcne and Billiez (1986:316). On the other hand, some older studies, most notably Berk-Seligson (1986) on Judeo Spanish-Hebrew CS in Jerusalem, which also included data from different generations, found no variation in CS patterns. All the studies that found variation found it was of a similar kind. Speakers who are dominant in one of the languages mainly engage in insertional CS with that language as the Matrix Language (ML). The generation with better command of the original Embedded Language (EL), a command that is more or less equal to their command of the original ML, tends to produce intersentential CS more than anything else. Jacobson (1989) compared Spanish-English CS data from people of differing age and social background in San Antonio, Texas. With due reservations, he concludes (Jacobson 1989: 122-23) that speakers can be categorized by their preferred bilingual mode of speaking. Either Spanish or English may function as the ML, or in other cases, there is no such asymmetry. He did not find, however, any clear overall preferences for either insertional or alternational CS. The general picture which emerged is that male speakers mainly produced English-dominant speech, with CS generally more symmetrical in higher social classes. Women tended to switch more asymmetrically, with younger women predominantly using English as the ML and older women

Bilingual Proficiency and Code-switching/Mixing Patterns m 63

preferring Spanish-based speech. All these differences could be related to the distribution of language proficiencies, since in general men speak English more often than women, and younger people speak it more often than older people. Bentahila and Davies (1995) found considerable differences in CS styles between their older and younger informants. In general, older informants in Morocco speak more French, since it has played a more important role in their lives. Younger informants know French mainly as one of the languages they use in school. As for CS, in which older informants switch more alternationally, with fairly equal roles for the two languages, while younger speakers use more asymmetrical CS. Moroccan Arabic functions as the ML and there are many French insertions. Remarkable flexibility is reported by Nishimura (1995) for second generation Japanese Canadians, who use the appropriate style of CS that goes with the specific interlocutor they are talking to. The styles they can choose from are basically monolingual Japanese with English cultural borrowings (for Japanese-dominant interlocutors), a style characterized by frequent alternational CS (for bilingual interlocutors who do use a lot of Japanese, for example, because they were sent to Japan for schooling), and basically monolingual English with Japanese interjections and other ethnicity markers (for interlocutors of their own group). Similar gradations were also found by Li Wei and Milroy (1995) for a Chinese community in Britain. The main exception to this general pattern of differences between the generations is the Judeo-Spanish-Hebrew speaking community studied by Berk-Seligson (1986). She also has informants from different generations, and moreover, her informants also differ in proficiency, but they all produce roughly the same kind of bilingual Speech. Community norms are apparently more strongly focused in Jerusalem than in the communities studied by the other authors mentioned earlier. All of these studies answered Poplack's (1980) call for empirical research into the correlation between bilingual proficiency and CS patterns. None of them completely affirmed her predictions, which were themselves generalizations made on the basis of the New York Puerto Rican data, the details of which we will get to later on. The troublemaker was usually the relatively high incidence of intersentential switching in the speech of the more proficient bilinguals, which ran counter to Poplack's conclusion that intrasentential mixing requires most bilingual competence. A more detailed description of the Dutch Turkish data studied by Backus will provide us with a more concrete empirical basis.

• 2. Dutch Turkish Data The data reported on in Backus (1996) consists of the transcripts of seven recordings of spontaneous conversations among bilingual peers, all members

64 · Rajendra Singh and Ad Backus

of the Turkish community in Holland. All speakers that took part in one recording belonged to the same social network, so speakers always knew each other well. Four conversations included only first or only second generation speakers; three others featured speakers of two or three generations, including of what will be called the intermediate generation (cf. Tosi 1984). First generation speakers immigrated sometime after their 12th year, intermediate generation speakers sometime during their elementary school years, and second generation speakers were born in Holland, or were at least brought there before they went to school. In total, 27 informants, both men and women, between 16 and 30 years of age, contributed to the study. The informants were asked to speak as they always do, and participant observation confirmed that they did. The most important aspect for our present purposes is the intergenerational type of variation that was found. The different generations within the immigrant community use their bilingual repertoire in different ways, along the same lines as in the cases of variation in immigrant and other bilingual settings reported on earlier. With these weapons in hand, we will then return to the proficiency issue, raised in Section 1. In in-group spontaneous conversations at least three different styles of CS are used. These three different bilingual lects are spoken by speakers from the three generations that were distinguished. The cut-off points between these were of course not arbitrarily chosen (they concur with the configuration of speakers that speak the three lects), but they are nevertheless simplifications: they impose a trichotomy on what is essentially a continuum with three clearly recognizable prototypes as its regions. Differences seem to reside especially in the relative distribution of different kinds of CS. To what extent they correlate with other kinds of contact phenomena, such as structural change and use of loan translations, has not been analyzed yet. Several means of quantification could be used to compare the relative distribution of Dutch and Turkish in the recordings. Here, we have used the roughest possible type. Table 1 gives, per generation, the number of speech turns taken by the various informants of that generation, and the percentage of those which are all Turkish, all Dutch or mixed. This is an oversimplification, but we believe these figures will suffice here, as they certainly illustrate the general picture. Figures for the number of Turkish, Dutch and mixed Ibble 1: Generational Distribution of Speech lurns

Turkish Dutch Mixed Total

First

Intermediate

Second

503 (79%) 31 (5%) 106 (16%) 640 (100%)

46 (22%) 64(31%) 93 (46%) 205 (100%)

175 (16%) 684 (61%) 258 (23%) 1117(100%)

Bilingual Proficiency and Code-switching/Mixing Patterns m 65 utterances, for example, would yield comparable distributions. Further details can be found in Backus (1996). • 2.1. First Generation Speech In the first generation vernacular, the role of Dutch is very limited. Only a handful of content words, and 75 per cent of them proper nouns, as in 2a, appear in what is pretty much monolingual Turkish. Furthermore, if the Dutch element is not a name of some sort, it is almost always a cultural borrowing. All of these are inserted into the standard mixed constituents that are so familiar from CS studies around the world (cf. Myers-Scotton 1993). Since virtually all inserted Dutch words are nouns, most mixed constituents are NPs (cf. 2b). A few minor patterns appear which will turn out to be more frequent in later generations. For example, Dutch reported speech occurs, embedded with a TUrkish verb of saying. The reported chunks are usually no longer than one or two words, as in 2c. (2) a. Ayhan, biz Ajaxli dogduk! A. we Ajax-ADJ born-lpl Ayhan, we were born as Ajax fans!' NIB: Ajax is a Dutch soccer team. b. evet, terras-ta oturuyorlar yes cafe-LOC sit-PROG-3pl 'yes, they are sitting at the outdoor cafe' c. welk diyor which say-PROG-3sg 'he said, ' • 2.2. Intermediate Generation Speech The second type of bilingual lect, associated with the intermediate generation, includes all the kinds of intrasentential CS generally encountered in the literature. For the Turkish-Dutch context, this lect was described earlier in Backus (1992). The examples in 3 briefly illustrate the range of mixing structures this generation engages in. Overall, CS is very frequent. Table 1 shows that there are almost as many speech turns with switches as without. However, and this is an important point to which we will get back, modifications are made to accommodate to specific interlocutors. There are equal shares of inter- and intrasentential switching. The latter is asymmetrical: Turkish consistently functions as the matrix language, while Dutch contributes content words as veil as longer constituents. Within Dutch stretches on the other hand, there is 10 intrasentential CS.

66 · Rajendra Singh and Ad Backus

(3) a. cogu student-far boyle ya, ooh, mesela bu sene afstuderen yap-arsa iyi baan-é var (Backus 1992) most student-pi such INT oh e.g., this year graduate do-AORCOND.3sg good job-POSS there.is 'for most students it's like this, you know, if they graduate this year for example, they'll have a good/06' b. iyi o zaman adam-dan roddelen yap-ar (Backus 1992) good then man-ABL gossip do-AOR.3sg 'there's a lot of good gossip about him then' c. ondan sonra lauw water -nan yika-yinca ... (Backus 1992) then after lukewarm water -with wash-while 'and then, when you wash it with lukewarm water' d. bir sene beraber tam?iyorduk ya, o-nun vriendin-i van mijn begeleider, die had een vriend in Turkije. her-GEN friend-POSS of my supervisor One year we met each other' [she said], a friend of my supervisor, she had a boyfriend in Turkey9 e. Bana bir$eyler soruyorlar: 'biz bunu fotograf $ekmek ßòßç dort ay met de begeleiding-le yap-tik with the supervision-with did-lpl 'they asked me a few things: "we worked with supervision for four months in order to make X-rays'" f. ja, ondan sonra bana s yledi ki:/i moet Nederlands gaan volgen. yeah, and then she said to me: you've got to go to Dutch class"9 g. Ja, birde onlar orda veya vast ofze werken langer yes suddenly they there either permanent or they work longer "yeah, and then they have a permanent job or they've been working therefor a longer time' h. ben konu$tugunuzu biliyorum, y z me s yle, kan ik meelachen (Backus 1992) º know what you guys are saying, say it to my face, so I can laugh too9 i. helede migrain-i varsa zou ik niet aanraden ... especially migraine-POSS there-is-if would INEG recommend 'especially if there's migraine, I wouldn't recommend ...' Insertional CS is very varied: single inserted nouns are again the most frequent (cf. 3a) and also vriendin in 3d and migraine in 3i, but virtually every other type of Dutch content word occasionally gets inserted too. Dutch verbs,

Bilingual Proficiency and Code-switching/Mixing Patterns · 67

as in 3a and 3b are always incorporated in a compound verb with the Turkish auxiliaryyapmak ('to do')· Inserted Dutch adjectives are always in predicative position, cf. vast in 3g. Some inserted Dutch elements are constituents or parts of constituents. Examples are the adjective + noun combination lauw water in 3c and the PP van mijn begeleider in 3d. Even the cross-linguistically rare 'ragged switch' occurs more than once, cf. the doubly marked PP in 3e, in which Dutch met and Turkish ~le mark the same concept. Inter-clausal CS mostly concerns Dutch reported speech, cf. 3f. However, structures that are more alternational in character, such as switched coordinated clauses, are fairly common too, cf. 3g. Note that with alternational types of switching, also exemplified in 3d, 3h and 3i, the asymmetrical situation where Turkish clearly is the ML and Dutch clearly the EL, starts to disappear. • 2.3. Second Generation Speech As Table 1 shows, Dutch is used more often than Turkish in second generation speech. CS is again very frequent, but now it is mostly intersentential. Interclausal and insertional CS are still striking aspects of the vernacular, but they occur far less often than in intermediate generation speech. Another difference is that there can be insertional CS when Dutch is the ML, cf. the insertion of Turkish key in 4a. Half of the insertions we do get are EL Islands (Myers-Scotton 1993), as in 4b, while in the earlier generations single word-switches far outnumbered switched constituents. In inter-clausal CS, switches of the alternational types, such as the coordinated clause in 4c and the copula complement in 4d, have gained in frequency. As for switched subordinated clauses, Dutch is now usually the language of the main clause. In 4e for instance, the dependent conditional clause is in Turkish. All in all, second generation in-group bilingual speech is mainly Dutch, with a fair share of alternational CS. (4) a. maar dat is weer koy, he ... but that's the countryside again, isn't it?' b. Oraya gittim dügüne helemaal niet opgemaakf ofzo. 'I went there, to a wedding, not made up at all or anything' c. bunlar oynamaya kalkinca sen de kalkman lazim onlarla en hoe moet je dan op de rest leiten? 'when they get up to dance, you must get up with them as well, and then how can you keep an eye on the rest?' d. .. .birde weet je wat de probleem is: birisi burda neyaparsa geriyanlisi da peqine gidiyor. 'and do you know what the problem is: whatever someone does here, the reactionaries come after her9

68 · Rajendra Singh and Ad Backus

e. zamanmda oynarsan, antrenmana gelirsen, dan moete ge lock's zaterdags wel in de basis staan he! 'if you play in your own time, and go to practice, then surely you should be in the starting line-up on Saturday, shouldn't you?9 To summarize, some level of language mixing characterizes all speakers studied. Different generations, however, have clearly different patterns of language mixing. As we move from first to second generation CS data, a general shift is visible from insertional to alternational CS as the dominant type. Along with this shift, the relative share of Dutch in this register becomes bigger and bigger. First generation speakers speak Turkish only, with some Dutch content words; intermediate generation speakers switch all over the place, and second generation speakers switch a lot too, but virtually always alternationally. This ties in very well with the results of most of the studies mentioned in the introduction.

• 3. Bilingualism: Proficiency in Two Languages The conclusions in 1 were based on these kinds of data, as far as they were published at the time, and a few additional observations: (5) a. The speakers in northern India whom Singh considered to be real bilinguals generally did not indulge in code-mixing, though they did switch their codes across entire discourses—'certainly not', as the Master said, Vithin a single sentence'. b. Speakers whom Singh considered to be perfect bilinguals seemed to indulge only in inter-discourse and what he calls special switches (to be discussed later). c. The Weberian pathos of Bloomfield's (1927) memorable comments on White Thunder could not possibly have been inspired by prescriptivism. d. Backus' second generation, Bentahila and Davies' older, urbane Moroccans, and Singh's metropolitan informants, who did not exhibit much intrasentential switching, had no difficulty in sustaining discourses in Dutch, Arabic, and English, respectively. e. The sort of question Grosjean (1991:255) asked Hyltenstam in 1991 comes back to haunt us more often than it should if we were on the right track. The first two observations suggest that there is yet another type of bilingual, not encountered in the studies of the Backus (1996) type. We will return to this

Bilingual Proficiency and Code-switching/Mixing Patterns m 69

in Section 6. Grosjean's question was: 'When it is mentioned ... that the patients violated the Equivalence Constraint, could this not be due not so much to their demented speech, but simply to their less proficient Swedish? ... how do linguistic constraints fare when one of the languages is not mastered very well?' (emphases ours) (Grosjean 1991). Notice that both the particular and the general problem raised by Grosjean's question underline the inevitability of questions of proficiency creeping back into places they had been ex cathedra pushed out of. Consistent with the fashion of the day, Singh had named this problem Grosjean's Problem because Grosjean has, as statements like the following reveal, repeatedly espoused a very liberal view of bilingualism (cf. Grosjean 1982 and forthcoming, for example): (6) According to us, bilinguals are those who use two, or more, languages in their everyday lives. Bilinguals have developed competencies in their languages to the extent required by their needs...(Grosjean forthcoming: 3). Definitions like 6 above are problematic not only because they do not require the fulfillment of even the three basic functions of language—pragmatic, mathetic, and aesthetic, as Halliday (1975) calls them—but also because they do not allow one to refer to circumstances when a language is, to use Grosjean's words, 'mastered only to a certain level of proficiency' (for some further discussion see Singh [forthcoming]). We would now like to add the following observations to the list just presented. Some of these will be illustrated and commented upon below. (7) a. Some of those who routinely switch and mix find it difficult, actually impossible, to sustain discourses in either of the two languages attributed to them. b. The population of students that provided Pandit (1986) with a veritable encyclopedia of counter-examples against all sorts of constraints cannot sustain discourses in any non-mixed language for more than a few seconds. c. The habitual non-mixer yields to the habitual mixer for Gricean maxims (cooperation, accommodation), though the latter often cannot follow these maxims for lack of linguistic competence. d. The Weberian pathos of Bloomfield's comments regarding White Thunder finds a sensitive echo as late as Bakker (1992:177). e. Drapeau (forthcoming) convincingly argues that what is sometimes construed as 'manifestation of balanced bilingual competence' is actually a rather strong index of 'lexical erosion' in Betsiamites, Quebec.

70 · Rajendra Singh and Ad Backus f. The dismissal of those who see 'a decline' of sorts in Betsiaraites, Delhi, Fes, and other places is so high-minded that one cannot but be suspicious of it. In many cases, it seems, it may actually be death rather than just decay, or at least a possible move towards languageshift. The first three are related to and support the aphorism given in Ib. They are, interestingly or even ironically, supported by observations like the following by Pandit (1986): (8) a. [These speakers] do not usually use either of these languages exclusively (p. 70). b. [These speakers] maintain that they themselves feel at ease with the mixed variety rather than with Hindi or with English (p. 71). c. Mixed Hindi English does not have a restricted function and is used in almost all situations (p. 73). d. One way this difference in responses to grammaticality could probably be explained is in terms of the extent of the informant's exposure to English (p. 88). The last three observations in 7 merely attempt to capture the Weberian pathos of Bloomfield's White Thunder comments. It is important to remember in this context that Bloomfield was no prescriptivist. The only mistake made by those who some of us hastily dismiss as linguistic neo-conservatives may be the diagnosis of death as mere decay. The ratatouille may be delicious, but it may also be a reflection of the inability to make either a decent curry or a decent pancake. It may well be a signpost on the road that leads to monolingualism, as Bakker clearly recognizes. The speakers commented on in 7 and 8 are no doubt highly competent monolingual speakers of their mixed lect, but that does not necessarily make them bilinguals. If we compare the Turkish case and Pandit's, the intermediate generation would be the one closest to her informants. However, the similarity is fairly superficial, as observations 8a and 8c do not apply to them (it is probable that 8b does). Both the intermediate and the second generation informants can yield to monolingual speakers of both languages and speak without mixing. The next section explores the degree to which 7c applies in the Turkish data.

• 4. Accommodation among Bilinguals with Varying Degrees of Bilingualism Whether a putative bilingual actually is one can be shown if she manages to use both languages when necessary. Accommodating to non-speakers of one

Bilingual Proficiency and Code-switching/Mixing Patterns m 71

of her languages would constitute evidence for this ability. More interesting for our investigation, however, is whether bilinguals accommodate to other bilinguals, in order to deal with perceived differences in bilingual proficiency in a pragmatically appropriate way. Theoretically, proficient bilinguals can accommodate to the Ll-dominants, but the reverse is impossible. The first generation Turkish-dominant speakers can certainly try to put as much Dutch as they can manage into their bilingual lect, but they can obviously not manage to speak the same lect as their second generation interlocutors. This follows straightforwardly from Grice's maxims and common sense: beginning L2 learners can only accommodate slightly more to proficient bilinguals than monolingual Spanish speakers can to monolingual Japanese speakers. But do the second generation informants in Backus' study actually accommodate? We have very little relevant data, but we do have some. The social networks of the second generation informants predominantly consist of other second generation members. Nevertheless, thanks to the maintained contacts with Turkey, and especially the strong tendency to many people from there, recent immigrants may play at least some role in these networks. We were fortunate enough to have a few 'mixed' networks represented on three of our tapes. All three figure one and the same intermediate generation informant, Ayhan. The other speakers are his first generation wife Bahar and his two second generation younger brothers on one tape, a second generation friend on another tape, and one first generation and three second generation friends on the third tape. Bahar is also one of the speakers on an all-first generation recording. Here is what happens: Bahar, who finds herself the object of incessant teasing by the two younger brothers, uses more Dutch than she does on the all-first generation tape. Her Dutch is also qualitatively different: she even attempts elliptical sentences now, as in 9, while on the first generation tape all of her Dutch elements are inserted proper nouns. (9) $ey bize, februari veertien vol... thing we-DAT february fourteen volleyball] 'Hey, we mil be playing a volleyball tournament at February fourteenth' Expected Dutch construction: 'He, we gaan op veertien februari volleyballen' Thus, one of our 'weak' or 'apparent' bilinguals only ventures out of the safe domain of proper nouns when under pressure to do so, and does that at the cost of exposing her L2 competence as 'a smattering' The younger brothers do their picking on Bahar in Dutch, as it helps their purpose if she has trouble understanding them. She clearly accommodates, the brothers do not. This is not supportive of 7c, which stipulates that it is the accomplished bilinguals who will do the accommodating. However, from the way the conversation goes, it is likely that the brothers do not accommodate

72 · Rajendra Singh and Ad Backus

because they do not want to. It is entirely possible that they cannot, but that is impossible to establish on the basis of the data. We will get back to this issue shortly. The other first generation speaker is Yusuf. He appears on tape with Ayhan and three second generation women, one of whom is his wife. He does exactly the opposite of what Bahar did: he speaks monolingual Turkish throughout, even though the conversation had been mainly in Dutch until the time he got actively involved in the discussion. Yusuf avoids exposing himself as a weak bilingual and chooses to remain a Turkish monolingual for all intents and purposes. Next, the intermediate generation speaker Ayhan. He clearly accommodates to first generation speakers, on both occasions, by speaking almost monolingual Turkish to them. That is, he does not modify the structure of his mixing: he simply dispenses with it almost entirely when talking to Bahar or Yusuf. In both cases, that may be because he wants to protect the first generation members (recall that Bahar is his wife) from the attacks of the second generation speakers. What little switching there is in his speech directed at Bahar and Yusuf is prototypical insertion of content words. There are no instances of the more elaborate switch types he uses with intermediate and second generation interlocutors. Those types, illustrated in 3, are the types generally talked about in the syntactically-oriented CS literature. They stand out as they are what makes this type of speech so obviously bilingual. However, it must be noted that Ayhan's data and those of him and three other informants in Backus (1992) do not give much evidence for this generation's ability to sustain long Dutch discourse. That they probbly can, follows logically from the fact that many intermediate generation members function normally in Dutch society, but they certainly do not flaunt that ability in bilingual speech, where Dutch stretches are never longer than two or three sentences. These sentences are, furthermore, often ungrammatical vis-ä-vis the norms of monolingual spoken Dutch, as in 10, where the article is wrong and the dummy daar does not sound right to native ears. This can spell one of two things: fossilized second language acquisition or an ethnic variant of Dutch. That the second generation, as we will see, does not share this feature suggests that the first option is the right one. (10) bir de $ey olsam bile daar was de risico dat ih ontslagen zou worden 'even if I had gotten it, there was a chance that I would get fired9 Expected:... was er het risico dat Of the two groups of second generation speakers that we have talking to first generation speakers, one accommodates and one does not. The three women

Bilingual Proficiency and Code-switching/Mixing Pat terns m 73

talking to Yusuf and Ayhan are engaged in a heated and serious discussion about Yusuf s ideas about male and female roles in the immigrant community. They limit the amount of Dutch in their speech to comply at least a little bit with Yusuf s language choice. What Dutch they use, however, still consists of entire sentences and clauses; the Turkish stretches are just longer than they were in the friendlier part of the conversation. The women, therefore, certainly do not accommodate completely. The two brothers engaged in the playful nagging of Bahar, do not accommodate at all. Again, the Dutch they produce when talking to her, is the same as what they produce when talking to Ayhan or to each other, and it consists mainly of full sentences and clauses, the sorts of things other native speakers of Dutch say. Among themselves, second generation members speak Dutch interspersed with some Turkish. A small percentage of their Dutch clauses and a higher percentage of their Turkish ones contain intrasentential CS (see Section 2.3.). Their discourse sustenance in Dutch is beyond doubt, but we may want to worry about their Tiirkish. Maybe their situation is the mirror image of that of the first generation, with good Dutch and little Turkish. Their use of Turkish is actually quantitatively quite similar to that of Dutch spoken by the intermediate generation. Qualitatively, however, there is a difference: there are not many signs of ungrammaticality vis-a-vis monolingual Tiirkish norms. Moreover, the second generation's normal functioning in the Turkish community illustrates their ability to sustain relatively monolingual Turkish discourse (as far as that is necessary in an immigrant community), an ability also shown by the three women when they are arguing with Yusuf. What can be gotten from these admittedly scanty results? Specifically, do they lend support to the idea that Ll-dominants switch when they have to and proficient bilinguals when they want to, as professed in la? To start with the former, Bahar does accommodate in that way and Yusuf does not. Bahar is under attack of the brothers, who use Dutch as a weapon because it allows them to say things to her which she has a hard time understanding. She rises to the occasion, however, and uses as much Dutch as her competence allows her to do. Yusuf uses the opposite strategy and does not give the others the chance to exploit the fact that he cannot keep up with their Dutch. He stubbornly keeps steering the conversation in the direction of Tiirkish, and, in the end, wins that battle. The point of all this: whether Ll-dominant speakers use CS or not when talking to proficient bilinguals, depends on their personality, their mood and the dynamics of that particular conversation, in the same way other communicative choices are made under the influence of such factors (such as whether to raise your voice, to use difficult words, to use slang expressions, etc.). How about the proficient bilinguals? Could it be that the younger brothers do not accommodate to Bahar because they cannot, because their Turkish 'isn't good enough'? We have already hinted that, even if that were the case, they have plenty of communicative reasons for their behavior. The three women

74 · Rajendra Singh and Ad Backus

locked in battle with Yusuf, at least show some evidence of their ability to shift from a predominantly Dutch bilingual mode of speaking to a virtually monolingual Turkish one when communicative motivations call for such a step. All this shows that 7c has got something going for it, but that it has to be taken as an idealization: the things people want to achieve in conversation might lead them to override the maxims of good behavior. • 4.2. Variation in the Turkish Community The remarks in this section pertain to the observations made in 7d, 7e, If. So far we have not addressed how conventionally established the patterns illustrated are for the Turkish community in general. Is the mixed lect of any one speaker, as it appears on the tapes, the same mixed lect the speaker in question uses with other bilingual speakers in his network? Even if we accept that the Observer's Paradox has not wreaked too much havoc (as Backus' participant observation gives him reason to think), how can we be sure that both the amount and the sort of CS would be even roughly the same for every speaker if we would tape them again, whether with the same or with different interlocutors? The unsatisfactory answer is, of course, that we cannot be certain, ever. But what we can reasonably put forward, is the following generalization: (11) The vernacular of a bilingual community characterized by CS, shifts from asymmetrical LI-based insertional CS to alternational types with growing degrees of bilingualism. It is obvious that the first generation does not have the necessary proficiency in Dutch to be able to use it a great deal in spontaneous speech. In fact, most of the first generation informants were attending a Dutch language class at the time the recordings were made. In addition, we saw that insertional CS of the intermediate generation tends to consist of one-word Dutch insertions, while the second generation often inserts longer chunks, such as constituents. It is again very likely, to say the least, that the second generation speakers have attained a much higher proficiency in Dutch than the intermediate generation. They have learned it from a much earlier age, and speak it much more often. Attributing both the higher incidence of Dutch material in second generation speech and the longer nature of their Dutch insertions to the second generation's higher proficiency in that language sounds plausible. The interesting result is that what is functionally the community's basic vernacular, is formally a continuum of more and less bilingual lects, following the cline of CS patterns just outlined. We tend to think of a community's vernacular as having both functional and structural cohesiveness (cf. Kachru 1982). In this case, these two do not go together. Three fairly recognizable lects, each with its own structural cohesiveness, share the same function. It is especially

Bilingual Proficiency and Code-switching/Mixing Patterns · 75

this combination of 'same function, different structure' and its plausible link with degree of bilinguality, which offers strong support for a link between bilingual proficiency and associated CS patterns. Two explanations could be given for the Dutch Turkish data, with equal face value validity. This paper deals mainly with the first one, and largely ignores the second one, for lack of analyzed data. They are: (12) a. The shift in CS patterns is the way it is because the Dutch proficiency of the second generation is higher. b. The shift in CS patterns is the way it is because the Turkish proficiency of the second generation is lower. Code-switching is often seen as an indicator of pending language shift. It is indeed tempting to see the ever-increasing proportion of Dutch used in the bilingual vernacular as reflecting the psycholinguistic balance of the two languages in the heads of these speakers. However, the logic behind such an assumption is ill-informed: the intergenerational differences reflect a shift in what is the 'unmarked choice' for in-group talk. By itself, they say nothing about proficiency levels. That speakers choose not to speak a language when talking to peers, does not mean that they are not capable of speaking that language. One could argue OK, the point is well taken, but on the other hand, there is a natural correlation between the amount of time you spend speaking a language and your proficiency in it'. Statistical correlations of this sort were for instance found by Mougeon and Beniak (1991). We think the answer to this should be 'yes, but we do not know whether the second generation speaks less Turkish. We only know that "monolingual" Turkish is not their unmarked choice for in-group speech. There may be many other situations in their daily lives in which they do choose to speak Turkish.' In fact, this is very likely. Members of the Turkish community in Holland still have sufficient use for Turkish, as was pointed out earlier. Talking to recent immigrants, visitors from Turkey or people in Turkey, as well as watching Turkish TV and reading Turkish newspapers, all require a 'Tiirkish only' strategy. In general, differences between subgroups are related to both proficiency and language choice. The speech styles adopted by the different generations in the studies mentioned in Section 1 as well as in Backus' study, are what one would expect given their level of proficiency in what was originally the second language. However, the flexibility in language choice shown by the second generation speakers (the older speakers in the Moroccan case), indicates that the EL-dominant style they adopt for their vernacular, is only the preferred way of speaking. It is motivated, but not dictated, by their distribution of proficiencies. The TUrkish data do not show this flexibility as well as some others though, such as Nishimura's.

76 · Rajendra Singh and Ad Backus

High proficiency in the second language naturally leads to greater use of it in the mixed lect, but what may ultimately be the most important reason for its increased use in the second generation, is that generation's greater use of it in other spheres of daily life. Second generation Turks simply use more Dutch, which in turn ensures that Dutch units become better and better entrenched in their idiolects, with an indirect beneficial effect on overall proficiency. Generally, this aspect of entrenchment receives little attention, since for monolingual speakers it is by definition very high. Whether the explanatory factor is greater proficiency or increased use of the language, is a difficult question to answer. It is somewhat of a 'chicken and egg' problem, since the first factor promotes the second and vice versa. Note, however, that use in daily life is likely to be primary, because disuse of a language leads to attrition of certain patterns (cf. Mougeon and Beniak 1991), and increased use to increased entrenchment of patterns of that language. From here on, we will assume that the second generation members qualify as proficient bilinguals, the intermediate generation as somewhat less proficient bilinguals, with Turkish being their stronger side, and the first generation informants as weak bilinguals. Contrary to Pandit's informants, the non-first generation Turkish bilinguals can use the languages in a diglossic way, despite their attested lack of willingness to accommodate in certain communicative settings. Mixed Dutch Turkish has a restricted function as the in-group vernacular. Its speakers can choose to speak it, whereas language choice is not an option in the speech of Pandit's informants, as 8, given earlier, testifies.

• 5. Transplantational and Indiginizational Bilingualism Before we get to the heart of the matter, a few clarifications may be in order. We wish to stress that we are not suggesting that there is a logical entailment relationship between low proficiency and high frequency of switching. There are two reasons for this. First, a speaker who habitually switches intersententially is almost by definition in control of the two systems (that is: we are not aware of any conventionalized language that consists of alternating sentences in two languages). That is not to say, by the way, that in particular instances, an intersentential switch may not be motivated by an inability to come up quickly enough with the right words in the language spoken before the switch. But even if we discard intersentential switching from our considerations, a second, and more important, reason remains. Mixed lects are definitely capable of conventionalizing so much that eventually they may serve all or most of the communicative needs of the community. In such circumstances, observation 7c becomes irrelevant, as there is no habitual non-mixer in such a community. When a mixed code is the normal means of communication in the community, and the remarks in 8 are meant to illustrate this state of affairs for mixed Hindi English, and an unmixed variety of LI is not in use, degree of

Bilingual Proficiency and Code-switching/Mixing Patterns m 77 proficiency in that unmixed variety is simply not a relevant factor for the sort of questions dealt with here. It would be as relevant as asking the speaker whether he also speaks some other foreign language.1 This is where observation Id comes in, about the differences between transplantational and indiginizational settings. For immigrant languages, the following general and interrelated tendencies may be given: (13) a. They tend to become mixed, in any case lexically, to a high degree. b. They do not tend to be subjected to much focusing. c. They tend to be situated somewhere on the road to language shift. All three of these may apply to indigenous languages as well, but tend to do so to a much less extreme degree in such settings. For instance, Myers-Scotton (1993: 71) remarks that there are some slight differences in CS behavior between higher and lower educated Kenyans (that is, between people with more and less English). The differences are similar to those reported for the Turkish-Dutch case and the other settings mentioned in Section 1. Nevertheless, in general, the reader of Myers-Scotton's work gets the impression there is just one, fairly recognizable, focused, bilingual lect in Nairobi. However, the more threatened indigenous languages become, the more they tend to resemble the prototypical immigrant language, as observation 7d testifies. As is usual in matters linguistic, this is no law, only a general tendency. An exception, for instance, would be a dying language that does not go through a stage of heavy lexical borrowing. The labels 'indiginizational' and 'transplantational' are obviously not used as theoretical constructs, they are simply suitable labels for types of language contact settings that tend to have slightly different contact outcomes. They could be expanded as 'maintenance of bilingualism, with possible addition of mixed lect as a register of LI' and 'change in vernacular towards mixed lect, with bilingualism usually involving this lect and L2', respectively. Again, these are just the prototypical cases. Nothing keeps speakers in the first setting from, for instance, shifting to L2 or from not creating a mixed lect, and nothing keeps speakers in the second setting from maintaining a monolingual version of LI, as yet another register in their linguistic repertoire. The latter is, for instance, common in immigrant communities that have extensive ties to their homeland, where, obviously, nobody would understand their mixed lect.

• 6. The Relationship between Code-switching and Proficiency The 'heart of the matter' referred to earlier is this: given what we have said above, one may legitimately want to be cautious about making statements of the following sort:

78 · Rajendra Singh and Ad Backus

(14) a. True bilinguals exhibit a very high incidence of intrasentential switching, b. Intrasentential switches require most skill. (15) a. Code-switching/mixing has been established as a legitimate manifestation of bilingual competence, b. Code-mixing is very common amongst fluent bilinguals. Despite a clear difference in the empirical import of the statements in 14 and 15, both types have become part of the folklore about code-switching/mixing. One has to be particularly cautious about statements like IS because (a) they do not appear to be hypotheses, (b) they do not invite what might be called a "degree interpretation'. Sure, code-switching/mixing is a manifestation of some sort of bilingual competence but it is really rather uncommon amongst at least some fluent bilinguals if by fluent one means very proficient. A plausible reason for this may be that they have no need for a mixed language. Not as an act of identity, nor because their ML needs the EL lexicon very much, except to fill obvious lexical gaps. However, as argued above, heavy borrowing is a very natural phenomenon in immigrant languages, so that fluent bilingualism may mean 'fluent in the dominant language of society and in the mixed register of LI' in such a context. Again, the sociolinguistic differences between transplantational and indiginizational contexts prove relevant here: it is easier, and, more importantly, more useful, for urban Indians to be fully proficient in Hindi (meaning: 'have' all registers) than for Dutch Ibrks to be similarly proficient in Turkish. They receive less monolingual input and have less need for monolingual output. The kinds of things we have mentioned so far raise at least the following questions: (16) a. If there is a correlation between patterns of code-switching/mixing and degree of bilingual proficiency, what does it look like? b. Why was not Poplack's (1980) invitation to use code-switching behavior to measure bilingual ability taken up earlier and more systematically? c. Given the fact that Asian linguists and Asianists have generally been critical of the monolingual bias of Western analyses of bi- and multilingualism (c/. Pandit 1977; Pattanayak 1981; Srivastava 1997 [1977]), why is the Indian-born Canadian linguist Singh asking these questions? As the last two questions are easier to handle, let us dispose them off first. Without distracting anything from P.B. Pandit's (1969) classic remarks to the effect that the Indian question would be: "why should any one give up a language?', it is important to point out that the virtual non-existence of the

Bilingual Proficiency and Code-switching/Mixing Patterns m 79 question of language maintenance in an agrarian society like traditional India or Asia sharply contrasts with its threatening presence in the West and in industrialized metropolitan India. Language-shift, in other words, does not occur in Gumperz' villages in Karnatak, but it does in Delhi and in Bombay. As far as the second question is concerned, we would like to suggest that the Brahmin's natural predisposition to please the ruling Ksatriyas is at least partially responsible for the fact that Poplack's invitation was generally taken as a very polite way of stating a conclusion.2 From now on, intrasentential mixing could be safely taken as a reflection of high bilingual proficiency. This somewhat unwarranted extension of our entirely justified political liberalism of the 1960s and the 1970s is generally left untouched in North America. Hymes is among the few linguists we know of who have not forgotten either the content or the pathos of Bloomfield's comments about White Thunder. Be that as it may. Let us turn to the difficult question. The question itself was first posed by Poplack. Her answer has many parts, some of which are reproduced in 17 below: (17) a. Code-switching, then, rather than representing deviant behavior, is actually a suggestive indicator of degree of bilingual competence (Poplack 1980:616) b. The cline of bilingualism follows the percentage of intrasentential switches (59 per cent for 'true bilinguals'). c. Expanding bilingual grammars can be represented as follows:

Intersentential switching

Tkg-switching

Intra-sentential switching

We do not mean to quibble, but it is difficult to refrain from noting that the set for non-overlapping figures above 'intersentential switching' do not prohibit the interpretation that if you do know the two languages only separately, you are not a bilingual because you do not have any intersection and the last possible figure in this series will spell not the death of two languages but the birth of true bilingualism. It may actually, as Bakker sensitively and tellingly shows, announce the birth of a new language (but we shall return to this later in this section). Perhaps the difficulties presented by Poplack's pictures are inherent to the metalanguage she employs. Despite its caveats regarding nominal, nonparametric, ordinal, and knock-outs, the metalanguage she uses is known to do well only in the middle ranges (cf. Rao: 1992). Our own observations, and some of the others mentioned in Section 1, add up to a picture at odds with the one presented by Poplack. A few additional comments can be made regarding Singh's data, the full details of which he must reserve for a longer piece. They

80 · Rajendra Singh and Ad Backus

concern a type of speaker not (yet) found among the bilingual Turks in Holland. These speakers, who we would classify as 'true bilinguals', do not generally code-mix at all and when they do in order to accommodate those they consider to be their country cousins they stop at parentheticals, mostly things like acdfla Ok', and ha 'yes*. They do, however, indulge in what may be called special switches, switches that require accessing their two grammars in very skillful ways. They come in two varieties. The first one is illustrated in 18 below: (18) a. Dhanyvad...(pause) or yeh angrezwala dhanyavad nahin he Thank you and it English thank you not is 'Thank you and it is not an English "thank you".' b. Acha bhai, jese urdu me kahate he bona pati. Well, now, as urdu in said bon appetit 'Well, as we say in Urdu, "bon appetit'" The second variety of special switches, the invisible one, is hard to illustrate without going into a detailed discussion of syntactic matters but easy to point to. Any one who has ever read Achebe, Raja Rao, or Rushdie would know how the boundaries of what we call English can be gently and entirely legitimately pushed in the direction of African topicalization, Sanskrit tatpurusha compounds, or Urdu semi-free word-order. These switches remain invisible to the ordinary eye and inaudible to the ordinary ear and constitute the true bilingual's revenge on interference enthusiasts. Special switches require more investigation, for which there may be some methodological problems. First, not every bilingual may feel the need to exert her revenge on the interference enthusiasts. That is, some may be too nice to engage in it. If that is correct, special switches cannot be a defining characteristic of proficient bilinguals, though they certainly remain indicative of proficient bilinguals. Second, these same people may do it when they are in a position in which they would want to take revenge, but in the in-group type of data that most of us have collected, these situations will not occur. Third, how sure can we actually be in individual cases whether something is an invisible switch or a case of interference after all? How should we go about falsifying the hypothesis that every interference is a special switch, in order not to become ultra-liberal 'special switch enthusiasts'? We will leave this problem for future work. On the other side of the continuum, the speakers that we would classify as "apparently bilingual' do the sorts of things that fly in the face of 'government', 'equivalence' and functional head constraint, and anything else one can think of. They are quite like Bakker's Michif speakers who speak neither French nor Cree. The reason that Pandit was able to find counter-examples to every constraint ever proposed is because her population was made up of a group of metropolitan teenagers unable to use either monolingual Hindi or mono-

Bilingual Proficiency and Code-switching/Mixing Patterns m 81

lingual English beyond a few seconds. Theirs is a language that is, in the words of Adorno (1992:291) 'not open to any calculus', a language that 'arises only in pieces and out of the disintegration of the existing languages'.3 At the end of that road, we may find a fully conventionalized mixed language, such as Michif. Perhaps Berk-Seligson's (1986) finding that there was little interpersonal variation between her informants indicates that the form of JudeoSpanish spoken in Jerusalam is also relatively far advanced on the road towards full conventionalization. Similar considerations may apply to the Maori-English morphological grasshopping from which Eliasson (1990 and 1991) gets his violations of the Free Morpheme Constraint. The violations of the Free Morpheme Constraints, in other words, may spell attrition as well as trouble for Poplack, a possibility actually suggested by Eliasson (1990), though he feels that Eruera's (1975) informant provides evidence for word-internal switches in highly accomplished bilingual speakers. Experience tells us that although the Free Morpheme Constraint is violable, even reasonably competent bilinguals do not violate this admittedly doubly redundant constraint—doubly redundant because, as Eliasson (1991) shows, it trivially follows from what might be called lexical integrity, subsumed by Giesbers' (1989) razor. The middle range, obviously, does the sort of things Kachru (1978), Bhatt (1994), and Singh (cf. Singh 1985, for example) have reported on—namely, mostly code-switching but also some intrasentential mixing. The switching is generally governed by topics and the mixing by the sorts of considerations Poplack, Muysken and Myers-Scotton have brought to our attention. We have already seen that these patterns are roughly compatible with the TurkishDutch data.

• 7. The Language Change Trap We would like to join Eliasson (1991) in his plea for a more thorough examination of interference, something almost invariably interpreted in a unidirectional way. If switches violating the Free Morpheme Constraint were indeed examples of 'creative and skilled utilization of the structural resources of two languages', one would expect intramotic switches in both directions. In north India, this is certainly not the case, and we are not fully convinced that such examples as have been gleaned from the Maori-English context exemplify creative and skillful utilization of the structural resources of two languages. English morphology is certainly not used by the speakers who say /§apa/ and / toermo/ for 'shops' and 'terms' to show their bilingual skills in the Indian context, and the unidirectionality of such switches in the Maori case is no less striking, as is the degree of asymmetry in CS in most reported cases of it. This raises a very important point. 'Highly bilingual' usually means 'highly proficient in L2'. Without the obvious emancipatory justification, however, there is no reason to equal these two concepts.

82 · Rajendra Singh and Ad Backus

As far as Adorno's sensitive—the arrogant amongst us will say 'pre-theoretical'—characterization is concerned, consider the following observation by Pandit (1986: 29): (19) This is where lies the crux of the problem. There are difficulties in arriving at the system of MHE [Mixed Hindi English] because unlike the languages it comprises it does not have any 'native-speakers'. It has only 'users' as distinct from 'native-speakers'. Despite her confusion at the possibility of the apparent birth of a new language, she is able to capture both the disintegration and the defeat of the calculus, albeit clumsily. We say confusion because ethnographically she goes out of her way to show that MHE is, in fact, an emerging new code and yet she insists on a literal interpretation of the epithet 'mixed' in the name she gives this code. She is a bit like us all when we talk about borrowed words. The Varanasi taxi-driver's answer to an American linguist: 'ap universal jaege, pahale kyo nahi kaha'—he had first used the High Hindi word vishwavidyalay for university—should serve as a warning to us all. She does, however, later indicate that her MHE may have rules of its own, providing part of the answer demanded by the status of hybrid forms, illuminatingly discussed in Eliasson (1990). Bhatia (1989) provides further evidence for the fact that Pandit's MHE does, in fact, have rules that belong neither to English nor to Hindi. It exhibits, for example, a special process of deriving the infinitival verb forms from nouns and verbs which is taken neither from English nor from Hindi. Even if Bhatia's examples and others can be shown to be completely independent developments, they are still contactinduced. A typical example of what tends to happen in contact settings is the fate of the auxiliary yapmak in the Turkish data. This verb can combine with nouns and verbal nouns to form compound verbs. In Holland, it also routinely combines with Dutch infinitives in similar constructions (cf. 3a and 3b). It has undergone change in the immigrant context, as it is now more widely used as an auxiliary than it was or is in Tlirkey. However, the concomitant semantic bleaching in these constructions (in kijkenyapmak, 'look do', it obviously does not mean 'to do' anymore), is in harmony with an internal development in Turkish, which is only speeded up in the contact setting because of the new or extended uses the verb is put to. Language change is part and parcel of language contact and makes some sort of description of the lects one claims the speakers to have a certain degree of competence in, absolutely indispensable.

• 8. Bilingual Competence One is, obviously, entitled to know what the objective correlates of our labels such as 'true bilingual' and 'apparent bilingual' are. Some of them are listed in 20 below.

Bilingual Proficiency and Code-switching/Mixing Patterns m 83

(20)

Bilingual Perfect (true)

1. Almost unlimited domain competence 2. Almost unlimited discourse sustenance 3. Mathetic and creative ability

Less than perfect

Very competent Weak Competent 1. Very large 1. Rather limited 1. Limited domain domain competence competence domain 2. Almost competence 2. Very limited unlimited dis2. Somewhat discourse constrained sustenance course sustenance discourse 3. Almost no 3. Mathetic and sustenance Mathetic and Aesthetic 3. Limited Mathetic Aesthetic functions ability and Aesthetic available ability

Apparent

Smattering

One might object, as Backus (1993) did, that this taxonomy may not work because of a community's ability to set its own norms. This objection was motivated by the fact that Dutch Turkish simply has a lot of Dutch words in its lexicon and a large range of permissible CS structures. The mixing and switching is part of the 'norm' or 'language average' (cf. Weinreich et al. 1968) of the vernacular. Whether its speakers also speak monolingual Turkish would be irrelevant if that language had no function in the community; it would never get chosen as the means of communication for an interaction. Such a community of Turkish speakers does probably not exist in Holland at the moment, but that such things are possible is proven by the existence of, for example, Michif. Singh, in an earlier version of this paper, pointed out that the relativization implied would make it impossible to compare one bilingual community with another. It would mean that Butler English in the India of the nineteenth century would be as much an instance of bilingualism as Standard Indian English in contemporary India (compare the problems with Grosjean's definition in 6 given earlier). The problem can be solved if it is understood that bilingualism may involve a mixed lect that is conventionalized enough to have taken over as the main register of its ancestor language, as long as there is another language, presumably the one that functions as the EL in the mixed one, that is also known and habitually used by the same speaker. But both need to have a reasonable set of suitable contexts, in order to allow for enough domain competence. This view does add an uncertain dimension to the whole enterprise, since conventionalization is, as Mackey (1970) pointed out, by nature impossible to pin down exactly. Clearly, more research is needed in order to identify

84 · Rajendra Singh and Ad Backus

when we are dealing with spur-of-the-moment code-mixing, when with 'monolingual' production in the mixed lect, and, most important of all, when the former develops into the latter. Presumably, at least some of the parameters mentioned in 20, such as domain competence and unlimited discourse sustenance, will feature in the answer. A further hypothesis may be that with on-going conventionalization, influence of L2 proficiency on the make-up of the mixed lect, will gradually decrease. What is at stake is essentially *when does a language have native speakers?', or, alternatively, *what does it take to count as a native speaker of a language?'. The formal answer Singh (1994) provides is: a native speaker of a language is a speaker who shares relatively stable grammaticality judgments on utterances said to be from his language with other speakers. Notice that the need to play the grammaticality game arises only when the pragmatic function is supplemented by the mathetic and aesthetic functions. We will return to this question in the next section. This definition excludes, in our view, not only the tourist guides of Varanasi but also Pandit's undergraduates who have remarkably inconsistent judgments on utterances in both Hindi and English, as well as, we may presume, the first generation speakers Bahar and Yusuf in the Turkish-Dutch study. Besides, Pandit's informants do not really ever use either language exclusively. Again, we are not aware of any mathetic or creative activity in German in Varanasi or in Hindi or in English amongst Pandit's undergraduates. They seem to be poised at the brink of a possible monolingualism. Be that as it may, the point is that both monolingual and bilingual proficiency is measurable across communities despite differences in norms because these cannot alter what it takes to count as a speaker. If your cxommunity requires you to do only this or that with Li or Lj you may count as fl pragmatic user of Li or Lj within and outside that community, but 'He knows some French' is not the same thing as 'He knows French'.

• 9. Code-switching and Proficiency Revisited If the resolution of the apparent contradictions in the results of Backus, Bentahila and Davies, and Singh on the one hand and Pfaff (1976; 1979), Poplack and Nortier on the other is not to be found in the infinite multiplicity of highly relativized individual "bilingual settings', what options do we have. Ttoo, we would say: (a) one of these groups of researchers is wrong and (b) they are both right for different ranges of bilingual proficiency. We think the second option not only has considerable empirical support (cf. Sections 1 and 2) but is also the right one for another reason. Let us explain why. Poplack's work, for example, is to be situated within the context of combating unwarranted allegations against non-upper class Puerto-Ricans. It is worth remembering that 'decay' and 'deviant behaviour' provide the constant

Bilingual Proficiency and Code-switching/Mixing Patterns m 85

counterpoint to her defense of the bilingual ability of her informants. Singh, on the other hand, deals with highly educated upper class metropolitan Indians. Singh's informants have never been subjected to such criticism. If anything, they have labeled such criticisms against, for example, Pandit's lost generation'. Singh's true bilinguals must count as native speakers of Hindi and English, though we are not sure if Poplack's or Pandit's can count as native speakers of the two languages attributed to them. The argument is the same as has been advanced above: these speakers have a native language, their bilingual lect, but that does not make them native speakers of both the contributing languages. It is just their bad luck (or rather their lack of means and/or wish to accommodate to the middle class norms) that their only or main language is not sanctioned as one by those in power. The claim that Singh's true bilinguals must count as native speakers of English and Hindi is supported by (a) their ability to use either language exclusively in almost any domain or register for hours at a time and (b) their habitual use of them for mathetic and creative functions (cf. Prabhu 1989). Whether the same holds true for our most bilingual Turkish informants is, as indicated in Section 6, unclear. They can certainly talk Tbrkish for hours at a stretch, but whether that will be in monolingual Turkish has never been investigated. Anecdotal evidence from the community suggests that they can, but also that their Turkish is immediately perceived as 'strange' by people in Turkey, because of caiques and syntactic transfer. Their ability to sustain long discourses in Dutch is beyond doubt, though the same proviso applies. Since there is some debate, created mostly by the TESL industry in Britain and the United States, about the existence of such true bilinguals, a few things need clarification. The absurdity of the TESL-industry inspired paradoxical claim that although Indian English is a variety of English, IE speakers are not native-speakers of English should be obvious. Although it makes sense to say that a speaker of variety i of language L is not a speaker of variety j ofthat language, it makes no sense to say that although i is a variety of L speakers of i are not speakers of L. Suppose it is possible to establish what it takes to count as a speaker. Now, the question of relativization of norms can be handled straightforwardly. Once Americans, Englishmen, Canadians, Indians, and Singaporeans are counted in, we can of course talk about American, Canadian, or Singaporean English. The fact that some of us do not count some of these in has to do with the extralinguistic fact that they do not own enough shares in the stock called English and nothing whatsoever to do with any linguistic criteria we are aware of (cf. Singh 1994 and 1995b; and Singh et al. 1995). Notice that as opposed to speakers of Indian English, who live in an endonormative context, the transplanted communities studied by Backus, Nortier, and Poplack live in a doubly exonormative context: the norms of both the languages they are supposed to be speakers of reside outside these communities.

86 · Rajendra Singh and Ad Backus

As far as Singh's 'less than perfect bilinguals' are concerned, they clearly do not have anything like native or near-native competence in English, the first sub-type—Very competent bilinguals'—being an intermediate case. Whereas Singh (1992) drew its data from conversations and elicitations from the group labeled 'perfect bilinguals', his earlier work was based on data from this intermediate group. He had never worked with 'weak' or 'apparent bilingual' speakers till very recently. As we saw, the whole range is represented in the Turkish-Dutch data too, with the possible exception of the upper extreme of the region inhabited by the 'perfect bilinguals'. The Turkish and Indian communities seem very similar, and yet the subpopulations within them cannot easily be matched across the communities. This is because they are all situated at slightly different points along a continuum, but with the same community-internal ordering. Whereas Singh's perfect bilinguals do not indulge much in code-mixing, Backus' most proficient bilinguals engage in a lot of intersentential switching. Whether this difference reflects differences in the distribution of proficiencies or different community norms, is at present impossible to say, but it most likely reflects the differences between transplantational and indiginizational communities. Also note that the Turkish of these informants is, in its in-group vernacular form, a mixed lect, that is full of conventionalized code-mixing, the presence of which can be accounted for by either reduced proficiency or conventionalization. On the grounds of the speakers' ability to sustain relatively monolingual Turkish discourse when they need to, we would argue in favor of the second solution. The very competent bilinguals among Singh's informants and the intermediate generation in the Turkish community are fairly close to one another. They indulge in intersentential switching as well as in reasonably smooth intrasentential mixing. That again there seems to be more of it in the Turkish case, most likely follows from the sociolinguistic differences. Finally, both contexts include what we have called 'apparent bilinguals'. They apparently mix languages, but do so only in the eyes of the beholder. In fact, Pandit herself speaks of her informants as users of M(ixed) H(indi) E(nglish). The longer duration of the Indian contact setting has brought about more conventionalization than for the newly arrived TUrkish immigrants, who simply speak Turkish and are slowly embarking on the road that might lead them to the level of bilingual ism entertained by, say, the intermediate generation, whose mixed vernacular they may add to their monolingual Turkish register at a later stage in life. Language contact situations are very complex, and invite reflection that goes beyond the construction of a typology of code-mixing communities because within these communities themselves three subcommunities can be isolated relatively easily—the one that does not mix or does so only intersententially, the one that does mix, and the one that only appears to or that tries to. What needs to be constructed is not a typology of code-mixing communities but a methodology for identifying comparable populations across communities because most dynamic settings (i.e. excluding fossilized ones

Bilingual Proficiency and Code-swtching/Mixing Patterns · 87

such as Michif), it seems to us, contain these three types. To construct typologies of the former sort is to abandon the hope of finding the similarities that cut across the differences functionalist categories are bound to hide. The fact that many communities contain various types is itself evidence for the dynamic, diffused nature of these contexts. However, we do not deny that at some level all Turks in Holland might form one speech community, as that is a notion which is relatively independent of proficiency (cf. Dorian 1982). Be that as it may. Let us get to the scale of bilingual proficiency. It is given in 21 below: (21)

High • Discourse and invisible switches • Discourse switches • Speech act switches • Adjacency pair switches • Parenthetical switches • Intrasentential switches • Single content word switches Low

Although 21 above seems essentially correct to us, the lower range clearly needs more work and refinement. The possible challenge from the behavior of a very weak bilingual speaker who knows that his interlocutor does not know his preferred and dominant language does not, however, constitute a threat to it. For reasons of cooperation and communication, such a speaker will use the other's language even if he has only a smattering of it. His accommodative talk may create the impression that he is not switching or mixing at all, leading the analyst to classify him, according to the criteria enshrined in 20, as very proficient in his L2. This accommodative/cooperative moment can be easily safeguarded against by adding the obvious provision that when the proficient bilingual switches his productions are grammatical vis-ä-vis the language he is using at a given point of time. How do we reconcile the scale with the facts reported in Poplack? We think the answer is easy. Poplack is dealing with the middle and lower ranges where the inability to mix, as opposed to the habit of not mixing, is indeed indicative of the lack of proficiency in L2 and the ability to insert only tags is indeed indicative only of rudimentary control of the other language. Her 'true bilinguals' are, in other words, the only ones that have any bilingual ability worth writing home about. We do not know what sort of competence in English and Spanish can be attributed to these speakers. The information that Poplack provides is somewhat circumstantial, though quite useful for other research questions—we know a lot about their age, sex, background, and attitudes but not much about their ability, for example, to sustain discourses in the two languages in question. Unfortunately, this is easier said than done, as

88 · Rajendra Singh and Ad Backus

measuring proficiency is fraught with pitfalls. Note that we have not provided very much either in the way of empirical data about separate monolingual proficiencies in the two languages. It would not have escaped the reader's attention that there is a sort of flipflop here—both the perfect bilingual and the almost perfect monolingual do not indulge in any mixing, the former obviously because he can afford not to and the latter equally obviously because he cannot. This is also the case with pause fillers and parentheticals—in the former case, they are put in for Gricean reasons but in the latter, for entirely linguistic reasons. This flip effect shows that it may not be necessary to go as far as Bentahila and Davies (1992) are willing to go in search for appropriate explanations, though they are certainly right in drawing attention to what they call communicative functions (ibid.: 456). It has been suggested that the differences between Poplack's informants and Singh's reside not where we think they do but in factors such as social class, education, and metalinguistic awareness. This translation of proficiency to what might loosely be called 'class' may be valid but it must be remembered that in the context in question it straightforwardly implies longer exposure to and proficiency in English. We have, in other words, no objection to adding the noun 'class' in conjunction or disjunction with the adjectives 'high' and 'low* in 21. Neither Backus' second generation nor Bentahila and Davies' informants could object because that is where they belong. As we have no intention of defending the scale in 21 to the last letter—we have not done the work necessary to propose a scale the particulars of which we would want to defend—let us just say that we hope to have persuaded you that such a scale is not only possible but also perhaps necessary not only to reconcile apparently contradictory results but also to integrate our findings into some general theory of bilingualism. The apparent contradictions we have looked at follow partly from the difficulty of using language to describe linguistic ability and partly from an entirely understandable liberal desire to project partial failure as partial success. An instrumentalist who is as bad at the sitar as at the tabla is after all quite balanced, and it is this rather unusual sense that the expression 'balanced bilingual' seems to have acquired in the current literature. This literature also seems to want to dismiss those who see mixing as evidence of disintegration. In setting linguists up against 'intellectuals'— and we hope she was not speaking for us all—all that Poplack (1980:588) may have been able to accomplish is to encourage the interpretation of 'a large degree of competence' as 'competence' itself. It is, therefore, difficult to see what is meant by saying that code-switching requires more linguistic competence in two languages than is generally assumed. Given that those who bemoan relexification and linguistic disintegration do not make quantitative statements, the liberal linguist's response is comparable to the assurance that partially skimmed milk contains more fat than is generally assumed. It is not a matter of more or less but of the fact that now plain, ordinary milk has to be

Bilingual Proficiency and Code-switching/Mixing Patterns · 89

labeled in a marked fashion—100 per cent milk. The lax defense offered by some linguists seems to us to be wide off the mark. Even the assurance that a new language may yet be born is a bit like telling a mother whose child is about to die that she is fertile enough to have another one. The solution is, in other words, not to pit linguists against intellectuals but to hope that observation and making sense can proceed together, as they did in the case of Weinreich, even if the sense making requires some counterfactual idealizations such as his ideal bilingual. • NOTES 1. The fact that questions of synchronic bilingualism are not relevant for situations the ethnographic description of which requires reference to only one lect is often ignored. Thus Agnihotri (1997:99) seems to claim (a) that the various constraints on code-mixing proposed by researchers are 'frequently violated in what we now call "Just a Normal Way of Talking"' and (6) that this 'Normal Way of Talking' is, in fact, an instance of bilingualism. Precisely because this 'Normal Way of Talking' observes no definable constraints, it provides evidence against the existence of bilingualism in the communities he is talking about. We must also add that appeals to the LePagian dichotomy 'focused/fluid' cannot help. If the 'Normal Way of Talking' is the only way of talking, it throws no light on proficiency in the two codes not even available for the simple reason that it can't. We have no quarrel with the claim that in 'fluid' communities 'there is a large part of human discourse that does not obey" any constraints. We are simply saying that that discourse is not bilingual discourse. 2. The metaphor is Singh's and he would like to leave the unpacking of it to his American colleagues. 3. The citation here has sometimes been misinterpreted, as in Annamalai's (forthcoming) review of Singh (19%). It is important to remind the reader that this is a citation and whilst we agree with the spirit of Adorno's sensitive observation the words are his and not ours.

