The Yearbook of South Asian Languages and Linguistics: 1999 9783110245240

Citation preview

EDITORIAL BOARD CHIEF EDITOR: Rajendra Singh, D6pt. de Linguistique, Universite* de Montreal, C.P. 6128 Succ. Centre-ville, MontrSal 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: (91^0) 258-500; Fax: (91-40) 258-120; e-mail: pdgalts@ uohyd.ernet.in K.P. Mohanao, Department of English, National University of Singapore, Singapore 119260. Voice: (65) 772-6042; Fax: (65) 773-2981; e-mail: [email protected] REGIONAL EDITORS: RJC. Agnihotri, University of Delhi, India; A.P. Andrcwskutty, University of Kerala, Kariavattam, Thiruvananthapuram, Kerala, India; Tfej K. Bhatia, Syracuse University, NY, USA; W.S. KaninatUlake, University of Kelaniya, Sri Lanka; Baber S.A. Khan, UAE Air Force Academy, Al-Ain, United Arab Emirates; Rajend Mcsthrie, University of Cape Town, South Africa; Tsuyoshi Nara, Seisen University, Tokyo, Japan; John Peterson, Universität Zurich, Switzerland; Tariq Rahman, National Institute of Pakistan Studies, Islamabad, Pakistan; A.P. Saleemi, Aarhus University, Denmark; Udaya Narayana Singh, University of Hyderabad, India; E. Tifffou, Universito de Montreal, Canada; Yogendra R \adav, Royal Nepal Academy, Kathmandu, Nepal. EDITORIAL ADVISORS: E. Annamalai, Max Planck Institut für Psycholinguistics, Nijmegen, The Netherlands; Ron E. Asher, The University of Edinburgh, UK; Bernard Comrie, University of Southern California, Los Angeles, USA; Wolfgang U. Dressler, Universität Wien, Vienna, Austria; Aravind Joshi, University of Pennsylvania, Philadelphia, USA; Ashok R. Kelkar, Pune, India; Paul Kiparsky, Stanford University, USA; E.E 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 ASSISTANTS: Luc V. Baronian and Noureddine Kahlaoui, University de Montreal, Canada. 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 them. 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.

The Yearbook of South Asian Languages and Linguistics 1999 Chief Editor

RAJENDRA SINGH

Available From:

SERIALS SUBSCRIPTION SERVICE (INDIA)

SUBSCRIPTION AQENT · BOOK SELLERS PUBLISHERS · STOCKISTS OF BACK-VOLUMES 4, B.S.8. HALL, KHYBER PASS MARKET, CIVIL LINES DELHI-110054 (INDIA), PH.: 3811659 FAX; 9M1-3812676. E-mail: »arHitOMtytm.net.In BRANCH OFFICE: 4830/24, AN8ARI ROAD, DARYA QANJ NEW DELHI-110002 (INDIA). PH.: 3245225

Sage Publications New Delhi · Thousand Oaks · London

Copyright © Rajendra Singh, 1999 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 1999 by Sage Publications India Pvt Ltd M-32 Market, Greater Kailash-I New Delhi-110 048 Sage Publications Inc /^W\ 2455 Teller Road 13k I Thousand Oaks, California 91320 ^æ/

Sage Publications Ltd 6 Bonhill Street London EC2A 4PU

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

ISSN: 0971-9539

/UNIVERSITATS (^ BIBLIOTHEI BIBLIOTHEK \ LEIPZIG

ISBN: 0-7619-9339-8 (US-hb) 81-7036-^04-9 (India-hb) Sage Production Team: Abantika Chatterji, N.K. Negi and Santosh Rawat

Contents Commentum Editons

7

A. INVITED CONTRIBimONS The Urdu and Hindi Ergative Postposition ne: Its Changing Role in the Grammar ELENA BASHIR What Has Bhartrhari Got to Say on Language?

37

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes

53

ASHOK R. KELKAR

MARIO PALASCHKE and WOLFGANG U. DRESSLER

11

B. OPEN SUBMISSIONS Specificity in the Bangla DP TANMOY BHATTACHARYA The Syntax of Defmiteness in Oriya KALYANAMAUNI SAHOO Expressives in Tamil: Evidence for a Word Class CAROLINE R. WILTSHIRE The Complexity of Maithili Verb Agreement

YOGENDRA P. YADAVA

71 101 119 139

C. REGIONAL REPORTS, REVIEWS AND ABSTRACTS Regional Reports 1. Europe

JOHN PETERSON

155

2. India: Indie

161

3. India: English in India

184

4. Nepal

198

UDAYA NARAYANA SINGH R.KL AGNIHOTRI

YOGENDRA P. YADAVA

6 · Contents Review Article The Widening Gyre of Englishness: Reversing an felitist Myth, a review of A Parakrama, De-hegemonizing Language Standards: Learning from (Post)Colonial Englishes about English OTTO M. IKOME Reviews Rofend J.-L. Breton, Atlas of the Languages and Ethnic Communities of South Asia by E. ANNAMALAI Venceta Dayal, Locality in WH-Quantification: Questions and Relative clauses in Hindi by AYESHA KIDWAI Lachman Khubchandani, Revisualszing Boundaries: A Plurilingual Ethos by CATHERINE MONTGOMERY Rajendra Singh (ed.), Grammar, Language, and Society: Contemporary Indian Contributions byP.G.PATEL Rajendra Singh and R£L Agnihotri, Hindi Morphology: A Word-Based Description by PETER GAEFFKE Kashi Wali and Omkar N. Kool, Kashmiri: A Cognitive-Descriptive Grammar

by RAKESH M. BHATT

211

225 230 236 242 245 247

á DIALOGUE Linguistic Revolutions: A Reply to Thnmoy Bhattachaiya JOSEF BAYER On Purism and Idealizations in Linguistic Theory: A Response ANJUM P. SALEEMI On So-called Compounds