• REFERENCES Adorno, T.W. 1992. Notes to literature. New York: Columbia University Press. Agnihotri, R.K. 1997. Review of R. Singh, Linguistic Theory, Language Contact, and Modem Hindustani. South Asian Language Review. 7(1). 93-101. Annamalai, E. (forthcoming). Review of R. Singh, Lectures Against Sociolinguistics. Journal of Pragmatics. Backus, A. 1992. Patterns of language mixing. A study in Turkish-Dutch bilingualism. Wiesbaden: Otto Harrassowitz. Backus, A. 1993. Bilingual Norms and Linguistic change. Paper presented at NWAVE 22, Universito d'Ottawa, Ottawa. . 19%. Two in one. Bilingual speech of Turkish immigrants in the Netherlands. Tilburg: Tilburg University Press. Backus, A. and R. Van Hout 1995. The distribution of codeswitches in bilingual conversations. Papers from the summer school on code-switching and language contact, pp. 16-28. Leeuwarden: Fryske Akademy. Bakker, P. 1992. A language of our own. Amsterdam: Universiteit van Amsterdam. Bentahila, A. and E. Davies. 1991. Constraints on code-switching: A look beyond grammar. Papers for the Symposium on Code-Switching in Bilingual Studies. Vol. 2, pp. 369-405. Strasbourg: E.S.F.

90 · Rajendra Singh and Ad Backus Bentahila, A. and E. Davies. 1992. Code-switching and language dominance. Cognitive processing in bilingual*, ed. by RJ. Harris, 443-58. Amsterdam: North Holland. . 1995. Patterns of code-switching and patterns of language contact. Lingua 96.75-93. Berk-Seligson, S. 1986. Linguistic constraints on intra-sentential code-switching: A study of Spanish/Hebrew bilingualism. Language in Society 15.313-48. Bhatia, T.K. 1989. Bilinguals' creativity and syntactic theory: Evidence for emerging grammars. World Englishes 8 (3). 265-76. Bhatt, R.M. 1994. Code-mixing and linguistic theory: Some preliminary remarks. Paper presented at the South Asian Linguistic Analysis (SALA) XVI Meeting, University of Pennsylvania, Philadelphia, May 20-22,1994. Bloomfield, L. 1927. Literate and illiterate speech. American Speech 10.432-39. Dabene, L. and J. Billiez. 1986. Code-switching in the speech of adolescents born of immigrant parents. Studies on Second Language Acquisition 8.309-25. Di Pietro, R. 1978. Code-switching as a verbal strategy among bilinguals. Aspects of bilingualism, ed. by M. Paradis, 275-82. Columbia, CS: Hornbeam Press. Dorian, N. 1982. Defining the speech community to include its working margins. Sociolinguistic variation in speech communities, ed. by S. Romaine, 25-33. London: Edward Arnold. Drapeau, L. (forthcoming). Code-Switching in Caretaker Speech and Bilingual Competence in a Native Village of Northern Quebec. IJSL. Eliasson, S. 1990. English-Maori language contact. Languages in contact: Proceedings of the Symposium 16.1 of the 12th International Congress of Anthropological and Ethnological Sciences, ed. by R. Filipovicand M. Bratanic, 33-49. Zagreb: Institute of Linguistics, University of Zagreb. . 1991. Models and constraints in code-switching Theory. Workshop on constraints, conditions, and models, pp. 17-50. Strasbourg: E.S.E Eruera, T. 1975. An explanatory syntactic analysis of grasshopper and Maori-English utterances. Auckland: Department of Anthropology, University of Aucklands. Giesbers, H. 1989. Code switching Tussen Dialect en Siandaardtaal. Amsterdam: PJ. Meertens Instituut voor Dialectologie. Grosjean, E 1982. Life With Two Languages. Cambridge, Mass.: Harvard. . 1991. Comments on Hyltenstam. Papers for the Workshop on Constraints, Conditions, and Models, pp. 255-58. Strasbourg: E.S.F. . 1992. Another view of bilingualism. Cognitive processing in bilinguals, ed. by R. Harris, pp. 51-62. Amsterdam: Elsevier. . (in press). The on-line processing of speech: Lexical Access in Bilinguals. The Bilingual Brain, ed. by P. Bhatt and R. Danes. Toronto: C.U.P. Halliday, M A.K. 1975. Learning how to mean—Explorations in the development of language. London: Edward Arnold. Hyltenstam, K. 1991. Language mixing in Alzheimer's dementia. Papers for the Workshop on Constraints, Conditions, and Models, pp. 221-54. Strasbourg: E.S.E Jacobson, R. 1989. Socioeconomic status as a factor in the selection of encoding strategies in mixed discourse. Codeswitching as a Worldwide Phenomenon, ed. by R. Jacobson, 111-39. New York: Peter Lang. Kachru, B. 1978. Toward structuring code-mixing: An Indian perspective. IJSL 16.27-46. . 1982. The bilingual's linguistic repertoire. Issues in International Bilingual Education, The role of the vernacular, ed. by B. Hartford, A. Valdman and C. Foster, 25-52. New York: Plenum Press. U Wei and L. Milroy. 1995. Conversational code-switching in a Chinese community in Britain: A sequential analysis. Journal of Pragmatics 23.199-281. Mougeon, R. and E. Beniak. 1991. Linguistic consequences of language contact. Oxford: Clarendon Press. Myers-Scotton, C. 1993. Duelling languages. Grammatical structure in codeswitching. Oxford: Clarendon Press.

Bilingual Proficiency and Code-switching/Mixing Patterns m 91 Nishimura, M. 1995. A functional analysis of Japanese/English code-switching. Journal of Pragmatics 23.157-81. Nortier, J. 1990. Dutch-Moroccan-Arabic code switching. Dordrecht: Foris. Pandit, E. 1986. Hindi-English code switching, mixed Hindi-English. Delhi: Datta Book Center. Pandit, P.B. 1969. Comments on 'How we measure the effects which one language may have on the other in the speech of bilinguals'. Description and measurement of bilingualism ed. by L.G. Kelley. Toronto: University of Toronto Press. Reprinted in R. Singh (ed.), 1997, Grammar, Language, and Society: Contemporary Indian Contributions. New Delhi: Sage Publications. . 1977. Language in a plural society: The case of India. Delhi: Delhi University Press. Pattanayak, D.P. 1981. Multilingualism and mother tongue education. Delhi: Oxford University Press. Pfaff, C. 1976. Functional and structural constraints of syntactic variation in code-switching. Papers from the Parasession on Diachronie Syntax, Chicago: CLS. . 1979. Constraints on Language-Mixing: Intrasentencial Code-switching and Borrowing in Spanish/English. Language 55.291-318. Poplack, S. 1980. Sometimes I'll start a sentence in Spanish. Linguistics 18.581-618. Prabhu, N.S. 1989. The mathetic function of English as a world language. Paper presented at the TESL Conference, Islamabad, Pakistan, January 1989. Rao, C.R. 1992. Statistics and truth: Putting chance to work. Bartonsvilles, Maryland: International Cooperative Publishing House. Singh, R. 1985. Grammatical Constraints on Code-switching: Evidence from Hindi/English. Canadian Journal of Linguistics 30.33-45. . 1992. Code Mixing: A comparative study. Paper presented at the Max Planck Institute, Nijmegan. . 1994. Indian English: Some conceptual issues. Second language acquisition, ed. by R.K. Agnihotri and A.L· Khanna, 369-81. Thousand Oaks: Sage Publications. . 1995a. Hindi-English Code-Mixing: Some Implications for a General Theory. Indian Linguistics. (Revised version of the paper presented at Max Planck, 1992.) . 1995b. 'New/Non-Native' Englishes Revisited: A Reply to My Colleagues. Journal of Pragmatics 24 (3). 323-33. . 1995c. Code-switching/Mixing patterns and bilingual proficiency. Paper presented at the Leeuwarden Summer School on Code-switching and Bilingualism Leeuwarden: Fryske Akademy. . 1996. Lectures against socioünguistics. New York: Peter Lang. . (in press). On Bilingual Lexical Access. The Bilingual Brain ed. by P. Bhatt and R. Davies. Toronto: C.U.P. Singh, R. et al. 1995. On 'New/Non-Native' Englishes. Journal of Pragmatics 24 (3). 283-94. Srivastava, R.N. 1997 [1977]. Indian bilingualism: Myth and reality. Grammar, language, and society: Contemporary Indian contributions, ed. by R. Singh, 218-46. New Delhi: Sage Publications. Torres, L. 1989. Code-mixing and borrowing in a New York Puerto Rican community: A crossgenerational study. World Englishes 8 (3). 419-32. Tosi, A. 1984. Immigration and bilingual education: A case study of movement of population, language change and education within the EEC. Oxford: Pergamon Press. Weinreich, U. 1953. Languages in contact. The Hague: Mouton.

Syntactic Typology and South Asian Languages I K.V. SUBBARAO I

This paper attempts to highlight some of the contributions that the study of typology of South Asian languages can make for a better understanding of the nature of human language. We wish to demonstrate how evidence from the study of lexical anaphors, conjunctive participles and pro drop in South Asian languages has implications to issues discussed in syntactic theory. In the end we present evidence that can be adduced from South Asian languages concerning the mental organization of language which reflects cognitive capabilities of the human mind. We intend to show that it is the cognitive capability of the human mind that enables it to mentally classify phenomena that share cognitive/semantic capabilities and to assign a single lexical category for such seemingly unrelated phenomena transcending genetic barriers.

• 1. Introduction There are more than a thousand languages spoken in the South Asian subcontinent and they belong to four genetically different language families: IndoAryan, Dravidian, Tibeto-Burman, and Austro-Asiatic. It is rare that one finds such a conglomeration of languages in one 'area'. Coupled with intense bi/ multilingualism that has been in existence in the subcontinent for thousands of years, the study of the syntax of these languages becomes much more fascinating.

• 2. Polymorphemic Anaphors South Asian languages have both a nominal and a verbal device for the interpretation of the lexical anaphors. The nominal device is either a 'simplex'

94 · KV. Subbarao

monomorphemic form as in Gujarati or a 'complex' polymorphemic form as in Hindi-Urdu, Marathi, Oriya, Mizo, Hmar, Telugu, Kannada, etc. A language may have a monomorphemic and a polymo hemic form as in Dravidian languages, some of the Tibeto-Burman and Indo-Aryan languages. Just as in Chinese, Korean, Italian (Cole and Sung 1994), many of the South Asian languages permit long-distance coindexation of the simplex monomorphemic anaphor with an antecedent in a higher clause (Davison 1999; Subbarao 1971). For example: Hindi-Urdu (1) ashoki ne saritaa, se apne liyel}j caay banaane ko kahaa Ashok erg Sarita to self for tea to make said 'Ashok told Sarita to make tea for him/herself In one interpretation apne liye 'for self can be coindexed with the embedded subject PRO1 which in turn, is coindexed with the matrix object sarita, thus permitting binding of an anaphor which obeys Principle A. However, apne liye 'for self can also be coindexed with the subject of the matrix sentence ashok which is a 'long-distance' antecedent. Such interpretation of the anaphor is an apparent violation of the Principle A of the binding theory which states that an anaphor must be bound in its governing category. One of the ways to account for the long-distance binding of the simplex anaphor apne liye 'for self without violating Principle A is to move the anaphor by head to head movement rule (Baker 1987) as a simplex anaphor is an X° level category. At the outset, such a solution may not sound very appealing but considering the fact that head to head movement is one of the permissible options of the Universal Grammar UG and such option does the job of accounting for the long-distance binding of a simplex anaphor, such a solution is welcome. There are, however, some South Asian languages such as Marathi (Wali 1999), Oriya (Ray 1999) and Hmar (Subbarao 1998) which permit longdistance coindexation of a complex anaphor. As an illustration, we shall consider the case of Hmar. Hmar is a Tibeto-Burman language spoken mainly in Assam and Manipur. Hmar has a simplex monomorphemic anaphor ama 'self and a complex polymorphemic anaphor ama leama 'self and self in 3rd person. It also has a verbal reflexive marker m which cliticizes onto the verb. Both the simplex as well as complex anaphor permit long-distance binding. We shall focus our attention on the long-distance binding of a polymorphemic anaphor in 2 in Hmar. Hmar (2) lalirn tluangij [[PROj amaa le airway thingpui ini/r siem- ding] Lali erg Tluangi self and self tea VR make PUT in] a- ti comp 3SG asked 'Lalij asked Tluangij to make some tea for herself^'.

Syntactic Typology and South Asian Languages · 95

In one interpretation, ama le ama 'self and self and m, the verbal reflexive, are coindexed with PRO, the subject of the embedded clause, which in turn is coindexed with tluangi which is the indirect object of the matrix clause. Since the polymorphemic anaphor and the VR are coindexed with PRO and PRO binds them, it is a case of local binding and Principle A is obeyed. In the second interpretation, ama le ama 'self and self and in, the VR, are coindexed with tali which is the subject of the matrix clause. Since the nominal anaphor and the VR occur in the embedded clause and the antecedent in a higher clause, it is a case of long-distance binding. Just as the long-distance binding of simplex anaphors is subject-oriented (Cole and Sung 1994), the long-distance binding of the complex anaphor too is subject-oriented. To account for the long-distance binding of a polymorphemic anaphor, Subbarao (1997) uses cross-linguistics evidence which is as follows: Languages such as Kannada, Telugu, Mizo, and Hmar have a nominal device and a verbal device as lexical anaphors. In these languages, the presence of the nominal device which is morphologically complex is optional whereas the occurrence of the verbal device is obligatory. The verbal device in these languages is monomorphemic. It does not have any independent existence of its own and hence, it 'cliticizes' onto the main verb. Since the occurrence of the verbal device is obligatory and that of the polymorphemic anaphor is optional, it is hypothesized that in languages such as Hmar, it is the clitic as a zero level category that moves upwards and not the nominal polymorphemic form. Clitics according to Chomsky (1995) exhibit the properties of X° and XP level categories. The clitic as a zero level category is moved by Jhe head to head movement rule (Baker 1987) at LF. This suggestion encounters a problem when confronted with data from Telugu, a Dravidian language, where the presence of VR permits only local binding and strictly prohibits long-distance binding (Subbarao and Lalitha Murthy 1999). Telugu (3) kamalui shaguftoj tana· ni - tanu*nj puguDu- koo- waDam windi Kamala Shagufta self ace self praise VR nominalizer heard 'Kamalaj heard Shaguftaj praising herself-i/j'. tana-ni-tanu 'self-acc-self—which is a reduplicated anaphor in Telugu (Subbarao and Saxena 1987) is coindexed with the embedded subject shagufia and it cannot be coindexed with kamala, the subject of the matrix sentence. It is the verbal reflexive kon which is a clitic that insulates the anaphor from coindexation with a long-distance antecedent. The solution proposed to account for long-distance binding of polymorphemic anaphors in Hmar should be applicable to Telugu polymorphemic anaphors too. If it is applied to data from Telugu, polymorphemic anaphors in Telugu too should permit long-distance binding. The fact that it is not the case

96 · K.V. Subbarao shows that further research needs to be done on long-distance binding of polymorphemic anaphors.

• 3. Control Theory The second issue that we consider concerns control theory. The notional subject that occurs in the subject position of a non-finite clause is governed by the PRO theorem which states that PRO is ungoverned and uncasemarked. (4) Johnj wants [PROj to meet Mary] PRO which is the notional subject of the embedded clause is neither casemarked nor governed as infinitives are non-finite and cannot assign case to their subject. In the conjunctive participial construction in South-Asian languages (Abbi 1984; Davison 1981; Masica 1991) the conjunctive participle (CP hereafter) in Hindi-Urdu, Punjabi and Kashmiri is devoid of tense and hence, non-finite. In contrast, in languages such as Telugu, Tamil, Kannada, Oriya and Bengali, the CP carries the past tense marker and hence, is finite. Irrespective of the fact whether the CP is finite or non-finite, the notional subject is not overtly present in the following examples from Hindi-Urdu and Telugu. Hindi-Urdu (5) PROj HaNs haNs kar baccci paagal ho-gae laugh laugh cp mkr children crazy became 'Having laughed a lot, the children went crazy/ Telugu (6) PROi now -i naw -i pillala -kit kaDupu noppi laugh cp mkr laugh cp mkr children dat stomach pain waccindi came 'The children almost got a stomach ache due to excessive laughing.' Lalitha Murthy (1994) points out that while PRO in Hindi-Urdu occurs in an ungoverned uncasemarked position, PRO in Telugu occurs in a casemarked, governed position because the conjunctive participial marker is also a past tense marker in Telugu. Thus, the CP is marked [+ tense]. It is in languages such as Telugu, Tamil, Kannada, Bengali, Oriya that a lexical subject occurs as the subject of a CP clause. The following example from Tfelugu is illustrative. Telugu (7) waanalu paD-i paD-i panTalu paaDu ayyeyi rains-nom fall cp mkr fall cp mkr crops-nom spoiled got

Syntactic Typology and South Asian Languages m 97

'Rains having fallen, the crops got spoiled' (literal) 'The crops got ruined due to excessive rains.' waanalu 'rains' is the subject of the conjunctive participial clause zndpanTalu 'crops' is the subject of the matrix clause and they are not identical. Lalitha Murthy (ibid.: 46) convincingly argues that the finite past tense marker -/ in Telugu assigns nominative case to the lexical subject waanalu 'rains' in the CP clause. Thus, the occurrence of the lexical subject in the CP constructions in languages such as Telugu, Tamil, Kannada, Oriya, Bengali, and Manipuri and the non-occurrence of the lexical subject in the CP constructions in Hindi-Urdu, Punjabi and Kashmiri is accounted for. Lalitha Murthy (ibid.: 46) points out that the PRO theorem cannot be a universal as PRO is governed and casemarked in Telugu, Kannada, Bengali, Oriya, Manipuri, etc. and hence, it is a parameter. Thus, the analysis of data concerning the occurrence of notional subjects in subordinate clauses has implications for the PRO theorem which is a putative universal.

• 4. Principle C The third issue that we discuss concerns the three principles—Principle A, B, and C of binding theory which deal with the classification of nominal expressions. The three principles are stated below: Principle A: An anaphor must be bound in a local domain. Principle B: A pronoun must be free in a local domain. Principle C: An r-expression must be free (Chomsky 1995: 96). According to Principle C, a referential expression must be free. Jayaseelan (1999) and Subbarao and Lalitha Murthy (forthcoming) provide data which clearly show that Principle C is violated in Malayalam and Telugu. We provide examples from Telugu. Telugu

(8) raadha/i amma -ki raadha/i- ni /danni/i baagu ceyyaali ani Radha mother dat Radha ace her must set right comp aasalkoorika desire/wish 'Radha's/i mother's desire was to set her/j right' (9) raamuDu/i raamuDi/j inT- loo unDaka inka ekkaDa Ramudu Ramudu's house in if not staying else where would unTaaDu? stay 'If Ramudu/j does not stay in Ramudu's/j house, where else would he stay?'

98 · KV. Subbarao The referential expressions raadha in 8 and raamuDu in 9 in Telugu are coindexed with an antecedent in a sentence which is a clear violation of Principle C of the binding theory. In Rabha too, a Tibeto-Burman language spoken in Assam, a referential expression is coindexed with an antecedent. The dative experiencer subject parmai in 10 is coindexed with u 'she' the subject of the matrix sentence in 10 and the experiencer/wnrwi in 11 is coindexed vnlh parmai, the subject of the matrix clause, which is a violation of Principle C. Rabha (10) parmaif. na slmar -e ur be nOkO dp- ba Parmai dat hot cp mkr she nom door close PST 'Having felt hot, Parmai closed the door' (Subbarao et al forthcoming). (11) parmai- na slkham -e parmai -be nOkO dp- ba Parmai dat cold cp mkr Parmai nom door close PST 'Having felt cold, Parmai closed the door' (ibid.). The question that arises is: Do these violations of Principle C in Malayalam, Telugu and Rabha force us to reconsider the validity of Principle C? I do not have any answer except to say that further research might shed light on such issues.

• 5. Agreement and Verbal Anaphoric Device The fourth issue concerns the phenomenon of subject-verb agreement vis-avis the occurrence of a verbal anaphoric device. The verbal anaphoric device is a clitic which can be a verbal reflexive or a verbal reciprocal or both in a language. The question we wish to address is: Is there any correlation between the occurrence of subject-verb agreement and verbal reflexive/reciprocal in a language? Are they mutually dependent? Is it the case that only those languages which exhibit subject-verb agreement are expected to have a verbal clitic as an anaphoric device? Prima-facie, it appears that such a correlation exists. Literary Dravidian languages such as Telugu, Tamil, Kannada and non-literary Dravidian languages exhibit subject-verb agreement and they also have a verbal reflexive/ reciprocal. The following example from Telugu is illustrative: Telugu (12) aa pillalu okaLLa too okaLLu debbalaaDu hon -Tunna those children one with one PL fight VREC PROG -ru 3PL 'Those children are fighting with each other.'

Syntactic Typology and South Asian Languages m 99

In 12 the verbal reciprocal kon as well as the subject-agreement marker -ðé occur. Malayalam, a Dravidian language, does not exhibit subject-verb agreement nor does it have a verbal reflexive/reciprocal (cf. Jayaseelan [1999] for further details). (13) raam tann-e tann -e aDicc - ø -u Ram self-ace self + EMPH hit 'Ram hit himself.' The verb aDiccu 'hit' (past) remains the same in all three persons, both numbers and three genders. The absence of VR is indicated by ö in 13. On the basis of data from several languages such as Telugu, Kannada, Mizo on the one hand and Malayalam on the other hand, we might be tempted to arrive at a tentative conclusion that only those languages which exhibit subject-verb agreement have a verbal anaphoric device. However, such a conclusion is not tenable when we look at Tibeto-Burman languages such as Manipuri and Rabha. Languages can be classified into four types with regard to the occurrence of a verbal reflexive/reciprocal vis- -vis verb agreement. We shall provide examples of languages from the subcontinent. Type 1: Those languages which exhibit verb agreement and have a verbal reflexive/reciprocal as in Telugu, Kannada of the Dravidian family, Mizo and Hmar of the Tibeto-Burman family. Type 2: Those languages which do not have subject-verb agreement and a verbal reflexive/reciprocal as well. For example, Angami and Tangkhul of the Tibeto-Burman family, Malayalam of the Dravidian family. Type 3: Those languages which do not have subject-verb agreement but have a verbal reflexive/reciprocal. Those include Manipuri and Rabha of the Tibeto-Burman family. Type 4: Those languages which have subject-verb agreement but no verbal reflexive/reciprocal. These include Hindi-Urdu, Bengali, Assamese, Gujarati, etc. The above classification may be presented in the following table: Verbal Agreement

+ Verbal Reflexive

Telugu Kannada Mizo Hmar

Manipuri Rabha

100 · KV. Subbarao Verbal Reflexive -

Hindi-Urdu Bengali Assamese

Malayalam Angami Tangkhul

Since our discussion concerns languages of Type 3, we provide examples from these two languages. The verbal reflexive in Manipuri may optionally be dropped in 14 which we have indicated in parentheses. In contrast, the occurrence of the verbal reciprocal in 15 is obligatory. Manipuri

(14) thombi -na mahak -na mass-bu- u- (je) -i Thombi nom she nom self ace see VR PRES 'Thombi looks at herself (15) nupimaca- sing (adu) -na makhoi masel thagat- na ~i girl PLDET nom each other praise VREC PRES 'The girls praise each other' In Rabha there is a polymorphemic nominal reflexive and reciprocal and there is a verbal reciprocal as a lexical anaphor. It does not have a verbal reflexive. Rabha

(16) OrOng (be) ina una nukbar -jln -nata they nom each other fall in love VREC PST PERF 'They had fallen in love with each other' (Subbarao et al. forthcoming). Both Rabha and Manipuri have verbal reflexive/reciprocal for anaphoric interpretation and they do not exhibit subject-verb agreement. It would be interesting to investigate if Proto Tibeto-Burman had agreement and a verbal device for anaphoric interpretation. It could be the case that it did because languages such as Mizo, Hmar, Aimol, Pake have a verbal and nominal device for anaphoric interpretation and subject verb agreement too and in others, it is the case that they have one or the other. It appears that in languages such as Manipuri and Rabha the feature of agreement got lost and the verbal device for anaphoric interpretation remained intact. To conclude, there is no universal correlation between the occurrence of verbal reflexive/reciprocal and subject-verb agreement.

• 6. Pro-Drop The next issue concerns the occurrence of pro, the null pronoun, in subject position in South Asian languages. It is generally agreed that null subject pro

Syntactk Typology and South Asian Languages m 101

occurs in languages which have 'rich' subject-verb agreement. Thus, 'it is the special status of the inflectional system of a language and its agreement markers that allows null subjects' (Jaeggli and Safir 1989). It is the phi-features (person, number and gender features) which permit the identification of the empty category pro and its 'recoverability'. Chinese, for example, does not permit pro in the subject position of a simple clause as it is not identifiable but does permit pro in the embedded clause provided the embedded subject is coindexed with the matrix subject as 17 and 18 illustrate: (17) zhangsan qi ma qi de [PRO hen lei] Zhangsan ride horse ride till very tired 'Zhangsan rode the horse until he got very tired.' (18) [PROyi hui dao jia] zhangsan jiu ku one return to home Zhangsan then cry As soon as he arrived home, Zhangsan began to cry' (Jaeggli and Safir 1989). As expected, languages like Hindi-Urdu, Kashmiri, Punjabi, Telugu, Kannada that exhibit a rich subject-verb agreement phenomenon permit pro drop of the subject. However, Tibeto-Burman languages like Angami, Manipuri, Rabha, Boro which do not exhibit subject-verb agreement permit pro drop of the subject of a simple and an embedded sentence as well. The following examples are illustrative: Angami (19) pro vO · tyo go will 'I/You/He/Shewillgo.' Rabha (20) pro riba -nO go will 'I/You/He/Shewillgo.' Manipuri (21) pro cat - kani go will 'I/You/He/She will go.' Boro (22) pro thang -In go will º/You/He/She will go.' When the subject of the matrix sentence is identical to the subject of the embedded sentence, the embedded subject is pro-dropped in Rabha and Angami.

102 · K.V. Subbarao Rabha (23) dapheng - be gaphung [ pro shilOng ina rengO Dapheng nom tome Shillong to will go 'Dapheng told me that he would go to Shillong.' Angami (24) dapheng -e [ pro shilOng- nu vO- tyoDapheng nommkr Shillong to go PUT 'Dapheng told me that he would go to Shillong/

Ine] kaniba that told I-di] aki pu that me told

In Angami pro drop of the embedded subject is obligatory. If an overt pronoun is present, it cannot be coreferential with the matrix subject as in 25. (25) daphengi -e \puo^ shilOng -nu vO -tyo -I-di] aki pu Dapheng nom Shillong to go PUT that me told 'Dapheng told me that he would go to Shillong.' In Boro a pronoun that occurs in subject position of an embedded sentence can be co-referential with the matrix subject. However, the preferred option is either to have a pro or a pronoun or a bound pronoun gau 'he/she'. Boro (26) thomba -ya [ projbiryl/ gaui -na shilOng - ao Thomba NOM he nom he (bound) nom Shillong to khlnthayl thang-gln hanna] ang-nl told will go that me 'Thomba told me that he'd go to Shillong.' In Manipuri on the other hand pro-drop of the embedded subject is optional. Manipuri (27) thomba -na aingonda [ pro/mahak shilOng - da cat -kani Thomba nom I + dat he Shillong to go will haina] hairammi QUOT told 'Thomba told me that he'd go to Shillong.' To summarize the above discussion, pro-drop which otherwise is optional in languages with 'rich agreement' becomes obligatory in Angami and a preferred option in Boro which exhibits 'zero/null agreement'. We shall now examine the occurrence of pro in left-adjoined complement clauses. In Rabha, Angami, Manipuri, and Boro, the embedded sentence can be left-adjoined to the matrix clause as 28-30 illustrate: In 28 in Rabha u (be) 'he' the subject of the left-adjoined clause, cannot be pro dropped.

Syntactic Typology and South Asian Languages · 103 Rabha (28) [ui (be) shilOng ina rengO Ine] daphengi be gaphung he nom Shillong to will go that Dapheng nom tome kaniba told 'Dapheng told me that he'd go to Shillong.' As shown in 29, in Angamipuo 'he' the subject of the left-adjoined clause cannot be co-referential with the matrix subject. And hence it cannot be prodropped. Angami (29) \puo-e*iQ shilOng -nu vO -tyo- I-di] thombai-e pu he nom Shillong to go FUT that Thomba nom said 'Thomba said that he'd go to Shillong.' In Manipuri, mahak 'he', the embedded subject in 30, is not co-referential with the matrix subject. If co-reference of the matrix and embedded subject is intended, pro has to occur as the embedded subject and the embedded subject pronoun cannot be overtly present as in 31. Manipuri (30) mahak*nj imphal -da cat-kani haina thombai na he Imphal to go willQUOT Thomba nom aingonda hairammi I + dat told 'Thomba told me that he'd go to Imphal.' (31) proiimphal -da cat-kani haina thombai -na aingonda Imphal to go will QUOT Thomba nom I -I- dat told hairammi 'Thomba said that he'd go to Imphal.' The case of Boro is slightly different as Boro has a bound pronoun occurring in the embedded sentence. Boro (32) [bfyli/j shillong -ao thang-gln hanna] thombai -ya ang-nl he nom Shillong to go will that Thomba nom me khlnthayl told 'Thomba told me that he'd go to Shillong.' The embedded subject bi 'he' is co-referential with the matrix subject as well as a discourse antecedent. If co-reference of the embedded subject with the

104 · KV. Subbarao

matrix subject alone is intended, bi 'he' the subject of the embedded subject has to be obligatorily PRO-dropped organ, a bound pronoun, occurs in Boro as in 33 (see Wali and Subbarao 1991 for a discussion of the bound pronoun). Boro

(33) [projgaunaj shillong -ao thang-gln hanna] thombai -ya ang-nl he (bound) Shillong to go will that Thomba nom me khlnthayl told 'Thomba told me that he'd go to Shillong.' To summarize the above discussion, while in Rabha the embedded subject cannot be pro-dropped in left adjoined clauses, in Angami and Manipuri it has to be pro-dropped for intended co-reference. In Boro it is the preferred option. Thus, the issue of pro-drop in left adjoined clauses appears to be quite complex and needs to be worked out in detail. The crucial point that needs to be underscored is that pro-drop that normally is permitted in languages with rich agreement is permitted in languages with no subject verb agreement in Tibeto-Burman languages and pro-drop that is otherwise optional is obligatory in languages in Angami and Manipuri in embedded clauses.

• 7. Mental Organization of Language In the concluding part of this paper we wish to present evidence that can be adduced from South Asian languages concerning the mental organization of language that reflects cognitive capabilities of the human mind. There are grammatical facts in languages which outwardly/externally appear to be unrelated though they share a common cognitive link. It is therefore reasonable to expect that such grammatical facts are grouped together under a single head in the mental lexicon. We wish to demonstrate that it is the cognitive capability of the human mind that enables it to mentally classify phenomena that share cognitive/semantic commonalities and to assign a single lexical category for such seemingly unrelated phenomena across languages transcending genetic boundaries. The first piece of evidence concerns the occurrence of a marker which functions as a verbal reflexive/reciprocal and as an inchoative marker in languages which are genetically unrelated. In South Asian languages there is a nominal device and a verbal device that functions as a lexical anaphor. This lexical anaphor (nominal and verbal) is coindexed with an antecedent that is local or long-distance. In some South Asian languages there occurs an inchoative marker that detransitivizes a transitive verb. At the surface it appears that the verbal

Syntactic Typology and South Asian Languages m 105 reflexive/reciprocal and an inchoative marker have nothing in common. However, Telugu, Tamil, Kannada from the Dravidian language family and Mizo, Hmar, and Manipuri from the Tibeto-Burman language family have a single clitic which performs both the functions, namely, a verbal reflexive/reciprocal and an inchoative marker. The verbal reflexive/reciprocal clitic and the inchoative marker in Telugu is kon, is koL in Tamil and Kannada and in Mizo and Hmar. In Manipuri, the verbal reflexive and reciprocal clitics are different in form: ja is the verbal reflexive and na is the verbal reciprocal. In Manipuri it is the verbal reflexive that functions as the inchoative marker and not the verbal reciprocal (Sarju in progress). The verbal clitic in all these languages functions like a detransitivizer. Hence, in Mizo (Lalitha Murthy and Subbarao 1999), Hmar (Subbarao 1998) and Dyirbal (Dixon 1980) the ergative marker does not occur when a verbal reflexive occurs. In French and Romanian too the reflexive clitic and the inchoative marker are homophonous (Sabina Poparlan, in personal communication). In languages belonging to four different language families, it is not fortuitous that the same marker functions as a verbal anaphoric device and an inchoative marker. The common feature that is shared between a verbal reflexive and an inchoative marker is intransitivity. This, in our opinion, reflects the cognitive capability of the human mind to classify phenomena which share cognitive commonalties under a single head and to assign a single lexical unit to capture it. The second piece of evidence comes from a clitic which functions as a linker in relative clauses which links the embedded relative clause with the matrix clause in languages which are genetically unrelated. There is a clitic in such languages which expresses doubt, apprehension or fear. It is the same clitic which occurs in embedded questions linking the embedded clause with the main clause. This marker is labeled as a dubitative marker in the grammars of Dravidian languages. It is interesting to note that the same clitic performs all the three functions namely, as a linker in relative clauses and embedded questions and as a dubitative marker in languages belonging to different language families. The clitic in Dravidian languages is oo, in Dakkhini Hindi-Urdu it is ki (Subbarao and Arora 1989), in Manglore Konkani it is ki (Nadkarni 1975), in Angami it is si (Kevichusa 1997; Kevichusa and Subbarao forthcoming). And in Sema it is kienO. Thus, languages which belong to four different language families have the same lexical item which performs a variety of functions. At the surface, these functions appear to be quite unrelated. The question that needs to be answered is: Is it an accident that there is a single clitic which performs all these functions, or is it the innate mental capabilities of the human mind transcending genetic family divisions that enables it to connect such phenomena in the mental lexicon? We think that latter is the case.

106 · KV. Subbarao

m 8. Conclusion «» I attempted to bring in crucial data from South Asian languages which have implications for the universal principles proposed in the theory of Government and Binding framework and I also hinted at how seemingly unrelated phenomena are mentally organized and how such organization reflects the cognitive capabilities of the human mind in grouping such phenomena together under a single head in the mental lexicon. • ACKNOWLEDGEMENTS

I am grateful to Ramakant Agnihotri for his valuable comments and suggestions. I wish to thank following language consultants who have always been very enthusiastic and cooperative in providing us data: Ms. Mimi Kevichusa for Angami, Ms. Deepamani Boro for Boro, Ms. Sarju Thokchom for Manipuri, Ms. Sabina Poparlan for Romanian and Dr. Upen Rabha Hakacham for Rabha. I thank Mr. Vijay K. Thakur and Ms. Mimi Kevichusa for their valuable help in the preparation of the manuscript. • NOTES 1. Abbreviations used are: ace: accusative, comp: complementizer, CP: conjunctive participle, dat: dative, DET: determiner, EMPH: emphatic, erg: ergative, PUT: future, mkr: marker, nom: nomionative, PERF: perfect, PL· plural, PRES: present, PROG: progressive, PST past, QUOT: quotative, SG: singular, VR: verbal reflexive, VREC: verbal reciprocal. • REFERENCES Abbi, Anvita. 1984. The conjunctive participle in Hindi-Urdu. InternationalJoumal ofDravidian Linguistics 13.252-63. Baker, Mark C. 1987. Incorporation: A theory of grammatical function change. Chicago: The University of Chicago Press. Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press. Cole, Peter and Li-May Sung. 1994. Head movement and long-distance reflexives. Linguistic Inquiry 25 (3). 355-406. Comrie, Bernard. 1998. South Asian languages and linguistic typology. The yearbook of South Asian language and linguistics, ed. by R. Singh, 237-46. New Delhi: Sage Publications. Darison, Alice. 1981. Syntactic and semantic indeterminancy resolved. Radical pragmatics, ed. by Peter Cole, 101-28. New York: Academic Press. . 1999. Lexical anaphors and pronouns in Hindi. Lexical Anaphors and Pronouns in Selected South Asian Languages, ed. by Barbara, C. Lust, Kashi Wali, James W. Gair, and K.V. Subbarao. Berlin: Mouton de Gruyter. Dixon, R.M.W. 1980. The languages of Australia. Cambridge and New York: Cambridge University Press.

Syntactic Typology and South Asian Languages m 107 Jaeggli, Oswald and Ken Safir. 1989. The null subject parameter. Dordecht: Kluwer. Jayaseelan, KA. 1999. Lexical anaphors and pronouns in Malayalam. Lexical Anaphors and Pronouns in Selected South Asian Languages, ed. by Barbara C. Lust, Kashi Wali, James Gair and KLV. Subbarao, 113-68. Berlin: Mouton de Gruyter. Kevichusa, Mimi. 1997. Relative clause formation in Angami. University of Delhi: M.Phil dissertation. Kevicbusa, Mimi and K.V. Subbarao. Forthcoming. Relative clauses in Angami. Indian Linguistics. Lust, Barbara, C, Kashi Wali, James W Gair and K.V. Subbarao (eds.) Selected lexical anaphors and pronouns in South Asian languages. Berlin: Mouton de Gruyter. Masica, Colin R 1991. The Indo Aryan languages. Cambridge: Cambridge University Press. Murthy, Lalitha B. 1994. Participle constructions: A cross-linguistic study. Delhi: University of Delhi, Ph.D. thesis. Lalitha B. Murthy and K.V. Subbarao. 1999. Lexical anaphors and pronouns in Mizo. Lexical Anaphors and Pronouns in Selected South Asian Languages, ed. by Barbara C. Lust, Kashi Wali, James Gair and K.V. Subbarao, 777-838. Berlin: Mouton de Gruyter. Nadkarni, M.V. 1975. Noun phrase embedded structures in Kannada and Konkani. Language 51 (3). 672-83. Ray, Tapes. 1999. Lexical anaphors and pronouns in Oriya. Lexical Anaphors and Pronouns in Selected South Asian Languages, ed. by Barbara C. Lust, Kashi Wali, James Gair and K.V. Subbarao. Berlin: Mouton de Gruyter. Sarju, Thokchom. In progress. Lexical anaphors in Manipuri. University of Delhi: M.Phil dissertation. Subbarao, K.V. 1971. A note on reflexivizition in Hindi. Studies in the Linguistic Sciences 1, ed. by Yamuna Kachru, 180-214. Champaign-Urbana, IL: University of Illinois . 1997. Linguistic theory and syntactic typology: A proposal for a symbiotic relationship. Keynote address delivered at the International Conference on South Asian Languages. Moscow: Moscow State University. . 1998. Lexical anaphors in Hmar. Delhi: University of Delhi, ms. Subbarao, K.V., and Anju Saxena. 1987. Reflexives and reciprocals in Dravidian. Studies in the Linguistic Sciences 17.1. Urbana-Champaign, IL: University of Illinois. Subbarao, K.V., and Harbir Arora. 1989. On extreme convergence: The case of Dakkhini HindiUrdu. Indian linguistics. . 1999. Lexical anaphors and pronouns in Telugu. Lexical Anaphors and Pronouns in Selected South Asian Languages, ed. by Barbara C. Lust, Kashi Wali, James Gair and K.V. Subbarao, 217-373. Berlin: Mouton de Gruyter. . Forthcoming. Aspects of Mizo syntax. Delhi: University of Delhi. Subbarao, K.Y, Manashi Gogoi Dutta and U pen Rabha Hakacham. Forthcoming. Ë grammatical sketch of Rabha. Guwahati, Assam: ABILAC. Wali, Kashi. 1999. Lexical anaphors and pronouns Marathi in Lexical Anaphors and Pronouns in Selected South Asian Languages, ed. by Barbara C. Lust, Kashi Wali, James Gair and K.V. Subbarao. Berlin: Mouton de Gruyter. Wali, Kashi, and K.V. Subbarao. 1991. On pronominal classification: Evidence from Marathi and Telugu. Linguistics 29.1093-110.

Tesniere Indicators and Indian Languages* IPROBALDASGUPTAI

Generative grammar has been moving closer to the tools of some type of dependency grammar since the 1980s, a point that has been explicit since Chomsky (1986: 208, n.37). But traditions do not live by tools alone. In this paper we offer reasons why generative grammar should borrow, not just the notion of indices in the sense of the dependency grammarian Lucien Tesniere (Tesniere 1959) (a tradition which shares only some traits with the sort of dependency grammar Chomsky [1986] had in mind), but also part of its methodological habitat. Since current generative usage reserves the word 'index' for reference marking devices, our proposed adaptation introduces the term Tesniere Indicators, not "indices', to avoid confusion. Beginning with conceptual considerations, we gradually move into the empirical basis for our proposals. What may be called the double infinity of syntax is a useful point of departure for the study of how best to represent the active exercise of freedom in language. The first infinity, or the familiar unboundedness of the set of syntactic objects that speakers of language L know, elicits research question one, RQ1: Given the infinity of sentences in L, how can a finite grammar adequately specify the membership of this infinite set? There is also a second infinity of syntax at the level of multiple syntaxes for L. McCawley's book Thirty million theories of grammar had only near-standard transformational versions of generative theory in mind. The number of reasonable candidate * I gratefully acknowledge Shastri Indo-Canadian Institute's 1997 Social Sciences and Humanities Fellowship and the help of Rajendra Singh at the Universito de Montreal in research that has contributed to the outcome presented here. Standard disclaimers apply.

110 · Probai Dasgupta

grammars for L is infinite. This point takes on theoretical significance in a dependency grammar context, as we shall see. Now, the corresponding research question RQ2 is: Given the second infinity, how can the finite community of linguists agree on ways of conducting the grammatical enterprise that reliably promise convergence? Generative grammarians ordinarily focus only on the first infinity. They consider the young child, YC, acquiring L. She must face finite data, consider finitely many grammar-options and quickly settle for specific outcomes. If we linguists keep YC in view, so the default generative doctrine runs, we will converge on a reconstruction of the language faculty enabling her to achieve these outcomes. The second infinity, a mere artifact of the youth of formal linguistics, will wither away as we advance. We propose to reset the default generative assumptions. In particular, we suggest taking both RQ2 and RQ1 seriously. Then we correspondingly iconize an older child OC. Having finished her basic acquisition of L, OC is explicitly learning about the heterogeneity of the variants of L she needs to comprehend, if not produce, across communicative space. We extend the notions of literacy and literature to cover oral heritage transmission and argue that she is also learning about heterogeneity within L across literary time. As part of this process, in many societies, OC is also learning how to use cognitive means invented by grammarians to cope with this heterogeneity of varieties of L. To sum it up, OC is acquiring language proper, whereas YC was a figure acquiring grammatical form alone. The relation between the two idealizations, YCI and OCI, is asymmetric. From OCI as a point of departure we can credibly derive YC-focused work as a special case. This derivation makes YC-focused research rational but YCI an inappropriately overextended idealization. Notice that no natural route leads from a YCI starting point to OCI reasoning. Given the availability of OCI, those who keep using YCI may be expected to meet this obvious challenge within their perspective. Pending such discussion, the task is to elaborate the grounds and consequences of OCI. Here we make our moves slowly to maximize explicitness and refutability. Generative research of course tries for international convergence on a viable parametric account of the typology of grammatical options. But success in this endeavor is seen as a matter of contingent luck. Normal generative discourse takes it that the infinite range of solutions available to the linguistic researcher is not a scientifically relevant infinity, not an object deserving theoretical attention that might shape our thinking about the forms and possibilities of convergence. Suppose we supplement the generative machinery by borrowing devices from some source, relational or Päninian or hermeneutic. Even if we borrow plenty of tools, say, from Tesniere's dependency grammar, such a retooling need not affect generative attitudes to the question of convergence in grammatical inquiry.

Tesniere Indicators and Indian Languages · 111

But what we suggest here is a more serious modification of the generative program: (a) phenomenal: introduce Tesniere Indicators; (b) tactical: propose a Theory of Receding; (c) conceptual: merge with Dependency Grammar Tradition. Our argument, responding to some compulsions on the ground, is that this triple move continues and extends core generative concerns and at the same time enables us to forge conceptually interesting links between the two questions RQ1 and RQ2 on the basis of Tesniere's fundamental insights. What are the compulsions on the ground? India is given to what Dasgupta (1993) calls 'FESH' (Formal Elaboration of Social Hierarchy): we encounter and use language as an inescapable social reality here. A sustainable linguistic enterprise in this region cannot imagine mimicking the dissociation of grammatical inquiry from socioHnguistics that some 'northern' or 'metropolitan* countries seem to go in for. Although Indian linguists thus obviously need a rapprochement between microlinguistic or grammatical tools and macrolinguistic concerns of a psycho-social sort, such a dialogue has proved difficult to initiate. So has the applied linguistics take-off that would enable our work to meaningfully address both the urgent social needs of marginalized population segments and our colleagues in other disciplines who would work closely with us if we facilitated the cross-disciplinary traffic. What tools might help us? In this paper, we unpack the triple move (a)-(c) that embodies our proposed appropriation of Tesniere, showing how it meets some of these needs. This package, whose sustainability remains to be tested, is at least new in suggesting that the remedy lies not in issuing standard exhortations that grammarians should learn more socioHnguistics and vice versa, but in modifying the content of grammatical research itself to bring internal factors into wellmotivated closer alignment with extragrammatical concerns. Our unpacking begins with the double infinity of syntax, placing this issue in the mind of the older child OC as in the idealization OCI introduced earlier. Aware of the way her teachers provide her with grammatical tools to help her cope with a heterogeneous and stratified linguistic environment, OC moves back and forth between one subcode and another. Crucial to the way we visualize her moving back and forth is the concept of Receding—which, as factor (b) of our triple move, reconstructs the heart of Tesniere's message. We shall now argue that, at Recoding, the two infinities meet fruitfully. The notion of Recoding reconstructs what Tesniere's metataxis across languages (Tesniere 1959: 283-319) and transference within a single language (ibid.: 361-664) have in common. Recoding or generalized 'structural change' (ibid.: 283) covers cases where two languages find different syntactic structures natural for the same meaning, as in Latin Docco pueros grammaticam ºteach children-Ace grammar-Ace' vs French J'enseigne la grammaire awe

112 · Probai Dasgupta

enfants º teach the grammar to-the children' (ibid.: 289), as well as syntactic alternation within one language: 'Soit par exemple une proposition independante: Alfred frappe Bernard, eile subsiste integralement et en bloc et son verbe garde toutes ses caractoristiques verbales si on le transffere en substantif par l'emploi du translatif que:je crois qu 'Alfred frappe Bernard" (ibid.: 544). In cases of both sorts, Tesniere postulates a content-preserving change of syntactic form. This mechanism, in its generality, is Recoding. What does Recoding have to do with the two infinities of syntax? We leave it to others to seek any authentic answers to this question for Tesniere's historical context. We cleave to our period in which Abel (1998), extending earlier work, argues that diglossic code-cleavage is universally part of any principled portrayal of language. Suppose we visualize every OC, for these or related reasons, as learning language L under a principled and universal stratification distinguishing its natural basis, call it the Informal Kernel IK(L), from its codified Formal Extension FE(L). Suppose FE(L) is or dominates the literate overlay on OC's primary linguistic basis. Finally, suppose that FE language use constitutively involves Recoding, that all recursion is subject to this, and that, once she has become a self-conscious user of relatively complex sentences, OC's language use is seldom free from the FE(L) shadow of care or memory-laden attentiveness, and never free when she is using recursion. Now we can argue that the two infinities of syntax meet at Recoding, as follows. The second infinity, of grammar-candidates, reduces to that of alternative syntactic descriptions for a given construction. This in turn reduces to the fact that any given construction is related, in Recoding terms, to many cognate constructions. Thus the complement clause that the arrow hit the branch in They perceived that the arrow hit the branch is equivalent to the complement clause the arrow hitting the branch which can take its place. Now, for Tesniere, the transference markers (the translatifs) involved here are the primary transference marker ing and the secondary transference marker that. Primary transference makes the word hitting from the word hit. Secondary or analytic transference uses the particle that to turn a clause—a verb-headed construction, in Tesniere's dependency grammar tradition—into a nominal. The relation between primary and secondary transference is conceived in terms of a typical diachronic trajectory: 'Destinoe a suppleer la translation du premier degre, la translation secondaire devient nocessaire lorsque le vieillissement de la translation primaire (et la substantiation de l'infinitif, par exemple, avec la perte des caractiristiques verbales qui s'ensuit) empeche celle-ci de fonctionner a plein. La translation secondaire est done une translation jeune par rapport a la translation primaire qui est plus vieille' (Tesniere 1959:543; emphases omitted). A generative reworking of this material may begin with the point that a 'young' syntactic mechanism of receding represents the speaker's freedom to pack sentence into sentence more directly. An Older', morphology-bound mechanism inherits the vagaries of history and does not so completely belong

Tesniere Indicators and Indian Languages m 113 to the syntax proper. Acceptability drops more rapidly when you continue ingembedding than if you have a string of paniculate complementizers: (d) "Harry considered the possibility of John perceiving Mary noticing Bill gushing over Tom listening to the guitarists playing (e) Harry considered the possibility that John might perceive that Mary noticed that Bill was gushing over the fact that Tom was listening to the guitarists playing Despite some experiments with local gadgets like a Double ing Filter, inapplicable here, generative grammar has never come close to a principled account of the (d)-(e) contrast. Such an account will need to take seriously the distinction between the receding device ing in (d)'s noticeably repetitive words perceiving, noticing, gushing, listening, and the less noticeable repetition of the recoding particle that in (e). The point made by Tesniere, who notes that a word modification device like ing is less "young' than a phrase or clause formative such as that, speaks directly to generative grammar's concern with what is living (connected to UG-given vitality) and what is dead (merely historically residual) in a language. To touch base with the infinities again, a recursive use of the particle that, as in (e), is a path to a true infinity, a domain where the speaker is free to extend the structure. But the recursive use of ing affixation, as in (d), sounds unnatural, which shows that here the speaker is not that free. As OC learns her FE(L), her recursion route is drawn by the formally manifest freedom of a syntactic particle to vote for (e) and against (d). That she locates her freedom in this fashion is part of the reality of the infinity of syntax in OC's mind. But recall that we have chosen OCI as our operative idealization to define the reality of language. Thus grammarians too, in an ideal world of perfect linguists feeding perfect education systems to help the perfect OC into her full mental equipment, converge on an account that reveals (e) as the locus of recursive freedom in English, and exposes (d) as a fake mimicking (e) only at a toy level. Our ideal figures then converge on a choice, within the second infinity of grammar-fragment candidates, which helps OC make sense of her preference for the that particle and against ing affixation as a recoding device in her practices at the level of the first infinity of indefinitely stretchable sentences. This first infinity level becomes seriously, productively available to OC only at her sophisticated, FE(L) phase of language acquisition, where she is in touch with older users of L. These older users are making their literacy, their long-term shared and structured memories, available to her through an account that OCI represents as a grammatical codification of L, GC(L). In variants of OCI that one will wish to devise for other purposes, one can enrich GC(L) to extend this point in, say, a literary or political direction. This, in detail, is the way the two infinities meet at Recoding. We continue the unpacking of our triple move (a)-(c) by focusing now on the phenomenal move (a), introducing Tesniere Indicators. The term Indicator,

114 · Probai Dasgupta

as explained earlier, translates his indice. At the operative level, a Tesniere Indicator is a morphological or syntactic exponent of, typically, inflection: 'Les principales espfeces d'indices sont, outre les terminaisons des difforentes flexions, Particle et 1'indice personnel. [...] L'indice personnel a pour fonction d'indiquer la personne et le nombre dans le verbe: fr.j'aime, tu aimes, U aime. L'indice personnel est done I'equivalent syntaxique exact des desinences personnelles d'une langue teile que le latin' (Tesniere 1959:84). But Tesniere Indicators are not simply a device for handling inflection. They belong to Receding theory, being akin to Transference Markers. The difference is that in the case of Transference, 'la categoric de dopart, c'est-a-dire celle du mot plein avant translation, est nettement dofinie et bien distincte de la cat6gorie d'arrivee. [...] Au contraire, dans le cas de simple indication, [...] la categoric de depart [...] ne differe pas de la categoric d'arrivee. [...] les translatifs et les indices [...] sont au fond deux variotes d'une seule et meme espece de mots [...]. Soit par exemple 1'article le, dont le role translatif est evident quand il fait d'un adjectif un substantif, cf. fr. le bleu, il n'a plus qu'un role uniquement indicatif quand il se trouve devant un mot qui est dejä substantif par lui-meme, cf. fr. le Uwe* (ibid.: 84). Where and how do Indian languages make it appropriate to add Tesniere Indicators to the generative grammarian's tool-kit? As a preparation to answering this question, we need to refer crucially to Tesniere's context itself and not just to our reasons today for wishing to inherit his work. OCI is no phantom of the 1990s. Tesniere too, in his own way, had invoked the child as a regulative principle in the construction of a linguistic science. Via his invocation of the child, we may approach his version of the theory-practice duality of linguistics. In Tesniere, the child is not a toddler engaged in primary grammatical development, but a school-going child, typically a teenager, learning how to control and represent grammatical structure in exercises of monolingual skill as well as translation. Much of his book, especially the mechanics of drawing dependency trees, was tried out on actual school-children by Tesniere and his teacher-trainees. Their feedback admittedly helped shape the book over the decades it took Tesniere to write it. This was possible because it embodied a special kind of linguistics. We latter-day linguists, even those of us who write dependency grammars (cf. Dasgupta 1989) or reconceptualize the bare phrase structure phase as having merged generative linguistics with the dependency grammar genre, are committed to a very different approach to issues in the theory and practice of grammar. What little Tesniere we appropriate comes to us through many mediations. It takes an effort to look at, and a greater effort to take seriously, what his book was trying to do. Tesniere assumes that language teaching and grammatical theory, descriptive grammar and historical grammar, particular description and general typology are only so many convenient divisions. They are not indispensable bifurcations foreshortening the scope of our work as a linguist. All these

Tesniere Indicators and Indian Languages · 115

aspects are of equally legitimate interest for any inquirer, whatever her age. Linguistics is open and transparent enough in principle that a child in secondary school should be seen as paying attention to the same linguistics that adult minds on the job must be paying more attention to. We are suggesting that the relevant adults pay some attention to the standard assumption that a social notion of'language', as distinct from an individual-brain notion of 'internalized grammar based on UG', comes into play in some border area where the microlinguistics of morpho-syntactic objects negotiates purely external terms of trade with the more applied and psychosocial allied disciplines concerned with language and literature. The assumption that this social notion of 'language' comes into play only outside the realm of grammar, we suggest, needs to be revised. For the study of grammar, we argue, involves the study of OC's acquisition of language as a whole, including FE(L), not just of her internalization of grammar, if by that we mean an IK(L) purely determined by UG interacting with basic input from the milieu. Indian concern with the generative renewal of linguistics has fortunately never proceeded in utter detachment from the conditions under which our OCs live and learn. Throughout India, grammatical schooling codifies the structures of this region's languages in terms of the Päninian tradition plus various cognate and descendant traditions, supplemented by some opportunistic spice from traditional school grammars of English. Generativists of Indian origin have stressed, with welcome support from colleagues elsewhere, the close affinity between Päninian principles of grammatical conceptualization and their generative rediscovery. In this context, it seems fair to say that Indian generativists have been working towards a generative recasting of the school grammars of Indian languages in terms of which each community, via its OCs, shall inherit its language differently, more rigorously. Such a recasting visualizes a merger of Päninian (and cognate/descendant) principles into a serious generative tradition that can viably absorb such resources. Why are we suggesting that the generative tradition needs to join forces with Tesniere's dependency grammar tradition? It may suffice to offer a detailed example of the sort of difficulty that we face in working towards a grammatical tradition usable in Indian schools, and how the triple move (a)-(c) can help. For the Päninian framework to keep serving Indian schools, it is important for us to have an articulate account of the difference between the Case features of the Noun in Sanskrit and the Case Particles in a modern Indian language like Hindi-Urdu. For these two kinds of exponents of Case have markedly different side effects: (f) Sanskrit: uttamah baalakah good-Nom boy-Norn 'the good boy'

116 · Probai Dasgupta

(g) uttamam baalakam good-Ace boy-Ace 'the good boy' (h) uttamasya baalakasya good-Gen boy-Gen 'the good boy's' (i) Hindi-Urdu: acchaa laRkaa good-Norn boy-Norn 'the good boy* (j) acche laRke ko good-Obi boy-Obi Ace 'the good boy' (k) acche laRke kaa good-Obi boy-Obi Gen 'the good boy's' P ninian grammars postulate Cases as internal States of the Noun correlated with the way it externally depends on major heads such as Verbs, a straightforward matter in Sanskrit: nominative in (f), accusative in (g), Genitive in (h). An Adjective, as it is in concordant dependence on the Noun, copies this state together with Gender and Number, a straightforward copying in Sanskrit. This makes it hard to handle Hindi-Urdu, where the traffic of the internal and external does not look straightforward from a viewpoint shaped completely by the facts of Sanskrit. In Hindi-Urdu, too, the Noun is in one of several States: a Vocative State not exemplified here, the Nominative State of (i), and an omnibus Oblique State indicating dependence on any other Case Particle, as in (j), (k). And again the Adjective copies not only the Noun's Gender and Number, but also its State Nominative, Vocative, or Oblique. However, Nominal State in Hindi-Urdu is only a truncated Case subsystem; the Oblique State marks not the Noun's Case but its relation with the actual particulate Indicator of this true Case. How does our triple move (a)-(c) help address this problem? Given component (a) in our package, we can use the Tesniere Indicator (ºº) device to bridge the gap between fusional morphology in Sanskrit and fusionaided analytic syntax in Hindi-Urdu. Recall that Tesnierean devices come with conceptual baggage attached. When Noun talks to Verb, this shows up as a personal ¹ on the Verb word—fusional in Latin and analytic in French in the examples from Tesniere discussed earlier. When neighboring head talks to Noun, this shows up as a dependency TI on the Noun word—fusional in the heavily inflected words of Sanskrit (g) and (h) and analytic in Hindi-Urdu (j) and (k) with their Particles. This conclusion is unlikely to elicit descriptive controversy. But we need to resort to the ¹ device again. It is not just true

Tesniere Indicators and Indian Languages m 117 Case Particles like /ko, kaa/ that count as TIs in Hindi-Urdu. We must also postulate for this language a fusional TI Obi, for Oblique State. This fusional ¹, in (j) and (k), serves as the exponent, internal to the Noun word, of a dependency relation involving this Noun and its Particle colleague. Move (a) helps by allowing us to recompose the story in a format that enables school-children to listen to Sanskrit and Hindi-Urdu in the same breath. A given TI can work analytically, as a secondary TI, in a syntax-prone Hindi-Urdu and inflectionally, as a primary TI, in a fusion-prone Sanskrit but nonetheless count, in the metataxis, as translationally the same ¹ manifested differently. At the same time, the syntactic exponent in Hindi-Urdu, a Case Particle, forces its Noun to bear an internal ¹, a fusional underspecified archi-Case this time, which its Adjective dependents must copy as a Sanskrit Adjective copies the fully specified Case of its Noun. Readers who think this is just like the contrast between Latin and French may need to examine the data closely. The Dative Case that a French Pronoun or a German Pronoun wears in subservience to 'Case-equivalent' Particles such as German zu or French de is also available as a serious Case, to mark certain arguments of Verbs. But the Oblique Archi- Case of a Hindi-Urdu Noun or Pronoun uniquely marks dependency on a Particle, nothing else. The Oblique Archi-Case of Hindi-Urdu is not merely a problem for secondary schools in language pedagogy, but an issue in linguistic typology and, for any accountable grammatical theory, an issue in descriptive and explanatory syntax. And we do wish to stress that adding TIs to the generative tool-kit helps cope with this issue. But our focus here is not on Hindi-Urdu Obliques per se. The point is to see this material as an example of how the Tesnierean tools help us to understand and formalize certain mappings between syntactic and morphological devices that do similar jobs. This is where we make the transition from points (a) and (b) in our triple move to point (c), the proposal that we generative grammarians should merge our enterprise with that of dependency grammar. The principle that a TI can have primary or wordbound and secondary or participate variants is what both permits language types to fan out typologically and holds them on leash parametrically. In other words, it enables us to understand how there can be so many mechanically different ways of saying the same thing that all end up structurally equivalent. Tesniere's method of looking at recodings in terms of preservation or alteration of category within a fourfold of Verbs, Nouns, Adjectives, and Adverbs/ Adpositions, which anticipates the generative classification of lexical categories, is orthogonal to the traditional taxonomy of constructions in terms of inflection, derivation, compounding, and syntax. The Tesniere Indicator is a powerful device that enables us to generalize across these levels without needlessly slurring over the specific properties of words, phrases, and other formations. One might initially imagine that a TI is just a notational variant of the functional projections of contemporary generative work. But the notation of functional heads through which a multiple-featured word has to head-move

118 · Probai Dasgupta

forces an unexamined syntax-focused solution to the question of what is basic in a reasonable typology, or what the default settings are in a reasonable characterization of parameters: does the unmarked core grammar prefer the null subject setting (Borer 1986) or the overt subject setting (implied in Chomsky 1981) of the null subject parameter? The ¹, by making it formally as easy to expand Receding information into a phrasal construction with a secondary particle as it is to pack it right into a word's feature matrix, encourages linguists to pursue that question instead of sliding down their notation into a premature answer. A metalinguistically important byproduct of this or any other adoption of Tesnierean devices has to do with mutually convertible grammatical analyses. One way for us to begin to inspect and develop this byproduct, which in this form is relatively new in any paradigm, is to consider a typical fusional language F and an idealized analytic or isolating language I. Assume that Fspeaking scholars have come up with a grammar of F which builds the main body of the grammar into word structure and hand-waves about the syntax trivially following from this. The I-speaking community also, in our picture, has produced a grammar of I where the relatively rigid constituent order rules are taken to be the natural basis for a syntactically specified set of grammatical functions, and any vestigial inflectional morphology is viewed as reflecting these functions in a trivial way. Now we put the grammarians of the two communities in a contact situation shaped by the fact that OCs in their respective educational systems have grown up with these tools. Both sides become willing to consider new grammars for the two languages. The contact situation throws up an F-inspired grammar of I, call this FGI, and an I-inspired grammar of F, call it IGF. FGI postulates initial configurations for I-sentences with all words expanded into highly inflected forms bearing various function markings that F makes one expect, but accepts the phonological reality of outputs where these supplements become inaudible. In the other direction, IGF postulates initial configurations for F-sentences with all constituents occupying rigidly patterned positions in an I-like configuration, but then permits most constituents to move away and yield outputs where the postulated configurations become undetectable. There are now four grammar candidates: old FGF, IGI, and new FGI, IGF. This mutual convertibility of grammars is in the first place the equivalence of the two grammars for each language (the fact that FGF and IGF must generate the same set of sound-meaning pairings). But a closer look at the matter turns it into something more than just identity of weak generative capacity in the familiar context of the evidence underdetermining the choice of the right theory. The translation relation between F and I becomes clearer if one examines the relevant sentence-equivalences both in terms of the self-oriented grammars FGF and IGI and in terms of the other-oriented grammars IGF and FGI. In some broad sense that nobody has the tools yet to bring to the explicit surface, the structure of each language has something to

Tesniere Indicators and Indian Languages m 119 tell the rest of us about that of every other. To put it intuitively, each language X invokes a certain X-image of human language in general, and that image potentially induces a specific set of X-grammars for all other languages in the light of the particular X invoking thir image. Consequently, the infinity of possible languages implies a corresponding infinity of possible grammars for each language. For a grammar G-i of a language L-a amounts to a set of questions we ask about L-a; and our discussion suggests that there will be as many sets of questions about L-a, and therefore as many grammars of L-a, as there are distinct languages L-b, L-c, L-d, etc. (inspiring, alongside the initial grammar G-a (L-a), various new grammars of L-a that may be called G-b (L-a), G-c (L-a), G-d (L-a), and so forth). To return from matters formal to the matters substantive that drive our proposed shift from YCI to the OC idealization, India is the kind of society where this question translates at once into an active question of how theoretically interesting dialogue can be pursued for simultaneously social and intellectual ends. If we terminate this discussion here, it is because only alternating between theoretical and practical work, in the manner pioneered by Tesniere, is likely to prevent theory from becoming sterile or arid and practice from becoming repetitively frustrating because of refusals to learn from theory-mediated experience. Perhaps what has been presented here is explicit enough for our practitioners of language pedagogy, and other relevant enterprises, to pick up the thread and offer criticism based on present—or future—experience. We close with the hope that grammarians will also offer standard theoretical criticism, possibly based on some other account of the Obliqueness phenomenon which many Indie languages share with Hindi-Urdu but which seems to have been neglected in the generative literature. • REFERENCES Abel, P. Rekha. 1998. Diglossia as a linguistic reality. The yearbook of South Asian languages and linguistics 1998 ed. by Rajendra Singh, 83-103. New Delhi: Sage Publications. Borer, Hagit 1986.1-subjects. Linguistic Inquiry 17 (3). 375-416. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. . 1986. Knowledge of language: Its origin, nature and use. New York: Praeger. Dasgupta, Probal. 1989. A dependency syntax of Bangla, Metataxis in practice: dependency syntax for multilingual translation ed. by Dan Maxwell and Klaus Schubert, 89-113. Dordrecht: Foris. . 1993. The otherness of English: India 's auntie tongue syndrome. New Delhi: Sage Publications. McCawley, James D. 1982. Thirty million theories of grammar. London/Canberra: Croom Helm. Tesniere, Lucien. 1959. Elements de syntaxe structural. Paris: Klincksieck.

B

Open Submissions

Gerundial Aspect and NP Movement • TANMOY BHATTACHARYA · The central assumption of this paper that DPs exhibit similar or identical behavior to that of clauses is a continuation of the investigation begun by Abney more than 10 years ago. In my earlier work, I have shown that DPs too involve NP movement. This paper* is an attempt to investigate both the wider assumption of the similarity between clauses and DPs and the particular finding of DP-internal NP movement in terms of the structure of gerunds in Bangla. Towards the former goal, I show that certain sentential aspectual properties are reflected inside the gerund DP, proposing, in effect, that, the gerund head is a nominal aspectual head. Towards the latter goal, I show that the derivation of gerunds involves NP movement inside the DP.

• 1. Introduction In earlier work (Bhattacharya [1998a,b,c], Bhattacharya [1999a,b] and Bhattacharya [forthcoming]) I have shown clear instances of NP movement, rather than N movement, inside the DP in Bangla. This, I have claimed, derives by assuming an SVO underlying structure for a head-final language. This paper provides yet another evidence for such DP-internal NP movement. First, I claim that gerunds embed a nominal aspectual head inside the DP and secondly, gerund formation is shown to constitute evidence of NP movement.

• 2. Gerunds as Complex Event Nominals The place of nominalization in grammar revolved around the issue of the enrichment of one component of the grammar via a possible simplification in • Special thanks to Misi Brody, Rita Manzini, Ad Neeleman, Andrew Simpson, and Neil Smith for comments on earlier drafts. Thanks due also to one anonymous referee for detailed and helpful suggestions.

124 · Tanmoy Bhattacharya

another since Chomsky (1970:185). His study led to the generalization that regular correspondences between linguistic forms should be captured in the syntax (through transformations) and the irregularities in the lexicon. This in turn led to the lexicalist versus non-lexicalist debate. Thus we may derive from the verb give the derived nominal (DN) gift or the gerundive nominal (GN) giving, whereas the former is traditionally viewed as part of derivational morphology, the latter as inflectional or as part of syntax.1 The literature on nominalizations in English includes numerous arguments to show the difference between these two different nominalizations: differences which indicate, according to Chomsky, a transformational derivation of GNs (i.e. underlyingly GNs are sentences in this theory).2 Due to these differences, Chomsky argued, it would be wrong to derive both from the same source by applying different transformations. He concluded that 'derived' nominals are not derived at all but are rather listed in the lexicon. In this paper, I make use of a distinction of event classification put forward in Grimshaw (1990). I assume with her that gerunds denote complex event nominals (CEN) whereas gerundives and result nominals are simple event nominals (SEN). In this theory, CENs have an obligatory argument structure or an a-structure. Consequently, I argue that gerundial constructions (the ones formed with the gerundial suffix -(w)a/no, to be discussed in Section 3)— participles, gerunds, 'gerundives' and result nominals—can be distinguished by virtue of their event (or aspectual) properties. That is, Grimshaw's distinction between CENs and SENs translates in the current analysis as aspectual differences. In other words, gerunds in Bangla project an a-structure in the syntax with specific aspectual positions absent in other nominals. In sum, based on Grimshaw (1990), I assume that gerunds denote complex events and this is encoded in the aspect of the phrase. This suggests that gerunds contain a syntactic position for aspect. The structure of the Bangla gerund that I introduce in the next section (and discuss in the rest of the paper) makes use of this conclusion. Furthermore, I suggest that the gerund suffix, to be discussed shortly, heads this aspect phrase. • 2.1 Structure of the Bangla Gerund Based on the introductory discussion in the preceding section, I suggest that gerunds in Bangla contain a fully projected VP containing the functional projections of AspP as follows: (1)

Asp VP (w)a/no/\ V NP

Gerundial Aspect and NP Movement · 125

Gerunds in Bangla (like in English) behave like a noun phrase justifying the DP structure in 1. The nominal character of the gerund is encoded in the D° head which is nominal. I present further justifications for the above structure as we proceed. In Section 3,1 discuss the -(w)a/no gerund suffix in Bangla. The following section presents evidence in favor of the TP in 1.1 postpone the discussion of nominal aspect as represented by AspP till Section 4.

• 2.2 Tin DP The presence of tense inside nominals was first pointed out in Hockett (1958: 238).3 Lecarme (1996:162) shows that in Somali, tense morphology is associated with nouns. The distinction between past/ non-past (see 2), parallels the identical distinction in the VP: (2) a. sannad-ka dambe year-DET next 'next year' b. sannad-kü/*ka höre year-DET.PAST before 'last year' Lecarme points out that nominal tenses in Somali can have an independent reading. Additionally, I suggest that in the context of the analysis of adjectives as specifiers of NPs (see Bhattacharya 1999b for this proposal), tense in adjectives in Japanese is another piece of evidence in favor of a TP inside the DP. Nakamura (1994: 375), in discussing the tense system in Japanese in general presents the following data which shows that the tensed adjectives in the nonpast tense which end with -/ contrast with those in the past which end with -katta:* (3) a. aka-i kuruma red-PRS car 'a red car' b. aka-katta kuruma red-PAST car 'a car that was red' Most crucially, however, the presence of the TP, apart from the reasons given so far, solves the problem of some Bangla gerunds where the gerund subject bears Nominative Case. I discuss this in Section 3.1. Finally, in the analysis of van Hout and Roeper (1998) a TP in a nominalization structure is needed in order to get the right event entailment of the nominal. Since gerunds are, by the assumptions of this paper, complex event nominals, the nature of this event as expressed through aspect is discussed in detail (see Section 4) concluding that the event entailed by the gerund suffix is imperfect.

126 · Tanmoy Bhattacharya

• 3. The Gerund Suffix in Bangla Although the issue of what constitutes a gerund has been contentious, the study of gerunds in English within generative grammar, most thoroughly analyzed in Abney (1987) subsuming the work of Schachter (1976), Chomsky (1981) and Reuland (1983) among others, shows uniformity in the range of constructions considered to be gerunds, namely, POSS-wg, ACC-ing and Ingof.5 This has also been the case for studies in non-generative frameworks like GPSG/HPSG as in Pullum (1991), among others, in English, and Dasgupta (1980) and De (1984) in Bangla. I do not see any reason to question this conformity in identifying the construction. However, since out of these three types of gerunds ACC-ing and Ing-ofdo not obtain in Bangla, I will discuss the POSS-ing type as the only gerund type in Bangla in the rest of this paper. For the purpose of this study I will, therefore, assume any deverbal nominal which can have a possible Genitive subject (to be made explicit below in Section 3.1) and which denotes a complex event as per Grimshaw (1990), as a gerund in Bangla. Gerunds are formed in Bangla by adding a gerund suffix. There are four gerund suffixes in Bangla: (4) -a: -we: •no: -ba:

pOR-a ga-wa pala-no kha-ba

'reading' 'singing' 'escaping' 'eating'

I will treat the first three suffixes as one group, the -(w)a/no group which contrasts with -ba in its distribution:

(5) a. khai. khawa 'eating' ii. khaba6 'eating' (dialectal) iii. * khaa iv. * khano b. dEkhi. * dEkhwa ii. dekhba 'seeing' (dialectal) iii. dEkha 'seeing' iv. *dEkhno c. takai. * fakawa ii. takaba 'staring' (dialectal) iii. * takaa iv. takano 'staring' Let us briefly discuss these two groups in turn.

Gerundial Aspect and NP Movement · 127 (w)alno

The suffixes -a/-wa occur after monosyllabic verb roots, -a occurs after consonant ending verb roots while -wa occurs elsewhere. Their distribution could therefore be accounted for by a phonological feature. This is implicit even in Chatterji (1926) since he considers that suffixes like -aano (earlier form of -no), etc., must attach to causative and denominative verb bases, these being longer than monosyllabic forms. Although -wa and -no are in complementary distribution, diachronically they are from different sources and are phonologically distinct from each other. These two therefore form a suppletive morpheme -(w)a/no. -ba As shown in 4, this suffix contrasts with the other group in its distribution. This suggests that a verb root forming a gerund with -(w)a/no also has another form with -ba. This has two varieties among the speakers of the language—ba and -iba.1 An interesting observation which remains unexplained in De (1984) is that -ba gerunds, unlike -(w)a/no gerunds, do not occur independently (i.e., they must have a genitive marker), rather, they appear in the template 6a as inob: (6) a. V + ba + GEN b. ja- ba- r go- GER- GEN 'going'

The appearance of the Genitive marker is a consistent diagnostic for gerundives in Bangla. Following the Orientalists who study Sanskrit in English, there has been an attempt (e.g., in De [1984] and Dasgupta [1986]) to distinguish the terms gerund and gerundive, identifying the latter with forms V + nominalizer + adjectivalizer/Genitive marker. It is important to see that the term gerundive has been wrongly used as the simple adjective of gerund in Chomsky (1970) onwards. However, given that there are various opinions on the gerundive in the classical usage itself as in Vedic/Sanskrit (see Peterson [1997] for a review), it is not clear whether gerundive is the right term for these constructions in Bangla. However, pending a more satisfactory analysis, I will continue to use the classical term 'gerundive' for these constructions. That is, the term gerundive will be reserved for constructions where the -ib (or -ba) morpheme is used to form the gerund and furthermore is followed by the Genitive Case marker.8 It is important to point out that the morphological identification of the gerund suffixes is not sufficient to identify a gerund phrase. The following example shows constructions sharing the gerund suffix. (7) a. amar naTok lekha

holo

na (gerund)

my play writing happened not 'my play writing never happened'

128 · Tanmoy Bhattacharya

b. amar lekha naTok (participle) my written play 'a play written by me' c. khabar ghOr (gerundive) eating-GER room 'dining room' d. apnar duTo lekha dekhlam (result nominal) your two write-GER saw '(I) saw two of your articles' Based on the analysis of Bhattacharya (19995), I will consider, without argument, Gerundives and Result Nominals in Bangla to be pure nominals and will not discuss these in this paper any further. • 3.1 Case of the Gerund Subject: T in DP Revisited In this section, I show that the presence of a T head inside the DP additionally accounts for the subject case properties of the gerund. First, let us briefly review the subject Gase possibilities. The following example shows that the class of Ps which mark their complements zero, mark the gerund subject as a Nominative (i.e. zero) or Genitive: (8) [rukun-(er) baRi aS-a] matro Rukun-(GEN) home come-GER as soon as 'by the time of Rukun('s) coming home' The example in 9 shows that the Ps which mark their complement Genitive, can also mark the gerund subject genitive optionally: (9) [rakhal-(er) pOr-a]-r phOle Rakhal-(GEN) read-GER-GEN as a result of 'as a result of Rakhal('s) reading it' These examples show that the subject can optionally appear without the Genitive in all cases. Therefore the optionality of the subject appearing in Genitive has nothing to do with the Gase of the gerund phrase as a whole. Note that in 8-9 it is implied that the Genitive gerundial subject and Nominative gerundial subject do not differ in terms of interpretation of the gerund phrase as a whole. This has been standardly assumed to be the case (see De [1984] and Dasgupta [1994]). Looking closely, however, we find that though in both cases the gerund itself denotes an event, the interpretation of the gerund with a Nominative subject entails a temporal event or at any rate a 'sequential' interpretation of events as the salient interpretation. With a Genitive subject the event highlights not a sequential interpretation but a causal or at any rate

Gerundial Aspect and NP Movement · 129

an agentive interpretation. This difference is most clearly visible in the case of temporal Ps likepO 'after', SOnge SOnge 'immediately', etc., on the one hand and causal Ps like/o/i/io 'because of,phOle 'as a result of, dorun 'because of, etc., on the other. As a result, in the case of Nominative constructions a causal P behaves like a temporal/sequential P and in the case of Genitive constructions a temporal P behaves like a causal P. In 10, a and b are respectively Nom with temporal P and a causal P both resulting in a sequential meaning. In 11 a, b a GEN subject is used, where both the temporal P and the causal P result in a causal meaning: (10) a. rOmen aSa-r pOr kaj Sum holo Romen coming-GEN after work start happened 'after Romen coming, the work got started' b. rOmen aSa-r phOle kaj Suru holo Romen coming-GEN result-of work start happened 'after the event of Romen coming, the work got started*9 (11) a. rOmen-er aSa-r pOr kaj Suru holo Romen-GEN coming-GEN after work start happened 'because of Romen's coming, the work started' b. rOmen-er aSa-r phOle kaj Suru holo Roman-GEN coming-GEN result-of work start happened 'because of Romen's coming, the work started' This distinction, can now be captured in the gerund structure that I have proposed in 1 in terms of the aspectual properties of the gerund. Anticipating the discussion in Sections 4 and 5 somewhat, the difference between the two gerund phrases in 10-11 is based on the following argument structure of the gerund (repeated from 1 plus the external argument position shown): The temporal/sequential interpretation is due to the T head which the V moves to (via the Asp head) in the case of 10 since the gerund head contains an appropriate aspectual feature. This forces the internal argument to check Nom at [Spec, TP]. In the case of the 11 interpretation, Genitive Case is checked at the [Spec, DP] domain which supplies the agent/causer interpretation as well. Consider in this connection the observation that the light verb v in a vP-shell structure can have a limited inventory of meanings in Hale and Keyser (1993) and CAUSE is one of them. It is therefore possible for the NP to check for an appropriate aspect at the Spec of Asp at Merge before it moves to Spec of D. However, in the predicate-based theories of Tenny (1987) and Borer (1993) only objects can check for Case at [Spec, AspP]. The subject argument therefore moves up to [Spec, DP] and checks for Genitive. The fact that the subject in 10 does not go all the way up to [Spec, DP] is evident from the following contrast:

130 · Tanmoy Bhattacharya

(12)

D* D

TP

Spec T' T AspP [nom]/\ Subj Asp'

Asp VP (w)a/no/\ V NP (13) a.* rOmen ei baRi aSa-r pOr Romen this home coming after b. ei rOmen baRi aSa-rpOr After this act of Romen coming home* c. rOmener ei baRi aSa-r pOr Romen's this home coming after That is, the Nom marked subject in 13a cannot move across the Dem, which is lower than D since [Spec, DP] is the domain of the Genitive case checking. However when there is a Genitive subject as in 13c it must move up to this position. • 32 The Gerund and the Participle The construction which is most closely related in structure and—as I shall claim—derivation to the gerund is the participle. The gerund and the participle have been treated as two sides of the same coin in Bangla grammar. The morphological identity of the gerundial and participial suffix (both being -wal no) and the similarity of word order between the two forms cannot be accidental. In fact Dasgupta (1980) in his dissertation treats the participial as derived from the gerund. He proposes a gerund-participle rule (in the lexicalist framework) where the participle is derived by the addition of a null affix to the nominal head of the gerund. Thus, only gerunds are identified in the lexicon and not participles. I will argue for a syntactic derivation of gerunds from the same verbal source as the participle. First, let us see that Bangla gerunds and participles freely convert from their verbal source: (14) a. amar kobita lekh-a my poetry write-GER 'my poetry writing'

Gerundial Aspect and NP Movement · 131

b. amar lekh-a kobita my written poetry "poetry written by me' However, as pointed out in Dasgupta (1980:139) there are cases of gerunds which do not function as participles: (15) a. ramer baje kOtha bole bERa-no Ram's nonsensical saying say-and go around-GER 'Ram's going around talking nonsense' b. * ramer bole bERa-no baje kOtha "nonsense Ram has been talking' (16) a. ramer Sastrio Songit ERa-no Ram's classical music avoid-GER "Ram's avoiding classical music' b. * ramer ERano Sastrio Songit "classical music avoided by Ram' Dasgupta uses this fact to argue that it is therefore natural to expect participles to be derived from gerunds, since all participles can also function as gerunds but the reverse is not true. That is, gerunds constitute a superset, consequently, gerunds which undergo this rule are marked lexically as such. I will show, first of all, that consideration of a bigger set of predicates will lead us to interesting consequences for the theory of Bangla nominalization in general and of the exceptional cases in particular. A consideration of predicate types (i.e. unaccusative/unergative nature of the predicate) is needed for a fuller account (see Section 6). First, I will discuss the properties of nominal aspect to justify the Asp head in gerund DPs as in 1.

• 4. Nominal Aspect In this section I will look into the type of aspectual information instantiated by the -ing morpheme in English and the -walno gerund suffix in Bangla and show that both encode imperfective aspect in gerunds. By nominal aspect, I mean aspectual information available inside the DP similar to the aspectual information at the clausal level. In this section, I hope to show that a clear cut case for nominal aspect can be made. In particular, I propose that the gerund suffix in the case of true gerunds (i.e. in exclusion of gerundives and result nominals) carries aspectual features which must be checked in the overt syntax. In accordance with Grimshaw (1990), the difference between different nominals would seem to follow from their difference in aspectuality which in turn indicates their difference in event readings. I show that the -wa/no suffix in Bangla encodes imperfective aspect in gerunds.

132 · Tanmoy Bhattacharya

This demonstration rests on the theory that grammatical gerunds by their very function display event properties through morphological or abstract aspect. Such a putative Semantic universal', I claim, is derivable from Grimshaw's formulations mentioned earlier. Furthermore, notwithstanding the pitfalls of finding historical motivation for any aspect of synchronic grammar, it may be noted that at least one historical interpretation of the connection between the current progressive in -/(/)- in Bangla and the older verbal noun in the locative exists in Chatterji (1926: 1025). The well-attested similarity between the clausal and the NP structures is reflected in the fact that event verbs pattern with count nouns and state/ activity verbs pattern with mass nouns. Thus, as Brinton (1995) points out, event verbs can be counted (as in 17a) like count nouns which take number morphemes, and activities can be modified by mass adverbials (as in 17b) just as mass nouns are modifiable with adverbials like much, a little, etc. (17) a. John arrived three times/ *a lot b. John knew a lot/ *three times The analogy between event verbs and count nouns and between stative/activity verbs and mass nouns is supported by English deverbalising suffixes. The -ing suffix makes the situation atelic, durative and dynamic by converting the situation into an activity. This is a shift from count to mass interpretation: (18) Count + ing -» Mass This is called DEBOUNDING or the process of imperfectivising in the verbal domain (Jackendoff [1991]) by the progressive -ing. The process equivalent to debounding in the nominal domain has been identified as GRINDING in Jackendoff. This is shown for Bangla in 19a, b. Furthermore, 19c-g show that the effect of the gerund suffix on the aktionsart of the verb: (19) a. ghOre roSuner gOndho in room garlic's smell "there is a smell of garlic in the room' b. tOrkaRi-te murgi pelam curry-LOC chicken found '(I) found chicken in the curry' c. probiner kobita lekha eggocche Probin GEN poem write GER progress PROG.3 'Probin's poetry writing is progressing' d. bar bar ghOnla beje oTha again again bell sound INF rise GER 'ringing of the bell again and again' e. robiner mOraTa OSSabhabik Robin GEN dieGERCLA abnormal "Robin's dying was/is abnormal'

Gerundial Aspect and NP Movement · 133

f. baRi bhaNgaTa taRataRi holo house break GERCLA quick happened 'the breaking of the house was quick' g. SaStrio SoNgitSonaTa Sabhabik classical music listen GER CLA normal 'listening to classical music is normal' The analysis of the -ing (and the -i(f)") in the verbal domain with the -ing or the -(w)a in the gerund in the nominal domain is best expressed through different aktionsart effects like continuous activity in 19c, iterative in 19d, achievement in 19e, accomplishment in 19f, and state in 19g. The gerund focuses on the activity in 19c, d and on process in 19e-f. In 19g the gerund gives an activity a temporary sense. The nominal counterpart to perfectivizing has the effect of turning a mass expression into a count one.10 This process, analogous to the temporal bounding of verbal situation, is called PACKAGING and is shown by the following example: (20) a. amake mOd-Ta dao tome drink-CLA give 'give me the drink' b. jOl-Ta poriSkar kOro water-CLA clean do 'clean up the (spilled) water' In conclusion, packaging therefore constitutes the second instance of nominal aspect, i.e. the DP equivalent of the clausal aspect. As the preceding discussion shows, nominal aspect can be either in the form of packagers or grinders. However, keeping the topic of discussion in focus, this proposal translates into minimalism as the gerund suffix selecting a [-PERFECT] feature for the numeration whereas the participle selects a [+ PERFECT] feature. • 4.1 Syntactic Account of Aspect Given Grimshaw's (1990) position on CENs (see Section 2) to have an internal aspectual structure and thus an argument structure (which by definition denote thematic and aspectual properties of the predicate), I will consider event identification in terms of aspectual features. That is, I will adopt a feature checking approach to aspect in line with many predicate-based accounts of aspect (Tenny [1987] and Borer [1993] in particular). Such theories are based upon Verkuyl's (1972) suggestion that the semantic nature of the object determines the telicity of the entailed event. Telic and atelic predicates respectively require bounded and unbounded NP complements.

134 · Tanmoy Bhattacharya

Syntactically, telicity of the verb is checked by an NP (which is an event measurer or delimiter) in the specifier of an aspectual head in the theory of Borer (1993). In other words, the NP carries a feature of [DELIMITER] or [EVENT MEASURER] or [BOUNDED] which is checked against the telicity feature of the Asp head like [+PERFECT]. In Borer's account, Case and aspect are linked by postulating an Asp head over the VP. Borrowing from Tenny (1987), Borer assumes that a feature [DELIMIT] of the Asp head is responsible for checking the corresponding feature of delimited direct arguments which pass through its Spec in the process, checking an Accusative Case feature as well: (21)

AspP Spec NP Asp [± delimit] [± perfect]

y tNP

I have adopted a similar position for the gerund and the participle structure. In the case of nominalization the nominalizer - (w)alno projects an AspP which takes a VP as a complement. That is, I take the position that the Asp head is equivalent to the light verb as in Hale and Keyser (1993a) and Chomsky (1995). The internal argument of the verb is merged to the right. The nominal character of the construction is reflected in the presence of a nominal D and Asp heads. Based inter alia on de Hoop's (1992) treatment of resultatives in Finnish, I assume that aspectual information encodes argument structure as Case. This would suggest that the Asp head checks both aspectual and Case features of the internal argument. Later, a distinction between the gerund and the participle is made based on the trigger for the NP movement. I assume that the external argument of the gerund is generated at the [Spec, AspP] position. Note that Asp therefore shares the similarity with v in being both a functional head by checking the Case (aspect) of the internal argument and a lexical head by virtue of having an external argument merged at its specifier. In sum, I have provided motivation for the presence of an aspectual head Asp inside the gerund DP in this section. I have thus motivated the gerund structure proposed in 1 in Section 2.1 fully.

• 5. Deriving the Gerund and the Participle Based on the conclusion that gerunds and participles in Bangla are syntactically derived from the same verbal base, I will now consider the derivation of the following pair of gerund and participle from the VP structure as in 23:

Gerundial Aspect and NP Movement · 135

(22) a. kobita lekha poetry writing 'writing poetry' b. lekha kobita writing poetry 'written poetry' (23)

VP V

NP

I

I

lekh- kobita Given the preceding discussion, the verbal stem lekh- in 23 maps into a syntactic structure where the gerundial aspect feature of the gerund head and the aspectual and Case features of the complement NP are checked against an aspectual functional/lexical head Asp as in 24.1 will assume with Chomsky (1995) (and what we have adopted in preceding chapters) that there may be multiple specifiers of functional projections. The AspP shell structure above is essentially identical to the vP-shell structure in Chomsky (1995) which may therefore have the external argument generated at the outer spec of AspP which moves up to [Spec, DP] to check Genitive. Internal structure of the AspP is shown whenever required. I will further assume with van Hout and Roeper (1998) that event anchoring is established trivially through an empty T in cases where the verbal head does not carry a Nom Case feature, if it does then Nom Case is checked at [Spec, T] along with the event feature checking. I will show the TP projection only when it is needed. (24)

DP D

TP

/\AspP /\ Spec Asp'

T

kobita Asp

I

lekha

VP

S\NP

V

iCRnQ

KODltd

136 · Tanmoy Bhattacharya

In line with our account of argument generation, the NP kobita checks the Case feature of the Asp head at [Spec, AspP]. The head movement of the V to Asp also involves an aspectual feature checking, this time [- PERFECT] for a gerund. The participle in 22b above is derived as follows:

(25)

lekha Spec

Asp'

kobita Asp

I

tekha

VP

/\NP

V

Notice that the NP movement to [Spec, AspP] takes place in the same way as in the case of gerund head except that the relevant aspectual feature in this case is [+ DELIMIT]. The V head first checks [+ PERFECT] at Asp and then moves up to an Adj head. This is in keeping with the observation in Egerland (1996: 318) that adjectives in adjectival participles share the aspectual feature [PERFECT] with the verbal head.11 Comparing the derivation for a gerund in 24 and the one for the participle in 25, we can see that both involve NP movement inside the DP with an extra head movement in the case of the participle. Thus, gerund and participle formation is a matter of the argument structure that the LIs are mapped onto in the syntax. The difference between the two derivations lies in the fact that the participial -walno is unable to check for Case. Borer (1993) provides for this possibility in her theory. The delimited feature may be distinct from Case, as [Spec, AspP] may or may not be a Case position. Therefore, delimited arguments which do not carry accusative pass through a Spec which is specified as a place for checking [+ DELIMIT] but [- ACC] (see Egerland [1996: 111] on this point). This is consonant with the conclusion in Bhattacharya (1999b) that the participial walno absorbs Case. However, as I will show in Section 6.2, the mere presence of an object is not enough to guarantee participle formation.

Gerundial Aspect and NP Movement · 137

• 6. Unergative/Unaccusative Gerunds and Participles In this section I return to the data set presented in 3.2. If we consider a bigger set of predicates which fail to have a participial form corresponding to a gerund (as in 15-16), it will become immediately clear that there is a pattern among them. Let us first consider some clear-cut examples in this connection: (26) a. cheler kaSa boy's coughing b. radha-r douRono Radha's running c. rebar haMSa Reba's laughing (27) a. *kaSa chele coughed boy b. * douRono radha run Radha c. * haMSa reba laughed Reba That is, none of the predicates of 26 can have a corresponding participial reading. Notice that these verbs belong to the unergative class of verbs. Since unergatives' only apparent argument is an external one and they are marked by the apparent absence of an object, they are standardly assumed to have a structure where the verb does not subcategorize for an object argument position. Except certain unergative predicates which can take cognate objects (like sing, dance, talk, etc.) unergatives in general cannot have corresponding participles. Now let us look at another set of data: (28) a. chele-Ta-r baRi aSa boy-CLA-GEN home coming 'the boy's coming home' b. chele-Ta-ar pherot jawa boy-CLA-GEN back going 'the boy's going back' c. chele-Ta-r baRi pouMchono boy-CLA-GEN home arriving 'the boy's arriving home' (29) a.? baRiaSachele-Ta 'the come home boy' b.? pherot jawa chele-Ta 'the returned boy'

138 · Tanmoy Bhattacharya

c.? baRi pouMchono chele-Ta 'the arrived home boy' That is, unlike the unergative set, here the instransitive predicates marginally allow the corresponding participles. The verbs in 28, belong to unaccusatives whose only argument is the internal one or in any event not the external one. • 6.1. Deriving the Unergative/Unaccusative Gerunds and Participles Let us look at the derivation of the gerund from an unergative verb and the non-derivability of the participle from unergatives. Since unergatives project a structure with an external argument position, I will assume that the derivation for 26a starts off with 30a and forms a gerund as in 30b: (30) a.

b.

chclcr

Asp

VP

kaSa

V

I

Ir ng n nuvju

The V checks for its gerundial aspect feature [- PERFECT] by head movement to Asp. The subject moves to [Spec, DP] to check Genitive. The Asp does not select for a [DELIMIT] as unergatives do not select an inner argument.

Gerundial Aspect and NP Movement · 139

For the derivation of the participle in 27a, the perfective aspect of the participle is contingent upon the presence of an affected/delimited object. If this is so then given the structure in 30a, the absence of an internal argument prevents participle formation since the nominal aspect feature of [DELIMIT] of the Asp head remains unchecked and therefore the derivation crashes. The V may head move via Asp to A° checking [PERFECT] at Asp but the NP argument cannot check Case since participle wa/no absorbs Case. The difference between the two constructions therefore derive from the difference in their aspectual properties. All of this is visible in the structure below:

(31)

A AspP kaSa /\^ Spec Asp' chele Asp VP [delimit]l/\

-kaSa

V

For the unaccusatives, recall that they do not project an external argument position. Given the discussion of unaccusatives in 6, the base structure from which a gerund 32b and a participle 32c are derived is as in 32a:

(32) a.

VP V

I aSa

NP

I

chele

b. chele-r aSa (gerund) boy-GEN coming 'the boy's coming' c. ? aSa chele (participle) 'the having come boy' Accordingly the gerund and the participles are derived as follows:

140 · Tanmoy Bhattacharya

(33) a.

DP / cheler

D'

/\AspP

D

/\ /\

Spec

Asp'

Asp

VP

I /\NP

aSa V

oC ï

In the case of both 33a and b, the complement NP moves from an internal position to an external position (externalization of an argument) as per the nature of unaccusatives. In the present theory this is made possible by the presence of the [DELIMIT] feature on the Asp for the participle and [POSS] at D for the gerund and a matching feature on the argument chele. It is generally assumed that unaccusatives express a "change of state'. They refer to either "change of location' (arrive, go, run, etc.) or a "change of condition' (improve, increase, diminish, etc.). This semantic distinction is assumed to be captured by the [DELIMIT] in the present proposal drawing on a similar proposal in Tenny(1987). Additionally, since unaccusatives, by definition do not have an Accusative Case checking feature, the NP further moves up to [Spec, DP] to check Genitive in case of 33a. The derived position of the noun is different in 33b since chele is a full DP in the gerund and can therefore embed a Dem modifying the

Gerundial Aspect and NP Movement · 141

complement N, whereas a Dem can only modify the whole phrase in a participle. Recall that Dems are considered to be merged at a Spec position of a head which is lower than D (see Bhattacharya 1999b). • 6.2. Affectedness of the Object We have mentioned at the end of Section 5 that though both gerund and participle formation is a matter of NP movement, the mere presence of an object does not guarantee the formation of participles. Rather, given the aspectual account of gerunds proposed in this paper, the object must be affected. This accounts for the data is Section 3.2 repeated below: (34) a. ramer Sastrio Songit ERa-no (gerund) Ram's classical music avoid-GER "Ram's avoiding classical music' b. * ramer ERano Sastrio Songit (participle) "classical music avoided by Ram' I suggest that the gerund is derived as in 35b from the base structure as in 3Sa where the object NP is marked as an unaffected agent since the object of the activity of avoiding does not get affected in any way by the activity itself:12

(35) a. NP

Sastriyo Songit [-affected]

DP

Sastriyo Songit Asp

eRano

I

/^ o^»*»;* tJvPllgl I

142 · Tanmoy Bhattacharya

That is, although the object is marked non-affected, it still has to move up to [Spec, AspP] to check the Case feature. However, since the participle selects a [+PERFECT] Asp, it also requires an affected object. Thus, the derivation crashes due to reasons of feature mismatch with the result that participle formation does not take place:13

(36) Spec ramer

D' D

AP

Adj eRano

AspP

Sastriyo Songit [-affect] Asp [+affect] P.T! ç«é/> CIVUIIU

Sastriyo Songit To conclude, I have shown that a feature theory utilizing aspectual and Case properties of gerund and participle arguments can account for the data presented earlier. In particular, keeping the central theme of this paper in focus, I have shown that information about Case features in combination with the delimited/non-delimited nature of the arguments drive NP movement inside the DP in the case of both gerunds and participles. The absence of this movement in the case of unergative participles as opposed to unaccusative participles is accounted for by the absence of either an aspectual or a Case feature in the former. The various possibilities of gerund and participle formation is summarized below:

(37)

Object

Gerund

Participle

Transitive

[-»-affect]

/

/

Transitive

[-affect]

/

X

Transitive

nil

/

X

[-1- affect]

/

/

Predicate Type

Unaccusative

Gerundial Aspect and NP Movement · 143

• 7. Conclusions In conclusion, in this paper I have shown that gerunds universally exhibit events through an Asp head either morphologically or abstractly. In continuation with the central theme of my earlier work, I have also shown that either Case (for gerunds) or a [+DELIMIT] feature (for participles) of this Asp head drives NP movement inside gerundial and participial DPs in Bangla. The difference in the trigger on the other hand accounts for the unavailability of unergative participles in Bangla. • NOTES 1. There is no general consensus in using these terms in the generative tradition. For example, in the same volume as Chomsky's paper, Bruce Fr ser uses the term Factive Nominate for Chomsky's gerundive nominals and Substantive Nominate for Chomsky's Derived Nominate (Fr ser 1970: 84-85). Further, gerundive nominals are often identified as verbal gerunds or imperfect gerunds as opposed to nominal or perfect gerunds (as in John 's refusing of the offer) which in turn is also identified as Action Nominal (Fr ser 1970; Grimshaw 1990) or as mixed nominals (Chomsky 1970: 215). A distinction is also made in the literature between Action/ Event/Process nouns versus Result nouns (see Grimshaw 1990, Siloni 1997 among others). I discuss this terminological confusion further in Section 3. 2. Corresponding to the sentences in (i), there are GNs in (ii) and DNs in (iii) (from Chomsky [1970:187]): (i)

a. John is eager to please b. John has refused the offer c. John criticized the book

(ii)

a. John's being eager to please b. John's refusing the offer c. John's criticizing the book

(iii) a. John 's eagerness to please b. John's refusal of the offer c. John's criticism of the book 3. In Potawatomi, the tense morpheme-ý/é can appear on both event and common object nouns. It is the same morpheme which is affixed to verbs to express tense or aspect relations: (i) a. nos b. nosan c. nciman d. ncimanpan 'my father' 'my deceased father' 'my canoe' 'my former canoe' (ii) a. nkasatas b. nkasataspan º am happy' º was formerly happy' 4. I ignore the possibility that 3 may be considered to exhibit relative clause properties. 5. This is another name for the 'mixed form' in Chomsky (1970:215). 6. All the (ii) forms with -ba in 5 also act as a stem for the corresponding gerundive khabar Of eating', dEkhbar Of seeing' takabar Of staring' (more on gerundives shortly). 7. This is shown below: (i) a. lekh-bai Son-balkor-ba 'writing/hearing/seeing' (Standard) b. likh-bal Sun-bal kor-ba 'writing/hearing/seeing' (Dialectal)

144 · Tanmoy Bhattacharya In case of (b) the high vowel of -iba raises the preceding vowel and deletes (see Dasgupta [1980] for details). 8. The -iblba form is sometimes reduced to -a making it similar to the -wa/no gerund structure. However, a gerundive always takes a Genitive marker after the affix. Notice that the template suggested in Dasgupta (1986) conflates two identifiable forms of the gerundive: (i) So-ba-r ghOr sleep-GER-GEN house 'bedroom*

(ii) So-ba-r jonno sleep-GER-GEN for 'for sleeping'

The construction in (ii) is now classified according to Dasgupta (p.c.) as Dependent gerund. However, due to a lack of analytical work nothing definitive can be concluded from this distinction. Moreover, since this paper is not concerned with the correct analysis of the gerundive, I leave the investigation of this distinction for future research. 9. A sequential reading of a causal P is harder to get than a causal reading of a temporal P (as in lOb), but it improves with a temporal adverb. The point that this data establishes is that the agentive/ causal reading is stronger (more salient) with the Genitive. 10. This process is identified as 'packaging' in Jackendoff (1991) where a portion of the stuff is spatially demarcated by referring to a serving, a kind or a quantity of it. 11. Note that, therefore, the Asp head contains two aspectual features in case of participles, one to attract the delimited/ affected NP and the other to head attract the V with participial perfective aspect. 12. Anderson (1979:44) argues on a similar line for the following contrast: (i) a. b. (ii) a. b.

The Mongols'destruction of the city The city's destruction by the Mongols John's avoidance of Bill * Bill's avoidance by John

The difference in event types is responsible for the contrast above. In order to be affected an object must be changed or moved by the action of the head nominal. Tenny (1987) reinterprets this as the delimited/non-delimited dichotomy that I have adopted for this study. However, I will continue to use affected/delimited interchangeably in this and other sections. 13. So far, I have ignored in this discussion the other type of counterexample of Dasgupta (1980) in Section 3.2—bole bERano 'talk and going around'. At present, I have no clue as to the aspect of complex predicates. However, it can be argued that if the aspect of the complex is determined by the aspect of the head, then this particular complex predicate would pattern with unergatives explaining the absence of the participial form. • REFERENCES Abney, Steven. 1987. The English NP in its sentential aspect. Cambridge, MA: MIT dissertation. Anderson, Mona. 1979. Noun phrase structure. Storrs, CA: University of Connecticut dissertation. Bhattacharya, Tanmoy. 1998a. DP-internal NP movement. University College London working papers in Linguistics 10.225-52. . 1998b. Kinship inversion in Bangla. Papers in Linguistics from the University of Manchester, Volume 7.143-56. . 1998c. The nP-Shell in Bangla. Paper presented at the 19th roundtable meeting of the South Asian Linguistics Association, York. . 1999a. Specificity in the Bangla DP. Yearbook of South Asian language and linguistics, Vol. 2, ed. by Rajendra Singh, 71-99. New Delhi: Sage Publications. . 1999b. The structure of the Bangla DP. London: University College London dissertation.

Gerundial Aspect and NP Movement · 145 Bhattacharya, Tanrno^ (Forthcoming). In search of the vague 'one*. Proceedings of ConSOLE 7 (Conference of the Student Organisation of Linguistics in Europe). Leiden: University of Leiden. Borer, Hagit 1993. The projection of arguments. Functional projections. University of Massachusetts occasional papers 17, ed. by Elena Benedict and Jeffrey Runner, 19-47. Amherst, MA: University of Massachusetts. Brinton, Laurel, J. 1995. The aksionsart of deverbal noun in English. Temporal reference, aspect and actionality 1. Semantic and syntactic perspectives, ed. by Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham and Mario Squartini, 27-42. Turin: Rosenberg and Sellier. Chatterji,Suniti Kumar. 1926. The origin and development of the Bengali language, Vol 2. Calcutta: Calcutta University Press. Chomsky, Noam. 1970. Remarks on nominalization. Readings in English transformational grammar, ed. by Roderick Jacobs and Peter Rosenbaum, 184-221. Waltham, MA: Ginn and Co. . 1981. Lectures on government and binding. Dordrecht: Foris Publication. . 1995. The minimalist program. Cambridge, MA: MIT Press. DasgupU, Probal. 1980. Question and relative and complement clauses in Bangla grammar. New York, NY: New York University dissertation. . 1986. Gerunds, participles and the meaning of categories. Bulletin of the Deccan College postgraduate and research institute 45.7-18. . 1994. Verbless and nonfmite clauses. Pondicherry Institute of Linguistics and Culture (PILC) Journal ofDravidic Studies 4.2.57-68. De, Mina. 1984. Bangla gerunds. Hyderabad: Osmania University M.Phil dissertation. de Hoop, Helen. 1992. Case configuration and noun phrase interpretation. Groningen: University of Groningen dissertation. Egerland, Verner. 1996. The syntax of past paniciples: A generative study ofnonfinite constructions in ancient and modem Italian. Lund: Lund University Press. Fräser, Bruce. 1970. Some remarks on the action nominalization in English. Readings in English transformational grammar, ed. by Roderick Jacobs and Peter Rosenbaum, 83-98. Waltham, MA: Ginn and Co. Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press. Hale, Ken and Samuel J. Keyser. 1993a. On argument structure and the lexical expression of syntactic relations. The view from building 20, ed. by Ken Hale and Samuel J. Keyser, 1-52. Cambridge, MA: MIT Press. . (eds). 1993b. The view from building 20. Cambridge, MA: MIT Press. Hockett, Charles F. 1958. A course in modem linguistics. New York, NY: MacMillan. van Hout, Angeliek and Thomas Roeper. 1998. Events and aspectual structure in derivational morphology. MIT working papers in linguistics 32. Papers from the UPenn/MIT roundtable on argument structure and aspect, ed. by Heidi Harley, 175-200. Cambridge MA: MIT Press. JackendofT, Ray. 1991. Parts and boundaries. Cognition 41.9-45. Jacobs, Roderick and Peter Rosenbaum (eds). 1970. Readings in English transformational grammar. Waltham, MA: Ginn and Co. Lecanne, Jacqueline. 1996. Tense in the nominal system: The Somali DP. Studies in Afroasiatic syntax, ed. by Jacqueline Lecarme, Jean Lowenstamm and Ur Shlonsky, 159-78. The Hague: Holland Academic Graphics. Nakamura, Akira. 1994. On the tense system in Japanese. Japanese Korean linguistics, ed. by Noriko Akatsuka, 363-78. Stanford, CA: CSLI (Center for study of Language and Information). Peterson, John. 1997. Grammatical relations in Pali and the emergence ofergativity in Indo-Aryan. München: Lincom Europa. Pullum, Geoffrey. 1991. English nominal gerund phrases as noun phrases with verb phrase heads. Linguistics 29. 763-99. Reuland, Eric. 1983. Governing -ing. Linguistic Inquiry 14.1.101-36.

146 · Tanmoy Bhattacharya Schachter, PüuL 1976. A non-transformational account of gerundive nominate in English. Linguistic Inquiry 7.2.205-41. Siloni, TaL 1997. Noun phrases and nominalizations. The syntax ofDPs. Dordrecht: Kluwer. Jinny, Carol. 1987. Grammaticalizing aspect and affectedness. Cambridge, MA: MIT dissertation. Verknyt, Henk. 1972. On the compositional nature of the aspects. Dordrecht: Reidel.

The Passive and Related Constructions in Marathi* ι PRASHANT PARDESHI ι It is becoming increasingly clear among studies of languages of the world that the passive expression is related to other constructions such as the reflexive, reciprocal, spontaneous, potential, and honorific. The passive construction in Marathi is also a case in point. Presently available characterizations of the Marathi passive cannot offer a unified account, let alone its correlations with other constructions, owing to their discrete nature and narrow perspective. In light of this, this paper offers a detailed alternative analysis of the Marathi passive construction (a) accounting for a hitherto ignored twin passive, namely, the GO and COME passive, and (b) demonstrating that the passive construction in Marathi exhibits a continuum with other constructions along the lines suggested by Shibatani (1985).

• 1. Introduction The passive construction has occupied a prominent position in linguistic descriptions in the last few decades, and has been analyzed by modern linguists in various frameworks, such as transformational and generative grammar (Chomsky 1965; Kuroda 1979), relational grammar (Perlmutter and * I am greatly indebted to Prof. Masayoshi Shibatani and Prof. Yoshihiro Nishimitsu for their invaluable suggestions, criticisms and guidance. Thanks are also due to Prof. Mark Campana and Frank Owens for their stylistic improvements. The responsibility of the remaining errors and inadequacies, of course, lies with me. The following glossing conventions are used in this paper: ACC 'accusative marker', CAUS 'causative marker', DAT 'dative marker', EMPH 'emphatic marker', ERG 'ergative marker', F 'feminine', GEN 'genitive marker', INSTR 'instrumental marker', LOG 'locative marker', M 'masculine', N 'neuter', PAST 'past tense', PERF 'perfective aspect', PL 'plural', POT 'potential marker', PRES 'present tense', PTCPL 'participle', 3 'third person'.

148 · Prashant Pardeshi

Postal 1977; Pandharipande 1981; Rosen and Wali 1989), functional grammar (Givon 1979). As pointed out by Shibatani (1985), however, these formal as well as functional approaches are too restricted to account for the patterns of distribution which passive morphology exhibits. It has also been pointed out that various languages employ the same morphosyntactic properties as the passive in reflexive, reciprocal, spontaneous, potential and honorific constructions (Langacker and Munro 1975; Shibatani 1985). The Marathi morpheme V-PERF+GO is also a case in point, and is used in four construction viz., the passive, spontaneous, potential and honorific. Previous characterizations of the passive in Marathi (Berntsen and Nimbkar 1975; Damle 1911; Joshi 1900; Kher 1899; Pandharipande 1981; Rosen and Wali 1989; Tarkhadkar 1836) are not only unable to provide a unified account, but also too narrow in perspective to account for the correlation of the passive with the other constructions. In light of this, the goal of this paper is to (a) provide an alternative account for the passive in Marathi, pointing out the inadequacies in the previous descriptions, and (b) explore the correlations of the Marathi passive with other constructions such as the spontaneous, potential and honorific. As pointed out by Shibatani (1985), for a correct understanding of the correlation of the passive with the other constructions, a broader perspective is required. Accordingly, we will adopt the prototype framework proposed by Shibatani (1985) to explore the correlation of the Marathi passive with the other constructions. Let us first introduce the framework in which our analysis will be carried out. • 1.1. The Framework: Shibatani (1985) Shibatani adopts a traditional view of grammatical voice as a category that signals an unmarked vs. marked distinction of mapping between the basic syntactic functions of subject and object, and the basic semantic roles of agent and patient. In an active clause, the agent occupies the most prominent syntactic slot of subject, and thus cannot be deleted. However, in a passive clause the agent is removed from the most prominent syntactic slot, and is either demoted to the role of an adjunct or not encoded at all. Passive voice, therefore, can be seen as a means of demoting the agent or deleting it altogether. Thus, the prototypical passive is agentless. Shibatani (1985: 837) defines the passive prototype as follows: (1) Characterization of the passive prototype a. Primary pragmatic function: Defocusing of agent b. Semantic properties: (i) Semantic valence: Predicate (agent, patient) (ii) Subject is affected.