RAJENDRA SINGH and PROBAL DASGUPTA

253 257 265

E. ANNOUNCEMENTS The Chatterjee-Ramanujan Prize Housekeeping

279 279

About the Chief Editor Notes on Contributors

281 282

Commentum Editoris We are delighted to put the second issue of The Yearbook of South Asian Languages and Linguistics in your hands. Like its predecessor, it could not have seen the light of day without the help of our editorial advisors, contributing regional editors, authors, reviewers, and the members of the Sage team. We are grateful to them. We are also grateful to our referees, who must remain anonymous, and to Noureddine Kahlaoui and Luc V Baronian, our editorial assistants for this issue. We are happy to have been able to persuade Mario Palaschke and Wolfgang U. Dressler, Ashok Kelkar, and Elena Bashir to accept our invitation to contribute to this issue of The Yearbook. Both Bashir and Palaschke and Dressier deal with classic problems in South Asian linguistics. Whereas Bashir takes on the celebrated ergative marker ne in Hindi-Urdu and the new dimension it seems to have acquired in Pakistan, Palaschke and Dressier take on the wellknown problem of assimilation in Middle Indie. In his contribution, Kelkar presents the system and insights of Bhartrhari, a grammarian-philosopher of the 5th century AD modern linguists would do well to read. The Open Submissions section contains contributions on Bangla, Maithili, Oriya, and Tamil. Two of these (Bhattachaiya and Sahoo) deal with two manifestations of definiteness in Indie—Bangla and Oriya—and invite further explorations of that theme in South Asian languages. As contributions to the Dialogue section will start coming in in greater numbers only after the inaugural issue has reached our audience, two of us had to step in once again. From now on, we sincerely hope to be able to eliminate ourselves as contributors. Possibly because it was not available to some of our authors, our style-sheet was not adhered to in a rather large number of cases. To avoid that problem, we are reproducing it in this volume. Potential contributors can now consult both the style-sheet and The Yearbook: we will, in other words, not process future submissions requiring unnecessary editorial work. Although this is not quite the place for a campaign-plea, we must underline the fact that your support, in the form of individual and institutional standing

8 · Commentum Editoris orders for The Yearbook, is essential for its continued survival. We promised a world-class annual on South Asian languages, and we think we have kept that promise. You promised continued support, and we are confident that it will be forthcoming. We are, as always, grateful to Abantika Chatterji, N.K. Negi and Santosh Rawat, members of the Sage team responsible for the production of this volume. Without their help, we would not have been able to do anything. Rajendra Singh Probal Dasgupta K.P. Mohanan

Invited Contributions

The Urdu and Hindi Ergative Postposition ne: Its Changing Role in the Grammar* lELENABASHIR·

This paper explores one area in which Urdu spoken in Pakistan is diverging from Urdu as spoken in India and from Hindi: developments in the use of the ergative postposition ne. The primary use of ne is as the marker of the subject (agent) in tenses of transitive verbs based on the perfective participle. In Pakistan, however, a new function for ne is emerging in Pakistani Urdu, where the construction NP ne INF AUX has become widespread. A potential contrast has thus developed between sentences like mujhe karaaciijaanaa hai Ί have to go to Karachi", and maiNne karaacii jaanaa hai Ί am going to/want to go to Karachi'. Earlier studies proposed that ne was becoming a marker of purposivity or conscious choice; however, examining the NP ne INF AUX construction separately for 1st, 2nd, and 3rd person sentences leads to the conclusion that ne is being regrammaticized as a generalized marker for the archi-case category SOURCE, rather than, as has previously been suggested, as a marker of (specifically) CONSCIOUS CHOICE. Use of this construction in assertions with 2nd person subjects reveals particularly interesting interactions between the semantics of SOURCE, the 'potential/ prospective' semantics of the infinitive, and the effects of discourse role. For example, in a sentence like aap ne mere saath jaanaa hai 'You have to/will go with me', the semantics of SOURCE is manifested in the speaker's desire/anticipation that the action take place/happen, and the addressee's assumed willingness to do it. The systemic result of these developments with ne is that the functional territory of ko (DATIVE/ACCUSATIVE postposition) is being restricted to situations where unambiguous external compulsion is involved. In addition, ko is acquiring certain discourse and socio-linguistic markings, e.g. formality, deference, solemnity, seriousness. Sources of the reanalysis are examined and possible future scenarios laid out.