The Passive and Related Constructions in Marathi m 149

c. Syntactic properties: (i) Syntactic encoding: agent -» φ (not encoded) patient -» subject (ii) Valence of Predicate]: Active = P/n; Passive = P/n-1. d. Morphological property: Active = P; Passive = P[+ passive]. Shibatani points out that, rather than arguing whether a given construction should be considered passive or not, a description must be offered as to what extent the construction in question is similar to or different from the prototype. This view of grammar thus assumes that various constructions exist along a continuum. Some of them are prototypical, others are similar to the prototype to a limited extent, while still others share no similarities at all with the prototype. Such an approach is essential in understanding the relationships among various constructions within a single language, and is capable of providing a useful framework for cross-linguistic research. The rationale for adopting this framework comes from the fact that other frameworks cannot offer a unified account of the passive construction or its correlation with other constructions. The correlations discussed here are neither purely syntactic nor semantic in nature, but are rather pragmatic, i.e. all of them share a common pragmatic function viz., agent defocusing. In what follows we will demonstrate that the broader pragmatic notion of agent defocusing offers a unified account for the passive construction in Marathi, as well as the constructions related to it. We will start our discussion with the passive, which has been extensively discussed in earlier studies.

• 2. The Marathi Passive Marathi has a twin periphrastic passive construction which typically consists of an agent NP, if at all present, followed by a postposition kaDUn or dwAre, and a transitive verb with either a perfect participle marker /-/-/followed by the verb/jA-/'go' (hereafter referred to as the GO passive), or a participle/-A^4// followed by the verb/ye-/*come' (hereafter referred to as the COME passive). These are exemplified next.1 (2) polis-An-nl cor pakaD-l-A police-PL-ERG thief.M catch-PERF-M 'The police caught the thief.' [ACTIVE] (3) a. polis-An-kaDUn cor pakaD-ΙΆ ge-l-A police-PL-by thief.M catch-PERF-M go-PAST-M 'The thief was caught by the police.' [GO PASSIVE]

150 · Prashant Pardeshi b. polis-An-kaDUn cor pakaD-NyAt A-l-A police-PL-by thief.M catch-PTCPL come-PAST-M 'The thief was caught by the police.' [COME PASSIVE] In her extensive survey of passive constructions, Siewierska (1984) points out that different languages employ different verbs as passive auxiliaries, and that many languages possess more than one periphrastic passive—e.g. a BE and BECOME passive (Swedish, Latvian, Polish, Finnish, Nez Perce), a BE and GET passive (English) or a BECOME and GO passive (Bengali)—which are not freely interchangeable. The passive construction in Marathi is a case in point, which employs GO and COME as passive auxiliaries. Among the IndoAiyan languages, Maithili also uses GO and COME as passive auxiliaries. Siewierska further points out that the characteristics associated with passive clauses which contain particular auxiliary verbs appear to be language specific, and the use of a given auxiliary in languages which possess more than one such constituent is determined by a variety of semantic, syntactic, and stylistic factors. In earlier analyses of the Marathi passive construction, the COME passive was either not discussed at all, or was treated as a construction synonymous with the GO passive. We will demonstrate with ample illustration that this, however, is far from true. The GO passive and the COME passive are neither interchangeable nor mutually exclusive in all contexts. Moreover, they are not discrete entities but rather form a continuum. We shall begin with a review of earlier analyses of the Marathi passive in order to establish some background, and then argue that these are in fact inadequate to account for the Marathi passive—to say nothing of its correlation with other constructions. • 2.1. Review of Earlier Research Earlier analyses of the Marathi passive construction can be broadly divided into two categories, viz., traditional grammars (Damle 1911; Joshi 1900; Kher 1899; Tarkhadkar 1836), and modern linguistic descriptions (Berntsen and Nimbkar 1975; Pandharipande 1981; Rosen and Wali 1989). For the sake of convenience we will review them separately. 2.1.1. Traditional Grammars Traditional grammarians such as Tarkhadkar (1836), Kher (1899), Joshi (1900), and Damle (1911) confine their discussion to the morphological and syntactic aspects of the passive construction, and totally ignore the semantic and pragmatic aspects. As for the morphological aspect, the traditional grammarians discuss in detail the diachronic change from the suffixal to the periphrastic passive. With

The Passive and Related Constructions in Marathi m 151 regards to the syntactic aspect, they define the karmaNIprayog (passive construction) as one in which the verb agrees with the object (karma [lit. deed/ fate/object]) in number, person, and gender. According to this definition, whenever the subject is marked and the object is unmarked, the resultant construction would be the karmaNI, or passive construction. (4) rAmA pustak wAci-t-o Rama.M book.N read-PRES-M 'Rama reads the book.'

(Active construction)

(5) rAmA-ne pustak wAcI-l-e Rama-ERG book.N read-PAST.N (Passive construction with "Rama read the book.' instrumental subject) (6) ma-ΙΑ kAm kar-aw-t-e I-DAT work.N do-POT -PRES-N (Passive construction with dative Ί am able to work/I can work.' subject) This definition is incorrect in that in Marathi, the verb always agrees with an unmarked nominal (if any), and is thus independent of any construction. Example 5 is, in fact, an ergative construction, and should be distinguished from the passive construction.2 Example 6 is a dative subject construction which is void of an agent. The traditional analyses completely ignore the semantic and pragmatic aspects of the passive construction, yet—except for Joshi (1900)—all have discussed the GO as well as the COME passive. Their analyses, however, have treated these constructions as semantically identical, and thus mutually interchangeable in all contexts. We claim that this is not the case, and will discuss both of these constructions in greater detail later on. The traditional grammarians also fail to capture the correlation of the passive with other constructions such as the spontaneous, potential, and honorific. Having summarized the traditional analyses, let us now turn to modern linguistic descriptions pertaining to the Marathi passive. 2.1.2. Modern Linguistic Descriptions In this section we will review the modern linguistic descriptions pertaining to the Marathi passive and argue that they are incapable of providing a unified account of the passive construction in Marathi. Although, Berntsen and Nimbkar (1975) talk about the GO as well as the COME passive, they, however, remain silent about their distribution and thus tacitly treat them as synonymous. They also fail to recognize the correlation of the passive with other constructions. Pandharipande (1981) as well as Rosen and Wali (1989), on the other hand, completely ignore the COME passive. The relational grammar approach adopted by these authors is discrete in nature, and fails to capture the correlation

152 · Prashant Pardeshi

between the passive and related constructions like the spontaneous, potential and honorific. We will demonstrate in the next section that this failure to recognize the correlations has led them to make unwarranted claims regarding the Marathi passive. Before proceeding further, a discussion on the differences between the passive construction and the ergative construction is in order. It should be recalled that the traditional grammarians failed to make a distinction between the ergative construction and the passive construction in Marathi (cf. 5). It is thus appropriate to clarify that these constructions are not the same and therefore should be treated as independent. • 22. The Passive and the Ergative Construction In Marathi, the subject of a transitive clause (third person) in the past perfective is marked with the ergative marker ne (singular) or nl (plural). If the subject is a first or second person, there is no overt ergative marker; in either case, however, the verb agrees with the patient nominal, and not with the agent nominal. Note the following examples: (7) ml/tU/rAm-ne noTis wAc-l-I I/You/Ram-ERG notice.F read-PAST-F Ί/You/Ram read the notice'. The fact that the patient nominal of the ergative construction behaves like a subject (i.e. controls concord), makes it look like the passive construction. However, these two constructions exhibit fundamental differences. First, the ergative construction can be passivized, while the passive construction cannot: (8) rAm-kaDUn noTis wAc-l-I ge-l-I Ram-by notice.F read-PERF-F go-PAST-F 'The notice was read by Ram*. *(9) rAm-kaDUn noTis wAc-l-l ge-l-I ge-l-I Ram-by notice.F read-PERF-F go-PERF-F go-PAST-F 'The notice was read by Ram'. Example 8 is the passivized form of the ergative construction 7, while 9 is the 'passivized' form of the passive construction 8. Note that 8 is grammatical while 9 is not. Second, the agent nominal can be omitted in the passive, but not in the ergative construction. Deletion of the agent nominal in the ergative construction (viz., 7) turns out to be ungrammatical, as shown in 10. However, the deletion of the agent in a passive expression (viz., 8) does not affect its acceptability, as shown in 11:

The Passive and Related Constructions in Marathi m 153 *(10) noTis

wAc-l-I

notice F read-PAST-F '(Lit) Notice read.'

(11) noTis wAc-l-I ge-l-I notice.F read-PERP-F go-PAST-F 'The notice was read.' Third, in ergative constructions the agent nominal is typically human, while in the case of the passive there is no such restriction. This is exemplified in 1214: (12) a. rAm-ne phAndl toD-l-I Ram-ERG branch.F .break-PAST-F 'Ram broke the branch.' *b. waryA-ne phAndl toD-l-I wind-ERG branch.F berak-PAST-F 'The wind broke the branch.' (13) fulpAkharAn-kaDUn parAgkaN wAhi-l-e jA-tAt butterflies-by pollens.N carry-PERF-N go-PRES.PL.N 'Pollen is carried by butterflies.' (14) HrudayA-kaDUn sharlr-Atll wiwidh bhAgAn-nA rakta-purawathA heart-by body-in various organs-to blood-supply.M ke-l-A jA-t-o do-PERF-M go-PRES-M 'In the human body, the heart supplies blood to the various organs.' Ergative constructions thus display all the properties of active transitive clauses. The only difference between a canonical transitive clause and the ergative construction is in the pattern of agreement. In the former, the verb agrees with the agent nominal, while in the latter the verb agrees with a nominal other than the agent. In the absence of an unmarked nominal, the verb assumes a third person singular neuter form and does not agree with any of the nominals present in the clause. To sum up, in this section we have presented a critical review of earlier research regarding the Marathi passive, and pointed out the inadequacies therein. We have also demonstrated that the passive and the ergative are independent constructions in their own right, and that the traditional treatment of ergative constructions as passive is inappropriate. In the next section we will present an alternative analysis of the Marathi passive construction.

• 3. An Alternative Analysis of the Marathi Passive As pointed out earlier, the previous treatments—traditional as well as modera—are unable to account for the Marathi passive and its correlation with

154 · Prashant Pardeshi

other constructions. In this section we will present an alternative analysis of the Marathi passive. The analysis presented here addresses two issues, namely, the distribution of the GO and COME passive, and the correlation of the passive construction with other morphosyntactically related constructions which have not been taken up in earlier studies: the spontaneous, potential, and honorific. To begin with, we agree with Pandharipande (1981) that the passive in Marathi is a rule-governed, in that it applies to a semantically definable class of volitional transitive verbs. Thus in Marathi, intransitive verbs typically fail to yield passives, and agents permitted in the passive tend to be typically human. We do not agree with Pandharipande, however, concerning her claims regarding the functions uniquely performed by the passive, namely, expressing a capabilitative (ibid.: 123) and a prescriptive meaning (ibid.: 127). Rosen and Wali (1989) classify the Marathi passive into two categories on the basis of the meaning they convey: regular passive (RP) and capabilitative passive (CP). Thus, like Pandharipande, they also claim that the passive expresses a capabilititative meaning. It is true that in Marathi, the same morphology, namely, V-PERF+GO is employed for expressing (non)capabilitative meaning; however, we claim that this is not the passive construction per se but rather the potential construction which shares morphosyntactic similarities with the passive. If capabilitative meaning is peculiar to the passive construction, it follows that intransitive verbs marked with the V-PERF+GO morphology should fail to express it, since in Marathi intransitive verbs do not yield the passive. Contrary to Pandharipande's claim, however, intransitive verbs involving the V-PERP + GO morphology do express a capabilitative reading. Pandharipande also claims that the capabilitative meaning expressed by the passive results from the ex-subject/agent's effort, and is determined by agent-external conditions such as the weather (ibid.: 123). Again contrary to Pandharipande's claim, however, the capabilitative meaning expressed by a verb can be neutral with regards to agent-internal conditions (headache, pain, hatred, happiness, physical, and psychological pain, etc.) or agent-external (weather, etc.) ones. Note the following examples: (15) a. ghAN wAs yet as-tyAmuLe rAm-kaDUn tyA foul smell coming be-because Ram-by that kholl-t zop-l-e ge-l-e nAhi room-LOC sleep-PERF-N go-PAST-N not "Ram could not sleep in that room as it was stinking.' b. bhitl-muLe rAm-kaDUn tyA kholl-t zop-l-e fear-because Ram-by that room-LOC sleep-PERF-N ge-l-e nAhi go-PAST-N not 'Ram could not sleep in that room out of fear.'

The Passive and Related Constructions in Marathi m 155 (16) a. bhAtA-lA wAs ye-t as-fyAmuLe rAm rice-ACC smell come-PTCPL be-because Ram kaDUn bhAt khA-ll-A ge-l-A nAHI by rice.M eat-PERF-M go-PAST-M not 'Ram could not eat the rice since it had a foul smell.' b. poT bhar-l-e as-fyAmuLe rAm kaDUn bhAt stomach.N full-PAST-N be-because Ram by rice.M khA-ll-A ge-l-A nAhl eat-PERF-M go-PAST-M not 'Ram could not eat rice since he was full/ According to Pandharipande's analysis, transitive verbs which involve the VPERF+GO morphology and express a capabilitative meaning should be treated as passives, while intransitive verbs which involve the V-PERF+GO morphology and express a capabilitative meaning would be either barred or treated as constructions unrelated to the passive. Pandharipande as well as Rosen and Wali thus fail to recognize the distinction between the passive and the potential construction, and make unwarranted assumptions about Marathi grammar. In addition, if the claim is correct that the capabilitative meaning is peculiar to the passive, a modal verb which expresses capability, namely, ShakNe (can) should not be tolerated in a passive expression. Contrary to the claim, however, a capabilitative modal can occur in a passive expression as exemplified below: (17) sAdhAraN widyArthyA-kaDUn-hl he gaNit sahaj ordinary student-by-EMPH this.N math.problem.N easily soDaw-l-e jA-u shak-t-e solve-PERF-N go-PTCPL can-PRES-N 'This mathematical problem can be easily solved even by an ordinary student/ Furthermore, Pandharipande's proposal predicts that those constructions should be blocked which involve the 'passive' morphology and (a) express effortless, agent-internally-determined capabilitative meaning, or (b) indefinite agent and a capabilitative meaning. However, neither of these predictions is borne out, as exemplified below: (18) bharpUr jewaN zAla as-Una-hl mAjhyA-kaDUn too much meal became be-PTCPL-EMPH me-by don Ambe sahaj khA-ll-e ge-l-e two mangoes.N without effort eat-PERF-N go-PAST-N 'Even after a heavy meal I could easily eat two mangoes.'

156 · Prashant Pardeshi (19) bharpUr jewaN zAla as-la tarl Aiskrim t go much meal became be-PERF though ice-cream.N sahaj kha-ll-a jA-t-a easily eat-PERF-N go-PRES-N 'Even after a heavy meal one can easily eat ice-cream.' The above data thus undermines the claim that a capabilitative meaning is peculiar to the Marathi passive. As pointed out earlier, it is not the passive construction per se which expresses the capabilitative meaning, but rather a potential construction which exhibits morphosyntactic similarities to the passive. According to Pandharipande, another function uniquely performed by the passive in Marathi is that of expressing a social convention, and thereby prescribing a particular mode of behavior (1981: 127). Only passives without agent phrases are used in this way, and such passives differ from other constructions conveying prescriptive meaning in terms of the degree of politeness expressed by them. Pandharipande claims that such passive expressions convey the highest degree of politeness. Contrary to Pandharipande's claim, however, such passives do not obligatorily convey a prescriptive meaning as exemplified below: (20) japAn-madhye japAnl bhAshA bola-l-I jA-t-e Japan-in Japanese language.? speak-PERF-F go-PRES-F 'The Japanese language is spoken in Japan.' (21) mahArAshtrA-t gaNeshotsaw mothyA pramANA-war Maharashtra-in Ganesh festival.M big scale-on sAjrA ke-l-A jA-t-o celebrate do-PERF-M go-PRES-M 'The Ganesh festival is celebrated on a grand scale in Maharashtra state.' We treat these examples as indefinite/covert agent passives in which the agent phrase, despite being conceptualized, does not occur in the construction. The unwarranted claims made by Pandharipande and Rosen and Wali regarding the meaning/functions of the passive stem from their failure to recognize the correlations that the passive has with other constructions. The constructions expressing capabilitative meaning discussed above are claimed by Pandharipande and Rosen and Wali to be the passive, while in fact they represent a separate one: the potential. At the same time, the constructions expressing a prescriptive meaning are claimed by Pandharipande to be passive, yet they are, in fact, indirectly related to the honorific construction in that a particular mode of behavior is prescribed as a means of expressing respect/ honor, as exemplified below:

The Passive and Related Constructions in Marathi m 157 (22) booT ghAl-Un mandirA-t prawesh ke-l-A boot wear-PTCPL temple-LOC entry.M do-PERF-M jA-t nAhl go-PRES not One is not supposed to enter a temple with shoes on.' The potential, as well as the honorific construction shares morphosyntactic similarities with the passive. To conclude, the discrete approach adopted by Pandharipande and Rosen and Wall fails to recognize these correlations, and has led them to make unwarranted claims regarding Marathi passives. Let us now turn to another important issue related to the passive that has not been addressed in previous analyses, namely, the distribution of the GO and COME passives. • 3.1. The Distribution of GO and COME passives As mentioned earlier, most of the previous studies on the Marathi passive have not even mentioned the COME passive, while the few that have treat it as semantically identical with the GO passive and thus interchangeable in all contexts. In this section we will demonstrate that the GO and COME passives are not semantically identical, and that the notion of 'intention' plays a key role in their distribution. We claim that, in Marathi, COME passives are typically employed to depict a meticulously planned, highly intentional event in which the agent plays the role of the planner, and who brings about the event with the aim of achieving the desired outcome. The GO passive, on the other hand, is typically employed to depict a non-meticulously planned, less intentional event. It should be noted that the notion of 'intention' is not a matter of 'all or nothing' but rather a matter of 'degree'. The higher the degree of intention, the greater the chances of employing the COME passive by the speaker and vice versa. From the above explanation, one can predict that in the case of a meticulously planned (highly intentional) event, the GO passive will be blocked and—conversely—in the case of a non-meticulously planned (less intentional) event, the COME passive will be prevented from appearing. Both of these predictions are borne out, as exemplified below: (23) hA bomb sabhAsthAnA-pAsUn shambhar miTar antarA-war this bomb.M meeting place-from hundred meter distance-on ubhyA kelefyA ekA moTArl-t thew-NyAt A-l-A standing did one motor-in keep-PTCPL come-PERF-M ho-t-A l*thew-l-A ge-l-A ho-t-A become-PRES-M/ Keep-PERF-M go-PAST-M become-PRES-M 'This bomb was kept in a vehicle parked a hundred meters away from the venue of the meeting.' (Kesari, 15 February 1998, wruttawishesh column [internet edition].)

158 · Prashant Pardeshi

(24) uttarpradeshA-tll kafyANsing yanche sarkAr baDtarfa Utterpradesh-in Kalyansing his government.N dismiss kar-NyAt A-l-e l*ke-l-e ge-l-e do-PTCPL come-PERF-N/ do-PERF-N go-PAST-N 'The Kalyan Singh government in Uttar Pradesh was dismissed.' (Kesari, 22 February 1998 wruttawishesh column [internet edition].) Events like planting a bomb or dismissing a government depict a highly intentional activity involving a high degree of planning, and—as correctly predicted by our proposal—the COME passive is allowed while the GO passive is blocked. The following examples depict the reverse situation: (25) a. japAn-madhye japAnl bhAshA Japan-in Japanese language.F 'In Japan, they speak Japanese.' *b. japAn-madhye japAnl bhAshA Japan-in Japanese language.F ye-t-e come-PRES-F 'In Japan, they speak Japanese.'

bol-l-i jA-t-e speak-PERF-F go-PRES-F bol-NyAt speak-PTCPL

(26) a. dakshiN bhAratA-t prAmukyAne tAndUL khA-ll-A south India-LOC mainly rice.M eat-PERF-M jA-t-o go-PRES-M 'In South India, mainly rice is eaten.' *b. dakshiN bhAratA-t prAmukyAne tAndUL khA-NyAt south India-LOC mainly rice.M eat-PTCPL ye-t-o come-PRES-M 'In South India, mainly rice is eaten.' The events depicted in these examples are less intentional, and typically lack meticulous planning in that there is no definite entity that exercises conscious effort in bringing them about. Thus, as per our proposal, they can be couched only in the form of a GO passive, while the COME passive is barred. Native speakers of Marathi do make a distinction between GO and COME passives. Note the following contrast: (27) tyA-ΙΑ sewAjeshthate-nusAr badhatl di-l-I he-DAT seniority-as per promotion.F give-PERF-F ge-l-i l*de-NyAt ΑΊ-Ι go-PAST-F/ give-PTCPL come-PAST-F 'He was promoted on a seniority basis.'

The Passive and Related Constructions in Marathi m 159

(28) tyA-ΙΑ sewAjeshthatA DawlUn badhatl de-NyAt he-DAT seniority violating promotion.? give-PTCPL A-l-I l*di-l-i ge-l-i come-PAST-F/give-PERF-F go-PAST-F 'He was given promotion, violating the norms of seniority.' Example 27 depicts a event in which a controller who has the authority to award a promotion does so as a part of some routine procedure, while example 28 depicts a meticulously planned activity, overriding an otherwise routine procedure. The former depicts a typically less intentional activity, while the latter depicts a typical highly intentional one. In consonance with our proposal then, the COME passive is blocked in 27 while the GO passive is blocked in 28. • 3.2. Additional Evidence In this section we will provide some additional evidence to support our proposal that the notion of 'intention' dictates the selection of GO or COME as a passive auxiliary in Marathi. 3.2.1. Adverbial Modification If our proposal is correct, adverb cukUn 'mistakenly' should occur only with GO passives which typically depict non-meticulously planned, less intentional events and should be blocked in COME passives which typically depict meticulously planned, highly intentional events. Conversely, the adverb muddAm 'purposely' should occur only with COME passives and should be barred from GO passives. All of these predictions are borne out as exemplified below: (29) rAm-kaDun rAwaN cukUn mAr-l-A ge-l-A Ram-by Ravan by mistake kill-PERF-M go-PAST-M 'Ravan was killed by Ram mistakenly.' (30) rAm-kaDUn rAwaNA-lA muddAm mAr-NyAt A-l-e Ram-by Ravan-ACC purposely kill-PTCPL come-PAST-N 'Ravan was killed by Ram purposely.' *(31) rAm-kaDUn rAwaNA-lA cukUn mAr-NyAt A-l-e Ram-by Ravan-ACC by mistake kill-PTCPL come-PAST-N 'Ravan was killed by Ram mistakenly.' *(32) rAm-kaDun rAwaN muddAm mAr-l-A ge-l-A Ram-by Ravan purposely kill-PAST-M go-PAST-M 'Ravan was killed by Ram purposely.' The contrast exhibited among 29, 30, 31 and 32 lends strong support to our proposal. 29 depicts an event in which Ram did not intend to kill Ravan, but

160 · Prashant Pardeshi just by mistake happened to kill him. The killing of Ravan in this case is less intentional and involves no meticulous planning and hence the GO passive is used. The situation in 30 is just the opposite. In the case of 31, the COME passive is employed—which depicts a meticulously planned, highly intentional event. Nevertheless, the adverb cukun 'by mistake' which modifies it is not compatible with such an event, and consequently presents a contradiction. The situation in 32 is just the opposite. The adverb muddAm On purpose' creates a contradiction in the context of a GO passive, as they do not depict meticulously planned, highly intentional events. As predicted by our proposal, 31 as well as 32 turn out to be ungrammatical. 3.2.2. Co-occunance with a Modal Expressing Possibility Marathi has a modal expression that conveys the notion of possibility. COME passives depict meticulously planned, highly intentional events and thus do not leave any room for speculation. On the contrary, GO passives can accommodate a speculative meaning. In consonance with our proposal, GO passives can thus accommodate the modal's speculative/possibilitative meaning, while the COME passives can not: (33) tu-ΙΑ dilll-t fas-aw-la jA-NyAci you-ACC Delhi-LOC cheat-CAUS-PERF go-PTCPL shakyatA Ahe possibility BE 'There is a possibility that you may be cheated in Delhi'. *(34) tu-ΙΑ dilll-t fas-aw-NyAt ye-NyAd you-ACC Delhi-LOC deceive-CAUS-PTCPL come-PTCPL shakyatA Ahe possibility BE "There is a possibility that you will be cheated in Delhi'. It is interesting to note that if the speaker of the utterance 34 knows in advance about some plan to cheat, the utterance is perfectly grammatical.3 In such a case, the speaker uses the COME passive but at the same time does not want to commit himself, and thus warns the interlocutor in a non-committal way by using the possibilitative modal form. This shows that extra-linguistic/pragmatic information plays a crucial role in the grammaticality judgments of the GO and COME passives. 3.2.3. Non-human Agents Passive expressions with non-human agents are always couched in the form of a GO passive, and never in the form of a COME passive. This fact receives a natural explanation under our proposal, in that non-human agents lack intention and the ability to plan meticulous planning in order to bring about an event. Note the following contrast:

The Passive and Related Constructions in Marathi m 161

(35) a. fulpAkharAn-kaDUn parAgkaN butterflies-by pollens.N Tollen is carried by butterflies.' *b. fulpAkharAn-kaDUn parAgkaN butterflies-by pollens.N 'Pollen is carried by butterflies.'

wAhi-l-e jA-tAt carry-PERF-N go-PRES.PL.N wAha-NyAt ye-tAt cany-PTCPL come-PRES.PL.N

(36) a. hrudayA-kaDUn sharlr-Atll wiwidh bhAgAn-nA heart-by body-in various organs-to rakta-purawathA ke-l-A jA-t-o blood-supply.M do-PERF-M go-PRES-M Ίη the human body, blood supply to the various organs is done by the heart.' *b. hrudayA-kaDUn sharlr-Atll wiwidh bhAgAn-nA heart-by body-in various organs-to rakta-purawathA kar-NyAt ye-t-o blood-supply.M do-PTCPL come-PRES-M Ίη the human body, blood supply to the various organs is done by the heart.' Consider further the above-mentioned examples. The nominals in the agent slots (viz., hruday 'heart' andphulpAkhare 'butterflies') are non-volitional, yet potent entities. We regard these expressions to be ambiguous between the passive and spontaneous constructions. On the agent-volitionality parameter, they can be interpreted as spontaneous, while on the agent-potency parameter they can be interpreted as passive. To summarize, the evidence presented in the foregoing discussion lends strong support to our proposal that the notion of 'intention' plays a crucial role in determining the distribution of GO and COME passives. We will now turn to the issue of the correlations that the Marathi passive has with other constructions (the spontaneous, potential, and honorific), which had not been addressed in the previous research.

• 4. The Passive and Related Constructions Having discussed the passive construction in the previous section, we now focus on the constructions related to it, namely the spontaneous, the potential and the honorific constructions. Shibatani (1985:825) claims that various constructions can be related not just in morphosyntactic or semantic terms, but also in terms of common pragmatic function. In Marathi, the passive, spontaneous, potential and honorific constructions share morphosyntactic similarities in that the agent is consistently marked with an oblique marker, while the main verb is rendered in the V-PERF+GO form. We will demonstrate that

162 · Prashant Pardeshi

despite the morphosyntactic similarities these constructions are independent in their own right, and yet they are related to the passive construction through a pragmatic notion, namely agent defocusing. Let us begin with the spontaneous construction. • 4.1. The Spontaneous Construction A prototypical spontaneous construction depicts an event that occurs on its own without the intervention of an external agent. In other words, spontaneous events are void of agency and volition. Spontaneous expressions in Marathi share morphosyntactic similarities with the passive construction—even though they differ semantically in that the former typically lacks an agent, while in the latter an agent is always involved. This correlation finds natural explanation in the framework offered by Shibatani (1985) in that the spontaneous and the passive share a common pragmatic function, namely, agent defocusing. As Shibatani (1985: 838) states: Defocusing of an agent is highly germane to spontaneous events and states. An event predicated of an agent is basically causative; i.e. an event is brought about by an agent. But an event dissociated from an agent is one occurring spontaneously. Thus a sentence with a defocused agent may be utilized to describe a spontaneous event. In spontaneous expressions the agent is absent altogether, while in the passive the agent is posited and defocused syntactically. Let us take a closer look at the spontaneous construction in Marathi. 4.1.1. The Spontaneous Construction in Marathi In Marathi, passive morphology, namely, V-PERF+GO is employed to express a spontaneous event as exemplified below: (37) yA kAdambarl-d pahill don prakarNa ml jANiwpurwak this novel-of first two chapters.? I consciously lihi'l-L parantu nantar-cl prakarNe matra ml lihill write-PERF-F but after-of chapters.F however I wrote Ahet ase ml muLIc mhaNaNAr nAhl. tl Uhi-l-i BE like I never say not those write-PERF-F ge-l-i asa-c ml mAn-t-o go-PAST-F like-EMPH I.M believe-PRES-M Ί have consciously written the first two chapters of this novel. However, I would never say that I have written the subsequent chapters. I firmly believe that they got written themselves (Kher 1990: 7).'

The Passive and Related Constructions in Marathi m 163 (38) yA rAsAynik abhikriye-t urjyA bAher Tak-l-I this chemical reaction-in energy.F out throw-PERF-F jA-t-e go-PRES-F 'In this chemical reaction energy is given off.' (39) yA kyAmeryA-t rol ApoAp gundAL-l-A jA-t-o this camera-LOC film.M by itself wind-PERF-M go-PRES-M Ίη this camera, the film rewinds itself.' (40) sundar strl-kaDe tAbaDtob laksha wedh-l-a beautiful lady-toward immediately attention.N draw-PERF-N jA-t-e go-PRES-N 'Beautiful ladies are often noticed.' (Lit. One's attention unknowingly gets drawn towards a beautiful lady.) (41) bhukampA-t hajAro lok jiwanta gAD-l-e earthquake-in thousands people.N alive buiy-PERF-N ge-l-e go-PAST-N 'In the earthquake thousands of people were buried alive.' It is noteworthy that the main verbs involved in these expressions are all agentive and volitional; however, they are deagentivized by dissociating agency from them. Furthermore, all of the verbs lack corresponding intransitive counterparts, hence passive forms are employed to fill the gap. This is a manifestation of the tendency of languages to fill in lexical gaps through available means. Note the following contrast: (42) samudra-cyA lATAn-nl jahAj dor fek-l-e sea-of waves-INSTR ship.N far throw-PERF-N ge-l-e go-PAST-N 'The ship was thrown far off by the sea's waves.' (43) *a. wAdaLA-ne chappara uD-aw-l-I ge-l-I storm-INSTR roofs.F fly-CAUS-PERF-F go-PAST-F 'The roofs were blown off by the storm.' b. wAdaLA-ne chappara uD-Un ge-l-I storm-INSTR roofs.F fly-PTCPL go-PAST-F 'The roofs got blown off by the storm.' The verb fekNe 'to throw' lacks an intransitive counterpart, while the verb uDawNe 'to make fly' has a corresponding intransitive form, namely, uDNe 'to

164 · Prashant Pardeshi

fly'. In order to express a spontaneous event then, the former employs passive morphology to derive an intransitive counterpart, while the latter uses a corresponding intransitive form. In order to express a spontaneous event in Marathi, passive morphology is used only when the verb in question lacks an intransitive counterpart. This indicates that spontaneous expressions are close to prototypical non-volitional intransitive verbs, and are thus distinctly different from volitional passive expressions. Interestingly, only the GO passive and never the COME passive can be employed in Marathi spontaneous expressions. This is consistent with our proposal that the COME passive depicts meticulously planned events. It is thus incompatible with the spontaneous expression, which depicts events occurring on their own accord, as exemplified next: (44) a. bhukampA-t hajAro lok jiwanta gAD-l-e earthquake-in thousands people.N alive bury-PERF-N ge-l-e go-PAST-N 'In the earthquake thousands of people were buried alive.' *b. bhukampA-t hajAro lok jiwanta gAD-NyAt earthquake-in thousands people.N alive bury-PTCPL A-l-e come-PAST-N 'In the earthquake thousands of people were buried alive.' (45) a. pUl kosaL-un tyA-khAU dahA lok bridge collapse-PTCPL that-below ten people.N ciraD-l-ege-l-e crush-PERF-Ngo-PAST-N 'The bridge collapsed and ten people got crushed below it.' *b. pUl kosaL-un tyA-khAU dahA lok bridge collapse-PTCPL that-below ten people.N ciraD-NyAt A-l-e crush-PTCPL come-PAST-N 'The bridge collapsed and ten people got crushed below it.' From the foregoing discussion it is evident that the spontaneous and the passive are separate constructions in their own right. Semantically, they differ remarkably in that spontaneous events are void of agency, while passive events tacitly imply it. Nevertheless, this distinction is not rigid and cases of overlap are attested (cf. 3Sa, 36a). Still, passive and spontaneous constructions are marked morphosyntactically in a similar way as they share a pragmatic function, namely, agent defocusing. In the passive, agent defocusing is partial, while spontaneous expressions lack agents altogether.

The Passive and Related Constructions in'Marathi m 165 • 42. The Potential Construction Potential constructions express the ability/inability of the agent to perform the activity described by the main verb. In Marathi, the potential construction is morphosyntactically marked in a similar way to the passive in that the agent appears in an oblique form, and the main verb is marked with the V-PERF+ GO morphology. This is because the potential construction is very closely related to the passive—as well as to the spontaneous construction. Shibatani (1985: 839) observes that ... An event that occurs spontaneously has a strong propensity to happen. If this automatic happening is negated, then a reading of impotentiality is implied. In Marathi, potential constructions are more felicitous in negative contexts. Let us take a closer look at the potential construction in Marathi. 4.2.1. The Potential Construction in Marathi Marathi has the following three constructions that can be treated under the rubric of potential [X = agent, V = Main verb ( + volitional)]. Α. Χ-φ NP V + shakaNe'can'...CAN potential (46) rAm bhAt khA-U shak-t-o Ishak-at nAhl Ram.M rice eat-PTCPL can-PRES-M /can-PTCPL not 'Ram can/cannot eat rice.' B. X-oblique NP V-aw n/!A/(not)... AW Potential (47) rAm-cyA-ne bhAt khA-waw-l-A nAhl Ram-GEN-INSTR rice.M eat-POT-PERF-M not 'Ram could not eat rice.' C. X-oblique NP V +jANe(go) nAhI(not)... GO Potential (48) rAm-cyA-ne bhAt khA-ll-A ge-l-A nAhl Ram-GEN-INSTR rice.M eat-PERF-M go-PERF-M not 'Ram could not eat rice.' The periphrastic CAN potential construction is felicitous in both positive and negative contexts, while the AW construction is felicitous only in negative ones. GO potential constructions are generally felicitous in negative contexts, but they are not rejected altogether in positive ones. The CAN potential construction is an active construction in that the agent is unmarked. In contrast, the AW and GO potentials are non-active constructions, as evidenced by the fact that the agent exhibits oblique marking.

166 · Prashant Pardeshi

As pointed out earlier, the discrete approach adopted by Pandharipande (1981) and Rosen and Wali (1989) fails to make a distinction between passive and potential constructions. Note the following examples: (49) rAm-kaDUn tyA kholl-t zop-l-e ge-l-e nAhl Ram-by that room-in sleep-PERF-N go-PAST-N not 'Ram could not sleep in that room.' (50) rAm-kaDUn ajibAt poh-l-e ge-l-e nAhl Ram-by at all swim-PERF-N go-PAST-N not 'Ram could not swim at all.' (51) rAm-kaDUn don kUomiTar suddhA dhAw-l-e Ram-by two kilometer even run-PERF-N jA-t nAhi go-PRES not 'Ram cannot run even two kilometers.' (52) Ai-kaDUn mulA-lA ragAw-l-e ge-l-e nAhl mother-by child-ACC scold-PERF-N go-PAST-N not a. 'The child was not scolded by the mother.' b. 'The mother could not scold the child.' (53) mAzyA kaDUn tyA-ΙΑ fasaw-l-e ge-l-e nAhl I by he-ACC cheat-PERF-N go-PAST-N not a. 'He was not cheated by me.' b. Ί could not cheat him.' Among the expressions mentioned here, those involving a transitive verb and expressing a capabilitative meaning (viz., 52,53) would be treated as passives by Pandharipande, and as impersonal passives by Rosen and Wali. Those involving an intransitive verb and expressing a capabilitative meaning (viz., 49, 50 and 51) would be either barred or treated as something unrelated to the passive. We disagree with both of these analyses, as they are discrete in nature and fail to provide a unified account for the phenomenon under consideration. In the potential construction, the obliquely marked agent is obligatory, whereas in the passive construction it can be optionally deleted. In all the examples given earlier, the agent phrases marked with kaDUn cannot be deleted—except in 52 and 53, which present cases of overlap between the passive and the potential construction. Under the gloss (a) interpretations, they are passives and permit agent deletion, while under the gloss (b) interpretations, they are potentials and do not permit agent deletion. In Marathi, only transitive (volitional) verbs can yield passive expressions, whereas the potential construction can be formed with intransitive, as well as transitive verbs (exemplified above). Thus the domain of application of the potential and the passive construction is not the same, and as such they must be recognized as

The Passive and Related Constructions in Marathi m 167

independent constructions. What the potential construction in Marathi shares with the passive is the pragmatic function of agent defocusing, and is thus morphosyntactically marked in a similar way. Pandharipande (1981) also claims that AW potential constructions express (in)capability of the agent, determined by agent-internal conditions like headaches, pain, hatred, happiness, physical/psychological pain, etc. GO potential constructions, on the other hand, express an (in)capability based on the agent's effort, and is determined by agent-external conditions like the weather. CAN potential constructions are neutral to whether their capability is determined by agent-internal or agent-external conditions, and whether or not the agent expends any effort in accomplishing the task. As mentioned earlier, however, the claims made by Pandharipande regarding the conditions that determine an (in)capability reading are not empirically supported (cf. 18,19). The GO potential construction is the only form that is relevant to our analysis because it shares morphosyntactic similarities with the GO passive. In subsequent discussion we thus confine ourselves to an examination of the GO potential construction. First, GO potential constructions are generally felicitous in negative contexts, although they are not rejected altogether in positive ones: (54) bharpUr jewaN zAla as-Una-hl mAjhyA-kaDUn too much meal became be-PTCPL-EMPH me-by don Ambe sahaj khA-ll-e ge-l-e twomangoes.N without effort eat-PERF-N go-PAST-N 'Even after a heavy meal, I could easily eat two mangoes.' Second, the (in)capability of the agent expressed by the GO potential construction is related to an event occurring in the real world, they cannot express a stative/attributive ability. Note the following contrast: *(55) rAm-kaDUn bhAt khA-ll-A jA-t-o Ram-by rice.M eat-PERF-M go-PRES-M 'Ram can eat rice/ (56) rAm-kaDUn bhAt khA-ll-A ge-l-A nAhl Ram-by rice.M eat-PERF-M go-PAST-M not 'Ram could not eat rice/ Third, the presence of a definite agent is mandatory in the GO potential construction, although in some cases the agent can be indefinite. When the agent is indefinite, the construction is freed from eventive or negative contextual restrictions. GO potential constructions express a capabilitative meaning in the case of a definite agent, and a possibilitative meaning in the case of an indefinite one. Note the following contrast:

168 · Prashant Pardeshi

(57) bharpUr jewaN zAla as-la tarl Aiskrim too much meal became be-PERF though ice-cream.N sahaj kha-ll-a jA-t-a easily eat-PERF-N go-PRES-N 'Even after a heavy meal one can easily eat ice cream.' (58) bharpUr jewaN zAla as-Una-hl mAjhyA-kaDUn too much meal became be-PTCPL-EMPH me-by don Ambe sahaj khA-ll-e ge-l-e twomangoes.N without effort eat-PERF-N go-PAST-N "Even after a heavy meal, I could easily eat two mangoes.' Having examined the potential construction we now turn to the honorific construction. • 43. The Honorific Construction In the honorific construction, deference is expressed towards the agent of the action. In Marathi, the honorific construction is marked morphosyntactically in the same way as the passive. This too can be explained in terms of the pragmatic notion of agent defocusing. As pointed out by Shibatani (1985:837-38): A universal characteristic of honorific speech lies in its indirectness; and one of the clear manifestations of this is avoidance of the singling out of an agent which refers to the addressee, the speaker, or the person mentioned in the sentence. Defocusing of an agent in some way is thus an integral component of the honorific mechanism. Let us take a closer look at the honorific construction in Marathi. 4.3.1. The Honorific Construction in Marathi In Marathi, honorification in general is not widespread, and the construction is found only marginally. In the Marathi honorific construction, agent defocusing strategy is twofold: coding the agent as plural (rather than singular), and indirect reference of the agent effected by assigning it an oblique marker. Note the contrast in the following examples: (59) ???a. rAjyapAl-An-nl yashaswl widyArthAn-cA state governor-PL-ERG successful students-GEN satkAr ke-l-A felicitation.M do-PAST-M "The state governor honoured successful students.' b. rAjyapAl-An-cyA-hasfe yashaswl widyArthAn-cA state governor-PL-GEN-hands successful students-GEN.

The Passive and Related Constructions in Marathi m 169 satkAr ke-l-A ge-l-A felicitation.M do-PERF-M go-PAST-M 'Successful candidates were honoured at the hands of the state governor.' (60) ???a. yA prasangl mukhyAmantryAn-n-I raktadAn this occasion chief minister-PL-ERG blood donation.N ke-l-e do-PAST-N On this occasion the Chief Minister donated blood.' b. yA prasangl mukhyAmantryAn-n-kaDun raktadAn this occasion chief minister-PL-by blood donation.N kar-NyAt A-l-e do-PTCPL come-PAST-N On this occasion the Chief Minister donated blood.' Pandharipande cites the example in 61 and argues that the agentless passive construction in Marathi expresses a social convention—thereby prescribing a particular mode of behavior. Such expressions convey the highest degree of politeness, compared with other expressions that convey a prescriptive meaning (Pandharipande 1981:12&-29). (61) ApfyA shikshakA mruddha asa bolla jA-t nAhl our teacher against like speaking go-PRES not (a) 'It is not talked like this against one's own teacher.' (b) 'You should not talk like this against your own teacher.' Pandharipande provides the glosses in the example. Gloss (a) expresses a covert agent expression and—according to our analysis—is a covert indefinite agent passive. The gloss in (b) is incorrect in our opinion, and should rather be One does not speak against his own teacher like this'. Under this interpretation, it is clearly a polite indirect request. This is in line with the fact that indirect reference to the agent is a hallmark of polite expressions. On our analysis, a covert indefinite agent passive underlies the (a) interpretation of 61, while an honorific one underlies the interpretation discussed above. This example represents a case of overlap between the passive and the honorific construction. Moreover, Pandharipande's treatment is incorrect in that covert indefinite agent passives do not exclusively convey a prescriptive meaning (cf. 20 and 21). The honorific construction is an active construction, while the passive construction is not. The passive and the honorific are independent constructions and should be treated as such. They share a morphosyntactic similarity and are correlated through the pragmatic notion of agent defocusing.

170 · Prashant Pardeshi

m 5. Diachronie Development and Synchronic Distribution The correlations discussed in this paper are not isolated phenomena peculiar to Marathi, but are widely attested in various unrelated languages such as Japanese, Spanish, and Telugu. Moreover, the correlations are not accidental—both language-internally as well as cross-linguistically—but are rather systematic. Shibatani (1997) rightly points out that in order to understand the patterns of synchronic polysemy, we must know something about the historical development of the constructions in question. To substantiate this claim he discusses the diachronic aspects of voice constructions, and points out that historically middle voice forms are the major source of passive constructions. In Classical Greek the middle voice category used to express reflexive, reciprocal, spontaneous and passive meanings, while in Spanish, reflexive constructions have given rise to other constructions like the spontaneous and the passive. The evolutionary path for the development of the Spanish passive is then Reflexive > Spontaneous > Passive. The Japanese passive also developed from a spontaneous construction. Shibatani (1998) proposes the 'principle of maximization of contrast' as the driving force behind these diachronic changes. This principle motivates a language to develop voice constructions so that a meaning contrast is maximized. Although the spontaneous construction seems to be the major source of passive constructions, there may be other sources too (cf. Haspelmath 1990). It Marathi the passive, the spontaneous, the potential and the honorific constructions all exhibit synchronic polysemy in their morphosyntactic marking, namely, the main verb in the perfective, followed by the auxiliary GO. In order to offer an explanation for this synchronic polysemy it is necessary to probe the historical development of these constructions. A diachronic account is beyond the scope of the present study, however, and must be left for future research. 6. Summary and Conclusions In the foregoing sections we have addressed a language-specific issue, namely, the treatment of the Marathi passive, pointing out that currently available accounts are inadequate. The alternative analysis presented here throws light on issues such as the distribution of GO and COME passives, and the correlations of the Marathi passive with the spontaneous, potential and honorific constructions. Previous analyses failed to recognize these corelations, owing to their discrete approach and narrow perspective. Moreover, formal views of grammar (transformational, relational), have ignored the spontaneous, potential and honorific constructions from their scope of analysis, and have thus failed to account for the synchronic polysemy exhibited by the constructions in question.

The Passive and Related Constructions in Marathi m 171

To conclude, the pragmatic notion of agent defocusing holds the key to unraveling the mystery of the synchronic relationship of the passive construction with other constructions, and thus provides strong support for Shibatani's claim that various constructions can be related not simply morphosyntactically or semantically, but also in terms of common pragmatic function. • NOTES 1. In citing examples from other studies, transliteration and glosses have been slightly modified to maintain stylistic consistency. 2. Cf. Section 2.2 for a detailed discussion of the passive and the ergative construction. 3. I would like to thank Ms. Vaishali Vaidya for pointing out this subtle contextual interpretation.

• REFERENCES Berntsen, M. and J. Nimbkar. 1975. A Marathi reference grammar. Philadelphia: University of Pennsylvania. Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Damle, M. 1911. (Reprinted in 1970) Shastriya Marathi Vyakaran [in Marathi: A Scientific Grammar of Marathi]. Pune: Deshmukh and Company. Givon, T. 1979. On understanding grammar. New York: Academic Press. Haspelmath, M. 1990. The grammaticalization of passive morphology. Studies in Languages 14. 25-72. Joshi, R. 1900.X comprehensive Marathi grammar. Pune: Arya-Bhushan Press. Kher, A. 1899./i higher Marathi grammar. Bombay: Radhabai Atmaram Sagoon. Kher, B. 1990. Hiroshima. Pune: Mehta Prakashan. Kuroda, S-Y. 1979. On Japanese passives. Muraki(eds.) Explorations in linguistics: Papers in honor ofKazuko Inoue, ed. by G. Bedell, E. Kobayashi, and M. Muraki, 305-47. Tokyo: Kenkyusha. Langacker, R. and P. Munro. 1975. Passives and their meaning. Language 51.789-830. Pandharipande, R. 1981. Syntax and Semantics of the passive construction in selected SouthAsian languages. Illinois: University of Illinois Ph.D. dissertation. Perlmutter, D. and P. Postal. 1977. Toward a universal characterization of passivization. Berkeley Linguistics Society 3. 394-417. Rosen, C. and K. Wali. 1989. Twin passives and multistratalism in Marathi. Natural Language and Linguistic Theory 7.1-50. Shibatani, M. 1985. Passive and related constructions: A prototype analysis. Language 61.821-48. . 1997. Gengo no kinou to kouzou to ruikei [in Japanese: Linguistic function, structure and typology]. Gengokenkyu (Journal of the Linguistic Society of Japan) 112.1-31. . 1998. Voice parameters. Kobe Papers in Linguistics 1.93-111. Kobe: Department of Linguistics, Faculty of Letters, Kobe University. Siewierska, A. 1984. The passive: A comparative linguistic analysis. London: Croom Helm. Tarkhadkar, D.P. 1836. Maharashtra bhashece vyakaran [in Marathi: A grammar of the language of Maharashtra). Bombay: Education Society.

Language-teaching and World View in Urdu Medium Schools in Pakistan ι ΤARIQ RAHMAN ι This article seeks to raise a number of questions. Some of them are: what ideological biases are imparted through language-teaching texts? With what aim are they imparted? Do they tend to create people who will support peace and justice in Pakistan and, more broadly, in South Asia? Not all of these questions have been adequately answered and the last can only be answered by similar studies from other South Asian countries.

• 1. Introduction The ordinary state-run school teaches most subjects in Urdu in most parts of Pakistan. And even when it does teach in Sindhi or Pashto, Urdu is taught as an additional language. As such, the world view of the products of state schools is more influenced by Urdu than by any other language in the country taken as a whole. The state-schools covered in this study are meant for the education of the poorer sections of society and are fully controlled by the state unlike the elitist English-medium schools where state control is minimal. State control manifests itself, apart from other things, in the way state sponsored ideology is disseminated through textbooks in these schools. The purpose of this study is to investigate how the Pakistani state uses the language-teaching texts in Urdu medium schools to disseminate and reinforce the ideologies of nationalism and militarism. A number of other ideologies, like assumed male superiority inherent in the texts, have been touched upon in passing but are not the focus of this study.

174 · Tariq Rahman

• 2. Review of Literature The main ideology-burdened texts, or propagandist texts, are historical, political, and sociological in nature. The subject of Pakistan studies contains elements of all three and is meant for disseminating state-supported ideologies. Such a use of history is not new. All official historians were made to conceal those aspects of the truth which would annoy the powers that be. However, the power of the modern state is much more than the pre-modern one. This power is derived not only through a much more powerful coercive apparatus than before but also by the fact that the modern state can reach more people through the media and through school textbooks than its predecessors. In Pakistan, Islam was emphasized as the cornerstone of the Pakistani identity from the beginning. Hence education, at least in state vernacular-medium schools, included an Islamic component in the humanities texts from the beginning. However, in Zia ul Haq's regime (1977-88) Islamization of education became more imperative and thorough. In a description of the rewriting of history, Pervez Hoodbhoy and A.H. Nayyar make the point that the concept of the 'ideology of Pakistan', a phrase used by the Jamat-i-Islami in its manifesto in 1951, became the focal point of ideology. Part of this ideology was that Pakistan was made for Islam (not just for Muslims but in order to create an Islamic state and society); M.A. Jinnah, the founder of Pakistan, was not secular in his political views but really wanted a religious state; and the essence of the Pakistani identity was religion and not ethnic nationalism. The revised textbooks were even less liberal than those of previous regimes with Islam (of a ritualistic type), chauvinism and militarism as their major components. Hoodbhoy and Nayyar sum up the changes as follows: ... in Pakistan, because of the adoption of an exclusivist national ideology, there are no constraints on the free expression of communal hatred. Thus, the Hindu is portrayed as monolithically cunning and treacherous, obsessively seeking to settle old scores with his erstwhile masters. This Hindu is responsible for the breakup of Pakistan (Hoodbhoy and Nayyar 1985:175). A number of other researchers have carried out even more detailed and indepth studies of school textbooks with a view to determining what ideologies they are based upon. K.K. Aziz, the famous Pakistani historian, studied 66 textbooks on social studies, Pakistan studies and history in use in the schools and colleges from class 1 to the BA level. Among other things, he points out that these textbooks glorify wars and create hatred for India (Aziz 1993:19293). The 'bitter fruit' of all this propaganda is summed up by K.K. Aziz as follows: ... the textbooks are training and bringing up the students in ignorance, bias and false logic. Ignorance and bias travel together because one reinforces

Language-teaching and World View in Urdu Medium Schools in Pakistan · 175

and encourages the other. Through them the textbooks elevate the prejudices of the society into a set of moral absolutes (ibid.: 243). Rubina Saigol, in her study of social studies textbooks from a feminist perspective, agrees with the previous researchers that Pakistani textbooks create myths which help the state maintain a high level of militarization and aggressive nationalism. She further points out that the categories of thought which are reinforced in the textbooks help to support male dominance. Women and working class people are relegated to the same level since they are expected to perform manual, concrete, mechanical, and lower order tasks which do not involve original thinking (Saigol 1995). Sibt-e-Hasan, in his study of school texts, also points out that they create hatred for Hindus, blind and fanatical nationalism and the idea that women are only fit for non-intellectual, socially less prestigious, work (Hasan n.d).

• 3. World View and Language The social studies texts, as we have just seen, are the major ideologydisseminating texts. However, language texts too serve the same purpose. They do it in two ways: first through a choice of emotive terms and second by presenting ideology-laden items—poems, prose pieces, letters, conversations, exercises—in language-teaching textbooks. The first strategy, that of using value-laden diction, is an integral part of language even if it is not consciously being manipulated to reinforce certain values. Most adjectives concerning human action are valuational. Thus, in most cultures, being 'intelligent', 'fast', 'enterprising', 'bold', 'confident', 'charismatic', 'beautiful', 'rich', and 'respected' are terms of positive valuation. But it can be argued that they all come from subjective assumptions. One of these assumptions, however unconscious, is that power and its manifestation, dominance or rank-seeking, are justified. The qualities of having better cognitive abilities than other people; or of possessing greater courage; or of having other distinguishing characteristics finally make a person stand out and become more powerful and potentially more capable of dominating other human beings. Beauty and riches are also distinguishing features which everyone does not possess. Prestige, respect or 'recognition'—as Francis Fukuyama calls it (1992:146)—are consequences of being distinguished; of being better than others. But, whereas all these qualities are in the interest of those who can be distinguished, they are against the interest of those who are condemned never to be distinguished. All those who are 'stupid', 'slow', 'cowardly', 'diffident', 'dull', 'ugly', and 'poor' should logically regard all the qualities mentioned as negative attributes—as weapons with which a minority is endowed and which, very unfairly, gives it power over others. The overall framework of power, however, is so universally accepted that everyone, including those excluded from its realm, accept it as

176 · Tariq Rahman

being positive. Thus, our agreement to use words at all means that we accept a certain view of reality, a certain construction of normalcy, which is powerladen in the sense that it is not in everybody's interest. But, apart from such ideological subtleties of language use, Pakistani textbooks and official media use deliberately ideology-laden and emotive words. Among these, those in the first cluster revolve around the concept of martyrdom (shahadat) in a holy war (jihad). Another cluster of words concerns politics. Secularism has been translated as la diniyat (literally speaking "not having a religion' or "lack of faith'). This makes the reader of Urdu feel that those who support secular politics are atheists or, at least, not good Muslims. Democracy has been used by everyone but it has meant different things. Ayub Khan's 'basic democracy9 was control over electoral colleges while Zia ul Haq's Islamic democracy was a camouflage for his own rule. Similarly, all welfare programmes—the socialism of Bhutto and the welfare state of Benazir Bhutto and Nawaz Sharif—have been named Islamic to appease the clergy which would otherwise condemn them as being leftist philosophies. Other clusters of words refer to social and cultural aspects. The male dominating values of our culture are reflected in the way womens' sexuality is stigmatized while mens' is controlled at the expressed, public level but actually flaunted at a sub-level of male interaction. Thus, while women cannot confess to having male friends, let alone lovers, in all but highly Westernized social circles, men boast about their conquests to their friends. While it is a disgrace for a family if a woman has sex outside marriage, it is much less of a disgrace for a man. Words pertaining to honor—izzat, asmat, ghairat, sharm, haia—all refer much more saliently, much more seriously, to female waywardness than to male. Indeed, in most areas of Pakistan men's honor lies so much in the control of female sexuality that they kill their women for it. From this notion of sexuality comes the strong imperative of hiding away the female, seen primarily as a sexual object, from other males. Hence words like "the family', 'androone-khana', 'ghar wale' (both meaning those who live inside the house), bacche (children) are used for the wife in middle class and working class families. In Pakistani languages female cousins are called sisters and even in Pakistani English the term cousin sister is used for them (Rahman 1990:72). Literary and language-teaching texts use these clusters in varying degrees. Thus, in the vocabulary itself they reinforce a world view contingent upon male-dominating, sexuality-denying, and aggression-validating values in the social sphere; religious and nationalistic values in the political sphere and a definite bias towards the sacralization of war and the military in the sphere of foreign policy. These values are also reinforced in English medium schools but, since pupils of Urdu medium schools are less exposed to liberal-humanist values in English fiction, foreign TV. programs and at home, they are more exposed to them than their English-medium counterparts.