* This paper is a revised version of a draft originally presented at the Michigan Linguistics Society Symposium held at Central Michigan University, 18 October 1997.1 am grateful for comments on the paper by conference participants Anvita Abbi, Peter Hook, and S. Radhakrishnan. Since then, Miriam Butt, Vit Bubenik, G.C. Narang, and Ruth Schmidt have made valuable comments on the paper.

12 · Elena Bashir

• 1. The Prototypical Function of ne This paper discusses developments in the use of the postposition ne in the Urdu of Pakistan. Urdu is typically described as a split-ergative language, in which ergativity is conditioned by tense-aspect: subjects of transitive verbs in tense-aspect forms built on the perfective [(y)aa] participle take ergativemarked subjects, while subjects of intransitives and transitives in nonperfective tense-aspect forms are nominative. The primary use of ne is as the (ergative) marker of the subject of tenses of transitive verbs based on the perfective participle.1 A typical example ofne in this function is shown in la; while Ib shows a non-perfective form of the same transitive verb, and Ic a perfective form of an intransitive verb.2 (1) a. laRkii ne anDaa khaayaa girl ERG egg (m.sg.) ate (m.sg.) 'The girl ate an egg'. laKke ne roTii khaayii boy(OBL) ERG bread (f.sg.) ate (f.sg.) 'The boy ate bread/food'. b. laRkii anDaa khaayegii girl(NOM) egg will eat (f.sg.) 'The girl will eat an egg'. c. laRkaa baazaar gayaa boy(NOM) bazaar went (m.sg.) 'The boy went to the bazaar'. In la the finite verb agrees with the direct object in number and gender; while in Ib it agrees with the subject in person, number, and gender and in Ic with the subject in number and gender.

• 2. Semanticization of ne It has long been noted in standard pedagogical grammars and linguistic literature (e.g. Barker 1967; Tuite et al. 1985; Hook et al. 1987) that some intransitive verbs also optionally allow ne in the past tense.3 For example 2: (2) a. raam khaaNsaa Ram (NOM) coughed 'Ram coughed'. (Tuite et al. 1985:264) b. raam ne khaaNsaa Ram ERG coughed 'Ram coughed (purposefully)', (ibid.: 264) The option ofne with this set of verbs creates the possibility of a semantic contrast between pairs of sentences such as 2a and 2b, in which 2a, the usual case,

The Urdu and Hindi Ergative Postposition ne · 13

is neutral with regard to volitional meaning(s), while 2b carries a meaning of volitionality or purposivity.4 Butt (1995), arguing that case is semantically rather than structurally assigned in Urdu, states that the contrast between V-intransitive SubjectNOM and V-intransitive Subject-ERG for these verbs expresses a difference in semantics.5 Following Mohanan's use of this term for Hindi, she calls this parameter CONSCIOUS CHOICE, adducing examples 3a and 3b to illustrate the contrast. (3) a. woh ciikhaa he scream (PFV-m.sg.) 'He screamed (despite himself)'. (Butt 1995:110) b. us ne ciikhaa he ERG scream (PFV-m.sg.) 'He screamed (on purpose)', (ibid.: 110) Some transitive verbs also show occasional non-appearance of ne. For example: (4) a. maaf kiijiegaa maiN aapko pahcaanaa nahuN excuse me I (f.) NOM you ACC recognized (m.sg.) not 'Excuse me, I didn't recognize you'. (PTV drama, Kaharen) The sentence in 4a occurs when a character answers the phone to find someone unexpected whom she doesn't recognize; it reports an inadvertent, nonvolitional act of non-recognition. Usually pahcaannaa 'to recognize' takes a ne-marked subject in perfective past tenses. Example 4a is especially interesting in that the verb pahcaanaa 'recognized' (m.sg.) does not agree with the subject (f.sg.) as one would expect it to do in the absence of the postposition ne.6 What may be happening here is that although ne is dropped, there remains a 'memory trace' of its presence, indicated by the retention of the (default) m.sg. agreement that the /ze-marked subject would have had.7 In 4b a more typical appearance of a perfective tense-aspect form of pahcaannaa is shown. b. jii maiN ne aap ko pahcaan liyaa hai yes I ERG you ACC recognized (default m.sg.) AUX 'Yes, I recognize you (from a previous meeting)'. It seems that with those verbs that allow both the presence and absence of Aie-marking on their subjects in finite sentences, its PRESENCE with intransitives is marked for [+ CONSCIOUS CHOICE], while its ABSENCE with transitives is marked for [- CONSCIOUS CHOICE], or inadvertent action. The absence of ne with intransitives is unmarked with regard to the CONSCIOUS CHOICE parameter, as is its presence with transitives. That is, the morphologically atypical form encodes the semantically atypical situation for each type of verb.