Language-teaching and World View in Urdu Medium Schools in Pakistan · 177

• 4. Methodology Just how frequently language-teaching textbooks expose students to ideological messages is not easy to determine. A rough estimate can be obtained by counting the number of ideological items—poems, prose pieces, exercises— in every language-teaching textbook used in the Urdu-medium schools. This assumes that the more children come in contact with ideological lessons in the classroom, or in the course of their studies, the more are they likely to be influenced by the ideologies they propagate. There are, of course, other influences on children and much depends on the teacher, the atmosphere at home, the locality and the peer group to say nothing of the child's own temperament. However, other factors being equal, the assumption upon which the following study is based is that the greater the exposure to ideological items the greater the chances of a students' acceptance of them. For this purpose the percentage of ideological lessons in textbooks has been determined. Such an exercise has never been undertaken by researchers as far as language-teaching school texts are concerned. However, Sibt-e-Hasan, a Pakistani researcher, carried out an analysis in Urdu on these lines of Urdu textbooks from class 1 to 5. He divided lessons into those on nationalism and religious or moral indoctrination. He then subtracted the number of these ideological lessons from the total number of lessons in the books (Hasan n.d). Unfortunately, he neither took out percentages nor did he compare the percentage of ideological lessons in Urdu textbooks with textbooks in other languages or subjects (such as Pakistan studies or history). In this study too the number of ideological lessons out of the total number of lessons in language-teaching textbooks has been determined in order to understand how propagandist school textbooks are. For the purpose of this exercise, ideological texts were defined as texts pertaining directly or indirectly to Islam, nationalism, and the military. In the first category are texts pertaining to Islam, Muslim civilization, and cultural achievements and eminent personalities from the history of Islam. In the second category are texts pertaining to Pakistani nationalism, Pakistani identity, the movement to create Pakistan, eminent leaders of Pakistan and so on. In the third category are texts pertaining to the glorification of war, the armed forces of Pakistan or military heroes. In many cases these themes overlap but the text is classified under the dominant theme. The percentage of the ideological lessons is calculated on the basis of the total number of lessons in the textbook. These percentages are given in the boxes given later in this study.

• 5. Ideology and Language-teaching Texts The state's major objectives—creating nationalism and support for the military—are attained by repeating a few basic messages in all the books. First, the

178 · Tariq Rahman

non-Muslim part of Pakistan is ignored. Second, the borrowing from Hindu culture is either ignored or condemned. Third, the Pakistan movement is portrayed mostly in terms of the perfidy of Hindus and the British and the righteousness of the Muslims. After the partition, in which Hindus are reported to have massacred Muslims while Muslims are not shown to have treated the Hindus in the same manner, India is portrayed as the enemy which is waiting to dismember Pakistan. The separation of Bangladesh in 1971 is portrayed as proof of this Indian policy rather than the result of the domination of the West Pakistanis over East Bengal. Above all, the 1948, 1965, and 1971 wars are blamed entirely on India and Pakistan is shown to have won the 1965 war. The armed forces are not only glorified but treated as if they were sacrosanct and above criticism. All eminent personalities associated with the Pakistan movement, especially M.A. Jinnah and Iqbal, are presented as orthodox Muslims and any aspect of their thoughts or behavior which does not conform to this image is suppressed. Indeed, the overall effect of the ideological lessons is to make Islam reinforce and legitimize both Pakistani nationalism and militarization. Thus, the state uses the emotive power of religion, patriotism, and romanticized history to create a Pakistani identity which supercedes kinship, regional or ethnic identities. By making India the ever-threatening 'other' it also uses all these sentiments as well as fear to support a large military, occasional adventurism across the border and nuclear weapons. As mentioned earlier, the ideology-enforcing items in Pakistani textbooks pertain to Pakistani nationalism and the military or war. The main impression which is conveyed is that all three components belong to a unified, sacred, religious tradition. Even poets of Punjabi, Pashto, and Sindhi, whose sufi Islam was quite different from the official Islam endorsed by the textbooks, are put into the same ostensibly orthodox tradition. This whole tradition is then dovetailed into Pakistani nationalism. The demand for a separate nation is not made to appear as the product of the desire to escape possible (or perceived) Hindu domination for economic and political reasons but an effort to establish a country for the free practice of Islam. Since Jinnah and Iqbal are also converted into upholders of these ideals, Islam supports Pakistani nationalism. The third component of the ideology, the military, is also sacralized because it is contingent upon the first two. The nation, which is a fort of Islam, needs soldiers who are nationalists in as much as they are fighters in the path of Islam. With such a meshing of the secular and religious ideologies, the state ensures that the Pakistani citizen is not seduced by ethnic nationalism which bases itself on linguistic or cultural identities. It also ensures that support of the army and war mongering is seen as support of both the nation and Islam. In short, dissent can be equated with both treason and apostasy. The percentages in Box 1 pertain to the textbooks of Urdu, English, Pashto, Sindhi, Arabic, and Persian used in the schools of Pakistan. The textbooks in

Language-teaching and World View in Urdu Medium Schools in Pakistan · 179

J

'S J2 §

z .2

&

i «

zz

zz oo U.

é

es £

ι

o

§ 53

*C

.1

£ 3

•8 S •o « Λ (g 7 «s ^ ΌΟΟ»-ΙΓΟΟΟΟΝΓ^ΟΟ

^ Γ θ · ^ · Γ η ^ · » ο τ ί · -f, r^> m

'2 11 * 7. o .a o D*

««-l

r*

—·

|i|| S l !'*>

iiiliS 111 i · 5 o ^ ««H o Sfc*-g . υ S-g sZ J8· ω

UUUU

CJU

180 · Tariq Rahman

Sindhi are used in the schools of rural Sindh whereas those in Pashto are used in some schools of the Pashto-speaking parts of the North West Frontier Province (NWFP). The figures for the percentages of ideological items have been arrived at by the procedure described in Section 3. From the figures in Box 1 one can create a hierarchy of language-wise ideological content in Pakistani textbooks. This is given in Box 2 below. Box 2: Ideological Contents of Language Textbooks Expressed as Percentages of Total Items Language Arabic Urdu Pashto Persian Sindhi English

Content (in percentages)

71 40 43 31 29 8

The role of Arabic in this context is understandable. It was introduced in Zia ul Haq's period in government schools as part of the Islamization process. But, whereas Arabic textbooks mostly contain Islamic items, those in Urdu contain items on nationalism and the military also. Moreover, Arabic is already associated with Islam but Urdu is taken as a language. Also, while Arabic is only compulsory in middle school, Urdu is read throughout the school years and even in the intermediate classes (i.e. 11th and 12th). Thus, it is Urdu which is the primary ideology-carrying language, not Arabic. Persian is not compulsory in Punjab and Sindh and is, even in the NWFP and Balochistan, only a minor language which students study for three years. Thus, Persian too has little influence on world view formation. Pashto and Sindhi are taught only in some parts of the NWFP and rural Sindh. They are so ideologically burdened, one would surmise, because these areas have had ethnic, language-based movements which the state counters in schools through education. English, always associated with modernity and liberal values, is the least ideology-reinforcing language. However, the percentage of ideological items is quite high in books of classes 7 and 8. Why the overall percentages are less is because in Sindh the books for class 1 to 7 were written by ELT specialists such as Zakia Sarwar, Abbas Hussain, Ambreena Kazi, Kaleem Raza Khan, etc. All these writers are liberal academics and the textbooks they have produced are completely devoid of state-sponsored ideological content. One anomaly in all the English textbooks is that there are two systems of teaching English. According to the older system English began to be taught from class 6. According to the new one English begins in class 1. It was in 1989 that the first

Language-teaching and World View in Urdu Medium Schools in Pakistan · 181

PPP government led by Benazir Bhutto decided to introduce English as an additional subject from class 1. The decision was implemented in selected schools in Sindh and the NWFP in 1990. The other provinces too accepted the proposal in principle. In the NWFP, the hurriedly prepared textbooks were used only for one year and then had to be withdrawn. Later on, the National English Language Institute (NELI), established in 1987, submitted a new curriculum for English Language Teaching (ELT) for classes 1-12 (NELI 1992). However, as the NELI report had predicted, the experiment was not successful. Indeed, NELI itself was abolished and even now all schools do not teach English from class 1. The new system, therefore, runs parallel with the old one in Pakistan. In general, however, English still begins from class 6. Thus, not all children are exposed to very low ideological messages in early childhood. Those who take up English from class 6 are exposed to higher (though not as high as in the case of Urdu) ideological doses. The focus of the above analysis has been the ideology of the state rather than that of the society. The distinction between the two is merely the use to which ideas are put. Islam, nationalism and militarism create security for the state—or what the state regards as security—by motivating citizens against ethnic nationalism and India. Societal views, such as are implicit in the privileging of men and educated or dominant people, help to create a view of common sense and normalcy in which women are simply assumed to be inferior and hence to be dominated as are the working classes. A number of studies have analyzed images in textbooks which help us understand how these societal views are reinforced.1 Muhammad Anwar, for instance, analyzed 105 textbooks of the Punjab Textbook Board. His results are that 'the frequency of the appearance of males in 81 per cent while females appear only 19 per cent of the time. One reason of this could be that 78 per cent of the authors were male' (Anwar 1982: 12) and in our male-dominated world public interaction is predominantly among men. Another could be that womens' work is simply ignored. They do work in the fields, graze animals, fetch water and even carry bricks as laborers. However, as women activists often point out, even this substantial contribution to the economy is ignored (SΔHE 1997:55). Perhaps the construction of social reality in our male-dominated society is such that all this labor does not register itself as reality in the mind. One reason for this could be that women are supposed to be hidden and men are not used to considering them partners in public interaction. In the analysis by Anwar: Service occupations [cooking, nursing, washing dishes, etc.] seem to be associated more with females than with males, and such an association is more prominent in Urdu than in English books (Anwar 1982: 38). Working class characters, like women, are also less in number. The following table illustrates this:

182 · Tanq Rahman Social Class

Frequency of Appearance

Percent

High Middle Low

1552 1026 329

53.4 35.3 11.3

Total

2907

100

Source: Anwar 1982:15.

It may be said that this is only realistic. Any attempts at glorifying working class characters would appear as false and sentimental. What is possible, without inviting cynicism, is to endow marginalized characters with potential qualities which have not been realised as yet. Unfortunately, the qualities of 'learned', 'rational', 'progressive', and 'genius' 'have been minimally portrayed in female characters' (Anwar 1982:58) reinforcing the myth that only men can have these qualities. Similarly, qualities of the mind and good manners are not associated with working class people. In short, as some intellectual theorists would have it, women and the 'lower classes' correspond to the body; middle and upper class men to the mind. This binary classification creates, and reinforces, a view of reality which is unfair to the working classes and women to begin with. In this context, the society for the Advancement of Education (SΔHE) pointed out that women are ignored or negatively portrayed (this is true for the media too [PWI1982]).2 Those who are in prestigious and traditionally male professions—such as physicians, engineers, architects, etc.—are ignored. In short, the gender bias against women is evident and all the messages we get, whether from the formal lesson, fiction or the media, reinforces it (SΔHE 1997: 56-68).

• 6. Efforts to Remove Biases As far as women are concerned, a change is being made—or, at least, proposed—by some non-government organizations (NGOs). SΔHE, for instance, conducted workshops of school teachers to sensitize them to gender bias. In a book written specifically for teachers of primary and secondary schools in simple Urdu and aptly entitled Main Jaag Uthi (I woke up) (SΔHE 1997), a number of model lessons about women and the necessity of educating and empowering them have been given. The Simorgh Women's Resource and Publication Centre is also engaged in developing materials for schools with human and womens' rights as the focus. Other NGOs, such as Aurat Foundation, also support the production of teaching material without gender bias. Rubina Saigol, Fariha Zafar, Asma Ajmal, Shahla Zia, Asma Jahangir, and Neelam Hussain— to name only a few activists—are busy contributing in various ways to the development of such teaching material. It is difficult to know, however, how far the state schools will accept such material and whether it will bring about a

Language-teaching and World View in Urdu Medium Schools in Pakistan · 183

real change in male-dominant values. But whether it does or not, those who favor changes in the interest of women welcome such endeavors.

• 7. Conclusion The Urdu medium schools are meant to indoctrinate the masses into becoming religious, nationalistic, and militaristic. Language texts help these ideological biases sink in the impressionistic minds of students, especially middle school students. The school, therefore, endeavors to create a world view which rejects ethnic nationalism for Pakistani nationalism irrespective of the genuine rights of the ethnic nationalities of the country in the process. It also whips up hatred against India and, therefore, supports the maintenance of a very high expenditure on defence. However, since school texts are not the only influences on children nor does everyone become brainwashed, there are many products of the vernacular schools who are liberal and who wish for peace rather than war. Since the future of South Asia depends upon peace it would be interesting to know what kind of socialization language-teaching texts and practices exist in India and other South Asian countries? If this socialization is not conducive to peace, as far as the texts are concerned, can these texts be changed? If so, who would change them? These questions are not only of academic interest. They are questions of the survival of millions of human beings in South Asia. ι ANNEXURE Language-wise Average Ideological Content in Textbooks Expressed in Percentages of the Total Number of Lessons

Urdu

English

Sindhi

Pashto

Class 1

22

19

22

Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Class 8 Class 9 Class 10 Percentage of Average Content (Ideological)

31 38 47 47 49 59 50 30 30 40

Zero—or not taught (NT) 2/NT 5/NT 7/NT Nil 4 7 15 28 7 8

36 36 33 26 34 36 26 21 21 29

44 44 66 37 46 50 36 NT NT 43

Arabic

Persian

Not taught Not taught (NT) (NT)

NT NT NT NT NT 70 81 50 80 71

NT NT NT NT 15 47 38 28 28 31

Note: English is taught at the discretion of the headmaster in urban schools. It is not taught in rural schools in Islamabad and Punjab. Persian is compulsory in Balochistan (6th-10th); in the NWFP it is compulsory (6th-8th).

184 · Tariq Rahman • NOTES 1. Two studies show that in Urdu fiction, both for adults (Parween 1984) and children (Bano 1985), power-oriented values, themes and attitudes are more preponderant than achievement or affiliation oriented ones. Being so focused on power, macho and militaristic values are easily accepted by students. Indeed, Meher Bano also shows that the NWFP textbooks, like fiction, also show the highest scores for power followed by achievement followed by affiliation (Bano 1985: Table 19). 2. A recent journalistic article (Ahmar 1997) shows how Urdu popular fiction reinforces traditional values showing 'good* women as demure and docile drudges.

• REFERENCES Ahmar, lasneem. 1997. Pulp Fiction? Herald (December). 111-14. Anwar, Muhammad. 1982. Images of male and female roles in school and colleges textbooks Islamabad: Womens' Division, Government of Pakistan. Aziz, K.K. 1993. The murder of history in Pakistan. Lahore: Vanguard Press. Bano, Mehr. 1985. An analysis of themes of childrens' readers in Urdu in terms of achievement, affiliation and power motives in Pakistan. Vols. 1,2 and 3. Peshawar: Dept. of Psychology, University of Peshawar, Ph.D Thesis, Vols 1,2 and 3. Fukuyama, Francis. 1992. The end of history and the last man. New York: The Free Press. Hasan, Sibt-e. n.d. Darsi Kutab: Zaria-e-Taleem Ya Zaria-e- Jabr Lahore: Awaz Foundation Barae Taleem. Hoodbhoy, Pervez and Nayyar, A.H. 1985. Rewriting the history of Pakistan. Islam, politics and the state ed. by Asghar Khan, 164-77. London: Zed Press. NELL 1992. Curriculum for English language teaching classes I-V. Draft Manuscript. Islamabad: National English Language Institute, Ministry of Education, Government of Pakistan, ms. Parween, Shahida. 1984. A thematic analysis of popular Urdu short stories (for adults) in terms of certain social motives. Peshawar: Dept. of Psychology, University of Peshawar, M. Phil thesis. PWI1982. Portrayal of women in communication media. Islamabad: Pakistan Womens' Institute. Rahman, lariq. 1990. Pakistani English: The linguistic description of a non-native variety of English. Islamabad: National Institute of Pakistan Studies, Quaid-i-Azam University. Saigol, Rubina. 1995. Knowledge and identity: Articulation of gender in educational discourse in Pakistan. Lahore: ASR Publications. SΔHE 1997. Main Jaag Uthi [in Urdu: I Woke up]. Lahore: Society for the Advancement of Education.

Pragmatic Explanations for Expressing Obligations of the Agent Referred to in Hindi* ι GHANSHYAM SHARMA ι Hindi possesses three constructions of similar syntactic nature—with an agent referred to, an infinitival verb form and one of three modal markers—to express three kinds of deontic modality. These overtly similar constructions are, however, employed by the speaker to achieve different pragmatic goals. Their individual meanings derive from the various pragmatic strategies involved and from the type of mutual knowledge and beliefs shared by speaker and addressee. This paper is an attempt to describe these strategies in a formal way.

• 1. Introduction To express obligations on the agent referred to (hereafter Agref), a Hindi speaker makes use of, among others, the following three syntactic constructions employing three different 'modals' (or verbal markers expressing modality), namely, 1. a. Agref-ko + Infinitive 4- caahie b. Agref-ko + Infinitive + paRnaa c. Agref-ko + Infinitive -I- honaa Most Hindi grammars try to explain these modal constructions in terms of their corresponding constructions in English. Such studies, though based on * I am grateful to Richard Geiger, Elena Bashir and Michael Shapiro for useful comments on a previous version of this paper. All errors, however, are solely my own.

186 · Ghanshyam Sharma

very subjective approaches, are a useful tool for interpreting these constructions for language-teaching purposes, but they fail to point out the different pragmatic strategies the speaker employs in selecting one instead of another. They tend to classify the three constructions according to the degree of 'strength' they are supposed to carry in a conversational setting. But, as shall be shown later, that is not the subjective parameter which a speaker employs in order to weigh them before using them. Rather, they are used by the speaker to achieve certain pragmatic goals. At the surface-structure level, the three constructions under discussion are syntactically parallel, inasmuch as all three demand an agent1 NP in a dative construction signaled by the postposition ko and a VP which includes a verb in the infinitive plus one of the three verbal markers. Differences can be found among them with respect to the verbal agreement they show with the object of the verb. The form caahie shows no agreement with the verbal (i.e. infinitival) object.2 The other forms, paRnaa and honaa, always agree (with that infinitival object) in number and gender, however. Another important difference between caahie, on the one hand, and paRnaa and honaa, on the other, can be noticed in their uses in different tenses and aspects. The modal caahie does not inflect according to tense and aspect, while paRnaa and honaa can be found in different tenses and aspects. The latter differ one from the other with respect to their ability to express deontic modality in directive illocutions. The mooa\paRnaa is used exclusively in the future tense for this purpose, as its use in the present or any other tense does not express deontic modality. Honaa, on the other hand, can be used in the present as well as the future tense. As similar as these forms are, however, their respective uses in differing contexts of discourse depend upon the pragmatic goals of the speaker as well as the illocutions selected to achieve them. The present paper is an attempt to discover and formalize precisely the situations in which the use of these constructions is possible.

• 2. The Three Modals and Their Verbal Agreement As far as their syntactic structures are concerned, the modals show little variation as a group in that they all have the same word order and restrict Agref (the logical subject) to the dative. The infinitive of the verb, which expresses the action to be carried out by Agref, may in some cases show agreement with its object and in other cases it may not. Let us look at the three modals separately now. • 2.1. Caahie Caahie is an aspect-less and tense-less verbal marker except in the past tense where it takes the auxiliary thaa, as in:

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 187 2. John ko vahaaN jaanaa caahie thaa John-Dat there go should was 'John should have gone there.' Otherwise, to repeat, it is not inflected. The meaning of the constructions employing this modal verbal marker can be understood in general in the following way: '(Because I feel it is in Agref's interest to do so,) Agref should/ ought to VP.' 3. Agref + ko (i.e., 'dative nominal')3 + Infinitive + caahie a. twnheN Peru jaanaa caahie [without a verbal object] you-Dat Peru go is advisable 'You ought to go to Peru.' b. twnheN tun kele khaane caahie [with an object-m-pl] you-Dat three bananas-m-pl eat-m-pl is advisable 'You ought to eat three bananas.' c. tumheN tun rotiyaaN khaanii caahie [with an object-f-pl] you-Dat three bread-f-pl eat-f-pl is advisable 'You ought to eat three (pieces of) bread.' As these examples show, when there is no verbal object the infinitive has the impersonal form singular masculine 3a. Except with verbs of movement as in 3a, when there is a verbal object, the infinitive shows agreement with it in gender and number 3b and 3c. Some styles of spoken Hindi, however, tend to neutralize the gender and number agreement when the situational context is informal and select the uninflected infinitival form: 3. b'. tumheN tiin

kele

khaanaa caahie [with an object-m-pl] you-Dat three bananas-m-pl eat-0 is advisable 'You ought to eat three bananas.' c'. tumheN tiin rotiyaaN khaanaa caahie [with an object-f-pl] you-Dat three bread-f-pl eat-0 is advisable 'You ought to eat three (pieces of) bread.'

• 2.2. PaRnaa PaRnaa can be used in any tense and aspect, but to express deontic modality (directive illocution) only the future tense is employed 4a. In other tenses and aspects it is used in declarative utterances (assertive illocution), for example in 4b, and therefore these uses will not be discussed in the present paper: 4. a. John ko Peru jaanaa paRegaa [directive illocution] John-Dat Peru go must-fut-m-sg 'John will have to go to Peru.'

188 · Ghanshyam Sharma

b. John ko Peru jaanaa paRlaa hai [assertive illocution] John-Dat Peru go must-pres-m-sg 'John has to go to Peru (habitually).' This modal carries information about both aspects and tenses. Its meaning can in general be summarized in the following way: '[Even though it would be unpleasant/harmful/distasteful/annoying/... for you,] you will have to VP.' Let us look at a few examples. 5. Agref + ko (i.e., dative nominal or experiencer subject) + Infinitive + paRnaa a. tumheN Peru jaanaa paRegaa you-Dat Peru go be obligatory-fut 'You will have to go to Peru.' b. tumheN tun kele khaane paReNge you-Dat three bananas-m-pl eat-m-pl be obligatory-m-pl-fut 'You will have to eat three bananas.' c. tumheN tun rotiyaaN khaanii paReNgii you-Dat three bread-f-pl eat-f-pl be obligatory-f-pl-fut 'You will have to eat three (pieces of) bread.' In this case, the agreement between the infinitive and its object is required, most likely because the modal itself here shows the agreement, which it does not do in the case ofcaahie. • 23. Honaa The constructions containing this auxiliary can be found in any tense or aspect of the verb, although the most common to express the deontic modality in Hindi are the present and future tenses. With respect to this particular modality, the future tense of the auxiliary indicates actions to be carried out by Agref at a time after the time of speaking, and in the same way the present tense may also indicate a future action. Constructions with honaa are semantically and pragmatically intermediate between those ofcaahie and paRnaa in the sense that they may sometimes express the speaker's advice and sometimes the speaker's judgment about obstacles Agref might encounter while carrying out the action. As we shall see later, the variation in meaning in these constructions depends very much on the mutual knowledge and beliefs of speaker and addressee. The speaker's intended meaning in this construction is usually as follows: 6. Speaker's intended meaning using honaa a. '[On the basis of the information I have, I feel I should remind you that] you have to VP.' b. '[On the basis of my authority, I emphasize that] you have to VP.'

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 189 The different uses of this structure will now be illustrated. 7. Agref + ko (i.e. Dative Nominal) + Infinitive + honaa a. tumheN Perujaanaa hai/hogaa you-Dat Peru go be supposed to-pres/fut 'You are (supposed) to go to Peru.' b. tumheN tiinkele khaane haiN/hoNge you-Dat three banana-m-pl eat-m-pl be supposed to-pl-pres/mpl-fut 'You are (supposed) to eat three bananas/ c. tumheN tiinrotiyaaN khaanii haiN/hoNgii you-Dat three bread-f-pl eat-f-pl be supposed to-pl-pres/f-pl-fut 'You are (supposed) to eat three (pieces of) bread.' It should be noted here that in 7b and 7c only the future tense forms of the auxiliary show gender agreement between the auxiliary and the verbal object.

• 3. The Illocutionary Point of Deontic Modality In almost all directive illocutions with which a speaker expresses her desire or wish for Agref to do /, deontic modality is employed. The person(s) by whom such a desired action is to be carried out can be any of the following: second person(s), third person(s), or, in monologues, even first person(s). The deontic modality can also be used in the case of a natural phenomenon to express the desires and wishes of the speaker(s), in which case no action by the agent is envisaged: only a 'so-be-it' fact is hoped for. The following taxonomy will show the relationships between the presence or absence of an Agref in a conversational setting, Agref's reference, and the strength of the deontic modality involved in the various cases: 8. Conversational settings Speaker4 Addressee(s) Situation 1: Situation 2: Situation 3: Situation 4: Situation 5: Situation 6: Situation 7:

I(we) I(we) I(we) I(we) I(we) I(we) I(we)

you you you I I you I

[sg./pl.] [sg./pl.] [sg./pl.j [sg./pl.]

Agref

Strength of deontic modality you [sg./pl.] (strongest) t he/she/they I/we I/we he/she/they „ [None] (weakest) [None]

In situation 1 the speaker wants her addressee(s) to do /because the addressee in this case is also the intended agent of/. In situation 2 it is the third person(s)

190 · Ghanshyam Sharma

who is/are supposed to carry out the action, while in situations 3 and 4 the first person(s) is/are supposed to carry it out. Situations 4,5 and 7 are monologues (in which the speaker is also the addressee) while their respective Agrefs differ in having in 4 a self-reference (possibly including others), in 5 a thirdperson Agref and in 7 no Agref at all. Situation 6 has an addressee other than the speaker, but also has no Agref. For example, I may utter (to myself or to someone else) a sentence like "It must rain tomorrow' or "It should be a hot day tomorrow', without there being an Agref to carry out any action. Keeping in mind the above taxonomy, a full list of possible Agrefs in Hindi can therefore be presented in the following way: 9. Situation 1 a. tujhe5 Perujaanaa caahiel paRegaa/ you-sg-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'You [sg.] ought to/(will) have to/are supposed to go to Peru.' b. tumheM Perujaanaa caahiel paRegaa/ you-pl-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'You [pi.] ought to/(will) have to/are supposed to go to Peru.' c. aapko Perujaanaa caahiel paRegaa/ you-pl-hon-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'You [sg. polite] ought to/(will) have to/are supposed to go to Peru.' d. tumlogoNko Perujaanaa caahiel paRegaa/ you-pl-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'You [pi.] ought to/(will) have to/are supposed to go to Peru.' e. aaplogoNko Perujaanaa caahiel paRegaa/ you-pl-hon-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'You [pi. polite] ought to/(will) have to/are supposed to go to Peru.' 10. Situations 2 and S a. use Perujaanaa caahiel paRegaa/ he/she-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'He/she ought to/will have to/is supposed to go to Peru.'

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 191

b. unheN Perujaanaa caahiel paRegaa/ they-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'They ought to/will have to/are supposed to go to Peru.' c. unlogoNko Perujaanaa caahiel paRegaa/ theyall-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'They all ought to/will have to/are supposed to go to Peru.' 11. Situations 3 and 4 a. mujhe Perujaanaa caahiel paRegaa/ I-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'I ought to/will have to/are supposed to go to Peru.' b. hameN Perujaanaa caahiel paRegaa/ we-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'We ought to/will have to/are supposed to go to Peru.' c. hamlogoNko Perujaanaa caahiel paRegaa/ weall-Dat Peru go is advisable/ be obligatory-fut/ hogaa/hai be supposed to-fut/pres 'We all ought to/will have to/are supposed to go to Peru.' 12. Situations 6 and 7 a. uspeRko ab

caahie [with an anti-transitive verb] that tree-Dat now fall-antitrans is expected 'That tree ought to/will have to/should fall now.' b. kal tak kaam puuraa ho jaanaa caahie/hai [with the verb 'become'] tomorrow by work complete become is advisable/is supposed to 'The work ought to/has to be complete by tomorrow.' c. kal

girjaanaa

baarish honii

caahie/hai [with the verbs'take

tomorrow rain take place is expected 'It ought to rain tomorrow.'

place'/'be']

As can be seen from the examples above, the strength of the deontic modality conveyed by the speaker decreases from situation 1 through situation 7. An

192 · Ghanshyam Sharma

important aspect of this phenomenon can also be noted in the use of the kind of verb the modal construction can take. For example, caahie can very well be employed in expressing wishes where there is no Agref. Situations 6 and 7, in fact, do not have any Agref: only the speaker's desire is conveyed. In this case, though, only an anti-transitive or a 'becomeVbe' type of verb is normally found. PaPnaa, on the other hand, can never be used in situations 6 and 7; this means that it is not normally used with an anti-transitive verb. The case of honaa, however, seems to fall in both of the areas of modal constructions which contain caahie andpaRnaa. Causative verbs do not normally occur in situation 6 and 7.

• 4. The Three Hindi Constructions Expressing Deontic Modality The three Hindi expressions which appear to occur in the same syntactic construction differ greatly in meaning. This difference in meaning can be accounted for with reference to the pragmatic settings in which they occur. • 4.1. Caahie 13. a. S desires/? (p = a proposition anchored in a conceived world tv/, different from wo, in which Agref does / at time /, ^to). b. To fulfill her desire mentioned in 13a, i.e.,/?, S utters a sentence containing 1. [V (active verb) + caahie] in situations mentioned in 9-11, or 2. [V ('to be' or an anti-transitive verb) + caahie] in situations mentioned in 12. 14. S does so under any of the following conditions: a. The speaker believes that because of all she knows Agref's doing / would be useful/helpful/beneficial/gainful/advantageous or even necessary for Agref or would be in the interest of either addressee, speaker or an absent third party. b. On the basis of all she knows, she believes that Agref s carrying out / would be a right action according to moral obligations on Agref. The pragmatic constraints discussed in 14a and 14b, respectively, can be illustrated with the following examples 15 and 16: 15. a. aapko roz duudh piinaa caahie you-hon-Dat every day milk drink is advisable 'You ought to/should drink milk every day.' a'. ΊΊaapko roz duudh piinaa caahie. haalaaNki you-hon-Dat everyday milk drink is advisable. However yah laabhdaayak nahiiN this beneficial not (is)

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 193

?? Ύονι ought to/should drink milk every day. However, it is not beneficial for you.' b. aapko turant aspataal jaanaa caahie you-hon-Dat immediately hospital go is advisable 'You ought to/should immediately go to the hospital.' b'. llaapko turant aspataal jaanaa caahie. lekin you-hon-Dat immediately hospital go is advisable. But yah upyogii/zaruurii nah N this useful/necessary not (is) ??'You ought to/should go to the hospital immediately. But this is not useful/necessary for you.' c. tumheN hindi siikhnii caahie. you-Dat Hindi learn is advisable. 'You ought to/should learn Hindi.' c'. TitumheN hindi siikhnii caahie. lekin yah upyogii nahiiN you-Dat Hindi learn advisable. But this useful not (is) ?? 'You ought to/should learn Hindi. But it is not useful.' 16. a. tumheN gariiboN kii sahaayataa karniicaahie you-Dat the poor of help do should 'You should help the poor.' a'. lltumheN gariiboN kii sahaayataa karnii caahie. lekin you-Dat the poor of help do is needed. But dharm-granth aisaa nahiiN kahte religious books this not say ??'You should help the poor. But no religious books say so.' b. aapko apne maaN-baap kii sevaa kamii caahie you-Dat your parents of service do is morally required 'You should/ought to look after your parents.' b'. llaapko apne maaN-baap kii sevaa karnii caahie. you-Dat your parents of service do is (morally) required. haalaaNki yah tumhaaraa kartavy nahiiN However this your duty not (is) ??'You should/ought to look after your parents. However, it is not your duty to do so/ In example 16 the obligation imposed by caahie upon Agref is usually based on moral judgments made by the speaker, though sometimes, as in 15, it may be the expression of an assessment of a different kind. To prove the pragmatic constraints mentioned above, it would be sufficient to see the examples in 15 and 16. The acceptability of 15a', 15b', 15c', 16a' and 16b' is pragmatically invalid in that if information is added to them with certain conjunctions—15a, 15b, 15c, 16a, 16b—utterances arise which are grammatically correct but pragmatically infelicitous.

194 · Ghanshyam Sharma

The use of the construction caahie in the (past) imperfect tense requires the past form of the auxiliary honaa, i.e., thaa, and refers to unfulfilled actions such as in: 'You should have done that.' 17. a. aapko roj duudh piinaa caahie thaa. lekin you-Dat everyday milk drink advisable was. But aapne aisaa nahiiN kiyaa you-hon-Erg such a thing not did 'You were supposed to drink/should have drunk milk every day. But you didn't do so.' b. aapko apne maaN-baapkii sevaakarnii caahie you-Dat your parents of service do morally required ihii. lekin aapne yah nahiiN kiyaa was. But you-hon-Erg that not did 'You should have looked after your parents. But you didn't do that.' This use of caahie in the imperfect tense, however, given the proper shared knowledge and beliefs between speaker and addressee, may refer to an action still to be carried out by Agref, such as '[you haven't yet done so, but] you are still advised to VP'. In such cases, though, the utterance does not contain explicit information that the action was not fulfilled. • 4.2. PaRnaa The speaker, on certain grounds, considers it necessary for Agref to do / and also thinks that carrying out the action /would be a bit annoying for Agref, and knows further that Agref at least would not, if not obliged to do so, want to do/. 18. Speaker's strategy: a. S desires/? (p = a proposition anchored in a conceived world w, different from wo, in which Agref does / at time /, **t0). b. To achieve p, S utters a sentence involving paRnaa. There may or may not be an addressee, and the reference to Agref may be either the addressee, a third party, or the speaker herself. 19. S performs 18b under one of the following conditions: a. She believes that Agref will not, under normal conditions, carry out/. b. She believes that Agref will not willingly cany out /. c. She believes that Agref will have difficulties in or will feel uneasy about carrying out /. 19 is a statement about the speaker's beliefs concerning the likelihood of Agref's carrying out an action, while 20 explains the speaker's reasons for wanting Agref to carry out the action:

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 195

20. S performs 18b because: a. She believes that Agref's doing / would be useful either to S, to Agref or to a third party. b. She believes that not carrying out / would be harmful either to S, to Agref, to a third party. 21. Examples: a. tumheN duudh acchaa nahiiM lagtaa. lekin (wnheN duudh you-Dat milk pleasing not is. But you-Dat milk piinaa paRegaa drink be obligatory-fut 'You da not like milk. But you will have to drink it.' a'. TltumheN duudh acchaa lagtaa hai. lekin tumheN duudh you-Dat milk pleasing is. But you-Dat milk piinaa paRegaa. drink be oblig.-fut ??'You like mUk. But you will have to drink it.' b. tumheN Perujaane ketie vah mazbuur kar rahaa hai. you-Dat Peru go for he compel-prog is. islie tumheN vahaaN jaanaa paRegaa. Therefore you-Dat there go be obligatory-fut 'He is compelling you to go to Peru. Therefore, you will have to go there.' b'. TitumheN Perujaane kelie koinahiiN mazbuur kar rahaa. you-Dat Peru go for nobody compel-prog (is). islie tumheN vahaaN jaanaa paRegaa. Therefore you-Dat there go be obligatory-fut ??'Nobody is compelling you to go Peru. Therefore, you will have to go there.' c. tumheN Perujaane meNbahut pareshaaniyaaN hoNgii. you-Dat Peru goingin many difficulties be-fut. phirbhii tumheN vahaaN jaanaa paRegaa Nonetheless you-Dat there go be obligatory-fut 'You will face many difficulties in going to Peru. Nonetheless, you will have to go there.' c'. lltumheN Perujaane meNbahut pareshaanii hogii. you-Dat Peru goingin many difficulties be-fut. islie tumheN vahaaN jaanaa paRegaa. Therefore you-Dat there go be obligatory-fut ??'You will have many difficulties in going to Peru. Therefore, you will have to go there.' The utterances 21a, 21b and 21c become infelicitous if further information is added to them with one of the conjunctions given in 21a', 21b' and 21c'. To

196 · Ghanshyam Sharma

express unfulfilled actions under this kind of obligation, no past tense form can be employed. Instead, the past tense oihonaa is used to indicate such an obligation. Other constructions can also be used to do so: 22. a. twnheNharroz duudh piinekii zaruurat thii You-Datevery day milk drink of necessity was 'You had to/used to have to drink milk every day/ • 4.3. Honaa The speaker wants to remind Agref about a previous intention (or a normal obligation he has) to carry out a certain action: 23. a. S desires p (p = a proposition anchored in a conceived world w,, different from w0, in which Agref does / at time // **i0). b. To fulfill her desire mentioned in 23a, i.e.,/?, S utters a sentence containing 1. [V (active verb) + honaa] in the situations 9,10 and 11 2. [V (to become or an anti-transitive verb) + honaa] in 12. 24. S does so under any of the following conditions: a. The speaker is not quite sure whether Agref still remembers his duty to carry out the action / or she is under the impression that Agref may possibly have forgotten an action already planned by himself or required by some authority, internal or external. b. On the basis of all she knows, she believes that Agref's carrying out / would be a right action resulting from moral or other kind of obligations on Agref. In 25 the different situations mentioned in 24 are exemplified based on the mutual knowledge of the speaker and the addressee: 25. Agref + ko (i.e. Dative Nominal) + Infinitive + honaa a. twnheN Perujaanaa hai. tumhaare vahaaN jaane you-Dat Peru go is supposed to. of yours there go kevicaar haiN of plans are 'You are (supposed) to go to Peru. You have plans to go there.' a', Titumhaaraa Peru jaane kevicaar nahiiN haiN.islie of yours Peru go of plans not are. Therefore tumheN vahaaN jaanaa hai you-Dat there go is supposed to ??'You have no plans to go to Peru. Therefore, you are (supposed) to go to there.'

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 197

a". lltumheN Perujaanaa hai. lekinyah zaruurii you-Dat Peru go is supposed to. But this obligatory nahiiN not (is) ??'You have to go to Peru. But there is no obligation to do so.' b. tumheN Perujaanaa hogaa. yah toy hai you-Dat Peru go be supposed to-fut This decided is 'You will have to go to Peru. It is decided.' b'. lltumheN Perujaanaa hogaa. agarna jaao you-Dat Peru go be supposed to-fut. If not go-subj to koi boat nahiiN then some problem not (is) ??'You will have to go to Peru. If you don't go, there will be no problem.' b". lltumheN Perujaanaa hogaa. lekinyah zaruurii you-Dat Peru go be supposed to-fut. But this necessary nahiiN not (is) ??'You will have to go to Peru. But this is not necessary.' As can be seen from the examples above, the construction honaa expresses Agref's plan to cany out an action (in this case the speaker simply wants to inform the addressee of his duty to do so) or else, in the case of the use of the modal marker in the future tense, an obligation imposed by the speaker. While 25a is an acceptable utterance, 25a' and 25a" are infelicitous. The same can be said of 25b and its variants.

• 5. Communicative Leveling in Mutual Beliefs of Speaker and Addressee and the Three Hindi Constructions Expressing Deontic Modality As far as the use of the modal operator expressing necessity Π and possibility Φ is concerned, we find that each of the three modals expressing obligation imposed by the speaker is on a par with the other two. The only difference is that of the strength of the deontic modality they are to carry in a normal context and the different kinds of deontic modality they are intended to communicate in a given situation derived from the level of mutual knowledge between speaker and addressee. But their negative counterparts yield different formalisms. For example, using the modal operator for necessity (the only possibility in this case), 9b will have the reading given under 26: 26. tumheM Perujaanaa caahiel paRegaal hogaalhai you-Dat Peru go is advisable/ be obligatory-fut/ be supposed to-fut/pres

198 · Ghanshyam Sharma

'You ought to/(will) have to/are supposed to go to Peru/ = Dp (i.e. 'It is necessary that p.') The negative counterparts of these modals, however, indicate that they are not similar as far as their capacity to express deontic modality is concerned: 27. a. tumheM Peru nahiiN jaanaa caahie you-Dat Peru not go is advisable 'You ought not (to)/should not go to Peru.' b. tumheM Peru nahiiN jaanaa paRegaa you-Dat Peru not go be obligatory-fut 'You will not have to go to Peru.' c. tumheM Peru nahiiN jaanaa hai you-Dat Peru not go be supposed to-pres 'You are not supposed to go to Peru.' (i.e. 'No one expects you to go to Peru.') d. tumheM Peru nahiiN jaanaa hogaa you-Dat Peru not go be supposed to-fut 'You are not supposed to go to Peru.' (i.e. 'No one will expect you to go to Peru.') Both 27a and 27c yield Q->p, i.e. it is necessary that not p, while 27b and 27d do not give that kind of formalism, since the latter are not cases of directive, but rather of assertive illocution by means of which the speaker informs Agref about an exemption from obligation. 27c, however, being ambiguous, may yield another reading, since, as has been noted before, the present tense form ofhonaa may mean either the same as caahie or aspaRnaa. Keeping in mind the dual functionality ofhonaa, 27 can be formally represented in the following way: 28. a. D-φ (i.e. 'It is necessary that not p.') b. -ι Dp (i.e. 'It is not necessary that p.') c. D-φ (i.e. 'It is necessary that not p.') c'. -i Dp (i.e. 'It is not necessary that p.') d. -i Dp (i.e. 'It is not necessary that p.') Therefore, one reading of 27c, namely 27c', cannot be considered a case of deontic modality in which Agref is obliged to carry out an action: it is simply a case of informing Agref of the situation. Most studies of mutual knowledge time and again stress the infinite nature of the tables of mutual knowledge between speaker and addressee as well as the difficulties the tables present in formally judging the nature of utterance meaning.

29. LEVEL I6

Bs · Dp, and either {B, -JCh · Dp/

h - Dp (= + -Kh · Dp; Φ -JCn · Dp)}

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 199 In other words, the speaker believes that Dp (Dp = 'It is necessary for the addressee to go to Peru.'), and, in addition, either she believes that the addressee does not know that Dp or she does not know whether the addressee knows that Dp (i.e. it is possible in this case that the addressee knows or does not know that Dp). 30. LEVEL II Bs -Kh · Dp, and either {Bs -,Kh -Bs - Dp/

-Bs -Kb -B, · Dp (= Φ -Kh -Bs - Dp; * -iKh -Bs · Dp)} Thus, the speaker believes that the addressee knows that Dp, and, in addition, either she that the addressee does not know that Dp or she does not believe that the addressee knows that she believes that Dp. So it is possible that the addressee knows or does not know that she believes that Dp. This reflexive process of speaker/addressee knowledge/beliefs is considered by many researchers of cognition and communication to be infinite. At a certain point in the process, however, the information shared by the two communicating parties becomes redundant and therefore is of no particular interest as far as the use of the three modal particles discussed in this paper is concerned. The exact point at which the redundancy becomes evident will no doubt vary from situation to situation and will require further investigation in order to be understood properly. It will therefore not be discussed at this time.7

• 6. Conclusion As we have seen above, it is the pragmatic meaning intended by the speaker with these three Hindi modal constructions that makes them differ from one another. The constructions are not in a contrastive situation because one can be used in place of another given the right conversational settings. Their intended meanings, however, differ greatly and can be accounted for only through pragmatic explanations that require an in-depth understanding of the speaker's beliefs of the addressee's beliefs or knowledge. It is difficult to assign them all the possible meanings which they may have in different situations. It is possible, though, to arrange and list the contexts on the basis of which their intended meanings can be demonstrated. Caahie, in general, is used to communicate the speaker's wishes, whereaspaRnaa, to put it simply, communicates her awareness of the obstacles Agref might encounter in carrying out the action desired by the speaker or imposed on Agref by outside forces. Honaa, on the other hand, in the appropriate communicative settings can play a dual role, i.e. it expresses an obligation or informs Agref of plans already made but not yet carried out.

200 · Ghanshyam Sharma • NOTES 1. The NP being referred to here as 'agent' occurs of course normally with an action verb and, as far as I can see at this point, this represents the default case in such constructions. Other verb types are, however, possible and correspondingly sometimes demand non-agent NPs in the dative. Within the framework of this paper I will restrict myself to a discussion of the default case and defer discussions of the other cases to a later time. 2. According to some Hindi grammars (Shapiro 1989:132f.), the caahie form should agree with the gender and number of the object, but this type of agreement is not found in standard written Hindi, though it is possible to find it in some areas where, because of dialectal influence, certain forms showing this phenomenon can be found in informal Hindi. 3. For a general discussion of this term see the Introduction to Verma and Mohanan (1990), PP.2fr 4. In normal communicative situations it is only the singular T that utters any sentence, but the singular speaker may speak also on behalf of a number of persons. For example, a representative of a political party can speak on behalf of all the members of the party. 5. The pronominal system of address in Hindi has for a singular reference three forms, tu, turn, aap, and for a plural reference again three forms, turn, turn log, aap log. Of these only tu requires a verb in the second person singular; only turn requires a verb in the second person plural; the other three pronouns all require a verb in the third person plural. Aap and aap log are honorific forms used in situations where respect and politeness are being expressed by the speaker. Tu (always singular in reference) and turn (when it has a plural reference) are intimate forms used in situations of great informality. 6. The explanation of the symbols used here are as follows: B, = speaker believes BH = addressee believes KS = speaker knows Kh = addressee knows • = that ? = whether / = or D = it is necessary Φ = it is possible -ι = negation {} = alternation bracket 7. However, a formalization of the different levels of mutual knowledge and beliefs can be sketched (without explanatory discussion) in the following way: Knowledge:

LEVEL!

Κβ ·Ρ, and either {K. -^Kh -P/ ^Ks ?Kh ·Ρ (= Φ -Kh ·Ρ; φ -JCh ·Ρ)} LEVEL II Κ« · ΚΗ ·Ρ, and either {IQ -JCh ·Κβ · Ρ/ -,Κ. ?Kh ·Κβ ·Ρ (= Φ ·ΚΗ -Κ« Ρ; Φ --.Κ», Κ* ·Ρ)} LEVEL III Ke-Kh-Κβ'Ρ, and either {K. -JCh -Κ« -Kb ·Ρ/ ^Κβ · ?Kh ·Κβ -Kh ·Ρ (= + -Kh ^ -Kh ·Ρ; Φ ^Kh ·Κβ -Κ», ·Ρ)}

Explanations for Expressing Obligations of the Agent Referred to in Hindi m 201 LEVEL IV Ks -Kh -Ks -Kh -Ρ, and either {Ks -.Kh -Ks -Kh -Ks -P/ -.Ks - ?Kh -Ks -Kh -Κ* -Ρ (= Φ -Kh -Ks -Kh -Κ* -Ρ; Φ ^Kh -K. -Kh -Κ. -P)} LEVEL V Ks -Kh -Ks -Kh -Ks -P, and either {Ks -.Kh ·Κβ -Κι, -Ks -Kh -P/ -,Κ. · ?Kh -Ks -Kh -Κ« -Kh Ρ (= Φ -Kh -Ks -Kh -Ks -Kh ·Ρ; Φ -^Kh -IQ -Kh -IQ -Kh -P)} LEVEL VI Ks -Kh -Κ« -Kh -K« -Kh -P, and either -,Κ, - ?Kh -Ks -Kh -IQ -Kh -IQ P (= Φ -Kh -Ke -Kh -Ks -Kh -Ks; P)}

Beliefs: LEVEL! Bs -P, and either {Bs -^Bh -P/ -,Bs ?Bh -Ρ (= Φ -Bh -Ρ; Φ -^Bh · P)} LEVEL II Bs · Bh -E and either {Bs-nBh-Bs-P/

^Bs ?Bh Bs -P (= Φ -Bh · Bs Ρ; Φ --,Bh Bs -P)}

LEVEL III Bs -Bh -BS P, and either {Bs-^Bh-Bs-Bh-P/ -.BS - ?Bh -BS -Bh -p (= 4

-Bh -BS -Bh ·ρ; Φ --,Bh -BS -Bh -P)>

LEVEL IV Bs -Bh -BS -Bh -P, and either {Bs -.Bh -Bs -Bh Bs -P/ ^BS · ?Bh -BS -Bh -BS -Ρ (= Φ

-Bh -BS -Bh -BS -Ρ; Φ --,Bh -BS -Bh -BS -p)}

LEVEL V Bs -Bh -BS -Bh -BS -P, and either {Bs-iBh-Bs-Bh-Bs-Bh-P/ -,BS · ?Bh -BS -Bh -Bs -Bh P (= Φ

-Bh -BS -Bh -BS -Bh -Ρ; Φ --.Bh -BS -Bh -BS -Bh -p»

LEVEL VI Bs -Bh -BS -Bh -BS -Bh -P, and either {B s -,B h -B s -Bh-B s -B h -B s P/ -.BS · ?Bh -Bs -Bh -BS -Bh -BS ρ (= Φ

-Bh -BS -Bh -BS -Bh -BS; Φ --Bh -BS -Bh -BS -Bh -BS P)}

• REFERENCES AND SELECT BIBLIOGRAPHY

Chierchia, Gennaro, and Sally McConnell-Ginet. \993.Meaningandgrammar. Cambridge, MA: MIT Press. Geiger, Richard A. (ed.) 1995. Reference in multidisciplinary perspective. Hildesheim/New York: Georg Olms Verlag. Grice, Paul. 1989. Studies in the way of words. Cambridge, MA: Harvard University Press.

202 · Ghanshyam Sharma Hoe, Leo. 1997. Adverbs and modality in English. London: Longman. Hook, Peter E. 1979. Hindi structures: intermediate level. Michigan Papers on South and Southeast Asia, 16. Ann Arbor, MI: Center for South and Southeast Asian Studies, University of Michigan. Kachru, Yamuna. 1980. Aspects of Hindi Grammar. New Delhi: Manohar Publications. Kratzer, Angelika. 1981. The notional category of modality. Words, worlds, and contexts: new approaches in word semantics, ed. by H. Eikmeyer and Hannes Rieser. Berlin: Walter de Gruyter. Levinson, Stephen C. 1983. Pragmatics. Cambridge: Cambridge University Press. Lyons, John. 1977. Semantics. Vol. 2. Cambridge: Cambridge University Press. McGregor, US. 1972. Outline of Hindi Grammar. Oxford: Clarendon Press. Mascica, Colin P. 1991. The Indo-Aryan languages. Cambridge Language Surveys. Cambridge: Cambridge University Press. Palmer, Ralph. 1986. Mood and modality. Cambridge: Cambridge University Press. Perkins, Michael R. 1983. Modal expressions in English. Norwood, NJ: Ablex Publishing Corporation. Schiffer, S.R. 1987. Remnants of meaning. Cambridge, MA: MIT Press. Searle, John R. and Daniel Vandcrveken. 1985. Foundations of ittocutionary logic. Cambridge: Cambridge University Press. Shapiro, Michael C. 1989. A primer of Modem Standard Hindi. Delhi: Motilal Banarsidass Publishers. Smith, N.V. (ed.) 1982. Mutual knowledge. London/New York: Academic Press. Vmna, Manindra K. and K.P. Mohanan. (eds.) 1990. Experiencer subjects in South Asian languages. Stanford: CSLI.

• Regional Reports, Reviewsjand Abstracts

Regional Reports • AFRICA: MAURITIUS · RAJEND MESTHRIE This review offers a brief background necessary for the understanding of the presence of South Asian languages in Mauritius, and an indication of the research carried out on those languages till 1996. Anyone wishing to work on South Asian Mauritian languages would do well to make contact with the Mahatma Gandhi Institute at Moka, for some studies not mentioned here, and for more recent research developments. Mauritius was an uninhabited island until it was settled on a small scale by the Dutch in 1638. The French subsequently took over the island (in 1715) and soon imported a slave population mostly from Africa but also from India. The neighboring island of Bourbon (now Reunion) had 25 per cent of its slaves from India in 1709. From as early as 1729 many workers from Pondicherry (the French enclave in India) had been taken to Mauritius, some as skilled artisans, others as domestic slaves. The languages of these immigrants were most probably a form of Tamil or Malayalam (Barz 1980:3). The British captured Mauritius in 1815 and were soon shipping Indian convicts to the island (between 1815 and 1837) to work on the sugar estates. With the emancipation of slaves in the early nineteenth century, the British addressed the resultant labor shortage by importing laborers from India, under the system of indenture. This led to the biggest stream of immigration into the island, with Indians soon forming a majority of the populace. The British recruited workers from northeast and south India, from the areas which are today Bihar and Uttar Pradesh, Tamil Nadu and Andhra Pradesh. Addison and Hazareesingh (1984:47) estimate that at the time they were writing, people of Indian descent made up 65 per cent of the populace. This makes Mauritius a particularly significant area

206 · The Yearbook (2000)

in the history of transplanted South Asian languages. As early as the 1860s there was (for a brief time) a chair of Tamil held by Professor Rajaruthnum Mudaliar at the Royal College in Port Louis; and in the 1870s a chair of Hindi (then known as Hindustani) held by Professor Rev. W. Wright at the same college. The 1972 population census of Mauritius gave the following statistics for languages spoken: the figures appear to reflect ethnic identity and language group identification, rather than usual language of the home. That is, there is little doubt that Creole is the pan-Mauritian vernacular (except for some older people), the other languages are spoken to varying degrees, and in the cities not at all by the young. 1972 population census

Chinese 20608 Creole 272075 English 2402 36729 French Gujarati 2028 Hindi 320831 Marathi 16553 Tamil 56757 Telugu 24233 Urdu 71668 Other 1134 Not stated 1131 (cited by A.D. Nirsimloo-Anenden 1990:38). • Books and Articles Specifically on Language and Linguistics The earliest study of a South Asian language on the island that I have been able to trace is S. Buckory's Hindi in Mauritius, published in Port Louis in 1967 by the Royal College. It is what might be considered an 'external history' of Hindi teaching on the island, that also contains information on the other languages of the territory. Having a greater basis in academic linguistics is the article by Philip Baker (1969) entitled 'The language situation in Mauritius, with special reference to Mauritian Creole', published in the African Language Review (Vol. 8:73-97). Baker reports inter alia on the main South Asian languages Bhojpuri/Hindi/Hindustani/Urdu, Gujarati and Marathi, and includes notes on smaller languages often ignored by subsequent commentators: Kacchi, Panjabi, Bengali, and Konkani. The first detailed study of Mauritian Bhojpuri is that of Nicole Domingue, in an M.A. dissertation at the University of Texas (Austin) in 1971, entitled 'Bhojpuri and Creole in Mauritius: A study of linguistic interferences and its consequences in regard to synchronic variation and language change'. This thesis holds the distinction of

Reports m 207 being the first study of a transplanted variety of Hindi amongst descendants of indentured Indians. Aspects of the dissertation are summarized in Domingue's article entitled 'Internal change in a transplanted variety', published in Studies in the Linguistic Sciences (Vol. 4:151-59). R.K. Barz reported on 'The cultural significance of Hindi in Mauritius' in the Journal of South Asian Studies (1980, New Series, 3:1-14). He stressed the symbiotic relation between Bhojpuri and Hindi in Mauritius, the former in the colloquial domain, the latter in more formal and literary domains. Peter Stein's Connaissance et emploi des langues a rile Maurice (Hamburg: Helmut Buske, 1982) is a 661 page book which includes a fairly long account of the sociolinguistics of the various South Asian languages. At the conference on the history of overseas Indians held at the University of Mauritius in 1984, Surendhra K. Gambhir examined the features of Mauritian Bhojpuri from the international perspective of other transported Bhojpuri-Hindis, including his own comprehensive study of Guyanese Bhojpuri. Gambhir draws attention to the fact that Mauritian Bhojpuri has the greatest number of typical Bhojpuri features compared to the other overseas Hindis. A revised version of this paper was published in 1986 as 'Mauritian Bhojpuri: An international perspective on historical and sociolinguistic processes', in U. Bisoondoyal and S.B.C. Servansingh's (eds) Indian Labour immigration. A comprehensive 39 page overview of the various South Asian languages of Mauritius is given by Ramyead, entitled 'Indian languages in Mauritius: A perspective', in U. Bissoondoyal (ed.) (1984). His main headings are I 'Introduction'; II 'The establishment of Khari Boli Hindi between 1900 and 1935'; III 'The cultivation of Hindi in Mauritius, 1935 to 1950'; IV 'The progress of Khari Boli Hindi, 1950 to the present'; V 'Some highlights of Urdu, Tamil, Telugu, Gujarati and Marathi'. A revised version of this article, focusing specifically on Hindi and Bhojpuri appears as 'Hindi in Mauritius—a perspective' in R.K. Barz and J. Siegel's Language transported—the development of overseas Hindi (Wiesbaden: Harrosowitz, 1988: 23-41). In the same collection is an article by P. Baker and A. Ramnah entitled 'Recognizing Mauritian Bhojpuri' (pp. 43-67). The article presents a structural discussion of Mauritian Bhojpuri via texts and commentaries on two folk tales, one involving a translation from an original Bhojpuri source from the district of Champaran given in Grierson (1903: 309-10). The authors are able to substantiate the view that Mauritian Bhojpuri grew out of a blending of features from Indian Bhojpuri, and add the view that standard Magahi also seems to have been a major contributor to that blend. Earlier, Baker and Ramnah (1995) contributed an article in Papers in Pidgin and Creole Linguistics (4: 215-38; Series: Pacific Linguistics A-72), entitled 'Mauritian Bhojpuri: An IndoAiyan language in a predominantly Creolophone Society'. Baker's interest in creolistics is reflected in the dictionary of Mauritian Creole: Morisyen-EnglishFrancais: Diksyoner Kreol Morisyen, by P. Baker and V.Y. Hookoomsing (Paris: Harmattan, 1987) which contains some items from South Asian languages that have passed into the Creole. N.C. Neerputh's Le system verbal du

208 · The Yearbook (2000)

Bhojpuri de Vile Maurice, was published in c!986 by the same publisher. I am informed of a Ph.D. dissertation on Mauritian Bhojpuri-Hindi, completed in 1997 at Delhi University by R. Ramjan, but have not yet been able to track down the exact title. • Books and Articles on Culture of Specific Language Groups The above list shows that South Asian languages of Mauritius have not been studied in very great detail. A greater literature exists concerning the history, culture and identity of Indo-Mauritians, some of which contains material of sociolinguistic and linguistic interest. A brief representative list of this literature follows: Bissoondoyal, B. 1984. Life in Greater India: An autobiography. Bombay: Bharata Vidya Bhavan. (Contains accounts of the learning and teaching of Hindi and the speeches—in Hindi, Sanskrit and English—in India and Mauritius, by this prominent community leader.) Bissoondoyal U. (ed.) 1984. Indians overseas: The Mauritian experience. Moka: Mahatma Gandhi Institute. (In addition to the linguistic article by Ramyead, cited earlier, this collection contains articles on the ethno-history of Gujarati and Marathi speakers, of Christian Indians, etc.) Carter, M. 1996. Voices from indenture: Experiences of Indian migrants in the British empire. London and New York: Leicester University Press. (This historical work contains specimens of writing by and on behalf of the indentured labourers in Hindi and English.) Emrith, M. 1967. The Muslims in Mauritius. Port Louis: Regent Press. Misra, K. 1981. Morisasa ke Bhojapuri lokagitum ka vivecantmaka adhyayana. Varanasi: Sanjya Prakashan. (Contains texts and studies of Mauritian Bhojpuri folk songs.) Nirsimloo-Anenden, A. Devi 1990. The Primordial Link: Telugu ethnic identity in Mauritius. Moka: Mahatma Gandhi Institute. (This book has a primarily anthropological focus.) Sooriamoorthy, R. 1977. Les Tamouls a rile Maurice. Port Louis: Royal College.

• REFERENCES (exluding those fully cited above) Addison, J. and K. Hazareesingh. 1984. Ë new history of Mauritus. London: Macmillan. Bissoondoyal, U. and S.B.C. Servansing. 1986. Indian labour immigration: Papers presented at the International Conference on Indian Labour Immigration (23-27 October 1984). Moka: Mahatma Gandhi Institute.

INDIA: DRAVIDIAN* I BH. KRISHNAMURTI Twenty-four or so Dravidian languages are spoken by 189 million speakers (1991 Census) of which the four literary languages account for 182.2 million * I have considered here only those authors and works that either have or are expected to have an impact on historical and comparative Dravidian studies. I have also included the grammars and

Reports m 209 within India. We can add another 7 million speakers of the literary languages outside India if we take into account the immigrant population in Sri Lanka, Malaysia, Singapore, Fiji, Mauritius, and other countries. 1. T. Burrow and M.B. Emeneau brought out the 2nd edition of A Dravidian etymological dictionary (DED[RJ) in 1984 incorporating all supplements and new data from old and new languages. It has 5557 entries, a to v. It has an Appendix of 61 entries which constitutes a supplement to an earlier volume, Dravidian borrowings from Indo-Aryan (1962), followed by indexes of words of individual languages (pp. 515-820). Cross-reference to entries of the 1st edition (DED) and supplements (DEDS, DEN) occur at the end of the volume. Reconstructions are not given as in the first edition. Following the Devanagari alphabetical order (as adopted by the literary Dravidian languages), citation of comparative vocabulary starts with Tamil or any other language that has a cognate available, followed by the other languages. The languages are arranged in a fixed order, from south to north, Tamil to Brahui. This volume is an indispensable tool of research for any student of historical-comparative Dravidian. Emeneau's selected papers were brought out in a volume in 1994. A number of his papers (pre-1980 and post-1980) were reprinted in that volume. In phonology: 'Brahui laterals from Proto-Dravidian V (1980), TrotoDravidian *c and its developments' (1988); in comparative morphology: 'Indian pronominal bases: A revision' (1980); in word-studies: 'Kannada Kampa, Tamil Kampan: Two proper names' (1985), 'Some notes on Dravidian intensives' (1987), 'The right hand is the "eating hand": An Indian areal linguistic inquiry' (1987) have been reprinted in this volume. Krishnamurti reviewed these briefly in his 'Introduction' to the volume, xv-xxvii. Emeneau published his long awaited grammar of Toda in 1984. 2. Bh. Krishnamurti (1991) showed that PD (C)V: CCV changed to (C)V: C(V) thereby making this pattern complementary to (C)VCCV; a similar change in syllable structure also happened in Indo-Aryan approximately at the same time. In a paper on Proto-Dravidian laryngeal *// (1997), he developed an idea that he expressed in a footnote in an article as early as 1963. He observed that Tamil äytam [äydam] was a relic reflex of this PD phoneme. It has the properties of lengthening the preceding vowel in free forms, geminating a voiceless stop in the following position, and dropping when followed by a voiced apical sonorant. It was retained in Early Tamil in the deictic roots ah-tu, ih-tu and the numeral '10'pah-tulpat-tu. These properties are also found in other parts of the grammar and vocabulary whose aberrant phonology is explained more systematically in terms of a laryngeal. In another paper dictionaries/vocabularies published during this period. I have taken for review the authors' names alphabetically. A bibliography of significant books and papers during this period is given at the end.

210 · The Yearbook (2000)

(1998a) he summarized the major sound changes in Dravidian and noticed that typologically motivated sound changes tend to be more regular than simple historical changes. In yet another paper (1998b) he showed how Gondi s> h> 0 is a totally regular sound change in some of the regional dialects through lexical diffusion. This is a theoretical paper showing that lexical diffusion is not incompatible with regular sound change. In comparative morphology, he (with G.U. Rao) published a paper, which discussed the reconstruction of the 3rd person pronouns in different Gondi dialects with focus on the interaction between phonological and morphological rules in language change. In another paper (1997) he has traced the origin and evolution of formative suffixes in Dravidian with copious illustrations. It was established that the so-called formative suffixes, which currently signal intransitive vs. transitive in some of the languages of South and South-Central Dravidian, were originally tense and transition morphemes. Some languages lost tense but not voice; others have lost both the grammatical functions, thereby converting them into mere formatives. He published two papers in Encyclopedias, one on 'Dravidian lexicography' (1991) and the other on 'Indian names: Dravidian' (1994). 3. Sanford B. Steever who did his Ph.D. on Tamil linguistics in Chicago in 1983 published several studies on Tamil and Comparative Dravidian. His research into the formation of serial verbs in Kui, Konda, and Pengo published as papers and a book (1988) led to the publication of a major book Analysis to Synthesis (1993) in which he included five of his earlier papers with appropriate revisions. He could show how two verbs, of which the second was an auxiliary 'be, give, etc.', got fused into a single verb through contraction. This explained clearly the formation of object incorporation in Kui, past negatives in South, South Central and Central Dravidian languages, and that of the past perfect in different subgroups. His publications mark the beginning of research in the interface of comparative morphology and syntax in Dravidian. He edited a volume on the Dravidian languages, published by Routledge in 1998. It includes descriptive accounts of ten Dravidian languages. Since several authors could not write chapters assigned to them, Steever had to write these by consulting original sources. 4. P.S. Subrahmanyam published a book Comparative Dravidian phonology (1983) which brought together all that was available on the developments of PD phonemes in different Dravidian languages up to that point. It certainly made progress over Zvelebil's book (1970) or Emeneau's phonological sketch (1970), but in many places he took a subjective stand. He has not included the revised subgrouping, which had become standard knowledge by 1983, viz., four subgroups instead of three (see Appendix B). Since DEDR was published a year later (1984), his reference to etymologies still depended on DED of 1961 and its supplements. The book apparently got outdated soon after it was

Reports m 211 published. He published a short survey article in 1988, in which he has not said a word about the remarkable contribution of Sanford B. Steever to comparative Dravidian studies. He also questioned my revised subgrouping without giving any substantive arguments. The reflexes of the innovation *n n T occurs only in SDI and SDII languages. CD and ND have no doublets in n- for T beside those derived from *yan/*yan-. I answered his objection (Bh. Krishnamurti 1985:220). Without answering my arguments, he simply says, To maintain that Ma. 1st person singular nan and plural ÞáççáÀ vouch for *n in the inclusive plural is indeed a very strange way of argument' (1988; 64). What is strange must be spelt out. In my latest reconstruction I have a laryngeal in those personal pronouns which have unexplained length between nominative (free form) and oblique (bound form), i.e.yaHnfyan-. 5. Kami! V. Zvelebil published several papers and books on Dravidian during the past two decades. The significant ones are Parts II and III of his work on IruLa (1979,1982) and two articles on Kurumba (1982,1988). The Pondicheny Institute of Language and Culture published his book Dravidian linguistics: An introduction in 1990. It is a textbook, covering a brief history of research in comparative studies. Half of the book is devoted to speculative topics such as Dravidian-Uralic, Dravidian, and Mohanjodaro, etc. • REFERENCES Andres, Susie. 1978. A description of Muria Condi phonology and morphology. Poona: Department of Linguistics, Deccan College Ph.D thesis. Emeneau, M.B. 1980a. Brahui laterals from Proto-Dravidian *r.JAOS 100.311-12. (Reprinted in Emeneau 1994: 303-6.) . 1980b. Indian pronominal demonstrative bases—A revision. Proceedings of the Sixth Annual Meeting of the Berkeley Linguistic Society, 20-27. (Reprinted in Emeneau 1994:307-16.) . 1984. Tbda grammar and texts. Philadelphia: American Philosophical Society. . 1985. Kannada/Cem/w, Tamil Aflm/wrt: Two proper names./XOS 105.401-4. (Reprinted in Emeneau 1994: 317-22.) . 1987a. Some notes on Dravidian intensives. Festschrift for Henry Hoenigswald on the Occasion of his Seventieth Birthday, 109-13. (Reprinted in Emeneau 1994: 323-28.) . 19875. The right hand is the 'eating hand': An Indian areal linguistic inquiry. Dimensions of social life: Essays in honor of David Mandelbaum, 263-73. (Reprinted in Emeneau 1994: 329338.) . 1988. Proto-Dravidian *c and its developments.//iO5108.239-68. (Reprinted in Emeneau 1994: 339-86.) . 1993. Tamil expressives with initial voiced stops. BSOAS 56. 75-86. . 1994. Dravidian studies: Select papers. Delhi: Motilal Banarsidass. . 1997. Brahui etymologies and phonetic developments. BSOAS 60.44(M7. Gwynn, J.P.L. (Assisted by J.V Sastry). 1991. A Telugu-English Dictionary. Delhi: Oxford University Press. Haridas Bhat, K.S. and Upadhyaya, U.P. 1988-97. Tulu Lexicon (in six volumes). Udupi: Rashtrakavi Gonda Pai Samshodhan Kendra. Krishnaraurti, Bh. 1983. (with Lincoln Moses and Douglas Danforth). Unchanged cognates as a criterion in linguistic sub-grouping. Language 59. 541-68.

212 · The Yearbook (2000) Krishnamurti, Bh. 1985. An overview of comparative Dravidian studies since Current Trends (1969). (Presidential Lecture at the Decennial Meeting of the Dravidian Linguistics Association, July 10-12,1980). For Gordon Fairbanks ed. by Veneeta Z. Acson and Richard L. Leed, 212-31. Honolulu: University of Hawaii Press. . 1989. (With G.U. Rao) A problem of reconstruction in Gondi: Interaction between phonological and morphological processes. Osmania Papers in Linguistics (1987 Volume). . 1991a. The emergence of the syllable types of stems (C)VCC(V) and (C)V:C(V) in IndoAryan and Dravidian: Conspiracy or convergence? Studies in the Historical Phonology of Asian Languages ed. by William G. Boltz and Michael C. Shapiro (Current Issues in Linguistic Theory 77), Amsterdam: John Benjamins B.V. . 1991b. Dravidian lexicography. Wörterbücher, Dictionaries, Dictionnaires (An International Encyclopedia of Lexicography), 2521-34. Berlin: Walter de Gruyter. . 1992. Dravidian languages. International Encyclopedia of Linguistics 1. 373-6. Oxford and New York: Oxford University Press. . 1994a. introduction' to M.B. Emeneau, Dravidian studies: Selected papers, xv-xxvii. Delhi: Motilal Banarsidass. . 1994b. Indian Names: Dravidian. Namen Forschung, Proper Name Studies, Lesnomspropres, ed. by L. Zgusta, 665-671. Berlin: Walter de Gruyter and Co. . 1997a. The origin and evolution of primary derivative suffixes in Dravidian. Historical, IndoEuropean and lexicographical studies (A Festchrift for Ladislav Zgusta on the occasion of his 70th birthday), ed. by Hans Hcnrich Hock, 87-116. Berlin: Mouton de Gruytcr. . 1997b. Proto-Dravidian laryngeal *// revisited. PILC Journal of Dravidic Studies 7 (2). 14565. Pondicherry: Pondicherry Institute of Linguistics and Culture. . 1998a. Patterns of sound change in Dravidian. The Yearbook of South Asian Languages and Linguistics, ed. by Rajendra Singh, 63-79. New Delhi: Sage Publications. . 1998b. Regularity of sound change through lexical diffusion. Language variation and change 10.193-222. Krishnamurti, Bh. and J.P.L. Gwynn. 1985. A Grammar of Modem Telugu. Oxford: Delhi. Lehmann, Thomas. 1989. A Grammar of Modem Tamil. Pondicherry. PILC. . 1991. Grammatik des Alttamil: Morphologische und Syntaktische Analyse der CankamTamil des Dichters Käpilar. University of Heidelberg Ph.D. Dissertation. Mahapatra, B.P. 1979. Malto: An Ethno-Semantic Study. Mysore: Central Institute of Indian Languages. Natarajan, G.V. \9S5.Abujhmaria Grammar. Mysore: Central Institute of Indian Languages. Ranganathacharyulu, K.K. 1987. A Historical Grammar of Inscriptional Telugu (1401 AD to 1900 AD). Hyderabad: Department of Linguistics, Osmania University. Sambasiva Rao, Gali. 1991. A Comparative Study of Noun Derivation. New Delhi: Bahri Publications. Steever, Sanford B. 1980. The genesis of poly-personal verbs in South Central Dravidian. IJDL 9 (2). 337-73. . 1981. Selected Papers on Tamil and Dravidian Linguistics. Maturai: Muttu Patipakkam. . 1987a. Remarks on Dravidian complementation. Studies in Linguistic Sciences 17 (1). 10319. . 1987b. The roots of the plural action verb in the Dravidian languages. JAOS 107 (4). 581604. . 1988. The Serial Verb Formation in the Dravidian Languages. Delhi: Motilal Banarsidass. 1989. Review by Hans H. Hock. Language 65 (2). 393-405.1991. KA. Jayaseelan. Linguistics 29 (3). 543-48. . 1993. From Analysis to Synthesis. Oxford: Oxford University Press. 1997. Rev. by K.V Zvelebil.//iaS 117.696-98. . 1998. The Dravidian Languages. London: Routledge.

Reports m 213 Subrahmanyam, P.S. 1983. Dravidian Comparative Phonology. Annamalainagar: Annamalai Univ. . 1988. Comparative Dravidian studies from 1980. IJDL 17 (1). 59-91. . 1991. Tense formation in Kota-Toda: A comparative study. In B. Lakshmi Bai, and B. Ramakrishna Reddy (eds.), 49-72. Sumati, S. 1982. Comparative Phonology of the South-Central Dravidian Languages. Hyderabad: Department of Linguistics, Osmania University M.Phil dissertation. Suvarchala, B. 1992. Central Dravidian Comparative Morphology. New Delhi: Navrang. Umamaheshwar Rao, Garapati. 1987. A Comparative Study of the Gondi Dialects. Hyderabad: Department of Linguistics, Osmania University Ph.D. dissertation. Zvelebil, Kamil V. 1979. The Intla (Erla) Language, Part II. Wiesbaden. . 1982a. Bejta Kurumba: First report on a tribal language. JAOS 102.523-27. . 1982b. The Irula (Erla) Language, Part III. Irula Lore, Texts and Translations. Wiesbaden: Harosswicz. . 1988. Jenu Kurumba. Brief report on a 'tribal' language of the Nilgiri area. JAOS 108.197301. . 1990. Dravidian Linguistics: An Introduction. Pondicheny: Pondicherry Institute of Linguistics and Culture.

NORTH AMERICA I TEJ Ê. ÂÇÁ¹Á This is the last report of this millennium. It is a follow up to my earlier report in the 1998 Yearbook which covered the trends in South Asian Linguistics in North America through 1996. Therefore, it is natural that the focus here will be on those works which have been published during the past two years (199799). This survey restricts itself primarily to the published works during this period and will not attempt to survey dissertations and theses on South Asian linguistics. As we proceed, it will become self-evident that South Asian linguistic research continues to flourish and diversify in this region. The account of research presented here is classified into three (?) major sections: (1) Historical and Diachronie linguistics; (2) Theoretical linguistics; and (3) Sociolinguistics and Applied linguistics.

• 1. Historical and Diachronie Linguistics Issues of the P ninian linguistics continue to dominate the developments in historical linguistics. Deshpande's works (1997 a, b) shed further light on P ninian linguistics. In the former paper (1997a), Deshpande traces the history of the belief that P nini's grammar was inspired by Lord Shiva and then goes to explore the interaction of the divine and linguistic inspirations behind P nini's monumental work. It is claimed that Bhartrhari was the first to

214 · The Yearbook (2000)

associate divine inspiration behind the three ancient grammarians (Pänini, Katyayana, and Patanjali.) The later article (1997b) attempts to determine the system of phonetic features underlying Pänini's metalinguistic rules. The notion of natural class, articulatory and other phonetic concepts are examined and analyzed. It is concluded that despite some limitations, the postulated feature system is remarkably enduring in terms of modern phonological theory. Hamp (1997), on the other hand, deals with the problem of reconstruction. The paper accounts for the reconstructed stems such as *sVktHn 'thigh' in Indo-Aiyan languages. By way of analyzing cognate -1- stems from five IndoAryan languages and -d- stems from Western Marwari and Hindi language, this paper stresses the hetroclitic status of such stems.

• 2. Theoretical Lingusitics • 2.1 Syntax The studies devoted to syntax center on two main topics—relative clauses, negation, and case marking. Such studies show a diversity of theoretical and analytical frameworks. Ackerman and Webelhuth (1998) is of interest to those South Asian linguists who are interested in a lexicalist theory of predicates. Using data from Malayalam and Hindi among other languages, the theory is formulated as a variant of head-driven phrase structure grammar. Drawing data from South Asian languages, such as Tamil, Malayalam, and Hindi together with other non-South Asian languages, Narasimhan (1998) offers an alternative account of quirky case-marking (dative case-marking with subjects) which best describes dative case marking in Hindi. The alternative explanation is based on Foley and Van Valin's Role and reference grammar (1984). In Mahajan (1997), the phenomenon of scrambling is re-examined with special reference to rightward movement in Hindi. The paper goes on to show that an IP right-adjunction account fails to explain phenomena of variable binding, anaphor binding, condition C effects, and the absence of scope ambiguities associated with leftward scrambling. These syntactic effects preclude assignment of rightward scrambling to the phonological component. Rajesh Bhatt (1997) examines the syntax-morphology interface with reference to Hindi correlative (or free relative) construction. The analysis claims that Hindi correlative clauses are base-generated in the subject position. The article goes on to claim that Hindi has free relatives and that case matching parameter is not a part of its syntax. In addition, the matching violations are not accessible to syntax. The grammatical account of prenominal relative clause construction in the Gultari dialect of Eastern Shina is the main concern

Reports m 215 of Hook (1997). The paper examines the gaping and agreement behavior in the construction with cross linguistic comparison. Lahiri (1998) deals with a semantic analysis of Hindi negative polarity items. In addition to semantics, the morphological complexity of such items is accounted for and the relevance of the proposed Hindi analysis for English polarity items is discussed in this work. Guitierrez Rexach (1997), on the other hand, deals with Hindi negative polarity items in context of rhetorical reading, ambiguity status, and their distribution in interrogatives and yes/no type of questions. Gair (1998) is a volume which deals with a wide variety of topics such as syntax and other aspects of South Asian linguistics. Selected and edited by Barbara Lust, this volume represents a collection of articles written by Gair over the last 30 years on linguistics in general and south Asian linguistics in particular. The articles are grouped into six categories: (1) Background and description of basic categories; (2) Syntax: Configuration, order, and grammatical function; (3) Deixis, anaphora, and agreement; (4) Change, grammaticization and linguistic area; (5) Diglossia, and (6) The development of syntax. • 22. Phonology and Phonetics In the area of phonology and phonetics interface, the paper by Broselow, Chen, and Huffman (1997) is a notable one. This comparative study deals with moraic structure and its relation to syllable weight. Among other findings, this cross-linguistic study reveals that all codas head a mora in Hindi and share a vocalic mora in Malayalam. • 2.3. Morphology Two important monographs on morphology were published in 1997: A. Ford, R. Singh, and G. Martohardjono, Pace Panini: Towards a word-based theory of morphology and R. Singh and R.K. Agnihorti, Modem Hindi morphology: A word-based description. Both challenge the established Päninian view of morphology, and the latter provides a detailed description of a very substantial chunk of the morphology of Modern Hindi.

• 3. Sociolinguistics and Applied Linguistics • 3.1 Sociolinguistics The work in Sociolinguistics ranges from a personal account of Indian Sociolinguistics and ethnic profiling, to the characterization of native speaker, mother tongue, diglossia, and grammar of language mixing. Bright (1997) gives a nostalgic personal account of the early days of Sociolinguistics and the

216 · The Yearbook (2000)

influence of Emeneau, Mary Hass, and Einar Haugen on his research in linguistics in general and South Asian linguistics in particular. The paper reveals that the 1964 UCLA conference on sociolinguistics and the 1966 volumes on sociolinguistics are the milestones of collaborative research in the field of sociolinguistics. As mentioned in Section 2.1, another work with historical dimension is Gair (1998). Singh (1998a) is a collection of articles which revitalize the debate on the concept of native speaker and other related concepts critical to the notion of 'native speaker' in linguistics. Pieter Muysken's chapter (We are all native speakers, but of which language) challenges the distinction between Elanguages (e.g. Hindi and Urdu) and I-languages (e.g. Quechua and Media Lengua). Other chapters reveal that the concept of 'mother tongue' is a multidimensional one which includes criteria, such as: the language spoken from birth; parent tongue; language spoken in a household; language used in daily verbal interaction; language of wider communication; institutional language; and identities, such as language, religious, cultural, political, and social identity. In short, the notion of 'mother tongue' or 'native speaker' is subject to dynamic individual and community perception of the notion at hand. South Asianist would be pleased to know that Munshiram Manoharal, New Delhi, brought out an Indian edition (Singh 1998b) of Singh (1996), inadvertently left out from my last report. The volume, with the provocative title Lectures against sociolinguistics, is made up of only slightly edited transcripts of a handful of lectures Singh gave at the National University of Singapore. It constitutes a serious challenge to the autonomist pretensions of contemporary sociolinguistics. As is evident from the monograph-size article, ethnic profiles of South Asian languages (37 languages, 17 official languages) are embodied in Levitt (1997). The characterization of Diglossia with respect to relatedness of the High and Low varieties is still a subject of controversy. Paolillo (1997) examines this controversy in the context of two models proposed for Sihala: Discrete model (Gair 1968; 1992) and Continuum model (De Silva 1974; 1979). Using a computer-generated, multi-dimensional graph of relations between the varieties of Sinhala, it is concluded that the analysis favors the Discrete model over Continuum model. Three chapters in Gair (1998) are devoted to Sinhalese Diglossia. Turning to the grammar of language mixing, recent attempts to posit constraints on the grammar of bilingual language mixing have gained new dimensions in the past two years. Following an experimental and comparative approach, Rakesh Bhatt (1997) attempts to posit universal constraints on code-switching based on an optimality theory. By way of examining the grammaticality judgments of Kashmiri-English, Hindi-English, KashmiriHindi, Spanish-English, and Swahili-English spontaneous speech and printed data, constraints are posited and prioritized on one hand and differences in

Reports m 217 constraint-ranking configurations of individual grammars are accounted for on the other. • 3.2 Language Acquisition Two studies which are of particular interest to South Asian linguistics are Someshekar et al. (1997) and Ritchie and Bhatia (1998). The first deals with first language acquisition while the second paper deals with second language acquisition. The aim of Somshekar et al.'s (1997) study is to compare the results of the acquisition of Principle C (Chomsky 1981) in Hindi with previous results obtained in English acquisition. The study shows that pronominal interpretation in Jab clauses relies more on pragmatic factors in Hindi than in English. The cross-linguistic comparison supports the universality of Principle C. The results favor the structural dependency of Principle C. Precedence has a stronger effect on right-branching structures than on left-branching ones. Hindi-speaking children seem to violate Principle C with age and development. Three independent factors—pronominal type, pronominal direction, and branching direction of the sentence—were employed in testing children's interpretation. Ritchie and Bhatia (1998) explore the acquisition of constraints on codeswitching by second language learners. The study reported in their paper required 30 beginning and advanced students of Hindi at Brown University to give 30 preference judgments involving pairs of Hindi-English code-switched sentences with the functional category Hindi future tense marker -egii. The general applicability of the Functional Head Constraint (FHC) though has been called into question (e.g. Mahootian and Santorini 1996). Bhatia and Ritchie (1996) and Ritchie and Bhatia (1999) have argued convincingly that the FHC constraint on competence provides a more natural explanation for the grammaticality judgments than production accounts (e.g. Myers-Scotton 1993). This paper sets out to explore this question by way of testing the following two competing hypotheses: (a) The null hypothesis: The FHC is not accessible to adult second language learners (implication: the CM/CS behavior of adult second language learners will be random with respect to the FHC). (b) The FHC availability hypothesis: The FHC is accessible to adult second language learners (implication: learners' behavior will be parallel to that of balanced bilinguals with respect to the FHC). The results show that the FHC is, in this particular structure, accessible to adult second language learners. South Asian linguists working on bilingual children might find Bhatia and Ritchie (1999) of considerable interest. This chapter presents a comprehensive account of the complex nature of childhood bilingualism, ranging from its formal to its socio-psychological dimensions. The main trends in the study of early childhood bilingualism are accounted for, while taking note of the recent major theoretical, analytical, and methodological findings and issues

218 · The Yearbook (2000) concerning language development in the bilingual child on a cross-linguistic basis including South Asian languages. • 33 Discourse Analysis Kachru (1997) contains a handful of studies devoted to cross-cultural discourse in world Englishes. It presents a critical examination and empirical research on discourse difference in the inner circle and outer circle Englishes. It is argued that functions, domains, roles, and values which govern different discourse styles in world Englishes must be taken into account in the process of identifying norms of writing in English, rather than following a monolithic academic writing style in the inner circle of English. • 3.4 Writing System Daniels and Bright (1996) provide an excellent overview of the writing systems of the world. The section dealing with South Asian writing systems (ibid.: 371437) has been edited by Richard Salomon. In addition to Brahmi and Khroshthi, the writing systems of the major Indo-Aryan, Dravidian and Tibeto-Burman languages of South Asia are covered. • 3.5 Language Pedagogy T\vo works devoted to computer-based, multi-media language pedagogy are Vidya and CyberGiani. Vidya (Kumar, Anejaneyulu, and Gupte 1997) is available in Hindi and English and is in a shell format to facilitate lesson customization. It is based on situation-based contextual language learning and student modeling of language constructs. It is claimed that in addition to language learning, it is also useful for aphasic patients. CyberGiani is a Microsoft Windows-based interactive multimedia program for learning to speak, read, and write Punjabi as a second language. For more details, visit: http:// www.sonanet.com.

• 4. Conclusion Although North America is a vibrant area in terms of research on South Asian languages and linguistics, many challenges still await. Some of these challenges were systematically addressed at the 20th Conference on South Asian Language Round-table held at the University of Illinois (July 9-11, 1999). One of the main themes of the conference was SALA 2000: Priorities and Directions.

Reports m 219 REFERENCES Ackerman, Farrell and Gert Webelhuth. 1998.Ë theory of predicates. Center for the Study of Language and Information. Stanford: Stanford University Press. Bhatia, Tej K. 1998. and William C. Ritchie. 1996. Bilingual language mixing, Universal Grammar, and second language acquisition. Handbook of second language acquisition, ed. by William C. Ritchie and Tej K. Bhatia, 627-88. San Diego: Academic Press, Inc. . 1999. The bilingual child: Issues and perspectives. Handbook of second language acquisition, ed. by William C. Ritchie and Tej K. Bhatia, 569-643. San Diego: Academic Press, Inc. Bhatt, Rajesh. 1997. Matching effects and the syntax-morphology interface: Evidence from Hindi correlatives. ÌÃÃ working papers in linguistics 31,53-68. Bhatt, Rakesh Mohan. 1997. Code switching, constraints, and optimal grammars. Lingua 102 (4). 223-51. Bright, William. 1997. Reminiscenes: Beginnings of sociolinguistics. The early days of sociolingustics: Memories and, reflexions, ed. by Paulston Christina and Richard Tucker, 53-60. Dallas, Texas: Summer School of Linguistics. Broselow, Ellen, Su-I Chen, and Marie Huffman. 1997. Syllabic weight: Convergence of phonology and phonetics. Phonology 14 (1). 47-82. Chomsky, Noam. 1981. Lectures on government and binding: The Pisa lectures. Dordrecht: Foris Publishers. CyberGiani. n.d. Sona Systems Corporation. Daniels Peter T. and William Bright (eds). 1996. The world's writing systems. New York: Oxford University Press. De Silva, M.W.S. 1974. Convergence in diglossia: The Sinhalese situation. Contact and convergence in South Asian languages, ed. by Franklin Southworth and Mahadeo Apte, 60-91. Ernakulam, India: Dravidian Linguistics Association. . 1979. Sinhalese and other island languages of South Asia. TYbingen: GYnther Narr Verlag. Deshpande, Madhav. 1997a. Who inspired Panini? Reconstructing the Hindu and Buddhist counter claims. Journal of the American Oriental Society 117 (3). 444-65. . 1997b. Panini and the distinctive features. Journal of Indo-European Studies. Monograph series 22.72-87. Gair, James. 1968. Sinhala diglossia. Anthropological Linguistics 10 (8). 1-15. . 1992. AFR, INFL, Case and Sinhala diglossia, or, Can linguistic theory find a home in variety. Dimensions of South Asia as a sociolinguistic area, ed. by Braj Kachru, Edward Dimock, and Bh. Krishnamurti, 179-97. Delhi: Oxford University Press. . 1998. Studies in South Asian languages: Sinhala and other South Asian languages Oxford: Oxford University Press. Ford, Alan, Rajendra Singh, and Gita Martohardjono. 1997. Pace Panini: Towards a word-based theory of morphology. New York: Peter Lang. Guitierrez-Rexach, Javier. 1997. The semantic basis of NPI licensing in questions. MIT working papers in linguistics 31. 359-76. Hamp, Eric. 1997. CDIAL 13073 Sakthan, Norn. Sakthi. Indo-Iranian Journal 40 (3). 259. Hook, Peter. 1997. Relative clauses in Eastern Shina. Journal of Indo-European Studies. Monograph series 22.140-54. Kachru, Yamuna. 1997. Cultural meaning and contrastive rhetoric in English education. World Englishes 16 (3). 337-350. Kumar, Vivekanandan, K.S.R. Anjaneyulu, and Uday Gupte. 1997. Vidya: A situated CALL environment. Computer Assisted Language Learning 10 (2). 149-72. Lahiri, Utpal. 1998. Focus and negative polarity in Hindi. Natural Language Semantics 6 (1). 57123. Levitt, Jesse. 1997. Language and ethnic briefs. Geolinguistics 23.157-334.

220 · The Yearbook (2000) Mahajan, Anoop. 1997. Rightward scrambling. Rightward movement, ed. by Dorothee Beerman, David LeBlanc, and Henk Van Riemsdijk, 185-213. Amsterdam: John Benjamins. Mahootian, S. and B. Santorini. 1996. Code switching and the complement/adjunct distinction: A reply to Belazi, Rubin and Toribio. Linguistic Inquiry 27 (3). 464-79. Myers-Scotton, Carol. 1993. Duelling languages: Grammatical structure in code switching. Oxford: Clarendon Press. Narasimhan, Bhuvana. 1998. A lexical semantic explanation for 'Quirky' case marking in Hindi. Studio Linguistica 52 (1). 48-76. Paolillo, John. 1997. Sinhala diglossia: Discrete or continuous variation? Language in Society 26 (2). 269-%. Ritchie, William C. and Tej K. Bhatia. 1998. Acquisition of constraints on code-switching by second language learners. Paper presented at the 19th SALA conference. York: University of York. . 1999. Codeswitching, grammar, and sentence production: The problem of light verbs. The development of second language grammars, ed. by Elaine Klein and Gita Martohardjono, 27391. Amsterdam: John Benjamins. Singh, Rajendra. 1996. Lectures against sociolinguistics. New York: Peter Lang. . (ed.). 1998a. The native speaker: Multilingual perspectives. New Delhi: Sage Publications. . 1998b. Lectures against sociolinguistics. Delhi: MRML. Singh, Rajendra and R.K. Agnihotri. 1997. Modern Hindi Morphology: A Word-Based Description. Delhi: Motilal Banarsidass. Somashekar, Shamitha, Barbara Lust, James Gair, Tej Bhatia, Vashini Sharma, and Jyothi Khare. 1997. Principles of pronominal interpretation in Hindi Jab clauses: Experimental test of children's comprehension. Cornell working papers in linguistics 15. 65-87.

I SOUTH ASIA: HISTORICAL* I HANS HENRICH HOCK As we all know, South Asia, or ancient India, excelled in synchronic linguistics long before the West. To the present day, the Prätteäkhyas and especially Pänim'sAslädhyäyi are among the greatest monuments of human endeavor to gain insights and make generalizations about linguistic structure.1 The phonetic observations of the Prätteäkhyas and the later Siksas went far beyond pre-modern Western observations, both in terms of accuracy and in terms of systematization and theory. The enormous debt of modern Western articulatory phonetics to the Sanskrit tradition is reflected even in its terminology. Thus, English voiceless and voiced, German stimmhaft and stimmlos, and their counterparts in other European languages are perfect caiques of the * An earlier version of this paper was read as an invited special lecture at the Second International Conference on South Asian Languages, 9-11 January 1999, at Punjabi University, Patiala. I am grateful to the organizers of the Conference, especially K.V. Subbarao, for inviting me to present the lecture. I also have benefited from comments received at the Conference. As usual, the responsibility for errors and omissions rests with me.

Reports m 221

Sanskrit terms aghosa and ghosavat. The influence goes even further, when the term palatal is used by many modern Western phoneticians as an articulatory designation of front vowels and the glide y, even though these are articulated in the velar area. The terminology reflects a theoretical innovation in the PrätiSäkhyas, which was based on resonance, not just articulation, and which on that basis classified front vowels andy as palatal, back vowels and v as labial, and -vowels, velars, and glottals as kanthya.2 (See Cardona [1986] and Hock [1992].) Less well known than the influence of the Sanskrit phonetic tradition on Western phonetics is the fact that the Päninian tradition, too, has influenced Western linguistics. Beyond its indirect influence on western linguistics in general, note especially Bloomfield (1939) which employs Päninian conventions of abstract underlying forms and of surface forms derived by ordered statements of generalizations. Bloomfield applied the same approach to German morphophonemics in his Language (1933). His statement in this context that 'this same consideration often leads us to set up an artificial underlying form' is picked up by Chomsky in his Syntactic structures (1957: 81, fn. 6), with the added comment 'We have also found this insight useful in transformational analysis, as, e.g. when we set up the terminal string John—C—have + en— be + ing—read underlying the kernel sentence 'John has been reading.' For all its accomplishments, however, the Sanskrit grammatical tradition remained essentially a synchronic one. Even grammars of the Prakrits, which derived Prakrit from Sanskrit by rule, did not claim to be historical in nature. In fact, as Deshpande (1979b: 109, fn. 38) has observed, while the orthodox tradition would consider the Prakrits in some sense derived from Sanskrit (by deviation, not by historical change), the Jaina tradition tends to look at Ardhamagadhi as basic and to consider Sanskrit as secondary. (See also Deshpande [1993] and Dundas [1996].) Historical linguistics, then, constitutes the West's contribution to the history of linguistics. Although the Latin grammarians had speculated that their language might be derived from some Ancient Greek dialects, their approach was essentially ahistorical and thus entirely comparable to derivations of Prakrit from Sanskrit in India. It remained up to the Renaissance, with its keenly developed sense of historicity and its awareness of the chronological relationship between the original Latin and the later Romance vernaculars, to lay the foundation for historical linguistics, beginning with Dante's De vulgari eloquentia. In its early stages, Western historical linguistics remained fairly parochial, focusing on the known languages of Western Europe, with some consideration of Greek and Hebrew. As the cultural horizon of Europeans broadened, so did the familiarity with other languages, which laid a more solid foundation for comparative historical linguistics. But without a solid understanding of phonetics and linguistic structure, most of Western historical comparative linguistics remained unsophisticated and fairly ad hoc.

222 · The Yearbook (2000)

It was only after the 'discovery' of Sanskrit, of the Sanskrit phonetic and grammatical tradition, and of the relation of Sanskrit to Latin, Greek, and other Indo-European languages that comparative and historical linguistics acquired the necessary sophistication. Although discussions of the history of linguistics tend to focus on William Jones's famous statement of 1786 on the relationship between Sanskrit and other Indo-European languages,3 what was most significant for the development of the field was the improved understanding of linguistic structure and, somewhat later, of phonetics, both inspired by the Sanskrit tradition. Moreover, even though the traditional Indian grammars that derived Prakrit from Sanskrit were not originally conceived as historical, they could be interpreted that way and could thus influence historical linguistics, at least in the area of Sanskrit and Indo-Aryan.4 Beyond its contributions to the establishment of modern historical and comparative linguistics, India at first also greatly contributed to the conceptualization and reconstruction of the Proto-Indo-European (PIE) ancestor language. In fact, while Jones had considered the possibility that the ancestor 'perhaps, no longer exists', Schlegel claimed that 'comparison further shows that the Indie language is the older, the others however later and derived from it' (1808, emphases supplied). As a consequence, early reconstructions of the Proto-Indo-European sound system looked remarkably close to that of Sanskrit. A major change in thinking took place toward the end of the 19th century, with the discovery of the 'Law of the Palatals', which established that Sanskrit and its Iranian relatives are separated from Proto-Indo-European by major changes in the sound system, both in the vowels and in the consonants. (For a more detailed discussion with references see Collinge [1985:133-42].) S.S. Misra has recently (1992) attempted to refute the arguments in favor of this changed perspective and to argue that Sanskrit (Skt) is much closer to Proto-Indo-European. However, his argument can only be maintained at the cost of rejecting the well-established hypothesis that sound change is regular. It is possible to derive Skt ca 'and' from PIE *Jt*e and ka- 'who' from *£wo- by regular sound changes and thereby to account for the relationship between the Sanskrit words and their cognates in other Indo-European languages; see 1-4. This account explains the palatalized outcome c in 4a (since the stop was followed by front vowel at the time of palatalization), and the non-palatalized outcome in 4b (before original non-front vowel). By contrast, if we were to start with the uniform á-vocalism of Sanskrit, it would be impossible to state any phonetic conditions for the different outcomes, since, as the examples show, both palatalized and non-palatalized outcomes occur before á-vowels. Nor would it be possible to explain the difference in vocalism in the related Indo-European languages, or the conditioned outcomes of the labiovelars in these languages, for that matter. (See Hock [In Press b] for further details.)

Reports m 223 (1) Velar-Labiovelar merger: *k,g,gh }>*k,g,gh (2) Palatalization: *k, g, gh > *c, j, jh/—[+ voc., + front] (3) Vowel merger:

I

> I

*kwe 'and' (cf. Lat. que, Gk. te) *ke *ce ca *kwo- *who,what' (cf. Lat. quod, Gk. po-) Velar-Labiovelar merger *ko Palatalization — Vowel merger ka

(4) a. Proto-Indo-European Velar-Labiovelar merger Palatalization Vowel merger b. Proto-Indo-European

Beyond the role that Sanskrit has played and may or may not continue to play in Indo-European linguistics, the languages of South Asia have also been the focus of historical linguistics in their own right. The Sanskrit -> Prakrit grammars, mentioned earlier, laid the foundation for Indo-Aryan historical linguistics, a discipline which received a tremendous boost with the increased knowledge of Buddhist and Jaina Prakrit and of Apabhram£a texts, the decipherment of the ASokan inscriptions, and increased awareness of the earliest stages of the Modern Indo-Aiyan languages. Well-known reflexes of this work include the Sanskrit grammar started by Jakob Wackernagel (1896-) and still not quite completed;5 Pischel's work on the literary Prakrits (1909); Tfcgare's Historical grammar of Apabhramfa (1948); Jules Bloch's Formation de la langue marathe (1920); Chatterji's Origin and development of the Bengali language (1926), Saxena's Evolution ofAwadhi (1937); Jha's The formation of the Maithili language (1958); and much of Masica's recent massive volume on the Indo-Aryan languages (1991). Most of the work on Indo-Aryan has been restricted to the literary languages. But one Indo-Aryan 'tribal' language, Bangani, has recently stirred up broader attention, as well as controversy, following Claus Peter Zoller's observation (1988,1993) that certain words in Bangani bear a strong resemblance to Western Indo-European languages (e.g. dokro 'tear' with initial dental stop like GK. dakru, Engl. tear, vs. Skt. afru, Toch. B akrüna without the stop). Zoller's data have been questioned by Beekes (1995) and van Driem and Sharma (1996), in part with highly disparaging remarks about Zoller's method of data collection. Based on first-hand fieldwork in Bangan, Anvita Abbi has

224 · The Yearbook (2000)

recently shown that most of Zoller's data are, in fact, genuine (1997). What is especially interesting is that the words in question do not seem to be traceable to any specific known Western Indo-European language (see e.g. Hock 1997). If the relatively small number of such words that have been collected so far can be augmented by additional data, the non-Indo-Aryan component of the Bangani lexicon may raise interesting questions for comparative IndoEuropean linguistics. In addition there is the as yet unanswered question of how the Western Indo-European features may have found their way into South Asia. A common view is that some of Alexander's troops brought them to India; but without further evidence this remains idle speculation. Historical and comparative research on Dravidian started remarkably early. In 1816, Francis Whyte Ellis of the Indian Civil Service related Tamil, Telugu, Kannada, Malayalam, Tulu, Kodagu, and even Malto to each other. A solid foundation for comparative Dravidian was set by Robert CaldwelPs Comparative grammar of the Dravidian or South Indian family of languages in 1856. More recent work includes Burrow and Emeneau's monumental Dravidian etymological dictionary, second edition (1984); Emeneau's Dravidian studies (1994); Krishnamurti's Telugu verbal bases (1961); Steever's Serial verb formation in the Dravidian languages (1988); Subrahmanyam's£>r0v/a/0n comparative phonology (1983); and ZvelebiPs work of the same title (1970). In spite of the great progress made in comparative and historical Dravidian linguistics, a number of problems and challenges remain. First, as noted, e.g. by Steever (1988), in spite of the efforts of scholars such as Burrow and Bhattacharya (1953, 1970), Emeneau (1955, 1984), Hahn (1911), Israel (1979), Krishnamurti (1969), and Subrahmanyam (1968), much less is known about the so-called tribal languages than about the literary languages, Tamil, Telugu, Kannada, and Malayalam. While the same may be true for the Indo-Aryan tribal languages, the number and geographical/genetic range of literary Indo-Aryan languages is broad enough to give something close to a representative sample. In Dravidian, the literary languages are restricted to the extreme south, and information for geographically or genetically more northern languages is much more restricted. Second, the 'time depth' for Dravidian is much more limited than for IndoAryan. While Indo-Aryan can be traced back to Sanskrit, which in its earliest, Vedic stages dates to at least 1,500 BC (most likely, to about 2000 B.C.), the earliest Dravidian attestations come from 2nd-century BC Tamil; and information on the more northern languages and other 'tribal' languages does not become available before the A.D. 19th century. These problems, however, pale when we turn our attention to the other languages of South Asia. While for Tibeto-Burman we have medieval literary Tibetan, for all other languages we again have to depend on 19th- and 20thcentury sources. As in the case of Dravidian, comparative reconstruction can increase the time depth of our knowledge of these languages, but progress in

Reports m 225 this area is severely hampered by both the relative dearth of data and the small number of linguists doing comparative-historical work on these languages. Most problematic for comparative-historical linguistic work are languages without known relatives, Burushaski in the extreme north (for which see especially Berger [1998]), as well as Nahali in Central India (for which see, e.g. Kuiper [1962]). For these languages, we are not even able to increase the time depth of our knowledge through reconstruction. As we will see, these limitations in attestation in space and time of the nonIndo-Aryan languages of South Asia have significant consequences for the continuing debate over the historical origins of the well-known phenomenon of South Asian convergence. A number of South Asian historical phenomena and developments have given rise to investigations whose implications go beyond South Asia. One of these is the question of the historical origins of ergativity in the IndoAryaii languages, as in Hindi maim ne kitabparhi read a book'. It is clear that this type of ergative construction reflects an earlier Sanskrit mayapustakam pathitam lit. 'by me a book (was) read', and that this type of structure began to replace the old finite past-tense constructions, beginning with the later Vedic period. Where opinions differ is on the question of whether this type of Sanskrit structure was a passive construction (a commonly held view, recently reiterated in Anderson [1977]), or whether it was ergative to begin with (Klaiman 1978; Hock 1985,1986). The latter interpretation has the virtue of explaining not only the transitive type maim ne kitäb parhl/mayä pustakam pathitam, but also the intransitive counterpart maim dilli gaya/gai/aham dülinagaram gatahlgatä went to Delhi', as well as variation in construction for certain classes of verbs such as verbs of speaking or consuming (see especially Hock [1985]). The tendency for intransitive verbs to construe like actives, and transitive verbs like passives, is the classical indicator of an ergative system. The fact that this system is found not only in Hindi but also in Sanskrit suggests that ergativity is inherited, and not the result of grammaticalization of an original passive. In tact, Comrie (1978) has observed that ergativity is a cross-linguistic tendency in perfectives, with counterparts in the past participles even of English. The conclusion that the construction maim ne kitäb parhi/mayä pustakam pathitam was ergative from the start, if correct, has obvious implications for theories on the historical origins of ergative systems, as well as for grammaticalization studies. Tonal contrasts in Panjabi (as in /körä/ 'horse': /korä/ 'leper': /korä/ 'whip') were apparently first noticed by Bailey in 1913.6 Since then, similar tonal contrasts have been observed in Rajasthani, Lahnda, and several other languages of the area (see the discussion in Masica [1991]). The development of tonal contrasts in Indo-Aryan languages clearly raises interesting questions for the study of tonogenesis in general, especially since the Panjabi tones are limited

226 · The Yearbook (2000)

such that only one tone can occur on a given word, no matter what the syllable count. In this regard, then, Panjabi is similar to Norwegian and Swedish. Some early publications have attributed the feature of tone to TibetoBurman subversion;7 however, Masica (Indo-Aryan, p. 132) points out that 'those languages where the tone developments are strongest (e.g. Punjabi) are not in immediate contact with tone languages, while those that are (e.g. Assamese, Nepali, Kashmiri) have not developed such features at all.' A language-internal development is therefore the more likely explanation. So far, the best account that I have seen is that of Purcell, Villegas, and Young (1978) who link the Panjabi tones to the allotonic effects of voiced aspirates and voiced h in Hindi. Initial voiced aspirate or in Hindi correlates with lowered F0 on the following syllable. This in turn corresponds to the Panjabi rising tone. Evidently, the contour rise from the initial lowered F0 to the following neutral pitch has become reinterpreted as a rising tone in Panjabi, after the loss of the conditioning environment, voiced aspiration. Hence Hindi / ghora/ 'horse* corresponds to Panjabi /korä/. Non-initial Hindi voiced aspirate or h also correlates with lowered F0, but on the preceding vowel, and the degree of lowering is not as great. The Panjabi correlate is a falling tone. As I argue elsewhere (Hock 1986/1991:98-100), this falling tone is best explained as the result of polarization vis-ä-vis the rising tone, again after the loss of the voiced-aspiration conditioning environment. Hence Hindi /korhi/ 'leper', with a different suffix, corresponds to Panjabi /korä/. Both falling and rising tones then contrast with the neutral tone of forms without earlier voiced aspiration or A, as in Hindi /korä/ Svhip': Panjabi /korä/. As observed in Hock (1986/1991: 98), this analysis of Panjabi tonogenesis makes a significant contribution to tonogenesis studies in general, in that it confirms the view that both preceding and following consonants can affect tonal quality, but that the effect of preceding consonants is more robust than that of following consonants.8 In a number of articles, Bh. Krishnamurti has demonstrated that the study of sound changes in certain Dravidian languages can make significant contributions to theories about the nature of sound change. Especially well known is his early investigation (1977, 1978) of regional Dravidian 'apical displacement', i.e. the metathesis of word-initial vowel + alveolar or retroflex consonant which introduces alveolars and retroflexes into word-initial position—a position in which they were not permitted earlier. Krishnamurti argues that the manner in which the change has been unfolding lends support to Wang's theory of lexical diffusion, rather than to Labov's approach. While this is a possible conclusion, it must be kept in mind that metathesis is a notoriously irregular process and might therefore, in Labov's framework, be expected to proceed less sweepingly and systematically than many other types of linguistic change. What may in the long run turn out to be more significant is that, by focusing on this particular process, Krishnamurti has brought to our attention a change which is unusual enough to raise interesting questions for general

Reports m 227 historical linguistics. Moreover, it would be interesting to investigate more fully the implications of this change for South Asian convergence, especially in light of the fact that Indo-Aryan permitted initial retroflex obstruents much earlier, starting as early as Vedic sat 'six* and related words. Of even more direct relevance to South Asian convergence studies is a 1991 paper by Krishnamurti which argues that, like Middle Indo-Aryan, a certain historical stage of Dravidian was subject to a two-mora conspiracy which eliminated trimoraic syllables of the type CVC in favor of bimoraic CV or CVC. Findings of this sort suggest that South Asian convergence is a process continuing over many centuries and millennia and that the historical periods may offer a field for convergence studies which is at least as interesting and fruitful as the more controversial and difficult prehistoric period. Krishnamurti also is the author of a more recent study on "The origin and evolution of primary derivative suffixes in Dravidian'. (1997) In this paper Krishnamurti establishes that many of the Dravidian 'roots' listed in Burrow and Emeneau 1961 and 1984 can be derived from other, simpler roots if they are analyzed as reinterpretations of original root + suffix combinations. This explains, e.g., the coexistence of the root *o- 'be, happen' with *aku- 'make', *api- 'become', and *an-/an- 'be in place'. The significance of this work goes beyond etymologically relating such sets of roots to each other; it opens up completely new avenues for the entire reconstruction of Proto-Dravidian, including perhaps even the historical phonology of alveolar and retroflex consonants, if my speculations in Hock (1996b) are on the right track. Moreover, it lends indirect support to theories (e.g. Benveniste 1935) which derive ProtoIndo-European roots of the format CVC(V)C-from simpler CVC roots by similar suffix conglutination. On the syntactic side, Sanford Steever's Serial verb formation in the Dravidian languages (1988) constitutes an important contribution. Its somewhat misleading title notwithstanding, this is an important reexamination of the often-heard claim that Dravidian languages permit only one finite verb per sentence. As Steever shows, the claim runs into a number of difficulties, including Old Tamil serial-verb constructions of the type 5, with two juxtaposed finite verbs in the same clause and morphologically agreeing with each other, as well as the finite verbs of embedded quotative constructions, such as the Tamil structure in 6, with finite vantän 'he has come' of the embedded citation coexisting with finite ninaikkiren think' within the same sentence. Steever plausibly proposes that the quotative marker enru and similar marking devices serve to 'shield' the finite verbs of direct discourse from the rule which permits only one finite verb per sentence. As for the serial-verb type 5, he suggests that only the second of the two verbs is syntactically finite, and that the first verb has finite marking by morphological agreement. He further notes that serial verbs are archaic constructions which tend to be replaced by other structures in southern Dravidian but are more faithfully preserved in geographically more northern Dravidian. Finally, following Lakshmi Bai

228 · The Yearbook (2000)

(1985) and Ramasamy (1981), Steever argues that relative-correlative constructions are indigenous in Dravidian and not due to Indo-Aryan influence (as has often been claimed, e.g. Nadkarni [1970], Sridhar [1978; 1981]), and that here again, northern Dravidian languages tend to preserve such structures more faithfully. (5) celvem allem go-fut.-lpl./neg.-lpl. We will not go (6) nän [[avan inke vantän] enru] ninaikkiren I-sg.nom/he-sg.nom/here/come-past-3sg./quotativemarker/thinkpres.-lsg. think [(that) [he has come here]]' In a 1988 review article of Steever's important monograph I examined data from a large variety of Dravidian languages which suggested modification of Steever's rule governing fmiteness in early Dravidian. As a consequence I felt it necessary to argue that fmiteness in early Dravidian was governed by a Rule A formulated as follows:9 Rule A: One finite verb is permitted per sentence. However, in conjoined structures, each of the conjoined clauses is treated as a sentence in its own right and is therefore permitted to have its own finite verb. Relativecorrelatives syntactically are conjoined structures, permitting each of the two sub-clauses to have its own finite verb. In addition, cited discourse is treated as syntactically independent and thus entitled to its own finite verb(s). Elsewhere, i.e., in structures with syntactic embedding, only one finite verb is permitted per sentence, occurring in the 'matrix' clause; other verbs have to appear in non-finite shape (participles, absolutives, infinitives, or verbal nouns). Notice that under this hypothesis, Steever's "serial verb' constructions are considered conjoined in origin, and their double finite verbs thus to be licensed by Rule A.10 Whether the original formulation by Steever or my later modification is accepted, Steever's monograph presents the most systematic study to date into the question of fmiteness in Dravidian, and his conclusions provide significant challenges to the overly simplified traditional hypothesis of Dravidian as obeying a 'Strict OV Constraint', with only one finite verb permitted per sentence. Linguists interested in historical Dravidian syntax, general linguistic typology, or South Asian convergence would ignore these challenges at their own risk. Many other important studies of issues in South Asian historical linguistics have been published and will no doubt be published in the future. What has

Reports m 229 been most significant, however, in South Asian historical linguistics, and probably will continue to be so, is the issue of linguistic convergence—what are the features that define the South Asian convergence area and how did they come about in the different languages and language families? This is an issue of equal importance to South Asian historical linguistics, general historical and Indo-European linguistics, and general studies of language contact.11 Moreover, linguists who have staked their claim on this issue are among the most celebrated in South Asian linguistics, including Suniti Kumar Chatterji (1926), Madhav Deshpande (1979a), Murray B. Emeneau (1954; 1956; 1974; 1980), F.B.J. Kuiper (1967; 1991), Asko Parpola (e.g. 1988), and Franklin Southworth (1974; 1979). In addition, South Asian convergence is one of the major topics covered in Thomason and Kaufman's important and widely-read monograph on Language contact, creolization, and genetic linguistics (1988). As early as 1833, Pott considered the dental: retroflex contrast of Sanskrit to be at least partly due to the influence of the 'autochthonous' languages (ibid.: 78), later specifying these as Dravidian. Beginning with Campbell's comparative grammar of Dravidian, other Sanskrit features have also been claimed to result from Dravidian substratum influence or, as I now prefer to call it, Dravidian subversion. Over time, the most important of these turn out to be SOV word order and the use of absolutives (of the type Skt. krtvä 'having done') and of quotative marking (Skt. iti) after cited discourse. Since all of these features—retroflexion, SOV order, absolutives, and quotative iti—are present in the earliest attested stage of Indo-Aryan, in Rig-Vedic Sanskrit, it is assumed that Dravidian subversion took place in prehistoric, pre-Vedic times. The case for prehistoric Dravidian subversion of Sanskrit/Indo-Aryan in northwestern India has been considered supported by a number of other factors. First, the presence of modern Brahui in the northwest has been taken to support an early Dravidian presence in the area, on the assumption that Brahui is a relic of an originally much more widespread Dravidian presence. Second, a number of lexical items found in the Rig-Veda have been attributed to Dravidian subversion. While some linguists (e.g. Southworth) tend to find massive evidence for Dravidian lexical items, Emeneau has been more cautious and has claimed that only a few items are really probative. The two most relevant items in his opinion are the word for 'peacock', mayura- (compare Tamil and Malayalam may/7), and budbuda- 'bubble' (compare Telugu budabuda 'with a bubbling noise', Tulu budubudu 'in drops' and non-reduplicating Kannada buda, budu, etc.). Underlying the subversionist account is the assumption of unilateral influence of Dravidian on Indo-Aryan and of a highly unequal prehistoric social relationship between Indo-Aryans and Dravidians. Dravidian speakers therefore were required to speak Indo-Aryan and, just as happened to English in modern South Asia, in shifting to Indo-Aryan they transferred structural features of their own language(s).