14 · Elena Bashir

m 3. The ne + Infinitive Construction In the Urdu of Pakistan and the Hindi of Delhi, ne has acquired another function. A construction consisting of NP-ne + INFINITIVE + tensed AUX is common with both transitive and intransitive verbs.8 This usage appears to have been in existence for at least 30 years, since it is mentioned specifically in B. Singh (1966:27), who while discussing the Hindi of Delhi, gives the following examples, commenting that speakers 'replace ko by ne, influenced by Panjabi'.9 (5) a. maiN ne baazaar jaanaa hai Ί have to go to the market'. (B. Singh's glossary) b. ham'ne khaanaa khaanaa hai 'We have to take our meal'. (B. Singh's glossary) It is also mentioned by Nairn (1975, vol. II: 14): '...a dialectal construction in which/nee/instead of/koo/is used after the real subject [emphasis mine], has become acceptable in magazines and newspapers published in Pakistan, though it is still used only by those writers whose mother tongue is Punjabi'.10 Nairn's prototypical examples of this construction are: (6) a. maiN-ne jaanaa hai Ί have to go'. (Nairn's glossary) b. maiN-ne kitaab paRhnii hai Ί have to read the book'. (Nairn's glossary) The glosses of these early examples (5 and 6) use the phrase 'have to', which is ambiguous between the readings of internal or external necessity. Tiiite et al. (1985) first propose that ne does not simply replace ko but provides a semantic contrast, illustrated in 7a and 7b, (7) a. mujh ko jaanaa thaa I DAT go (INF) AUX (p.) Ί had to go'. (Tuite et al. 1985; their glossary) b. maiN ne jaanaa thaa I ERG go (INF) AUX (p.) Ί had to go'. (T ite et al. 1985; their glossary) stating that '7a implies that an external necessity made the subject want to leave; 7b implies that the subject intended to go because of his own plans or purposes' (ibid.: 265). They make the explicit proposal that: 'ne, at one time a true ergative case marker, has been refunctionalized as a marker of purposivity' (ibid.: 265). Butt and King (1991:34), also discussing this topic, give examples like 8, in which 'the pairs differ only in terms of the case on the subject, and this difference accounts for the difference in meaning'.

The Urdu and Hindi Ergative Postposition ne · 15

(8) a. anjum ne ghar jaanaa hai Anjum ERG home go (INF) AUX Anjum wants to go home'. (Butt and King's glossary) b. anjum ko ghar jaanaa hai Anjum DAT home go (INF) AUX Anjum has to go home'. (Butt and King's glossary) In the discussion that follows, I shall use the term 'conscious choice' in a sense equivalent to that of Tuite et al.'s 'purposivity'. The most frequently cited (constructed) example of this usage is maiN-ne jaanaa hai Ί want to/have to go', in which the agent is 1st person. The other constructed examples presented so far have either 1st person or 3rd person (full NP, proper name) subjects. Interestingly, the prototypical example, maiN ne jaanaa hai seems to have either dominated thinking about this construction or been simply reproduced by subsequent authors from the early mentions of the usage. Perhaps it has been focused on because it is intransitive while the original use ofne is with transitives. Restriction of discussion of this usage to the prototypical examples with jaanaa has resulted in lack of attention to the behavior of transitive sentences of this type, for example the fact that the INF agrees with non-&0-marked objects in the same way as it does in other constructions, e.g. caah- 'want to' constructions (see examples 17,18,19,33,34).11 Therefore, in the interest of exploring the behavior of ne more deeply, we need a wider variety of data which include the kind of context that can only come from connected discourse. The following collection of examples is the result of: (1) occasional noting of utterances from various Pakistan Television broadcasts and spontaneous speech occurrences; (2) listening to the full video recording of the television drama Tanhaiyan;12 (3) listening to the full video recording of the television drama/lfl/icA;13 (4) two episodes of the television senesAndhera Ujala.14 I have classified these utterances by person, in an effort to identify any differences in the semantics of 1st, 2nd, and 3rd person sentences, and to determine whether any clues about the development or extension of this usage through the paradigm can be found.15 • 3.1. 1st Person (9) aap se maiN ne ek sawaal karnaa hai you from I ERG one question do (INF) AUX Ί want to ask you a question'. (Pakistan Television [PTV] program, Cricket: The Real Thing, 18 March 1996 [+ CONSCIOUS CHOICE] (10) mujhe bhuulgayaa ki maiN ne aap ko I (DAT) forgot that I ERG you DAT kis taraf le jaanaa which direction take Ί forgot which direction I want to take you in'. (PTV program, Cricket: The Real Thing, 18 March 1996) [+ CONSCIOUS CHOICE]

16 · Elena Bashir

(11) ham ne Dalnaa hogaapyaaz we ERG put (INF) AUX(fut) onions 'We need to/will add onions'. (PTV program, Pot Luck, 16 September 1996) [+/- CONSCIOUS CHOICE] (12) saarii zindagii to maiN ne unglii pakaR-ke whole life TOP I ERG finger hold-CP nahiiN naa calnaa thaa, xud kabhii na not EMPH walk (INF) AUX(p.) self sometime or kabhii to other TOP calnaa siikhnaa thaa, to siikhliyaa walk (INF) learn (INF) AUX(p.) so learned couldn't go through my whole life being led by the hand; sometime or other I had to learn to walk by myself, so I have learned'. (Tanhaiyan) [+/- CONSCIOUS CHOICE] (13) haaaai, do baj gaye das baje maiN ne oh no, 2 o'clock went 10 o'clock I ERG aufis jaanaa thaa aaj merii jaub kaa office go (INF) AUX(p.) today my job of pahlaa din hai first day is Oh, no, it's two o'clock; I was supposed to go to the office at ten— today is the first day of my (new) job'. (Tanhaiyan) [+/- CONSCIOUS CHOICE] (14) Bibi: aur phir kaunsaa maiN ne us malkaa atizbaiT— and so what I ERG that queen Elizabeth wo kyaa naam us kaa us se guftaguu karnό hai. that what name her her with conversation do (INF) AUX 'And I'm not going to be talking/have to talk with that—what's her name—Queen Elizabeth'. [+/- CONSCIOUS CHOICE] Sania: aap ko inglii siikhnaa paRegaa you DAT English learn-INF will befall 'You will have to learn English'. (Tanhaiyan) [-CONSCIOUS CHOICE] (15) A: kyaa baat hai beTe, aafis meN koii jhagRaa hogayaal what matter is son office in any quarrel became 'What's the matter, son, was there a quarrel in the office?' B: nahiiN 'No' A: bidaa ne kuch kahdiyaahai Bida ERG something said 'Did Bida say something?' C: maiN ne kyaa kahnaa! do tun din se aise I ERG what say (INF) 2 3 days from like this

The Urdu and Hindi Ergative Postposition ne · 17

muuD oaf hai mood off is 'What would I say! (i.e. I didn't say anything.) He's been in a bad mood like this for two or three days'. (Tanhaiyan) [+/CONSCIOUS CHOICE] (16) A: maiN twnheN kaafii pilaataa baRii zabardast I you (DAT) coffee drink (CS) very good calo uTho come on get up 'I'll take you for coffee, come on, get up'. B: maiN nahiiN jaauuNgii 1 not will go 'I'm not going'. A: do ghanTe meN tumhaaraa daftar bhaag to nahiiN 2 hours in your office runaway TOP not jaaegaa will go 'Your office isn't going to run away in a couple of hours'. B: maiN ne nahiiN jaanaa I ERG not go (INF) '(I told you) I don't want to/won't go'. (Tanhaiyan) [+ CONSCIOUS CHOICE] In 16, maiNne nahiiN jaanaa is used as a more forceful variant of maiN nahiiN jaauuNgii (future). (17) omaaigauD is kaa to ek lafz mere palaa nahiiN paRaa oh my God this of TOP 1 word my 'lap' not fell ab maiN kyaa karuuNgii now I what will do aur DisiZan bhii maiN ne lenii hai and decision (f.) too I ERG take (INF-f. sg.) AUX o gaus oh gosh Oh my God, I don't understand even a word of this—what should I do now—and I am supposed to make .a decision too—oh gosh'. (Tanhaiyan) [+/- CONSCIOUS CHOICE] Example 17 is used in a situation where the character has to make a decision which she really doesn't want to make at the present moment, although she lad previously decided that she did want to do it. (18) Petitioner at police station: sar, maiNne bhii ek an kamii thii sir I ERG also one utterance (f.) do (INF-f.sg.) AUX (p.f.sg.) 'Sir, I also want(ed) to say something'. (Andhera Ujala, episode Ghar Kahani [ +/- CONSCIOUS CHOICE]

18 · Elena Bashir

(19) Police havaldar, Karam Dad: aw vaise, buzurgo, ham ne to apne kaaGzaat bhii to anyhow gentlemen we ERG TOP self's papers (m.pl.) also TOP puure karNe16 haiN naa complete do (INF m.pl.) AUX EMPH And anyway, gentlemen, we also have to complete our paperwork, don't we'. (Andhera Ujala, episode Ghar Kahani) [-CONSCIOUS CHOICE] (20) Gossiping relative: maiNne socaahaiaaj caaheNkuch hojaae I ERG thought today even if something happens aaj main ne zaruur jaanaa hai today I ERG definitely go (INF) AUX Ί decided that today, no matter what might happen, I am definitely going to go'. (Aanch) [+ CONSCIOUS CHOICE] (21) Father: beTe aap agar caaheN to aap bhii mere saath caleN son you if want TOP you also me with come 'Son, if you want to, you come along with me too'. Son: saurii paapaa ham ne aap ke saath nahiiN jaanaa sorry papa we ERG you with not go (INF) 'Sorry, Papa, we don't want to go with you'. (Aanch) [+ CONSCIOUS CHOICE] (22) Watchman: mis, chuTTii hogaii maiNne darvaaze miss, closing time became I ERG doors (m.pl.) band karne haiN close (INFm.pl.) AUX 'Miss, it's closing time. I have to lock the doors'. (Aanch) [- CONSCIOUS CHOICE] (23) A: aziiz, beTaa uTho khaanaa khaalo Aziz, son get up food eat (IMPER) Aziz, son, get up and eat your dinner'. B: maiff ne nahiiN khaanaa kuch I ERG not eat (INF) anything Ί don't want to eat anything'. (Aanch) [+ CONSCIOUS CHOICE] In eight of the 15 natural examples above, the semantics ofne turns out to be not as unambiguous as it appears in the constructed examples. In 11,12,13, 14, 15, and 17 either a [+ CONSCIOUS CHOICE] or a [-CONSCIOUS CHOICE] reading is possible; or, equivalently, the situations contain elements of both ability and inability to choose. In fact, in examples 19 and 22, spoken by employees referring to actions they have to take in connection with their duties, the meaning is [- CONSCIOUS CHOICE].