230 · The Yearbook (2000)

Finally, both Emeneau and Thomason and Kaufman claim that the Dravidian subversion hypothesis is the simplest account for the presence of retroflexion and the other features in both Vedic Sanskrit and Dravidian and therefore should be preferred to other, alternative hypotheses. Though the Dravidian subversion hypothesis has been widely accepted, from the very beginning a minority of linguists have questioned the cogency of the evidence adduced in its favor. The earliest scholar to have expressed such doubts seems to have been Bühler (1864). Jules Bloch, too, has on occasion expressed misgivings (e.g. Bloch 1925:16,29).12 And beginning in 1975,1 have repeatedly examined earlier arguments pro and con and concluded that the evidence is not sufficiently strong to establish the Dravidian substratum hypothesis beyond a reasonable doubt. (See Hock [1975] [1984] [1996a], and the literature cited there.) For instance, the lexical evidence of mayura and budbuda turns out to be problematic. While Tamil/Malayalam mayil may look very similar to mayura, so do Munda moral, and Eastern Iranian Saka mur-äsa. On the other hand, the more northern Dravidian languages have forms like Parji mo/7/7, which look much less similar to the Sanskrit word. Moreover, there are alternate forms such as Konda mrilu, some of which look very similar to words for 'cat', such as Kuwi S. mrfyuli. Given that Skt. marjära can mean both 'cat' and 'peacock' and that, moreover, the cry of peacocks has a certain similarity to the meowing of cats, the possibility cannot be excluded that all of these words are onomatopoetic in origin, may therefore have been created independently, and are of no value for attempts to establish borrowing and thus, linguistic contact. (See Hock [1996a] for details.) Onomatopoeia is an even more obvious explanation for RV budbuda 'bubble' and for the Dravidian words of the type budabuda.13 The lexical evidence, thus, is not sufficient to establish prehistoric contact between Dravidian and Sanskrit/Indo-Aryan. The present-day location of Brahui likewise cannot be considered to furnish unambiguous evidence for an early Dravidian presence in northwestern South Asia. As noted in Hock (1996a) (with references to earlier literature), according to their own traditions, the speakers of Kurukh and Malto, the other two North Dravidian languages, came to their present northerly locations from Karnataka, via the Narmada Valley, and this tradition is supported by other evidence. There is thus ample evidence for Dravidian migration to the north, which could also account for the present-day location of Brahui. In fact, the Brahui, too, believe that they have come from the outside, although their claim to have come from Aleppo in Syria must be considered fanciful. Neither of the two arguments for an early Dravidian presence in northwest South Asia, thus, is sufficiently cogent to establish prehistoric Dravidian/IndoAiyan contact in that area. It is of course still possible that the area was settled by Dravidian speakers in prehistoric times and that these speakers either were absorbed by the Indo-Aryans or migrated south; but that is merely one of

Reports m 231 several possibilities. Moreover, such a scenario would be quite complex, violating the subversionist claim that we should prefer the simplest account possible. There are problems, too, with the purely linguistic arguments. As has been pointed out repeatedly, other Indo-European languages, beside Sanskrit/ Indo-Aryan, have developed a dental: retroflex contrast or even a triple contrast dental: alveolar : retroflex (conservative Swedish and Norwegian dialects). The constituent order SOV appears to be a feature inherited from Proto-Indo-European and thus of no relevance. Similar arguments have been made for absolutives and quotatives.14 In fact, I have argued that ProtoDravidian and Proto-Indo-European had very similar syntactic typologies even before contact, which can be captured by our earlier 'Rule A, and that the syntactic similarities between Indo-Aryan and Dravidian thus are not the result of contact but of common inheritance. (See Hock 1996a and the references cited there.) Finally, as observed in Hock (1996a), there are problems both with the argument that the Dravidian subversion hypothesis is the simplest account and with the counterargument that the evidence for Dravidian subversion does not meet the 'reasonable doubt' test. For instance, the fact that both Portuguese and indigenous languages of Brazil have contrasting oral and nasal vowels might a priori be attributed to subversion or convergence, and this account could be claimed to be simpler than the assumption that the contrast arose independently. However, in this case we know that both Portuguese and the indigenous Brazilian languages had the contrast before contact. The 'simpler' contact account thus is historically inaccurate. As for the 'reasonable doubt' test, the major problem is to define how reasonable is reasonable. The major problem lies in the severe limitations on our knowledge of what languages were located where in South Asia during the early second millennium BC and of the structure of these languages at that time. As noted earlier, it is only for Sanskrit that we have direct linguistic evidence for this period, as well as textual evidence which makes it reasonably possible to locate its speakers in geographical space. Even for Dravidian, we lack such evidence. And the fact that the Dravidian languages appear to have undergone significant changes in their root structure (as shown by Krishnamurti [1997]) and in some important aspects of their syntax (see Steever [1988] and/or Hock [1988]) suggests that we need to treat with considerable caution any claims about the structure and vocabulary of Dravidian in the early second millennium BC. The case is even more problematic for Tibeto-Burman and Munda, which are also possible candidates for early contact with Sanskrit/Indo-Aryan; see especially Witzel (1995) for Tibeto-Burman, and recent unpublished work of his for Munda. The existence of isolates such as Burushaski and Nahali in present-day South Asia raises the additional possibility that there may have been a number of other languages in the area, some of which may have

232 · The Yearbook (2000)

disappeared without trace, others perhaps surviving in relic form in Burushaski and/or Nahali. Some of these issues probably can never be resolved, such as the question of whether Indo-Aryan may have been in prehistoric contact with languages other than Dravidian, Munda, or Tibeto-Burman. For other issues, however, it is to be hoped that continued, detailed work on the various language families of South Asia may yield insights that can bring us closer to a resolution. In fact, such detailed work is needed not only for evaluating the controversial issue of when and how South Asian convergence began and how it spread within South Asia, but also for more 'mundane' issues, such as the development of smaller convergence areas (e.g. the Nilgiri area discussed by Emeneau [1989], or the interactions between Dakkhini Urdu and Telugu studied e.g. by Kachru [1986] and Arora and Subbarao [1988; 1989]), or even such issues as the loss or reduction of ergativity in various Indo-Aryan languages (e.g. Stump 1983), or the question of the relationship, if any, between IndoAryan and Iranian, or even Armenian ergativity. As the work of scholars like Krishnamurti and Steever shows, such detailed studies can lead to important new insights, may challenge traditional views, may inspire additional work by other scholars, and thus may, directly or indirectly, enrich our understanding of South Asian and general historical linguistics. South Asia thus continues to hold great significance for historical linguistics and the promise of rewarding insights for those willing to take the challenge of following in the footsteps of scholars such as Btoch, Campbell, Chatterji, Emeneau, Jha, Krishnamurti, Saksena, Steever, Wackernagel, and Zvelebil. • NOTES 1. While the grammatical tradition of the Akkadians to teach Sumerian-as-a-second-language for literacy purposes may go back to an even earlier period (somewhere between 2100 and 1800 B.C.), its goal and its reach is much more modest. 2. Which, in turn, is the source for the early western term 'guttural'. 3. 'The Sanscrit language, whatever be its antiquity, is of a wonderful structure; more perfect than the Greek, more copious than the Latin, and more exquisitely refined than either, yet bearing to both of them a stronger affinity, both in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident; so strong indeed, that no philologer could examine them all three, without believing them to have sprung from some common source, which, perhaps, no longer exists: there is a similar reason, though not quite so forcible, for supposing that both the Gothick and the Celtick, though blended with a very different idiom, had the same origin with the Sanscrit; and the old Persian might be added to the same family, if this were the place for discussing any question concerning the antiquities of Persia.* 4. Grimm's pioneering formulation of the phonetic generalizations capturing the nature of the Germanic consonant shift does not seem to have been influenced by these Prakrit grammars. 5. The final volume, on the verb, is still missing—perhaps reflecting the verb-final syntax of both German and Sanskrit (?). 6. Reference in Bahl (1957).

Reports m 233 7. References in Bhatia (1975). 8. Note that this account only addresses the origin of the tonal contrast. Once in place, the contrast can be analogically extended. Moreover, as was pointed out at the Patiala Conference, foreign borrowings without voiced aspiration (such as Engl. bribe) may be assigned a tone in Panjabi that would normally be associated with earlier voiced aspiration, presumably because that tone most closely matches the intonational contour on the foreign word. 9. See especially Hock (1989). 10. Steever (1993) objects to my account by faulting me for not having expert knowledge in Dravidian linguistics. But this is really beside the point; what matters is the linguistic evidence. Steever's bibliographical reference suggests that he was only aware of my review in Language, not of my review article, which provides extensive evidence from a large variety of Dravidian languages in support of my proposal. 11. It is also an issue of great political significance in present-day India, in that different identity movements, especially the Dravidianist movement of Tamil Nadu and the Hindutva movement of the 'Sangh Parivar', embrace different interpretations of the evidence in their attempts to establish themselves as the original inhabitants of the Indian subcontinent. (On this issue see e.g., Hock [In Press a]). 12. He later changed his opinion; see e.g. Bloch (1929). 13. Moreover, note the difference between the dental of the Rig-Vedic word and the retroflex of the Dravidian words. If Indo-Aryan retroflexion had indeed resulted from Dravidian subversion, one would expect a borrowing from Dravidian to preserve the retroflex. 14. For the quotatives, however, the situation may be more complex, with true Dravidian and Indo-Aryan quotatives both being innovated; see Hock (1999). • REFERENCES Abbi, Anvita. 1997. Debate on archaism of some select Bangani words. Indian Linguistics 58.1-14. Anderson, Stephen R. 1977. On mechanisms by which languages become ergative. Mechanisms of syntactic change, ed. by Charles Li, 317-63. Austin: University of Texas Press. Arora, Harfoir and K.V. Subbarao. 1988. Convergence and syntactic change: The case of the conjunctive participle in Dakkhini Hindi-Urdu. Papers from the Conference on the Theory and Practice of Historical Linguistics. Chicago: Linguistic Society. . 1989. Convergence and syntactic reanalysis: The case of so in Dakkhini. Studies in the Linguistic Sciences 19 (1). 1-18. Bahl, Kali Charan. 1957. Tones in Panjabi. Indian Linguistics 17.139-147. Beekes, Robert S.P. 1995. Comparative Indo-European linguistics. Amsterdam: Benjamins. Benveniste, Emile. 1935. Origines de la formation des noms en indo-europeen, 1. Paris: AdrienneMaisonneuve. Berger, Hermann. 1998. Die Burushaski-Sprache von Hunza und Nager. (In three parts: Grammatik; Texte mit Übersetzungen; Wörterbuch.) Wiesbaden: Harrassowitz. Bhatia, Tej K. 1975. The evolution of tone in Punjabi. Studies in the Linguistic Sciences 5 (2). 1224. Bloch, Jules. 1920. La formation de la langue marathe. Paris: Champion. Engl. transl. by D.R. Chanana: The formation of the Marathi language. Delhi: Motilal Banarsidass. . 1925. Sanskrit et dravidien. Bulletin de la Socitte de Linguistique de Paris 25. 1-21. . 1929. Some problems of Indo-Aryan philology. Bulletin of the School of Oriental Studies (London) 5. 719-56. Bloomfleld, Leonard. 1933. Language. New York: Holt, Rinehart &Winston. . 1939. Menomini morphophonemics. Etudes phonologiques dediees a la momoire de N.S. Trubetzkoy. Travaux du Cercle Linguistique de Prague 8. 105-11. Bühler, George. 1864. On the origin of the Sanskrit \ingua\s.MadrasJoumalof Literature and Science, No. I, Third Series, 116-86. [I have not been able to see this article; reference in Wackernagel 1896: 165 and elsewhere.]

234 · The Yearbook (2000) Burrow, Thomas and S. Bhattacharya. 1953. The Parji language: A Dravidian language ofBastar. Hertford: Stephen Austin &Sons. . 1970. The Pengo language: Grammar, texts, and vocabulary. Oxford: Clarendon Press. Burrow, Thomas and Murray B. Emeneau. 1961. A Dravidian etymological dictionary. Oxford: Oxford University Press. . 1984.X Dravidian etymological dictionary. 2nd ed. Oxford: Oxford University Press. Caldwell, Robert 1856. A comparative grammar of the Dravidian or South Indian family of languages. First edition. London. (Third edition repr. 1974, New Delhi: Oriental Books.) Chatterji, Suniti Kumar. 1926. The origin and development of the Bengali language. Calcutta University Press. Repr. 1970, London: Allen &Unwin; distributed by Motilal Banarsidass, Delhi. Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton. Cardona, George. 1986. Phonology and phonetics in ancient Indian works: The case of voiced and voiceless elements. South Asian languages: Structure, convergence, and diglossia, ed. by Bh. Krishnamurti et al., 60-80. Delhi: Motilal Banarsidass. Collinge, Neville E. 1985. The laws of Indo-European. Amsterdam/Philadelphia: Benjamins. Comrie, Bernard. 1978. Ergativity. Syntactic typology, ed. by W.P. Lehmann, 329-94. Austin: University of Texas Press. Deshpande, Madhav. 1979 a. Genesis of Rgvedic retroflexion: A historical and sociolinguistic investigation. Aryan and non-Aryan in India, ed. by M. Deshpande and P.E. Hook, 235-315. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. . 1979 b. Sociolinguistic attitudes in India: A historical reconstruction. Linguistica Extranea, Studia 5. Ann Arbor, MI: Karoma. . 1993. Sanskrit and Prakrit: Sociolinguistic issues. Delhi: Motilal Banarsidass. Dundas, Paul. 1996. Jain attitudes toward the Sanskrit language. Ideology and status of Sanskrit: Contributions to the history of the Sanskrit language, ed.byJ.E.M. Houben, 137-56. Leiden: Brill. Ellis, Francis Whyte. 1816. Note to the introduction. A grammar of the Teloogoo language, by A.D. Campbell. Madras. Emeneau, Murray B. 1954. Linguistic prehistory of India. Proceedings of the American Philosophical Society 98. 282-92. (Reprinted in Emeneau 1980.) . 1955. KolamL· A Dravidian language. (University of California Publications in Linguistics, 12.) Berkeley: University of California Press. . 1956. India as a linguistic area. Language 32.3-16. (Reprinted in Emeneau 1980.) . 1974. The Indian linguistic area revisited. International Journal of Dravidian Linguistics 3. 92-134. (Reprinted in Emeneau 1980.) . 1980. Language and linguistic area: Essays selected by A.S. Dil. Stanford, CA: University Press. . 1984. Toda grammar and texts. (American Philosophical Society Memoir Series, 155.) Philadelphia: American Philosophical Society. . 1989. The language of the Nilgiris. Blue Mountains: The ethnography and biogeography of a South Indian region, 133-43. (Reprinted in Emeneau 1994.) . 1994. Dravidian studies: Selected papers, ed. by Bh. Krishnamurti. Delhi: Motilal Banarsidass. Hahn, Ferd. 1911. Kurukh grammar. Calcutta: Bengal Secretariat. Repr. (no date), Delhi: Mittal. Hock, Hans Henrich. 1975. Substratum influence on (Rig-Vedic) Sanskrit? Studies in the Linguistic Sciences 5 (2). 76-125. . 1984. (Pre-) Rig-Vedic convergence of Indo-Aryan with Dravidian? Another look at the evidence. Studies in the Linguistic Sciences 14 (1). 89-107. . 1985. Transitivity as a gradient feature: Synchronic and diachronic evidence from IndoAryan, especially Sanskrit. Proceedings of the Conference on Participant Roles in South Asia and Adjacent Areas, ed. by A.K. Zide et al., 247-63, Bloomington: Indiana University Linguistics Club

Reports m 235 Hock, Hans Henrich. 1986. 'P-oriented' constructions in Sanskrit. South Asian languages: Structure, convergence, anddiglossia, cd. by Bh. Krishnamurti et al., 15-26, Delhi: Motilal Banarsidass. . 1986/1991. Principles of historical linguistics. 1st ed. 1986, 2nd ed. 1991. Berlin: Mouton de Gruyter. . 1988. Finiteness in Dravidian: Review article on Steever 1988. Studies in the Linguistic Sciences 18 (2). 211-31. . 1989. Conjoined we stand: Theoretical implications of Sanskrit relative structures. Studies in the Linguistic Sciences 19 (1). 93-126. . 1992. Were r and / velar in early Sanskrit? Viaya-Vratin: Professor A.M. Ghatage felicitation volume, ed. by VN. Jha, 69-94. (Sri Garib Dass Oriental Series, 160.) Delhi: Sri Satguru Publications. . 1996a. Pre-Rgyedic convergence between Indo-Aryan (Sanskrit) and Dravidian? A survey of the issues and controversies. Ideology and status of Sanskrit: Contributions to the history of the Sanskrit language, ed. by J.E.M. Houben, 17-58. Leiden: Brill. . 1996b. Subversion or convergence? The issue of pre-Vedic retroflexion reconsidered. Studies in the Linguistic Sciences 23 (2). 73-115. . 1997. On Bangani. Indo-European list ([email protected]), 24 Feb. 1997. . 1999. Quotatives, quotativals, and the question of prehistoric convergence in South Asia. Second International Conference on South Asian Languages, Punjabi University, Patiala, January 1999. . In Press (a). Through a glass darkly: Modern 'racial' interpretations vs. textual and general prehistoric evidence on arya and däsa/dasyu in Vedic society. Proceedings of the Conference on Aryan and non-Aryan in South Asia: Evidence, Interpretation, and ideology (October 1996), ed. by Johannes Bronkhorst and Madhav Deshpande. . In Press (b). Out of India? The linguistic evidence. Proceedings of the Conference on Aryan and non-Aryan in South Asia: Evidence, Interpretation, and Ideology (October 1996), ed. by Johannes Bronkhorst and Madhav Deshpande. Israel, Michael. 1979. A grammar of the Küvi language. Trivandrum: Dravidian Linguistics Association. Jha, Subhadra. 1958. The formation of the Maithili language. London: Luzac. Repr. 1985, Delhi: Munshiram Manoharlal. Jones, Sir William. 1786. The third anniversary discourse, on the Hindus. Published 1788 in Asiatick Researches 1. Kachru, Yamuna. 1986. The syntax of Dakkhini: A study in language variation and change. South Asian languages: Structure, convergence, and diglossia, cd. by Bh. Krishnamurti et al., 165-73. Delhi: Motiial Banarsidass. KJaiman, M.H. 1978. Arguments against a passive origin of the IA crgative. Papers from the 14th Regional Meeting of the Chicago Linguistic Society, 204-16. Krishnamurti, Bh, 1961. Telugu verbal bases: A comparative and descriptive study. Berkeley and Los Angeles: University of California Press. . 1969. Konda or Kübi: A Dravidian language. Hyderabad: Tribal Culture Research and Training Institute. . 1977. Sound change: Shared innovation vs. diffusion. Phonologica 1977. 205-11. (Supplement to Innsbrucker Beiträge zur Sprachwissenschaft.) Innsbruck. . 1978. Areal and lexical diffusion of sound change. Language 54.1-20. . 1991. The emergence of the syllable types of stems (C)VCC(V) and (C)VC(V) in IndoAryan and Dravidian: Conspiracy or convergence? Studies in the historical phonology of Asian languages, ed. by W.G. Boltz and M.C. Shapiro, 160-75. Amsterdam and Philadelphia: Benjamins. . 1997. The origin and evolution of primary derivative suffixes in Dravidian. Historical, IndoEuropean, and lexicographical studies: A festschrift for Ladislav Zgusta on the occasion of his 70th birthday, ed. by H.H. Hock, 87-115. Berlin: Mouton de Gruyter.

236 · The Yearbook (2000) Kuiper, F.B J. 1962. NahalL· A comparative study. (Mededelingen der Koninklijke Nederlandse Akademie van Wetenschappen, Aid. Letterkunde, N.R., 25:5.) Amsterdam: NoordHollandsche Uitgevers Maatschappij. . 1967. The genesis of a linguistic area. Indo-Iranian Journal 10. 81-102. (Repr. in InternationalJoumalofDravidian Linguistics 3.135-53 [1974].) . 1991. Aryans in the Rigveda. (Leiden Studies in Indo-European, 1.) Amsterdam and Atlanta, GA: Rodopi. Lakshmi Bai, B. 1985. Some notes on correlative constructions in Dravidian. For Gordon H. Fairbanks, ed. by V.Z. Acson and R.L, Leed, 181-90. Honolulu: University of Hawaii Press. Masica, Colin P. 1991. The Indo-Aryan languages. Cambridge: Cambridge University Press. Misra, Sarya Swamp. 1992. The Aryan problem: A linguistic approach. New Delhi: Munshiram Manoharlal. Nadkarni, Mangesh V. 1970. NP embedded structures in Kannada and Konkani. Los Angeles: UCLA Ph.D. dissertation in Linguistics. Parpola, Asko. 1988. The coming of the Aryans to Iran and India and the cultural and ethnic identity of the däsas. International Journal of Dravidian Linguistics 17 (2). 85-229. Pischel, Richard. 1900. Grammatik der Präkrit-Sprachen. (Grundriß der indo-arischen Philologie und Altertumskunde, I: 8.) Engl. transl. by S. Jha: Comparative grammar of the Prakrit languages. 2nd rev. ed. 1981, Delhi: Motilal Banarsidass. Pott, August Friedrich. 1933, 1836. Etymologische Forschungen auf dem Gebiete der indogermanischen Sprachen, l and 2. Lemgo: Meyer. Purcell, E.T., G. Villegas, and S.P. Young. 1978. A before and after for tonogenesis. Phonetica 35. 284-93. Ramasamy, K. 1981. Correlative relative clauses in Tamil. Dravidian syntax, ed. by S. Agesthialingom and N. Rajasekharan Nair, 363-80. (Annamalai University Publications in Linguistics, 73.) Annamalainagar: Annamalai University. Saxena, Baburam. 1937. Evolution ofAwadhi. 2nd ed. 1971. Delhi: Motilal Banarsidass. Schlegel, Friedrich von. 1808. Ober die Sprache und Weisheit der Indien Ein Beitragzur Begründung derAUhertumskunde. Heidelberg: Mohr & Zimmer. Southworth, Franklin C. 1974. Linguistic stratigraphy of North India. International Journal of Dravidian Linguistics 3.201-23. . 1979. Lexical evidence for early contacts between Indo-Aryan and Dravidian. Aryan and non-Aryan in India, ed. by M. Deshpande and P.E. Hook, 191-233. Ann Arbor: Center for South and Southeast Asian Studies, University of Michigan. Sridhar, S.N. 1978. Linguistic convergence: IndoAryanization of Dravidian languages. Studies in the Linguistic Sciences 8 (1). 197-215. . 1981. Linguistic convergence: Indo-Aryanization of Dravidian languages. Lingua 53.199220. Steever, Sanford B. 1988. The serial verb formation in the Dravidian languages. Delhi: MotUal Banarsidass. . 1993. Analysis to synthesis: The development of complex verb morphology in the Dravidian languages. New York: Oxford University Press. Stump, Gregory T. 1983. The elimination of ergative patterns of case-marking and verbal agreement in modern Indie languages. Papers in historical linguistics, ed. by G.T. Stump, 140-64. (Ohio State University Working Papers in Linguistics, 27.) Subrahmanyam, P.S. 1968. A descriptive grammar of Gondi. Annamalainagar: Annamalai University. . 1983. Dravidian comparative phonology. Annamalainagar: Annamalai University. Tagare, Ganesh Vasudev. 194S. Historical grammar ofApabhramsO. (Deccan College Dissertation Series, 5.) Poona. Repr. 1987, Delhi: Motilal Banarsidass. Tbomason, Sarah Grey, and Terrence Kaufman. 1988. Language contact, creoUzation, and genetic linguistics. Berkeley and Los Angeles: University of California Press.

Reports m 237 van Driem, George, and Suhnu R. Sharma. 19%. In search of Indo-Europeans in the Himalayas. Indogermanische Forschungen 101.107-46. Wackernagel, Jakob. 1896· .Altindische Grammatik. Göttingen: Vandenhoeck & Ruprecht. Witzel, Michael. 1995. Early Indian history: Linguistic and textual parametres. The Indo-Aryans of ancient South Asia: Language, material culture, and ethnicity, ed. by G. Erdosy, 85-125. Berlin and New York: de Gruyter. Zoller, Peter. 1988. Bericht über besondere Archaismen im Bangani, einer Western PahariSprache. Münchener Studien zur Sprachwissenschaft 49.173-200. . 1993. A note on Bangani. Indian Linguistics 54.112-14. Zvelebil, K. 1970. Comparative Dravidian phonology. The Hague: Mouton.

I SOUTH-EAST ASIA AND THE SOUTH PACIFIC I FRANCE MUGLER The geographical area loosely designated here as South-East Asia and the South Pacific has two main foci of interest for linguists specializing in Indian languages. One is Singapore and Malaysia, a traditional area of migration from India, particularly south India, and home to a number of Indian languages, Tamil being the most prominent. The other, in the South Pacific, is the Fiji Islands, which at the turn of the century saw the arrival of Indian indentured laborers, and where the main language among Indo-Fijians (or 'Fiji Indians') is a variety of Hindi. The presence of Indian languages, of the two largest language families of India, Indo-Aryan and Dravidian, has added to the already extremely rich multilingual nature of societies in these two neighboring regions, which also have in common indigenous languages of the Austronesian family: Bahasa Malaysia in Singapore and Malaysia, Fijian in Fiji and other Oceanic languages in many other island nations of the South Pacific. The region is also home to a number of universities, many of which have at least a linguist or two—sometimes members of small South Asian language programs, sometimes of linguistics or other departments—who are involved in research on Indian languages, including but not limited to those spoken in their country of residence. This is true of universities in Malaysia and Singapore, of the University of the South Pacific, in Fiji, and of a few universities in Pacific Rim countries, such as New Zealand and Australia. In the Yearbook 1998, the report for this region concentrated on Southeast Asia and surveyed the work published or presented in the previous five years (Saleemi 1998). This included in particular research by David Gil, K.P. and T. Mohanan, Mukherjee, Nihalani, and Saleemi. Research on Indian languages in Southeast Asia (essentially Malaysia and Singapore) has generally been of two kinds. On one hand a number of South Asian scholars who are

238 · The Yearbook (2000)

based in one of these two countries have been working on various Indian languages, chiefly Hindi, Urdu, Punjabi, Malayalam, Sindhi, and Sanskrit. Most of that work has been in syntax, although phonetics and sociolinguistics are also represented. As for research on the Indian languages spoken in the two countries, it has focused on Tamil, by far the most important Indian language in the region in terms of speakers, and generally has concentrated on sociolinguistic or language planning issues and in particular the use of Tamil— along with Malay, Mandarin, and English—in education. Gil (1995) is an example of the first kind of research. It is a cross-linguistic study of universal quantifiers which includes Malayalam and Punjabi, along with English, Georgian, Hebrew, Japanese, Maricopa, Russian, Tagalog, and Warlpiri. Gil surveys universal quantifiers and analyses their meaning and their syntactic and morphological properties. He argues that these quantifiers vary according to the property of distributivity and proposes a number of linguistic universals. Kwan-Terry and Luke (1997) is an example of research on sociolinguistic issues related to language policy and use in education, with particular reference to Singapore and Malaysia, although China and Hong Kong are also examined. It examines the variability in notions of literacy, vernacular, and modern tongue. The authors point out, for instance, that in the complex multilingual setting of Singapore and Malaysia, it is sometimes difficult to establish a person's mother tongue. In Singapore, Tamil is one of the recognized 'symbolic mother tongues', along with Malay and Mandarin, which are encouraged by the government as part of an effort to maintain the cultural traditions of the three major ethnic groups, and all three languages are used in primary education in public schools. In Fiji and the South Pacific, most research on Indian languages has been on Fiji Hindi, until recently. Jeff Siegel, whose Language contact in a plantation environment (1987) is the classic sociolinguistic history of the development of Fiji Hindi, has continued his research. In "Indian languages in Fiji: Past, present and future', he discusses the development of Fiji Hindi, its relationship with Standard Hindi, the current use of Indian languages in Fiji, language attitudes, and prospects for language maintenance. A volume containing the proceedings of the sessions on language contact at the Second International Conference on Oceanic Languages, held at the University of the South Pacific in Fiji's capital, Suva, in 1996, (Tent and Mugler 1998) features three articles related to Fiji Hindi: Arms (1998) explores some phonological and grammatical characteristics of Fiji Hindi, including the elimination of certain suppletive morphemes and the disappearance of the distinction between the present simple and continous forms, and the past habitual and continuous forms. Mugler and Tent report on a sociolinguistic survey of language use and attitudes in Fiji. In spite of the high status of English, the ex-colonial language, which is the only official medium of instruction in schools, co-official language

Reports m 239 along with Fijian and Hindi, and dominates most of the media—not to mention its ubiquitous and much hyped pressure as 'global' language—both Fijian and Fiji Hindi are overwhelmingly used for most everyday functions by their respective speakers. The study also points to the paradox that Fiji Hindi still suffers from very low prestige in spite of being extensively used, while 'Shudh' or 'Standard' Hindi has a high status but is hardly ever used. Although there are correlations between age and education on one hand, and positive attitudes towards English on the other, English is not displacing Fiji Hindi (or Fijian), but is merely added to people's repertoire. The survey also reports on the continuing decline of Dravidian languages in Fiji. The shrinking of domains and loss of speakers by Tamil, Telugu, and Malayalam probably started very early on, for two reasons: indentured laborers started being recruited from south India only about 25 years after their north Indian counterparts, and south Indians have always been a minority in Fiji—about a quarter of all indentured laborers. The survey documents a continuing loss of speakers with each generation, to the point where, if current trends continue, Malayalam and Telugu are likely to disappear within a generation while Tamil, which has always had more speakers, may survive a bit longer, although it is losing speakers at the same rate as the other two Dravidian languages. Research on Dravidian languages in Fiji expanded, after the survey, to an analysis of the nature and structure of the varieties used in Fiji. This work has been conducted in collaboration with linguists from the Central Institute of Indian Languages in Mysore, India. An analysis of recorded data indicates that the three Fiji varieties share characteristics which one might expect of such transplanted and threatened languages: retention of archaisms, of regional dialectal features which help to identify the provenance in India of their speakers' ancestors, influence of the languages they have been in contact with in their new environment in Fiji, including Fiji Hindi, English, and Fijian (in that order of importance), and of a number of processes of morphosyntactic simplification which are likely to be the result of attrition, i.e., shrinking of domains, rare use and limited competence among many of the remaining speakers (Mugler forthcoming). Further research on Tamil, with Sam Mohan Lai, has revealed an influence on the script on those speakers who are literate in Tamil, leading to some reading pronunciations and the use of'high' forms, albeit alongside the features identified above as typical of the Fiji variety (Mugler and Mohan Lai, forthcoming). In recent decades and especially after the 1987 military coups, Indo-Fijians have left Fiji in significant numbers, migrating mainly to Canada, the USA and, closer to home, Australia and New Zealand. Shameem, whose Ph.D. thesis was on language maintenance among the Wellington (New Zealand) IndoFijians and is currently based at the University of Auckland, has continued her research on this immigrant community. In her article (1998) she reports on a performance test designed to validate self-reported first language proficiency

240 · The Yearbook (2000)

in Fiji Hindi among a group of Indo-Fijian teenagers in Wellington. In spite of the well-known problems associated with self-report studies, the results of the test correlated with the self-report data, demonstrating the validity of the selfreport scale. This was truer of aural than of oral data, however, and respondents tended to judge their oral proficiency at a level slightly higher than that where they were placed according to test results. In Australia, universities have been facing funding cuts for a decade or so, and the few South Asian language programs, typically small and with low enrolments, have suffered. Nonetheless, Sanskrit and Hindi are still taught both at the Australian National University (ANU) in Canberra, and Latrobe University in Victoria. Some scholars and students in South Asian studies or linguistics work on Indian languages. At La Trobe, for instance, David Bradley works on Tibeto-Burman languages. Stephen Morey, a Ph.D. student in linguistics at Monash University, has been collecting stories, songs and other texts in four Thai languages spoken in Assam—Khamti, Phake, Aiton, and Khamyang—the last two of which are endangered. He is also planning to write grammars, and has started designing software to represent and print the scripts. Peter Hohn Brown recently completed a Ph.D. thesis at La Trobe on the auxiliary verbs of Kannada, and Alec Coupe, of ANU, for his MA thesis, wrote a phonetic and phonological description of Ao, a Tibeto-Burman language of Nagaland. There is a great deal of scope for further research in Asian languages in Southeast Asia and the South Pacific. Both regions are rife with linguistic diversity and are a linguist's paradise. Scholars in this part of the world are very unlikely to run out of research topics any time soon, and further work is needed in just about any area of linguistics, although one might wish in particular for more research on the homegrown varieties of Indian languages used in the region. • REFERENCES Anns, David. 1998. Tendencies in Fiji Hindi. Proceedings of the Second International Conference on Oceanic Linguistics: Vol 1, Language Contact, ed. by Jan Tent and France Mugler, C-141:10. Canberra: Pacific Linguistics, ANU. Gil, David. 1995. Universal quantifiers and distributivity. Quantification in nature language, Vol 1 ed. by Emmon Bach, Eloise Jelinek, Angelika Kratzer and Barbara H. Partee, 321-62. Dordrecht, Netherlands: Kluwer. Kwan-Terry, Anna and K.K. Luke. 1997. Tradition, trial and error: standard and vernacular literacy in China, Hong Kong, Singapore and Malaysia. Vernacular literacy: A re-evaluation ed. by Andr£e Tabouret-Keller, Robert B. Le Page, Penelope Gardner-Chloros and Gabrielle Varo, 271-315. New York: Oxford University Press. Mugler, France. Forthcoming. Dravidian languages in Fiji: survival and maintenance. Proceedings of the South Asian Language Analysis (SALA) Eighteenth Roundtable, Jawaharlal Nehru University, New Delhi, India, 6-8 January 1997. Mugler, France and Jan Tent 1998. Some aspects of language use and attitudes in Fiji. Proceedings of the Second International Conference on Oceanic Linguistics: Vol. I, Language Contact, ed. by Jan Tent and France Mugler, C-141:109-34. Canberra: Pacific Linguistics, ANU.

Reports m 241 Mugler, France and Sam Mohan Lai. Forthcoming. Tamil in Fiji. International Journal of Dravidian Linguistics. Saleemi, Anjum. 1998. Southeast Asia. The yearbook of South Asian languages and linguistics ed. by Rajendra Singh, 202-6 New Delhi: Sage Publications. Sbameem, Nikhat. 1995. Hamai log ke boli: Language shift in the Wellington Indo-Fijian community. Wellington: Victoria University, Ph.D. thesis. . 1998. Validating self-reported language proficiency by testing performance in an immigrant community: the Wellington Indo-Fijians. Language Testing 15 (1). 86-108. Siegel, Jeff. 1987. Language contact in a plantation environment. A socioUnguistic history of Fiji. Cambridge: Cambridge University Press, . Forthcoming 1998. Indian languages in Fiji: Past, present and future. South Asia, ed. by Howard Brasted. Tent, Jan and France Mugler (eds). 1998. Proceedings of the Second International Conference on Oceanic Linguistics: Vol 1, Language Contact, C-141. Canberra: Pacific Linguistics, ANU.

Review Article Substantiating Blasphemous Claims for the Belabored Nativity of these other Englishes A Review of The Native Speaker: Multilingual Perspectives edited by Rajendra Singh. New Delhi: Sage Publications, 1998, pp. Price Rs. 375 Reviewed by Arjuna Parakrama, University of Colombo. This, the fourth volume in the Language and Development Series of Sage explores an important and theoretically vexed issue on the "native speaker', using perspectives and insights derived from the experience of language contexts in which English, of whatever kind, coexists with other languages in multilingual settings. Dispensing with matters of form first, it should be noted that the volume is well-produced and carefully thought out. It coheres as a volume, which is unusual for a collection of essays produced within different sub-disciplines by a variety of scholars who span the globe. The central thread of the many arguments, so to speak, unravels through the laying bare of the basic premise or presumption upon which the so-called native Englishes are privileged vis-a-vis the so-called non-native Englishes in mainstream linguistics. This is followed by separate accounts of the specific contexts of Cameroon, Singapore, and India which tease out different nuances of this problematic,1 with particular reference to the politics of native speaker speak, so to speak! These 'empirical' studies are supplemented by a discussion of the real and imagined competence of native speakers, the notion and implications of 'nativity' in language, a historical account of structural linguistics in its colonial garb and also both a neurolinguistic analysis of native speakers as well as a neo-Chomskian account of the problems related with the unquestioned privileging of native speakers over others.

244 · The Yearbook (2000)

The point of the book is its overall impact and collective presentation, spelt out, for instance, in the editor's virtuoso introduction and the series editor's critical commentary. Yet, the Introduction and Commentary should, in my view, best be read as Afterwords' as they constructively engage in critical debate with the contributions in this volume, and this is precisely what I propose to do in this review. What is clear even on a first reading, however, is that the individual presentations complement each other and, what is rare in contemporary scholarship, that there is a genuine dialogue among them, even in the face of substantive disagreement, which leads to a powerful process of dialectical learning. Rather than attempt to treat each individual intervention separately, I shall, therefore, endeavor to present a summarized version of the central thematic of the book, and then to analyze its implications. In doing so I shall also endeavor to sketch out the main variations and divergences in the text as well. The hierarchy and implicit value system in the classification of Englishes (or, for that matter, other colonial 'European' languages) as native/old/inner circle vs non-native/new/outer circle/nativized is critiqued from a number of perspectives. In fact, in its most radical form this critique questions not merely the hierarchy but also the very distinction itself (see Singh, Kandiah and Annamalai). Implicit in such a classification is a hierarchy which devalues the varieties of English spoken, for instance, in Singapore, India, Sri Lanka, and West Africa vis-ä-vis the varieties spoken in England, Australia and the USA. This pecking order is roundly condemned by the contributors, by and large, for linguistic as well as socio-political reasons. These sorts of bedrock questions are raised and answered by the contributors: In what fundamental ways do the so-called native Englishes differ from the so-called non-native Englishes? Are there 'neutral' measures of quality that can distinguish these categories in this way in terms of the structure/form of each variety? Is it not implicit in this typology that the so-called non-native Englishes do not serve a fair number of native speakers? Kachru's (and others', of course) position that these Other' Englishes do not count among their proficient users any real native speakers is refuted both directly and indirectly. This is the central issue that leads to many of the other questions in this volume. It can be asserted that, in general, for the contributors, though there are dissenters, the varieties of English described by some under the category of non-native, nativized, or new/peripheral Englishes can count upon sufficient (and ever-increasing) numbers of native-like users. The process by which these users become 'native speakers' is surely not fundamentally different to any other case, since they learn the language as their mother tongue during the 'critical period' (MufWene, 111). However, these processes of becoming/being 'native' speakers/users have been presented by various writers as being different to those of the other privileged (always-already 'native' to borrow, mischievously, from Althusser) language contexts, which I find difficult to accept. I suggest here and elsewhere that the processes are more similar than is comfortable for those on both sides

Review Article m 245

of the divide. The appearance (or simulation) of difference is what lends these varieties their theoretical importance since they are more amenable to study. I, for one, do not see empirically in the milieux that I have studied, any such qualitative difference from that which obtains in the more 'normal' context. For instance, the children who acquire Sinhala as a mother tongue do so in much the same way as those who acquire Lankan English as a mother tongue in Sri Lanka. There are, of course, the complexities and nuances that obtain in any fraught multilingual setting. However, it is hard to dispute that most contexts the worlds over, today, are, in fact, multilingual. I am willing even to risk the generalization that today's normative language context is one of multilingualism, and not the other way round. The relatively smaller numbers of native speakers in this sense of, say, Lankan English, may lead to many important distinctions and modifications in the processes by which this language variety develops. Yet, these are differences of degree and not of quality. This fragility on the one hand and porousness, on the other, is symptomatic of language, as such. As Mufwene has pointed out, even in the case of Standard British English, many crucial changes came from outside and were as a direct result of conflict within and without other languages. The argument I have made elsewhere perhaps needs repeating here. These so called non-native varieties of English/ Spanish/ French/ German are not special cases of languages in a special context, but rather better gauges of language as such than their dominant so-called native counterparts, because the very struggles and contestations that took place many centuries ago in these latter varieties, over which we have mainly conjecture, hypothesis, educated guesses and which took place over epochs is now taking place (over the last 150 years or so) before our very eyes/ears. Yet, since the processes are taking place contemporarily and since many of the changes are seen as "language interference', hypercorrection, schizoglossia, fossilization, etc. but are, in fact, also classed, gendered, 'raced' and regioned interventions, the changes are contested to the extent that language as the site of the struggles of these broad social and cultural categories, is thrown into relief, as it were. The sense then is that we have, in these languages, accelerated change while at the same time there is a sense of seeing this acceleration in slow motion. If I may use as concept/metaphor the image of a video cassette recorder, the paradox about the 'film' (as record) of the history of development of these languages (or varieties) is that it can be seen in fast forward mode but with the individual frames frozen in slow motion! Thus, a vast amount of ground can be covered in a short span, while at the same time sharp attention is drawn to minute detail. But, more about this later. One of the problems with such key texts rooted firmly as it is within traditional linguistics, and, in fact, with the disciplinarization of knowledge in general, even in its most exciting and innovative forms, is the inability to transcend, however much we try, the disciplinary boundaries and discursive space

246 · The Yearbook (2000)

allocated by the dominant disciplinary paradigms. One would have wished in this collection that the more important multidisciplinary work in the areas of post-colonial theory, cultural studies, discourse theory, post-structuralism, subaltern studies and so on would have crept in to sophisticate the analysis of the properly (read narrowly) linguistic. Or at any rate that there would have been a constructive engagement with these controversial fields of inquiry. However, even in the most challenging of positions directed against the old dispensation within linguistics, there is a careful avoidance of this other work. In other words, I am suggesting that the critics of the native speaker and its inherent value systems do not go far enough in this admirable intervention due perhaps to the fact that the contributors feel the need to salvage the known terrain of the discipline which they habituate. I am reminded of Wittgenstein's famous comment which I shall modify here: When we realize that the very foundations of a culture (here an intellectual culture) are cracked, what we do is patch this with straw; but in order to appease our collective conscience we use the very best straw that money can buy! This is, of course, merely a tendency in this text. Singh's trenchant introduction and Kandiah's filigreed intervention are obvious exceptions, albeit in a different register. Thus, the demystification of the native speaker here takes the form of his (this must mark the utter gender blindness of the text under review, which does not even provide the now mandatory 'table manners', much less engages seriously with the gendering of the native speaker as norm-maker) reincarnation as proficient speaker. We need to consider whether this is substantially more than a cosmetic change of label. However, many of the problems associated with the native speaker can be adequately dealt with in this new formulation. For instance, that not all native speakers are equally proficient, that native speaker competence varies widely, that in certain special cases language acquired natively may go into disuse, can be taken care of now. We have replaced a magic category with a more intellectually defensible one. Yet, though the concept of the rational judgment of a proficient speaker is able to clear up some of the problems, there are others that remain. For instance, the native speakers who remain in situ, native speakers who are adults and so on are still labeled arbiters of a language. This means that native speakers are, in fact, all proficient speakers, unless something exceptional has happened to them (such as leaving the area for an extended period of time). This still does not grasp the nettle of the problem which is that there are differences/discrepancies among native/proficient users—that even the concept of'proficiency' must allow for the same variation as did 'native speaker' status. In addition, the determination of the proficiency of native speakers and proficient users still remains problematic. Who is to act as judge and jury in this context; who determines whether a speaker is sufficiently proficient to become a member of the select club of arbiters or not? By what processes does all this happen? In Mufwane's account and elsewhere in this text one gets the

Review Article m 247

impression that these judgments of value (of acceptance/inclusion or rejection/exclusion as well as a range of in-betweens) are well-made, consistent and entirely non-controversial. This predicates a kind of bonhomous egalitarian consensus which certainly has no parallel in the few linguistic contexts that I am qualified to comment upon. The volume is full of fulsome and heartwarming notions about 'sharing'—norms, values, standards—which dissimulates dominance with and without hegemony, which I find in my own work to be the operative condition of language in (an unequal) society. Mufwene suggests, for instance, that there are some native speakers whose language may become rusty through disuse. These exceptions are very clearly explained: either they acquire or learn some other language, they do not use their own, they move out of the location in which the language is spoken and so on. And yet none of the contributors are willing to say that a native speaker who uses the language, who has acquired it in early childhood during the critical stage, who speaks the language in adulthood and who remains within that language community, a native speaker in the clearest, least problematic sense of the literature, may not be in any position to arbitrate on his or her native language. However, I shall not be lapsing into a pernicious empiricism if I note the obvious that there are plenty of such native speakers who do not use the language in ways that other native speakers in the same community at the same time would accept, and vice versa. There are also contexts aplenty in which native speakers are unsure, 'inaccurate', or diffident about what constitute appropriate/correct/acceptable usage in a particular situation. There is also no appreciable pattern to all this, no 'standard deviation', so to speak. What is clear, however, is that the contestations and struggles for precisely such recognition are effaced, smoothed over, covered up, through a panoply of mechanisms such as standardization and the operation of hegemony which I have described elsewhere.2 To summarize so far, two issues of concern emerge: one is the absence of an account of language where the native, the non-native, the nativized, the central, the peripheral etc.—however, we may wish to describe it—is seen as a contestatory (and negotiatory, to coin a word) site, all the more insidiously so since this contestation is always covered over. Similar accounts of language as the very site of struggle obtain in postFoucaultian discourse theory: no account of hegemony at work and hence no understanding of resistance to this hegemony is discernible here. Thirdly, and this might have worked in the favor of looking at the first two in a more serious way, no real cross section or multi-disciplinarization of the study of linguistics has been attempted. It would seem that the space opened up by the so-called other Englishes would allow for a questioning not merely of the classification of English and other English, native English and nonnative English within the discipline but also fundamental issues about classification of language as such, the disciplinarization of linguistics and the very notion of scientific inquiry that linguistics is said to espouse.

248 · The Yearbook (2000)

Here, Michel Paradis' and Pieter Muysken's contributions are symptomatic. There is an underlying thread in these texts which are woven into a notion of science that seems to be outdated and untenable today. In fact, the kinds of questions raised about the Objectivity' of science, its 'scientific' basis, its discursive complicity with overarching structures of power and so on, have been the subject of debate at least for the last 50 years within the scientific community itself. There seems to be an attempt to salvage linguistics as a science (as the science, perhaps), whose privileged native speaker as the neutral, independent, arbiter of language must be enhanced. On the other hand, Alan Ford's intervention, deceptively, even mischievously 'technical', negates any neat classification for itself as it dramatizes the self-interested nature of the category of native speaker in its current (mis)use. It is here that one would wish for linguistics to take up the challenge of the broader cultural theories and other developments in social theory and philosophy. And for post-colonial linguistics to lead the way, dragging a reluctant discipline in its wake. The assumptions made by many of us are generalized by our own experience and need to be questioned fundamentally: to attempt to theorize this phenomenon through the use of a certain post-structuralist jargon would be to say that we fetishize concrete experience. For instance, we have one of the 'western' contributions (I use the word western loosely here because the point I am making is more important than the 'technical' or nominal origin of the writer) in which the native language is described as a language One acquires from one's crib'. I am reminded of the old feminist dictum which read something like this 'all the women of the world are housewives'. On the one hand this appears to be a very insightful and deep aphorism, which is critiquing the devaluing of women's domestic labor and suggesting that whatever else a women does she is also responsible for all the house work, and is subservient to a man. But I remember suggesting as a young graduate student that this also took many things for granted. To be a 'housewife' takes for granted the notion of a house, a nuclear family and all that comes with it, which may not obtain in many of the so-called third world contexts of poverty and radical cultural difference. Here, then, was the classic western (first world) generalization which homogenized 'all women' in the name of international solidarity. The end, alas, does not justify the means. So too with the idea that a native language is the dialect acquired from the crib as described by Michel Paradis. No crib, no native language, then, Sir? What about the Manger and all this nativity stuff that I have been playful about so far, or is that another special case? The example about women is my way to highlight again the lacuna in this text of any sense of male hegemony. If the reader were to perceive this as irrelevant to the argument, I rest my case. It is precisely the tired old argument that gender issues are not central to language norms and normativity, to issues of prestige and standardization, to extra-linguistic value embedded in language, and through all this to privileged native user status, that cannot be sustained

Review Article m 249 by a moments careful scrutiny. But, then, ideology never needed evidence in its support. Ideology unravels itself through its metaphors, and it is no accident that the concept metaphor of childbirth criss-crosses the entire debate in this volume. Anyway, this may sound trivial to some, but there is a certain therapeutic effect in questioning the kinds of assumptions and descriptions we make, to show that they come from a worldview which needs to be radically reexamined. We have in these accounts of language classification a privileging of science as neutral, self-evident and so on. For instance, the crib business keeps coming up, the native speaker is someone who has been speaking a particular lect of a given variety from his (of course!) crib. He has been exposed to it from birth, not acquired it accidentally, and has continued to speak it as an adult. Paradis defines proficient speakers as those who can easily be distinguished from other speakers, but admits that they show differences in test situations, and that they may also differ in their judgment on the acceptability of sentences and so on. Yet, these matters of cultural imperialism cut both ways, albeit unequally. The notion of'nativity' presented by Annamalai and echoed by the editors has no taint of irony or self-consciousness as it 'takes in vain' the name of the socalled divine birth! It is this relative luxury that I have chosen to foreground in my choice of title. Eliot, an utter reactionary, incidentally, wrote, 'after such knowledge what forgiveness?' in Gerontion. Is such heresy only possible with a straight face, a la Annamalai? The sheer gall in claiming for the 'illegitimate' variety an 'immaculate conception' is radically subversive. Yet, one suspects that this delicious irreverence, so symptomatic of the counter-attack from the margins, remains inaccessible. The epiphany of native speakership, so to say, just will not go away. In Kandiah's words, the deathless native speaker has been resurrected/reincarnated as an avatar. This I think is a crucial image. This is true of those native speakers who will then be acceptable to Mufwene as well. These proficient speakers are still, however, native speakers of the language in disguise! What gets swept under the carpet in these re-formulations is the very process standardization and the creation of language norms. Perhaps on the long term, non-elite speakers are, selectively and in certain circumstances, able to influence the broadening of norms, the extending of standards, the egalitarianizing of changes. But this takes place over a very very long period and is not always an intervention that comes out as planned. To be reiterated here is that making problematic of the competence of some native speakers and specifying the reasons and conditions of such problems as displacement, non-use of the language for an extended period, and so on, as well as the replacement of the native speaker as native speaker with the notion of the proficient (native) speaker is certainly an important qualification. It removes some of the lack of charity in the concept of native speaker but at the same time it salvages (a) another native speaker who has obtained the

250 · The Yearbook (2000)

necessary credentials, (b) it privileges speech over other forms of language, (c) it allows the classed, raced and gendered (ideological) basis of language standards and norms—in short, its self-interested nature—to remain intact. In other words, this reformulation does not address the crucial issue of ideology and hegemony in the control of language and through language in the exercise of power within 'multilingual' communities.3 In addition, this salvaging of the native speaker also attempts to carry linguistics as a discipline which is unbiased, scientific, apolitical, neutral and so on to conduct business as usual. Yet, as has been established elsewhere, the act of description is itself not a neutral activity. Selection of any kind, classification, analysis, exclusion, inclusion, all these are fundamentally political acts, they involve judgment, valuing, devaluing, revaluing and so on; there is no act of 'disinterested' description with no value judgments attached. The history of the reception of Daniel Jones'Pronunciation Dictionary, even its current status, should give the lie to such claims for pure description. That is perhaps the easy objection to live with: this notion of science (that linguistics in this version appears to be the paradigmatic species of) is one of the precritical positivist variety which steadfastly refuses to allow for the fact that language itself is already coded, the descriptive basis of it is also coded, the history of linguistics is implicated in the colonial enterprise. It is not the case merely that branches of structuralism (and thereby structuralist linguistics) are contaminated by colonialism. The power of this text is, however, the opening up of issues, the laying out of the biggest contention, that then leads to further modifications and debate. The unearthing of the native in its curious versions in which the first world comprises of native and the so-called third world non-natives—'paradoxical reversal of the colonial encounter'—and the unpacking of the native, the destabilizing of classification of languages has somehow clearly become imbricated in the hierarchy/worldview of dominant linguistics even in its radical corner. Thus, the examination and close scrutinizing of the history of linguistics as complicitous with the colonial enterprise, the relationship that language normativity bears to ideology and hegemony, all this is extremely timely and opens up a vast avenue of other issue which need further discussion and analysis. All credit to the editor and contributors for conceiving and presenting this text as a wholistic 'unity in diversity' in which different aspects of a crucial problem not merely in linguistics but in social theory in general is articulated and deconstructed, only to be re-assembled again anew. In this sense the challenge for the linguistic establishment has already been presented. My sense is that the very questions that are raised in the multilingual context of English can be argued with cogency in the dominant contexts in which these languages are said to be 'at home'.

Review Article m 251 I

II

III

IV

Non-standard Forms

Non-native Variety

A Variety

A Variety

(Interference Errors, Fossilized Interlanguage etc.)

(Inferior Special Case justified on affirmative action principles)

(like any other [But 'new' and 'different' in some weak versions])

(which is a better index of language-assuch [due to socio-historical conjuncture] Theoretically more important than S(B)E,etc.)

The figure above characterizes the trajectory of the debate on, say, Indian English. The majority of the interventions in this text fall clearly into the third category, and here too under the strong formulation. Kandiah takes up a position similar to position four (p. 107); Singh points out that the 'new varieties enterprise' has been unable to turn interesting facts into evidence, which at the very least leaves the door open for new theoretical formulations that transcend the specific context of this 'enterprise' (p. 61). I shall use an example to establish my argument here. Jean D'Souza writes in a deceptively simple-sounding intervention full of quiet insight that has the ability to get right to the heart of the matter: 'All varieties, other than BE, are transplanted and all are nativized' (40). All varieties are, in fact, transplanted and nativized, allowing no exceptions. All varieties of English are transplanted in the sense that they are the product of language contact, language conflict, language interference and so on. Is it not the case that British English which is said to be a non-transplanted variety in this sense is also the net result of a prehistory history of such transplanting in terms of the comings and goings of different groups of people to the land we now (mis)call Great Britain! Even within the island is not such a homogenization suspect? For instance, is 'transplantation' only a matter of crossing the seas? What of movements within the landmass? The point is that the absolute, unquestioned hegemony of British English for 'us' (as linguists, former colonial subjects, natives who are Englishspeakers by the grace of god and the empire but who are not quite native speakers) comes undone if one takes the case of Sri Lankan English (or Indian English or Singapore English, etc.) seriously. There can be no non-transplanted English (or any language for that matter), no purity of place/time of origin, no unmarked, uncontested sharing of standards/norms. The miraculous conception of English-at-home in the mother country becomes just as problematic and fraught as other cases of ordinary parturition elsewhere!

252 · The Yearbook (2000)

In this sense my conceptualization of the so-called other Englishes as a mise-en-scene in which a fast-forwarding takes place, but nonetheless where each individual frame is as it were frozen in slow motion. It is useful in the sense that it gives us an understanding into the processes of transplanting, I would rather call it translating the discursive and other registers/styles at play here. I would say that the fundamental nature of language is a kind of translating. It is only in the specific cases of translating from one language to another that we focus on the issue of translation, but translation is the general case which is covered over, effaced, even dissimulated in a context in which only one language is used. Similarly, I would say transplanting is the general case, the state-of-the-art, so to speak, of language. If you go far back enough in time, if you look deep enough into the interstices of the matter, there is no qualitative difference between the history of British English and the history of Sri Lankan English in this sense. A serious and thorough-going engagement with Lankan English should therefore lead to a fundamental revaluation of our understanding of British English. The processes by which language standardization takes place, the mechanisms by which linguistic norms are established is complex and varied. Proficient users of a language are determined in terms of the nexus of access to power through class, status, gender, ideology and regional affiliation. There is also an element of chance or randomness in sub-sub-areas in which no such contention exists. These proficient users arbitrate when called upon to do so, but more importantly they are users (and, hence, disseminators) of a particular variety (or, more precisely, particular variations). This constant use serves to habituate others who wish to be accepted by the elite group of proficient users. In short, replication takes place through acquisition, imitation/mimicry, learning and so on. In this context, rigid arbitration of the proficient user is seldom formally invoked. The irony, twice born, is that despite all the hullabaloo, the native speaker is back again, albeit in different garb. We have, indeed, killed the snake, not scotched it. The source of this misquotation is also apt, since it too deals with an act of tyranny, or in the privileged male metaphoric of almost parricide. If I am permitted to gossip about non-existent people in order to make a point, Macbeth's double trust as subject and as kinsman he broke in killing Duncan. This heinous crime led to others such as the murder of Banquo referred to in my (mis)quote. So too with these other upstart Englishes, or so the story goes. Not only have they illegitimately usurped the proper place of Standard (British) English (and, of course, God Bless the Queen and the American Way—apple pie, not cigars, please—but it's not cricket, all these snide remarks), they now refuse to acknowledge an unending subservience, a debt of gratitude and a self-evident inferiority. I salute the radical honesty and theoretical rigor of this endeavor which places "intellectual correctness' (33) above all else. It should be required reading/engagement for all those in the field, in every sense of that word.

Review Article m 253

NOTES 1. Used, after Althusser, to denote a set of inter-related problems taken together. 2. De-Hegemonizing Language Standards: Learning from (Post) Colonial Englishes about 'English' 1995. London: Macmillan. 3. Kandiah's tour de force in this collection is an exception, as it is to my charge against the blind acceptance of the scientific pretentions of linguistics or, indeed, of science itself! I regret that I have short-changed both the readers of this review as well as Professor Kandiah through not dealing adequately with this wide-ranging and brilliantly argued essay. I strongly recommend that this be required reading for anyone interested in these issues.

Reviews R.K. Agnihotri, A.L. Khanna, and I. Sachdev (eds), Social psychological perspectives on second language learning. New Delhi: Sage Publications, 1998. Pp. 239. Rs 375. Reviewed by Kathleen Connors, Universite de Montreal. This book is a collection of 12 articles on the role of attitudes and motivation in language learning, use, and evaluation. Nine of these papers directly concern Indian learners, in India or as immigrants in Britain. Most of the research reported here aimed to study the relation between attitudes and motivations, that between levels and types of motivation, on the one hand, and achievement and/or proficiency, on the other; that between attitudes and achievement and/or proficiency, or the impact, on language learning, of the presence/ absence of a target-language community or arena in the social environment of the learner; and the effect of social characteristics of the learner and his relationship to the target language and reference group on that learning. The main contribution of the papers about Indian learner groups is to explore and emphasize the role and importance of these last, purely social factors, beyond the social-psychological ones. In the contexts where a targetlanguage community or some other social arena for the L2 is present, moreover (especially early), exposure to and use of that language often turn up as correlates and predictors of proficiency. The editors, in their Introduction (p. 19), present a very helpful correlation matrix referring to earlier studies on Indian learners of an L2 (usually English). The most frequently significant correlates with achievement are type or amount of schooling and degree of exposure to the target language. Integrative and instrumental motivation both correlate significantly with teachers' evaluations, but only in the one study performed in the (British) target-language environment. Mass learning of an L2, such as English in our time, appears to respond to social forces independent of individual learner profiles.

256 · Wie Yearbook (2000)

The results of two studies in the present collection on L2 learning in India, but not of English, are of complementary significance. One (C. Sawhney and R.K. Agnihotri, Acquisition of Hindi as a second language by Tamils in Delhi. A social psychological perspective) concerned 66 grade nine Tamil pupils in Delhi learning and acquiring Hindi as an L2. The other (C. Sawhney, The role of attitudes and motivation in foreign language learning: The case of German in India) dealt with 29 students of unspecified age learning German in three institutes, also in Delhi. In the Hindi study, the only variables accounting for considerable portions of the variance in the results of a cloze test (13 per cent and 11 per cent, respectively) were parental encouragement and the socioeconomic status of the family. The only ones contributing comparably to the variance in reading comprehension test scores were positive evaluation of the Hindi teacher (10 per cent) and type of school previously attended (9 per cent). In the German study, a cloze test score was mainly predicted by the student's educational background (33 per cent of the variance), and an examination score by the teacher's evaluation of the student (34 per cent of the variance). These non-English L2 studies remind us that school language-learning, and evaluations of that learning in a scholastic context, have a life of their own. Despite the obviously different significance, in India, of learning Hindi, on the one hand, and German, on the other, the dominant predictor variables in both studies were tightly bound up in the socio-educational situation of the learners, and, again, were not personal attitudinal or motivational factors. In this connection, one large study (356 subjects) (R.K. Agnihotri, Predictors of achievement in English tenses: A socio-psychological study), the only one to focus in detail on the learning of a linguistic sub-system, showed that evaluations of the learning even of English can be determined by factors which are almost purely scholastic. Among 19 social and socio-psychological variables whose correlation with achievement in the use of English tenses was studied, only five were picked out by the step-wise multiple regression analysis as significant predictors of that achievement: the type of school the subject had attended, his or her marks in English, the choice of an honours vs a pass course, the degree of claimed use of English outside the English class, and relative agreement with stereotypes about the English language. Of these, the first, a specifically social scholastic variable, determined in turn by economic factors, was picked out as the main predictor, accounting for 29 per cent of the variance. A secondary message emphasized by several of the papers is the relative prominence of instrumental as opposed to integrative motivation, notably for South Asian learners (Introduction: 12; Khanna, A.L. et al. Teacher evaluation of Asian ESOL learners in Britain and its social psychological correlates; Khanna, A.L. and R.K. Agnihotri. Some predictors of speech skills: a socialpsychological study; and Khanna, A.L. et al. Immigrant identity and language proficiency: a sociolinguistic study of ESOL). Though motivation measures

Reviews m 257 were not picked out as predictor variables by the step-wise multiple regression analysis in the above-mentioned German study, the subjects indicated that their reasons for learning German were mainly instrumental (p. 128). One of the contributions, in turn, of this kind of finding is a corrective to the long-standing association in the literature of instrumental motivation with 'subtractive' (as opposed to 'additive') bilingualism. The British ESOL study cited above demonstrated (p. 75) that in a context where the learners showed little desire to integrate into the host (and target-language) society and strongly maintained the use of their LI, instrumental motivation was second only to self-evaluation in English as a highly significant correlate with the teacher's evaluation of their English. The face validity of this finding is supported by the realism of instrumental motivation: The results of the study in this collection dealing with British teachers' evaluations of Indian, among other, accents in English (Sachdev, I., Oral assessment and accent evaluation: Some British data) indicate clearly that Indian accents are regarded as foreign in Britain. A fortiori, results of studies done in India cannot indicate an association between instrumental motivation and subtractive bilingualism, since the latter is a stage in linguistic assimilation (R. Landry and R. Allard, The use of minority and majority group languages in minority education: A theoretical model), not a foreseeable possibility in India (with English—much less German—as the assimilating language). In the end, the chief effect of these studies, which started out in search of socio-psychological factors in L2 learning, is to reinforce doubts that have often surfaced in the literature as to the importance of such factors. For Indian learners at least, and one suspects, for large populations of L2 English learners in (or from) other parts of the world, broadly social forces and factors trump personal psycho-social ones in the determination of L2 proficiency.

M. Hariprasad, H. Nagarajan, P. Madhavan and K.G. Vijayakrishnan (eds) Phases and Interfaces of Morphology. Hyderabad: CIEFL Publication. 1997. Pp. ix, 362. Price: not available. Reviewed by Sylvain Neuvel, University of Chicago. Imagine a child visiting his grandmother, sneaking up to the attic and finding an old dusty wooden chest. The rusty hinges squeak as he slowly opens it and his face lights up for he finds a myriad of toys waiting for him to play with. This is pretty much how I felt when I first flipped the cover of this very modest looking book. The copy I got was barely holding together. It had its tan jacket cut too wide and the title underneath was printed at a very peculiar angle. But once I started reading it I realized that it gave a whole new meaning to the French proverb L 'habit ne fait pas le moine (or to its more to the point English equivalent Don't judge a book by its cover.)

258 · The Yearbook (2000)

Prefaced by K.P. Mohanan and Tara Mohanan, this collection of 16 articles is divided into four sections. The first one deals with core elements of morphological theory. Parts II and HI are dedicated to the morphology-phonology and morphology-syntax interfaces, and the last section handles issues of developmental and computational morphology. Part I—Approaches to Morphology contains: (a) Postlexical morphology in parametric grammar by Probal Dasgupta (b) Morphological analysis in Russian by J.P. Dimri (c) Allomorphy: A systemic viewpoint by V. Prakasam and (d) Kannada prefixation by K.V. Tirumalesh. Set in a Minimalist framework, Probal Dasgupta's proposal expands the boundaries of morphology and makes use of such notions as topic, focus, given vs novel (what he calls iota roles, by opposition to theta roles) to account for a variety of phenomenon ranging from stress patterns to the behavior of NP's in wh- movement. The non-Minimalist reader may have a hard time calling everything here morphology, but at the very least it paves the way for some interesting syntax, and the idea that communicative properties like novelty or topic may be involved 'further' down the grammatical road is also quite interesting. J.P. Dimri reviews some of the methods of morphological analysis formalized over the years by Russian linguists and proposes his own. The historical references in this article are quite interesting. I also found many of Dimri's observations stimulating. As for his proposal, it is obviously circular (his 'three-way' comparison method used to identify a suffix works by comparing words derived from the same underlying base or derived using the same word formation model). Prakasam creates new terminology, defining the terms allofonn, allolexe and allomorph to identify different types of variants, these three terms corresponding to three different facets of the minimal sign. I now turn to Tirumalesh's contribution. I shall take the liberty of spending a little more ink on this article because I believe it has substantial ramifications; I will ask the reader to forgive the unevenness that ensues. This paper by Tirumalesh is a reply to an analysis of Kannada prefixes by M. Aronoff and S.N. Sridhar that tackles the distinction between prefixes and the first elements of compounds (M. Aronoff and S.N. Sridhar. 1988. Prefixation in Kannada. Theoretical morphology: approaches in modern linguistics ed. by M. Hammond and M. Noonan, pp. 179-91, San Diego: Academic Press). Aronoff and Sridhar describe prefixes as 'a small set of bound forms which are tightly tied phonologically and morphologically to their heads', while first members of compounds are 'members of an open major lexical category and have looser phonological and morphological heads' (ibid.: 180). On that basis, Aronoff and Sridhar identify a small set of about a dozen Kannada prefixes that combine with stems of different categories. These prefixes trigger both forward and backward assimilation, with a few exceptions due to the properties of the prefix. Kannada compounds, on the other hand, do not involve backward assimilation, forward assimilation being optional and related to properties of the stem. Tirumalesh challenges this analysis. He notes that

Reviews m 259 Kannada words all share the property of having the weight of at least two matras (or moras). Tor a string to qualify as a lexeme in native Kannada, it is necessary that it has at least two matras' (ibid.: 76). Tirumalesh calls this the matra requirement. Interestingly enough, there are no prefixes of less than two matras in Kannada. Furthermore, whenever resyllabification leaves the first syllable (or prefixal root) with the weight of only one matra, either the prefixfinal consonant is geminated or the prefix vowel is lengthened to meet the matra requirement. This leads Tirumalesh to believe that these prefixes are derivatives of existing lexemes. Some of them correspond to a truncated form of existing lexemes. These lexemes can serve as the first member of what Aronoff and Sridhar call a compound, but only if the following stem begins with a vowel. If the following stem begins with a consonant, then the truncated form is used. There is then no need for speakers to store these prefixes in their lexicon since they can be derived with general principles. When the prefix cannot simply be obtained by truncation, Tirumalesh suggests a morpho-syntactic rule stored in the lexicon pairing the lexeme to its prefixal form. This analysis is not only simple and elegant, it is also in concordance with Kannada tradition; grammarians as ancient as Keshiraja of the llth century A.D. treated lexemes and prefixes together as a single phenomenon. Although Tirumalesh very convincingly cripples Aronoff and Sridhar's distinction between prefixing and compounding, he is not willing to give it the coup de grace. He states:'[...] In such cases, the difference between prefixation and compounding is a matter of terminology only and not substantive' (p. 83); one might actually be tempted to simply deny the existence of compounds as a special class of words (see R. Singh and P. Dasgupta. 1999. On socalled Compounds, The Yearbook of South Asian Languages and Linguistics 1999 ed. by R. Singh, P. Dasgupta and K.P. Mohanan, pp. 265-75, or see Neuvel; this volume, for another type of distinction between the two). Part II—Morphology-Phonology Interface contains: (a) Consonant deletion rules in the Naokhaii dialect of Bangla: Their lexical status and mode of application by Shyamal Das (b) Word formation and vowel quality in Bangla: A lexical phonological account by Tanmay Ghosh (c) Some salient properties of echo word formation in Indian languages by Chandrakant Jha, Kamlesh Sadanand and K.G. Vijayakrishnan (d) Phonological rule application at the word Level by K.G. Vijayakrishnan. Shyamal Das borrows from different theories and provides a thorough account of the very interesting phenomenon of consonant deletion in the Naokhaii dialect of Bangla. He argues against the phonemic status of a particular segment and is able to provide a unified account of the entire phenomenon under a set of word level rules. Look for his autosegmental version of metathesis. In continuation, Tanmay Ghosh calls upon a simple constraint on the moraic weight of minimal words to account for vowel length differences in Bangla within the framework of the Optimality Theory; the prerequisite postulate being that coda consonants do not contribute to syllabic weight in

260 · The Yearbook (2000)

Bangla. Jha et. al examine echo word formation in Indian Languages. Echo word formation, or as Yip calls it 'Reduplication with Fixed Melodic Material' is a distinctive phenomenon in Indian Languages. McCarthy and Prince (John McCarthy and Allan Prince. 1994. The Emergence of the Unmarked, Proceedings ofNELS 24, 333-79) state that the overwritten melody, for example /gi-/ in the Tamil echo word kiLaigiLai 'branch', is always the unmarked member in the segmental inventory of that language and need not be specified. Evidence from a variety of Indian languages is used by the authors to convincingly disprove this assumption. Data from Punjabi is also used to support the claim that the overwritten melody is on the same plane as the reduplicated melody. A very interesting article, whether you agree with the conclusions or not. Lexical phonology assumes that, at the word level, morphological operations feed phonology. Having offered contradictory evidence, Borowsky proposed to reverse this ordering. In his article, K.G. Vijayakrishnan argues that the morphology-phonology order is universal at the word level and that the phenomenon evidenced by Borowsky can be explained with a multi-planar representation in which different morphemes occupy different planes. He then proposes that the ordering of tier conflation and phonology must be parameterized for in English and German phonological rules must apply before tier conflation, while in languages like Tamil, Bangla and Punjabi, tier conflation must precede phonology. Part III—Morphology—Syntax Interface includes: (a) A Kannada perspective on morphological causatives by R. Amritavalli (b) How do languages create quantified expressions: A cross-linguistic investigation by P. Madhavan (c) Deverbal compounding in Oriya by Kalyanamalini Sahoo (d) Agreement morphology in Chukotan by Andrew Spencer. Amritavalli proposes a syntactic analysis of morphological causative constructions in Kannada. She argues that these constructions are monoclausal, contrary to a long tradition in generative grammar. In her view, 'derivational morphology, no less than inflectional morphology, can be done in syntax, "upto monoclausality'". Assuming that the morphology of natural languages somehow mimics formal logic, Madhavan takes a look at quantifier expressions from English, Malayalam, Hindi and Sanskrit and suggests that they are made up of a pronoun, either + wh or - wh, and a logical operator. He also examines reduplication as a mean of creating quantifiers. K. Sahoo proposes an analysis of de-verbal compounds in Oriya. In her view, de-verbal (or synthetic) compounds are formed at the level of syntax and involve head movement. Andrew Spencer's thorough article is an analysis of the argument system of Chukotan that spans over 45 pages. He highlights some of the problems it poses for any morpheme-based theory of morphology, and more specifically Halle and Marantz' Distributed Morphology (c/. M. Halle and Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. The view from building 20, ed. by K. Hale and S.J. Keyser, 111-76. Cambridge: MIT Press.) To solve these problems, Spencer proposes to expand on Stump's theory of

Reviews m 261 rules of referrals and allow for what he calls 'broad' referrals under which entire word forms can be deferred to other paradigms. Part IV—Developmental and Computational Morphology contains: (a) Morphological causatives in children's speech: Mono- or bi-clausal by Hemalatha Nagarajan (b) Reduced and simplified morphology in some French and English pidgins and Creoles by Priya Hosali (c) Morphology-syntax interface: A note on Tamil-Telugu acquisition by B. Lakshmi Bai (d) Recognition of tense in negative verb phrases by M. Hariprasad. Telegu has both bi-clausal periphrastic causative constructions and monoclausal morphological causatives involving a suffix. H. Nagarajan examines morphological constructions produced by Telugu native children. While these children are using the causative suffixes, they also use a double accusative case marking typical of periphrastic causatives, suggesting a bi-clausal structure. She offers an analysis of these transitional constructions and proposes an explanation based on general economy principles. Hosali looks at different examples of morphological simplification in French and English pidgins and Creoles. B. Lakshmi Bai investigates the expression of instrumentality by Tamil and Telugu native children at different stages of their language acquisition. She is able to show that prior to expressing instrumentality through a morphological case marker, children use participial sentence constructions from which instrumentality is inferable. Tamil and Telugu instrumental markers can express other argument-predicate relations, and children will show a preference for participial constructions as a mean for expressing instrumentality even after having learned the morphological case markers. Finally, turning to computational morphology, M. Hariprasad proposes an algorithm for interpreting time in Telugu negative phrases, which do not carry tense markers. As the preceding overview shows, this book is a rather heterogeneous mixture of proposals set in a variety of theoretical frameworks. One may have wished for a common thread, a certain sense of harmony (it certainly would have helped with this review), for this book is only as good as the sum of its parts. Fortunately, most of its parts are great and what it lacks in unity, it easily makes up for in content. I found most of the articles to be insightful and stimulating, regardless of one's beliefs or theoretical allegiance. The preponderance of Indian linguists amongst the contributors is felt throughout the book; the data is well chosen and presented in a way that reflects a unique understanding of these languages.

262 · The Yearbook (2000)

King, Robert D. Nehru and the language politics of India. New Delhi: Oxford University Press. 1997. Reviewed by Jayant K. Lele, Queen's University. This book offers a defense of Nehru's 'feet dragging' on two major language policies while he was the Prime Minister of India: linguistic reorganization of the states and adoption of Hindi as the national language. Nehru's ambivalence on these issues is well known and his delaying tactics were widely criticized. King, however, claims that through these 'Nehru did a far better job of joining battle in India's language conflicts than he is customarily given credit for' (p. xiii). As evidence, he offers a brief, selective summary of India's history and of the major events surrounding the 'language polities' in India. He also has a theory about language politics. The first chapter is a short historical account of language development in India. By the end of the nineteenth century, says King, English was as firmly entrenched in India as the English themselves. It became the language of negotiations between the leaders of the emerging independence movement. King also claims that for Indians the ubiquity of language leads to their (including Nehru's) heightened sensitivity to its uses, misuses, mysteries and promises. Despite this, says King, nothing in Nehru's background prepared him to realize that language as a political force could engulf him and India so deeply up until his death in 1964. In the second chapter, King wants to look at the science and the history of language so as to gain an insight into Nehru's mind. He divides language into two categories: language-as-object and language-as-means. The latter is further divided into linguistic and non-linguistic uses of language. The nonlinguistic use is subsumed under the category of 'language as icon'. This becomes King's way of comprehending the relationship between language and nationalism. With this promising beginning one would expect an attempt to integrate the various dimensions of individual-society-language relationship into a theory of identity formation. The instrumental use of the icons of identity, such as language, should form part of such a theory. For the language politics of formal democracies, innovative insights already exist in critical socio-linguistics. Jürgen Habermas has, for example, produced ideas about communicative competence, an unabridged view of rationality (as simultaneously instrumental, communicative and strategic) and the hollowing out of the public sphere. They are equally salient for making sense of language politics in India under conditions of elite pluralism. Were King to write a monograph or an article on language and the politics of identity, he might have felt compelled to confront some of these issues. In this book he restricts himself to defending Nehru's inaction. For Ordinary people', says King, language is a means (of communication and expression) but for linguists it is an object of analysis. For Nehru, it was

Reviews m 263 both. He used it with emotion as a writer but viewed it like a linguist and hence saw nothing in it to be emotional about. Therefore, according to King, it took Nehru a long time to recognize that language is also a political force and can be used effectively in the tug of war of domestic politics. Ordinary people, on the other hand, treat 'language as icon', as an object that is made to stand for something else, such as a nation. It serves as a badge of membership or as a means of exclusion and exile. People are thus driven to murder or suicide in defense of such icons. This is King's simplified theory of identity politics. Language is seen as a 'handmaiden' of nationalism. For him nationalism was a product of Europe and washed up on the shores of the colonial empires only in the twentieth century. With it came language politics. King makes no attempt to distinguish between those who manipulate language as an icon 'to goad people into murder and fury' and 'the ordinary people', whose mediation of their surrounding universe, through their native language, is normally both rational and emotional at the same time. King makes only cursory references to the ways in which memories of past struggles and unfulfilled aspirations become encapsulated in people's icons. He sees this only as a source of primordial-emotional attachment, without a rational content. He lauds the revival of Hebrew and its power as the icon of Jewish identity, with the holocaust as the determining factor, but treats its survival as a rare miracle. He does not seem to realize that contradictions and exceptions, as explanations of events, can only work if they are anchored in a cogent general theory. For King, the relationship between language and nationalism (demonstrated with the examples of Hebrew, Irish and Turkish) is, otherwise, strictly instrumental in nature. 'There is a "free-market" principle at work in regard to language. People will speak and use the language that they feel will be of the greatest advantage to themselves and their children' (p. 39). The sage conclusion of the chapter is that governments should, as with other markets, keep only a light hand on the linguistic reins of a country. The third chapter opens with the possibility of a discussion of a critical issue: how to ensure India's unity, without resorting to an administratively forced homogenization or integration of its population. Or, in other words, how to introduce authentic federalism to a sub-continental polity. The Congress, in the era of its realization under Gandhi, made an attempt to meet this challenge. But it paid no attention to the preconditions of an authentic federalism of equals. Caught between Gandhi's Utopia of trusteeship and the Western ideas of capitalist modernization, the Congress party merely adapted the late colonial politics of interest group pluralism to the needs of state-directed capitalist modernization. It fashioned an agriculture-industry compromise that left state-level politics in the hands of dominant caste elites. Not surprisingly, during the Nehru era, language politics thus became intertwined with the politics of region, caste, religion and ethnicity.

264 · 77* Yearbook (2000)

The substantial literature on political and economic exploitation of identities, by local and national elites, receives scant attention in King's iconic framework. Polemical or analytical, it could throw much light, as data if nothing else, on the real politics of language. King dismisses it as "partisan, tendentious, badly argued, and intellectually provincial9 and of "dubious value in a scholarly discussion of the subject' (p. 92). King sees caste and communal rivalries as the "ulterior motives' (p. 70) behind language conflicts rather than as icons put to instrumental use in the same way as language. No distinction is made between the authentic urges of community identity that hark to the principles of justice and equality and their instrumentalist manipulation. In the post-partition context, Nehru's fears, that the demand for linguistic states could feed disruptive and separatist tendencies, was understandable. Seeing it as inherently detrimental to the security, unity, and economic prosperity of India results from a lack of vision. Like other Congress leaders, Nehru also failed to see that nurturing diversity would require an authentic, democratic federation of fully equal citizens in fully equal states. During the era of the nationalist struggles, open debate about the possibilities for such a federation and of its socio-economic prerequisites did not seem possible. King refers to the superficial and paternalistic comments of the Simon Commission, on the need for the protection of minorities in language majority provinces, but also cites Ambedkar's fear of the dominant caste elites in the new states. However, like the Congress leaders before him, he pays no attention to what this means in terms of class-caste-elite domination of a formally democratic polity. King divides the world of language politics into two parts. The one of modern leaders like Nehru is seen as sober, rational, and dispassionate. The other consists of the local leaders who, intensely passionate about their caste and religious interests, can incite the masses to suicide and murder in the name of linguistic states. The orientalist vision of the Indian masses, as easily excitable in defense of their icons, is unmistakable. In such a world of primordial passions, the only rational course is to delay action until passions are spent and the masses are pacified. A rational dialogue can, once again, be initiated between "interested' elites. Nehru's recalcitrance and caution, like that of his colonial predecessors, can thus be proved to be rational. King's real attraction for Nehru results from his belief that Nehru, like him, was a man imbued with a western sense of rationality about language that allowed him to separate language analysis from language use, linguistics from language politics and reason from emotion. I would rather admire Nehru for his genuine ambivalence towards the forceful use of central powers in suppressing authentic popular urges, even though he often found it difficult to restrain himself.

Reviews m 265 Bh. Krishnamurti. Language, Education and Society. Series Language and Development, Vol 7. New Delhi: Sage. pp. 328. Rs 475. Reviewed by Robert D. King, University of Texas at Austin. This volume is an anthology of a selection of Bh. Krishnamurti's publications, 13 in all, from between 1962 and the mid-1990s. Almost half deal with Telugu on topics ranging from its social and regional dialects to matters of style and modernization. Others papers deal with English in the Indian context, with spelling pronunciations in Indian English, with the competition between English and Indian regional languages as the preferred medium of instruction in schools, with literacy models in developing nations, with the politics of language in India. The articles, even those written in the 1960s, hold up remarkably well at this distance. (Would that I could say the same of mine!) Which is to say that they are, as one expects from their author, solid, well argued, reasonable, never trendy, never too-clever-by-half. Very little seems dated about them. There is something useful here for almost anyone with an interest in language in modern India: for the dialect geographer, for the student of journalistic and literary style, for the sociolinguist qua language-planner, for the educator. Different as the articles are and written as they were over a period spanning some thirty years, they are thematically related (as Krishnamurti says in his Introduction) by a concern with 'problems of language in relation to the Indian society and the educational system as and when they became burning topics of the day'. How to bring Telugu and other vernaculars up to the speed our age thinks it requires, whether to teach in English or in Telugu or Hindi, how much standardization is a good thing and how much a bad thing, above all what is best for India regarding language policies—these are the things that have moved Krishnamurti to write over a long career. Telugu is Krishnamurti's native language and—one sees it immediately—his truest language love. Any linguist working on Telugu who ignores Krishnamurti's work on Telugu places himself at disadvantage. On the 'big' issues of language in India, I find almost everything Krishnamurti says agreeably tough-minded and commonsensical. His article Official Language Policies: The Eighth Schedule of the Constitution of India' (1995) is must reading for anyone who writes on language politics in India. I do, however, have a caveat. If his recommendation that 'The Eighth Schedule should include all languages which have above one lakh speakers' were ever adopted, there would without doubt soon be reason to regret it. Adding more languages to those listed in the Eighth Schedule will not diminish the potential for mischief that inherently lurks in language on the subcontinent, quite the contrary. I think B.R. Ambedkar and his team got things about exactly right when they listed in the Eighth Schedule of India's Constitution 14 languages, affording them official status. Other languages have been added to the list since 1950,

266 · The Yearbook (2000)

bringing the total to 19 today (the last time I checked). Fourteen official languages, it seems to me, is enough for any country. The editor of the Sage series in which this volume appears, Udaya Narayana Singh, supplies a very useful overview of Krishnamurti's work, stressing the easy balance he has struck between narrow linguistic interests and the broader view of language and life that university administration (Krishnamurti was Vice-Chancellor of the University of Hyderabad) forces on one. Those others of us who have tried to straddle the two must admire his accomplishment. George Orwell,—and no English stylist has ever measured his words more carefully—would say of a writer whose work he really liked, I mean really liked, that 'he is more than ordinarily good'. It was Orwell's highest praise. Well, let me take a phrase from the master's book and say of Bh. Krishnamurti as linguist that he is more than ordinarily good. This handsomely produced compilation of the best of Krishnamurti is a fitting memorial to his achievement.

Annie Montaut (ed.), Les langues d'Asie du sud. Faits de langues 10. Paris: Orphys, 1997. pp. 260 Price not available. Reviewed by Luc V. Baronian, Stanford University. As the title of the book and the name of the collection indicate, this work discusses linguistic facts from South Asian languages. A total of 27 articles are presented, displaying facts from all the domains of grammar in the various languages spoken in the area defined as 'South Asia': Bangladesh, Bhutan, India, Nepal, and Pakistan. For obvious reasons of space, we will not have the time to discuss in depth each and every one of these articles, but we will try to cover as much as possible. My first general remark (or perhaps one should say warning) about the book is that although the idea of presenting raw linguistic facts in no defined theoretical scheme might be seducing to some, it almost always turns out that some theoretic choice must be made at some point in presenting the facts. Of course, I am aware that these articles do not pretend to give the reader purely objective facts. In fact, many of them use facts of languages to show support for one or another (very general) theoretic notion. My point is only that some notions which are currently debated on the theoretic scene are taken as a point de depart in some articles (like composition, markedness, zero morphemes, or even morpheme structure itself) to analyze the facts at hand. Some theoretic stand is, of course, necessary in most cases, but this should be kept in mind while reading this type of book. Overall, the book is very well organized, the general presentation nicely sums up the subject of each article, allowing the reader to browse through it

Reviews m 267 according to his needs, be they phonology, morphology, language, language family, or other. Montaut starts off the book with a general perspective of multilingual South Asia, which includes a history of Indian linguistics and the political history of India. Pan-Indian features such as retroflex consonants, reduplication, two orders of causative derivation, etc. are listed and the discussion on the emergence of a new 'Indian' linguistic family is introduced. (A thorough article on this subject would have made a nice conclusion to this book, especially since at least two articles—Michailovsky and Toumadre—question the Pan-Indian nature of one of these features.) Only the section discussing Hindi dialects seemed somewhat strange to this reviewer. The author explains how nonlinguistic factors such as one's pride in his idiom or the socio-political power of Hindi can affect the claims of which language one speaks. She then writes: "the extreme variability of non-standard Hindi does not help to straighten things out' (my translation). I am not a specialist, but I would have thought that the reasons behind the Variability of non-standard Hindi' are precisely those nonlinguistic factors she mentions, meaning that the line between non-standard Hindi dialects and non-Hindi dialects must be arbitrary. The article by Abbi on reduplication in South Asian languages shows that this morphological process can be used for many purposes: forming nouns from verbs, emphasis, distributivity, etc. One is only tempted to add: 'just like any other morphological process.' The article by Bhat on the verb/noun distinction in the Munda languages is very convincing. (This is one of the articles using data to support a theoretic notion). Bhat argues, mainly, that facts from Munda languages are insufficient to justify a verb/noun distinction in these languages. His main argument is that all words, irrespective of their correspondence to 'noun-like' or Verblike' concepts, take not only the same aspects, tense and mood, but also affixes. Montaut's article on ergativity in Indo-Aryan languages suggests that its historical causes are to be linked with phenomena occurring in the future tense in Bengali and some Eastern Hindi dialects. (In fact, these were ergative at some point in history. Hindi and Bengali, therefore, represent two stages of evolution: ergativization and de-ergativization). The core of the argument is that the agreement pattern is the same in the accomplished present of the ergative Indo-Aryan languages and in the future tense of Bengali and the Eastern Hindi dialects. Some interesting semantic arguments from Benveniste and Kurylowicz are referred to here. (See also Tournadre's article, which argues that ergativity in Indo-Aryan and Tibetan languages are really not the same phenomenon.) Zvelebil signs two brief notes, the first on the history and divergences of the epenthetic vowel of the Dravidian languages. He takes a 'morphous' morphology as his point de depart, which leads him in the first article to say that the epenthetic vowel varies according to the weight of the base. In the second

268 · The Yearbook (2000)

article, this position leads him to describe Dravidian languages as agglutinative and identify some empty morphemes and morpho-phonological phenomena. Krishnamurti argues that of the three identifiable gender/number patterns in the Dravidian languages, the original one is: (a) a masculine singular (human) pronoun; (b) feminine/non-human singular pronoun; (c) a masculine or masculine and feminine plural pronoun; and (d) a feminine/nonhuman plural pronoun. He then provides a useful summary of gender and number identification in 17 Dravidian languages. The two papers by Pilot-Raichoor discuss respectively the negative zero morpheme in south Dravidian languages and the verbal system of Badaga. The first argues that the phenomenon of having a 'bare root', with no tense morpheme attached to it to identify the negative in south Dravidian languages is linked with the general negation process in Dravidian languages. This process is identified as 'suspensive-reassertive', i.e. a weakening of the assertion, as opposed to Indo-Aryan languages, which have a recusative-type process, i.e. adding something to negate. The zero morpheme of negation is then simply the ideal realization of this process. Murugaiyan illustrates how the distinction between existential and equative sentences has weakened in Dravidian. However, the distinction is not lost, especially in negative sentences. There is also a discussion of the verb -/r, which has become a copula, and some new verbs, which require existential status. Courthiades wants to describe the phonological mechanism that gave the current dialectal system of Rromani. He believes it is better to classify these languages according to phonology and morpho-phonology, instead of phonetics and the lexicon, because the latter are influenced by the milieu of contact. Be that as it may, the author really loses me when he says that the Rromani writing system is 'diaphonological, noting the deep entities of the system9 (my translation and my emphasis). And that the diaphonemes are 'the deep structure underlying the different realizations'. To be charitable, it is true that the author only applies these notions to writing and inter-dialectal comprehension, but even so, this is a strong psychological claim that would need at least some support. (As a native French speaker, I can understand some Spanish, but does that mean that my underlying representation of a Spanish word is the same as the underlying representation a Spanish speaker would have of the corresponding French word?) Michailovsky argues that rhotacization in some Himalayan languages is not the same phenomenon as retroflexes in Indo-Aryan languages, which, except in two dialects, because of borrowing, does not exist in Himalayan languages. He admits that it is possible to consider a series of retroflexes, but they have a different origin, he argues. He has convinced me that the origin is different and it is true that some dialects did not even undergo the process, but it seems to me that whether this development was or was not influenced by Indo-Aryan languages remains to be seen.

Reviews m 269 Dasgupta illustrates how definitude is triggered in Bengali: (a) the choice between determiner and quantifier, some of the latter making the noun indefinite; (b) the choice between deictic and relative determiner being responsible for definite vs. indefinite; (c) numerals with a classifier make the noun definite when preceding it; (d) quantifier/classifier sequences always exclude definitude; and (e) case markers play a minor role, but it is said that their presence sometimes induces definitude. Anti-definitude, on the other hand, has a strange behavior: (a) 'about one' and 'about two' require different orders between classifier, numeral and noun; and (b) collective and aggregation markers are neutralized. Chevillard signs a paper on the -taan and -ee in Tamil. These are identified under the generic term 'particule' and it seems that they are to be interpreted as clitics. According to the examples, -taan is used to identify the theme of a statement (as opposed to the rheme) when it is not the first part of this statement, or to add emphasis when it is the first part or, again, to introduce a restriction to a statement. The use of ~ee is more problematic for the author, seemingly having many idiomatic interpretations, but a general 'identification' role of the secondary part of the sentence, he believes, can be extracted. The other articles included in the book are: Composite verbs in South Asian languages by Nespital; two other articles by Murugaiyan—Verbal locutions in Tamil and To possess and to feel in Tamil; Jacquesson's Morphologic expression of person in north-eastern Tibeto-Burmanese languages: A systematic use of the possible techniques; Tournadre's Particularities of Tibetan ergativity in respect to Indian languages; Gair's Realization in Singhalese; Kshirsagar's Nominal sentences in Ocjia and Marathi; and Paquement's Phonetic evolution and differentiation of Indo-Aryan dialects: Retroflex consonants J and LiebesX ('love' as in 'love letter'), based on a large number of models, is not complicated by the fact that -5 is not and never has been a genitive form of Liebe. Another problem of German compounding which becomes trivial by an affixational treatment is the adverb + noun type. Normally a compound with an adverb as left constituent is acceptable but not established and perspicuous as an adhoc formation, for example the words in 10: (10) Noch-Mitglied

'still + member', One who is still a member but will resign soon' Jetzt-Zeit 'now + time', 'present times' Irgendwie-Verständnis 'somehow + understanding', *vague understanding'

However, there is a small number of adverbs which have developed some productivity and which appear in a number of lexicalized compounds: (11)

Soforthilfe Sofortmaßnahme Sofortprogramm Nicht-Fachmann Nicht-Raucher Nicht-Linguist

'at-once H- help', 'emergency aid' 'at-once + measure', 'immediate measure' 'at-once 4- program', 'emergency-aid program' 'non + professional' 'non + smoker' 'non + linguist'

Despite the productivity of the pattern, the compounds in 11 still are not quite normal. Another very small set of compounds is perfectly normal and inconspicuous. When the first constituent means 'inside' or Outside', the adverb is used and the corresponding adjective is impossible; when it means 'front' or 'back', the reverse is the case: (12) Adv + N Innentemperatur Außentemperatur Außentemoeratur *Vomachse *Hintenachse

Adj + N *Innertemperatur *Äußertemperatur *ÄußertemDeratur Vorderachse Hinterachse

'inside temperature' Outside temperature' 'front axle' 'back axle'

Compound members can differ in productivity and can block each other—just like affixes.

292 · The Yearbook (2000)

To sum up: at least in the German language, there is no borderline between compounding and derivation, a fact which is perfectly accounted for in the theory of Singh and Dasgupta. • NOTES 1. Cf. Becker (1999) for a more detailed criticism of word syntax. 2. The word Donaudampfschiffahrtsgesellschaft ('Danube + steam + ship + voyage + company') is not a joke (as its extensions '+ captain's + widow...') but the official name of that shipping company. 3. Only suffixes beginning with a vowel are resyllabified. A few of those suffixes behave like stems because they are recently reanalyzed compound constituents (e.g. gut[.?]artig 'good + kind + y', 'benign'). 4. One could add that in German only one of the last three syllables of simplex words can be stressed; as endocentric compounds are stressed on the first constituent and compounding is recursive, an arbitrary number of syllables can follow the stress of a compound word. 5. Constants like BundesX are stressed when combined with a simplex noun or a common, lexicalized compound. 6. What has been said about German compounding in this paper is far from an adequate treatment of the problems, which cannot be achieved in a short article. The most thorough survey of the data is Ortner and Ortner (1984), the most thorough treatment of stress Doleschal (1988), who, in her conclusion, raises the question if it makes sense at all to assume stress rules for German compounds (ibid.: 25). She also points out that the regional variants differ considerably. My own idiolect is much closer to hers than to Stotzer's (1989). 7. Cf. Stepanowa and Fleischer (1985:143ff.). 8. See Becker (1992:10-16) for details.

• REFERENCES AronorT, Mark. 1976. Word formation in generative grammar. Cambridge, Mass.: MIT Press. Becker, Thomas. 1992. German compounding. Rivista di Linguistica 4.5-36. Becker, Thomas. 1999. Paradigmatic morphology. To appear in: Explorations in seamless morphology, ed. by R. Singh. New Delhi: Sage Publications. Doleschal, Ursula. 1988. Zum deutschen Kompositaakzent. Tema con variazioni. Wiener Linguistische Gazette 40-41.3-28. Giegerich, Heinz J. 1985. Metrical phonology and phonological structure. Cambridge: Cambridge University Press. Hockett, Charles F. 1955. A manual of phonology. Indiana University Publications in Anthropology and Linguistics; UAL memoir 11. Baltimore: Waverly. Moulton, William G. 1947. Juncture in modern standard German. Language 23.212-26. Ortner, Hanspeter, and Lorelies Ortner. 1984. Zur Theorie und Praxis derKompositaforschung. Mit einer ausführlichen Bibliographie. Tübingen: Narr. Singh, Rajendra, and Probai Dasgupta. 1999. On so-called compounds. The yearbook of South Asian languages and linguistics 1999 ed. by Rajendra Singh, 318-32. New Delhi: Sage Publications. Stepanowa, Marjja D., and Wolfgang Fleischer. 1985. Grundzüge der deutschen Wortbildung. Leipzig: VEB Bibliographisches Institut. Stötzer, Ursula. 1989. Zur Betonung dreiteiliger Substantivkomposita. Deutsch als Fremdsprache 26.263-65. Sturtevant, Edgar H. 1947. An introduction to linguistic science. New Haven: Yale University Press.

Second Degree Morphology: A Difficulty for the One Variable Constraint? ISYLVAINNEUVELI

• 1. Introduction The purpose of this intervention is to argue that although the one variable constraint of Ford Singh and Martohardjono (1997) used in Singh and Dasgupta (1999) is eminently reasonable, perhaps the theory within which it is embedded should be modified to account for what I shall refer to as second degree morphology. Compounds provide some of the more convincing arguments for the existence of word internal constituency structure. For the naive speaker, they seem clearly to be made up of two or more words. As a result, most theories of word-based morphology have struggled to provide an explanation for them. Some, having the courage of their convictions, have chosen to deny the existence of compounds as a separate class of words and to treat them like any other complex word. Others have tried to provide a different account of compounds, but in doing so have also forgotten the very principles on which word based morphology is supposed to be built. I will not, in these pages, argue the merits of word-based morphology. I will propose an analysis of compounds derived from very general cognitive principles that is both sufficient and in concordance with the idea that words have no internal structure other than a phonological one. I shall draw on some facts of French to support my argument. Last, but not least, I will show that, without having recourse to any theory dependent principles, we can make some predictions regarding possible diachronic changes in the morphology of compounds.

294 · The Yearbook (2000)

m 2. The Premise • 2.1 Reasoning on Morphology Because of its simplicity, I will use the formalism of Ford and Singh (1995) to illustrate my point but I believe that the same conclusions apply in amorphous morphology or in any theory of morphology in which words do not have (or at least are not supposed to have) any internal constituency relations. Both Anderson and Ford and Singh would accept that in acquiring its language, the child first learns words, then creates relations between those words and that is all that is necessary for my demonstration to hold. Besides being more restrictive than traditional Päninian models (cf. Ford et al. 1997), one of the biggest advantages of any word-based morphology is the fact that its acquisition is easily explained as it relies on positive evidence alone. In fact, in a word-based model of morphology, acquisition boils down to garden variety induction, to use Peirce's denomination, or what I prefer to call generalization. This parallel with other spheres of cognition can be illustrated with simple inferences. Consider the following statements: (1) a. Dressier likes apples. b. Kiparsky likes apples. Just as anyone knowing facts a and b may conclude that any linguist likes apples, a child knowing the French words facile 'easy', facilement 'easily', rapide 'quick' and rapidement 'quickly', may infer that for any French adjective of the form X, there exists an adverb of the form Xment. In Ford and Singh's morphology, the morphological relation between two words of a language can be represented by a generative rule of the following form (cf. Ford and Singh 1995): where [X]A and [X']B are semantically related words A and B are categories ['] is the formal difference between both words In this formalism, X is referred to as the VARIABLE, and ['] is referred to as the CONSTANT which brings us to one of the necessary principles of their theory of morphology: (2) any morphologically related pair of words can be analyzed in terms of only one variable and only one constant at a time. Once again, this can be equated with any non-linguistic generalization. In example 1, Dressier and Kiparsky are reanalyzed as the variable 'any linguist'

Dialogue m 295 and 'to like apples' is reanalyzed as the constant. Principle 2 is easily justified for if we do not limit ourselves to only one variable, we could conclude from la and 'Kiparsky drives a car' that anything can be said of any linguist.

• 3. The Problem • 3.1 To Compound or Not to Compound It is obvious why some would analyze a word like can-opener as being made of two words. Words that appear to be made up of two nouns are quite common in English, and French also has a rather large number of words of the type ouvre-bolte 'can-opener', coupe-papier 'paper knife', couvre-chaussure Overshoes'. The 'combination' of a verb and a noun seems productive enough in French to suggest a rule of the type: (3) [X]V,3SING[Y]N«->[XY]N with [XY]N given an interpretation similar to 'tool used to do X to Y' Besides the fact that a rule like 3 seems in direct conflict with principle 2, Ford and Singh note the following (cf. Ford et al. 1997): (4) Compounds like the French ouvre-bolte or English doghouse, belong to paradigms where only one 'part' of the so-called compound can be freely replaced. In every case, the replacement of the other part is restricted to a certain degree (ouvre-boite, ouvre-bouteille, but not ferme-boite 'can-closer'). Because of empirical observations like 4 and, I would assume, because principle 2 is vital to their theory, Ford and Singh's theory of morphology does not recognize compounds as anything special. In their model, a word like ouvrebolte 'can-opener' is instead morphologically é elated to the noun bolte 'can' by a rule like 5 that contains only one variable and in which the 'verbal' part of the word is a constant that looks like a word, but is not: (5) [X]N NN where the underline indicates the head of the compound

Dialogue m 297 The rules belonging to this set are syntactic in nature; they create complex words with internal structure and may, though they do not always, specify one component of such complex words as its head. Needless to say, there is nothing A-morphous about this. Anderson provides a few arguments to justify this treatment of compounds. First, he cites Selkirk's observation that 'the result [of compounding rules] belongs to some lexical category (not necessarily that of any of its parts) and it fills the same slots as non-compound members of the same category' (Anderson 1992: 294). The second part of this observation is hardly an argument in favor of internal structure. The idea that a noun, whether it is a compound or not, can fill the same slots as another noun should not come as a surprise. As for the first part ofthat statement, if we assume that compounds are words like any other, that is without internal structure, and that they are the product of WFR, the fact that they can belong to a different lexical category naturally follows, since that is precisely what WFR's do. If, on the other hand, they are the product of syntactic rules, then we must explain why this particular subset of syntactic rules can give a lexical category to a word. The French verb + noun word ouvre-boüe, for example, is a masculine noun, even if its noun component is feminine. More importantly, the meaning of this word is much greater than the sum of the meaning of its parts (i.e. the meaning 'tool' must come from somewhere). Adding meaning and changing lexical category is something we would expect to be done in morphology, not in syntax. Another argument used by Anderson is that of English words like scrubwomen or outdid in which, according to Anderson, the lexical items woman and do must be available for inflection. It is obvious that these forms must be Separate' at one point or another, but it does not have to be in the output of the WFR. Morphological rules using inflected forms as an input are not at all impossible, contrary to what is stated in A-morphous morphology. French adverbs like those in 8 are often 'made' from an adjective inflected for gender.2 (8) doucement 'gently', ouvertement Openly', franchement 'frankly', vivement 'sharply',grandement 'greatly', etc... The introduction of WSR in morphology not only goes against the very principles on which word-based morphology is built, it also does not provide any explanation for the facts described in 4,6b and 6c.

• 4. The Solution As we have said before, in a word-based model of morphology, acquisition is possible through generalization, a process already present in other spheres of cognition. In example 1, knowing that both Dressier and Kiparsky like apples, we inferred that this is true of any linguist. Let us now assume that

298 · The Yearbook (2000)

our knowledge has doubled since then and that we now know the following facts: (9) a. Kiparsky likes peaches, b. Dressier likes peaches. Once again, through generalization, we may reach the conclusion that any linguist likes peaches and add that conclusion to the bulk of our knowledge. This is precisely where Ford and Singh's theory leaves us. But the ability to generalize does not die there and because the conclusions we have reached are added to our knowledge, they are also available for further use, as in 10: (10) a. Any linguist likes apples. b. Any linguist likes peaches c. Any linguist likes any fruit. What is important here is that we did not break any law in acquiring the wisdom of lOc, we simply reanalyzed the facts in lOa and lOb as one constant ('any linguist') and one variable ('to like any fruit'). The fact that our constant in this generalization was our variable in the previous one does not violate principle 2 in any way and is, in fact, totally irrelevant. Most of the tasks we perform in our everyday life involve multiple degrees of abstraction. Second degree generalizations are not only possible, they are an integral part of our mental capacity to generalize. While I believe linguistics should not make use of'language organ' driven devices when general cognitive principles will do, I also believe that any linguistic theory that calls upon a cognitive capability should not impose language-proper limitations on that capability. At the very least, such theory should offer some explanation as to why our ability to generalize would somehow be diminished when it comes to language. Ford and Singh and Singh and Dasgupta refuse to consider compounds as a special class of words because it would allow for morphological strategies like 3 that include more than one variable, which, as we have shown, seems to violate principle 2, a principle also present in non-linguistic generalization. I have already shown that second degree generalizations are possible outside the realm of linguistics and that those generalizations do not violate principle 2, even if the constant in a second degree generalization may have been a variable in a first degree generalization. Let us now see how this translates in morphology. First degree generalizations are based on actual words. In 11, two different first degree generalizations are made from several pairs of words. The morphological strategies produced by these generalizations are not really compounding rules since the constant in these strategies is not an actual word but is best described as a suffix.

Dialogue m 299 (11) a. boite 'can'louvre-boite 'can-opener' bouteille 'bottle 'louvre-bouteille 'bottle opener' [X]N [uvrX]N,masc I '[uvrX]' = 'tool used to open X' b. ongle 'naiV/coupe-ongle 'nail clippers' papier 'paper5/coupe-papier 'paper knife' [X]N [kupX]N,masc I '[kupX]' = 'tool used to cut X' The morphological strategies produced in 1 la and 1 Ib not only include a form identical to that of the French verbs ouvnr Open' and couper 'cut', but the meaning of these verbs is also part of the semantic interpretation associated with those strategies. Because they are present in both form and meaning, the verbal predicate part of each compound can be reanalyzed as a variable. (12) a. [X]N [uvrX]N,masc 1'[uvrX]' = 'tool used to open X' b. [X]N «-> [kupX]N, ma« I '[kupX]' = 'tool used to cut X' c. [X]v,3rdsing [Y]N [XY]N I '[XY]N' = 'tool used to do X to Y' The second degree generalization formalized in 12c is made possible by three different factors: 1. The morphological strategies on which it is based are similar in form. 2. The semantic interpretations associated with these strategies intersect. 3. The meaning of each compound is 'transparent' to the meaning of each of its parts. It is important to note that a morphological strategy such as 12 is consistent with every principle of word-based morphology. The output of 12 can be described as one word, not two, it does not have any internal structure other than a phonological one and it was made using only one variable and one constant.

• 5. The Results Second degree generalizations may at first glance be taken to be far less restrictive than they actually are. Even if second degree morphology can account for the compounds actually found in languages, it must also make some predictions as to the ones that are not. For there to be a 'useful' second degree generalization, there must first be two different morphological strategies that share both form and meaning. For example, the English morphological strategies relating to doghouse and coffee table could not give rise to a second degree generalization like 12 since there is no significant semantic intersection between both rules: A doghouse is clearly not a house and the meaning 'coffee' is buried so far deep in that of coffee table

300 · The Yearbook (2000)

that the only possible generalization from the morphological strategies relating to these words would be one of form only, creating a sort of meaningless word pattern. I will assume here that such semantically empty word patterns can be created, and I will ask the reader to accept the following postulates: (13) a. The more abstract the reasoning, the more difficult it is. Second degree generalizations are more cognitively costly than first degree ones. b. Morphological strategies with meaning are more likely to be created and used than semantically empty ones. Having made the assumption that meaningless word patterns are possible, we must also accept that a third or fourth degree generalization in which the lexical categories are taken as a variable is also possible. The morphological strategy thus created would allow us to combine any two words, regardless of their category. We already know that this is possible since, as we noted in 6c, previously uncombined categories have been introduced in languages over time. If we take 13a and 13b to be true, we can then make some predictions regarding the introduction of new compounds in any given language. (14) In order of 'likeliness': a. First degree generalizations: compounds belonging to an existing paradigm as described in 4. b. Second degree generalizations: productive combinations with a precise semantic interpretation like the French verb + noun 'tool' nouns. c. Second degree semantically difficult or impossible to interpret word patterns: existing combinations without clear semantic interpretation. d. Third degree semantically difficult or impossible to interpret word patterns: previously uncombined categories. This ordering might seem so obvious as to not evoke much interest, but I believe that providing an explanation for the obvious is a prerequisite for any linguistic theory. It is precisely the fact that second degree morphology can offer an extremely simple explanation for extremely simple facts that makes it appealing.

• 6. Conclusion We have seen that Ford and Singh's and Singh and Dasgupta's treatment of compounds is too restrictive and cannot account for many of their properties; that A-morphous morphology's reliance on syntactic rules and semantically

Dialogue m 301 empty word formation rules goes against the core principles of word-based morphology. I have stipulated that morphology relies on generalization for its acquisition, a cognitive process present even outside the realms of linguistics. Since a second degree of generalization is present in other spheres of cognition, there is no reason to deny its availability to morphology. Second degree morphology can account in a very simple way for the many apparently conflicting properties of compounds, all in concordance with the principles on which word-based morphology is built. It allows us, as Ford and Singh (1995) argue, to include all phonic changes triggered in morphological context in the morphological strategies themselves, thus eliminating morphonology and allowing for an autonomous phonology. While I have not offered irrefutable proof of the existence of a second degree of abstraction in morphology, I believe, because of its simplicity and its cognitive justification, that second degree morphology must be taken as a default extension. Denying the existence of this second degree would require either a demonstration that morphology does not rely on generalization or some explanation as to why the ability to generalize is somehow handicapped when it comes to language. This being said, there are still many questions to be answered. For instance, is second degree morphology a part of grammar in every language or does its exercise constitute language change? Where does grammar end and creativity begin? Is there a defined boundary between the two? I will obviously not attempt to answer these questions here but I hope my response to Ford, Singh and Dasgupta and my proposal will at least succeed in provoking some discussion on the subject. • NOTES 1. While this attempt at a minimal definition may be criticized, what is important for our discussion is that this rule includes, without a doubt, the meaning of the verb ouvnr. 2. Note that the debate on what can serve as an input for a WFR disappears if, like Ford and Singh, we assume that morphological rules are bi-directional relations between words.

• REFERENCES Anderson, Stephen R. (1992). A-morphous morphology. Cambridge: Cambridge Studies in Linguistics. Ford, A. and R. Singh. (1995). Quelques avantages d'une linguistique dobarrassee de la morpho(pho)nologie, In Trubetzkoy's Orphan. Proceeding? of the Montreal Roundtable on Morphonology: contemporary responses, 119-39. Amsterdam: Benjamins. Ford, A., Rajendra Singh, and Gita Martohar^jono. (1997). Pace Panini, towards a word-based theory of morphology. New-York: Peter Lang. Singh R. and P. Dasgupta. (1999). On So-called Compounds. The yearbook of South Asian languages and linguistics 7999, ed. by Rajendra Singh, 265-75. New Delhi: Sage Publications.

In Praise of Sakatäyana: Some Remarks on Whole Word Morphology* • RAJENDRA SINGH and ALAN FORD · • 1. Introduction Pänini's Asiadhyäyi is, as Thieme (1971) points out, an extended argument, presumably contra Sakatäyana1 and others, that a grammar can be built up with small units, particularly in the domain of morphology. For Pänini, the word is an entirely derived entity, something made up of smaller pieces, put together according to the combinatorics he provides. The parts that enter into his combinatorics are according to him all real (cf. Deshpande 1997). Both his atomistic ontology and his methodology have been questioned in the immanent critique that begins with Patanjali's insistence on nityatva and ends with Bhartrhari's demonstration that words are seamless wholes and that the parts Päninians delight in coming up with are at best grammatical fictions (cf. Singh 1998). Bhartrihari's critique is, however, ignored by most—even his defenders, such as Kelkar (1999), who see him primarily as a philosopher of language, obviously making a distinction he would have abhored, and do not draw what seem to us to be obvious grammatical conclusions from his insistence, for which he provides several substantial arguments, on seamlessness (for some recent examples of insistence on [a somewhat differently motivated and grounded] seamlessness, see Starosta [1999 and forthcoming]). Post* We are grateful to Probal Dasgupta, Wolfgang Dressier, Ashok Kelkar, David Stampe, and Stanley Starosta for convincing us that it was better to show precisely how what we say can be said is in fact said with the 'minimalist' formalism we propose for morphology. We are also grateful to Sylvain Neuvel for drawing our attention to certain matters of exposition.

304 · The Yearbook (2000)

Renaissance grammatical practice in the West almost completely abandons the Greco-Roman constmal of morphology as a study of relationships of shapes of whole words, and ends up, perhaps aided in this transition by its increased exposure to Hebrew grammarians, adopting the Päninian position, later espoused by leading structuralists from Saussure (cf. Singh 1992 and Vajpeyi 1997) to Bloomfield. Despite some modern attempts to revive the ancient Greco-Roman practice (cf. Robins 1959 and Matthews 1974, in particular), the Päninian way of doing morphology has been dominant for centuries now, possibly because a fully formalized full-fledged alternative had not been made available until recently (neither fish nor fowl attempts such as Anderson 1992 actually end up supporting the Päninian view, as Sadock [1995] is happy to note). The purpose of this short dialogue-initiating note is to outline the alternative that has been available at least2 since Ford and Singh (1991), to show some of its applications, and to invite South Asianists to tell us why the Päninian view of morphology should be preferred. Hoping to shift the burden of proof, we shall concentrate not on the critique of that view, best characterized as morphemology (a la Janda 1983), but on the presentation of whole-word morphology.

• 2. The Theory All that needs to be said about word structure in any language (of any type whatsoever) can and must be said by instantiations of the schema in 1 below. We refer to these instantiations as W (ord) F(ormation) S(trategies) because as generalizations drawn from known particular facts, they can be activated in the production and understanding of new words (cf. Ford and Singh 1991; Ford et al. 1997)3: WFSs must be formulated as generally as possible, but, and this is crucial, only as generally as the facts of the matter permit. 1. /X/a/X7b where: a. /X/a and/X'/b are words and X and X' are abbreviations of the forms of classes of words belonging to categories a and b (with which specific words belonging to the right category can be unified or on to which they can be mapped). b. ' represents (all the) form-related differences between/X/and/X7 c. a and b are categories that may be represented as feature-bundles d. the