The Urdu and Hindi Ergatwe Postposition ne · 19

• 3.2 2nd Person Recently I have observed interesting instances of the NP-ERG + INF construction used with 2nd person subjects, i.e. addressees, in a quasi-imperative sense. Generally, unless we have been specifically told about them, we are not privy to the motivations or conscious states of others. Thus for a speaker to attribute conscious choice or voliiionality to a 2nd person subject (addressee) issemantically somewhat anomalous. Butt and King have argued that the NPERG + INF construction marks [+ CONSCIOUS CHOICE]. From this point of view, 2nd person constructions which are questions or conditionals are not surprising: an assertion using this construction raises new questions. However, such sentences do occur in assertions as well as in questions and conditionals. Examples 24-29 are questions, i.e. they question the intentions of the subject; 30-32 are conditionals; and 33-38 are assertions. 3.2.1. Questions (24) A: aur turn kyaa kar rahii ciRiyoN ke saath khel rahii and you what are doing birds with playing turn ne paRhnaa-vaRhnaa nahiiNhai? you ERG study-and-all not AUX And what are you doing—playing with birds; don't you want to study (i.e. pursue your education)?' [+ CONSCIOUS CHOICE] B: kaisepaRhuuN yuunivarsTii to band hai how study university TOP closed is 'How can I study; the university is closed'. (Tanhaiyan) (25) yeh baat to mujhe turn se kamii hai this word (f.) TOP I(DAT) you from do (INF f.sg.) AUX aaxir turn ne ghar sambhaalnaa nahiiN? finally you ERG home look after not Ί have to say this to you. After all, you are going to be responsible for the house, aren't you?' (Tanhaiyan) [+/- CONSCIOUS CHOICE] The following exchange illustrates very nicely the contrast between ne and ko, the DATIVE. (26) Zara: maiN bhuulgayii maiN ne Dinar pejaanaa I forgot I ERG dinner to go (INF) thaa saad hai AUX (p.) Saad is Ί forgot—I was supposed to dinner—it's Saad'. [-»-/-CONSCIOUS CHOICE] Zain: turn ne jaanaa? you ERG go (INF) Are you going to go?' [4- CONSCIOUS CHOICE]

20 · Elena Bashir

Zara: haaN,mujhe jaanaa to paRegaa 'Yes, I will have to'. [- CONSCIOUS CHOICE] Zain: phirjaanaa tojaao 'So if you are going, go!' (Tanhaiyan) In 26 note the contrast between Zara's first and second utterances; in the first, she presents the going as ambiguous for conscious choice, while in the second, by using paR—which requires DAT, as marked for [-CONSCIOUS CHOICE]. (27) Father: acchaa choTii aap ne calnaa hai mere saath? so little you ERG go (INF) AUX me with 'So, little one, do you want to go with me?' (Aanch) [+ CONSCIOUS CHOICE]17 The following two utterances occurred as translations of a sentence presented in English. When second-language Urdu-speaking students were asked to render 'Who do you want to give your book to?' and 'When do you want to go to the bazaar?' in Urdu, they spontaneously produced the sentences shown in 28and29.18 (28) Who do you want to give your book to? aap ne apnii kitaab kis ko denaa hai you ERG self's book who DAT give (INF) AUX [+ CONSCIOUS CHOICE] (29) When do you want to go to the bazaar? aap ne baazaar kab jaanaa hai you ERG bazaar when go (INF) AUX [+ CONSCIOUS CHOICE] 3.2.2. Conditionals (30) agar aap ne yeh nahiiN Dalnaa... if you ERG this not put (INF) 'If you don't want to add this...' (PTV program, Pot Luck, 16 September 1996) [+ CONSCIOUS CHOICE] (31) aur aaindaa agar turn ne derse aanaa ho and again if you ERG late come (INF) AUX to mujhe aafis se Telifon kar denaaTOP me (ACC) office from phone do (INF) maiN xud aa-kar abbaa ko khaanaa khilaa duuNgaa I self come-CP Father DAT meal will feed 'And if you are going to come late again, phone me from the office; I'll come myself and give Abba his dinner'. (Tanhaiyan) [+ CONSCIOUS CHOICE]

The Urdu and Hindi Ergative Postposition ne · 21

In 31, the addressee is presented as being fully in control of her actions. (32) calne sepahle ek baat karnii aapko going before one word (f.) do (INF f.sg.) you DAT maiN sirf yeh kahne kii kosis kar rahiihuuN I only this saying of attempt am doing ki agar turn ne wahaaN jaanaa hai to... that if you ERG there go (INF) AUX then 'Before you go I want to say one thing to you. I am just trying to say that if you are going there, then ...' (Tanhaiyan) [+ CONSCIOUS CHOICE] In these questions and conditional sentences with 2nd person subjects, the subject is assumed to have potential control of his actions. All three of the genuine questions assume (the option of) [+ CONSCIOUS CHOICE] on the part of the addressee. Example 25 is a rhetorical rather than a real question. 3.2.3. Assertions: Statements About Addressee's Probable/Expected Actions (33) jin jin se turn ne baateN karnii thiiN whom whom with you ERG words (f.p.) do (INF) AUX (p.f.pl) kar IHN have done 'You have spoken to whomever you wanted to/were supposed to'. (Tanhaiyan) [+/- CONSCIOUS CHOICE] (34) Father: puucho ask(IMPER) '(Go ahead and) ask (me)'. Daughter: aap mujhe DaanTeNge to nahiiN you me will scold TOP not 'You won't scold, will you?' Father: beTe, mujhe pataa hat child I (DAT) known is ke aap ne mujhe koii aisii baat nahiiN that you ERG me any such matter (f.) not puuchnii ask (INF) f.sg. 'Child, I know you won't ask me anything like that (i.e. for which I would have to scold you)'. (Aanch) [+ CONSCIOUS CHOICE] (35) Husband: maiN tumheN jo biyaa-ke laayaa thaa I you (ACC) REL-TOP married-CP brought woh un kii maaN kii haisiyat se laayaa thaa. CORREL their mother of status from brought ab turn ne saabif kamaa hai now you ERG prove (INF) AUX

22 · Elena Bashir

ki turn unkii maaN ho yaa nahiiN whether you their mother are or not 'When I married you and brought you here, I brought you as their mother. Now it's up to you to prove whether you are their mother or not/ (Aanch) [+/- CONSCIOUS CHOICE] (36) Husband: kal agar maiN xud twnheN saaping par yesterday if I self you (ACC) shopping to na lejaataa not take to meraa xyaal hai turn ne to nahiiN then my opinion is you ERG TOP not kahnaa thaa say (INF) AUX(p.) 'If I hadn't taken you shopping myself yesterday, I don't think you would have asked me to'. [+ CONSCIOUS CHOICE] (Aanch) 3.2.4. Assertions: Instructions, Injunctions (37) is ko xuub acchii tarah aap ne hilaanaa hai this ACC very well you ERG stir (INF) AUX 'You have to stir it well'. (PTV program, Pot Luck, 18 March 1996) [+/- CONSCIOUS CHOICE] In example 37, the 2nd person NP-ne + INF assertion takes the form of directions or instructions. The semantics of sentences like this is like that of English sentences like: You don't want to hear this, Counselor,19 or You want to go straight to the next stop light, take a right and go three blocks. In this mode of giving instructions, the use of want appears to attribute volitionality to the addressee, reflecting the mixed semantics of simultaneously giving instructions and attributing conscious choice to the addressee. (38) acchaa, baabaa, saat taariix ko nagiin kii saalgiraa hai all right, VOC, 7th date DAT Nagin of birthday is aap ne aanaa hai sab ko saath le-ke you ERG come (INF) AUX all ACC with bring-CP All right, Nagiin's birthday is on the seventh; you have to come and bring everyone with you'. (Aanch) [- CONSCIOUS CHOICE] Sentence 38 is a married daughter's invitation to her parents to come to her step-daughter's birthday party at the joyous conclusion of the story. In 38, which is a strong injunction, the SPEAKER'S wish for the action to happen and expectation that it will are projected onto the subject. The fact that the mood is happy and the relationship one of love and intimacy allows this identification of expectations to occur felicitously. The ne construction is characteristic of informal dialogue and tends to occur in happy parts of a story, but not in serious or 'high-style' speeches.

The Urdu and Hindi Ergative Postposition ne · 23

• 3.3. 3rd Person Now let us examine the semantics of NP-ne + INF sentences with 3rd person subjects. (39) Q: us ne kyaa kahaa he ERG what said 'What did he say?' A: kahnaa us ne kyaa? say (INF) he ERG what 'What was he to say?' (i.e. What could we expect him to say under the circumstances; therefore, he didn't say anything.) (PTV drama by Jamil Malik, 26 May 1996) [+/- CONSCIOUS CHOICE]20 (40) A: huaa kyaa thaa happened what AUX (p.) 'So what happened?' B: honaa kyaa thaa happen (INF) what AUX (p.) 'Nothing (unexpected) happened!' jis din us ne sahrse lauTnaa thaa which day he ERG city from return (INF) AUX (p.) mai ne sab intizaam kar liyaa thaa I ERG all planning did AUX (p.) On the day when he was to return from the city I had planned everything (to kill him)'. (Andhera Ujala, episode Tabiir [+/CONSCIOUS CHOICE] (41) Father: kal tumhaani mammii ne aanaa thaa yesterday your mommy ERG come (INF) AUX (p.) mujhe bahut afsos ho rahii hai I (DAT) much sorrow is happening ki abhii tumhaani mammii nahiiN aa saktii that now your mommy not come can 'Your mother was supposed to come yesterday; I am very sorry that now your mother can't come'. (Aanch) [+/- CONSCIOUS CHOICE] (42) us ne waapas aanaa hainaa he ERG back come (INF) AUX Q.TAG 'He is going to come back, isn't he', (example courtesy of Ruth Schmidt, p.c.) [+/- CONSCIOUS CHOICE]21 In all of these 3rd person examples, the semantics ofne is unspecified for the parameter of [CONSCIOUS CHOICE] of the subject. In fact, CONSCIOUS CHOICE is not the relevant semantic parameter at all. The meaning is rather one of an ACTION EXPECTED OR ANTICIPATED BY THE SPEAKER.22

24 · Elena Bashir

• 4. Interaction of ne with Speech-Act Roles and 'Prospective' Aspect The specific semantics contributed byne in this construction is a function of its SOURCE specification, its 'prospective' meaning and speech-act roles (grammatical person). Consider the following: compulsion, the negation of conscious choice, can be attributed felicitously to a 3rd person subject, since it originates outside the subject and is thus potentially observable by the speaker. Information about conscious choice, however, originates in the mind of the subject (agent) and is not accessible to an outside observer unless he has been specifically made privy to it. Thus attributing conscious choice to a 3rd person subject is semantically anomalous and requires supporting context. If we consider the functions of ne, both as marker of ergative subjects of transitive verbs in perfective tenses (whether agents or sources) and as marker of subjects in the NP-ne INF construction, as marking the archi-case role of SOURCE (as opposed to GOAL),23 then the SOURCE specification of ne interacting with the defining characteristics of the three speech-act participant roles (grammatical person categories) and the 'potential' meaning carried by the infinitive, gives rise to the different meanings seen in the examples above. This analysis assumes the following alignments of discourse roles and grammatical categories for each of the three persons. Observer' is taken here to mean the observer of the mental state of the (grammatical) subject, i.e. the initiating motivation for the act. By default, the observer of the subject's mental state is the subject itself. The speaker, on the other hand, is the observer of the resulting action or situation. Semantics otne (by grammatical person categories) 1st person: speaker = subject = observer (of agenf s mental state) Emergent meaning: conscious choice by subject 2nd person: speaker * observer (speaker has access to observer [by questions]) Emergent meaning: Questions and conditions: potential conscious choice by subject Assertions: action anticipated by speaker (potential conscious choice by subject) action desired/enjoined by speaker 3rd person: speaker Φ observer Emergent meaning: action expected by speaker (speaker typically has no access to mental state of subject/observer). Thus in the cases of assertions about 2nd person subjects and of 3rd person subjects, the SOURCE nature of ne is realized as a manifestation of this generalized category, i.e. the coding of an action as anticipated by the speaker.24

The Urdu and Hindi Ergatwe Postposition ne · 25

This perhaps explains the remarkable felicity of the English gloss 'is/was supposed to' for this usage. Given the above analysis, I propose that the semantic parameter associated with ne, which was initially called CONSCIOUS CHOICE, be generalized to SOURCE SPECIFIED, which would subsume the categories of CONSCIOUS CHOICE and ΑΝΉΟΡΑΤΕϋ ACTION for lst/2nd and 3rd person subjects respectively. Accordingly, from this point on, CONSCIOUS CHOICE should be understood as one subtype of SOURCE specification.

• 5. Markedness Relations of ne and ko With the spread of the NP-ne + INF construction, ne is encroaching on the territory of koy but with the innovation that the option of a semantic contrast between ko and ne along the parameter of source specification becomes possible. The function of ko is being restricted to contexts which are unambiguously [-CONSCIOUS CHOICE] involving strong external obligation or compulsion. The functional territory ofne is also expanding at the opposite end of a putative volitionality continuum, at the expense of NP-NOM INF cooA-'want to' constructions25 and of future and present continuous finite tense forms.26 INTENTION

DESIRE

NECESSITY

NECESSITY

NECESSITY

COMPULSION

*be going to*

*want to'

(internal)

(moral)

(external)

(external)

'have to, need to'

'should'

'have to'

'must*

future, present continuous

caahnaa

ko

ko

ko

ko

Situation future, after ne present change continuous ne

caahnaa ne

ne

ko ne

ne

ko

Previous situation

Anticipated action

Internal motivation

>

> > External compulsion

Figure 1: Volitionality continuum

The markedness situation for constructions involving ne seems to be as follows: Finite constructions: V-intr. - ne [0 CONSCIOUS CHOICE]27 + ne [+ CONSCIOUS CHOICE] V-tr. - ne [- CONSCIOUS CHOICE] + ne [0 CONSCIOUS CHOICE]

26 · Elena Bashir

Using the generalized parameter of [SOURCE SPECIFIED] allows us to express the markedness relations for the NP-ko/ne + INF constructions as follows: Infinitive constructions: NP-AU? + INF [+ SOURCE SPECIFIED] NP-fo + INF [- SOURCE SPECIFIED] In the variety of Urdu under discussion here, in the NP-ne + INF construction, the presence of ne is not marked for (in the sense of specifically coding) [+ CONSCIOUS CHOICE]; rather, it is the dative postposition ko that is marked for the absence of CONSCIOUS CHOICE, or equivalently for the presence of obligation or compulsion. The dative postposition ko also seems to be acquiring other types of specific discourse connotations (marking) as well. Consider examples 43 and 44: (43) ekskyuuz mil, sar, mujhe aap se kuch kahnaa hai excuse me sir I (DAT) you from something say (INF) AUX 'Excuse me sir, I would like to say something to you*. (Tanhaiyan) (44) agar aap ko unkaa dil jiitnaa hai, if you DAT her heart conquer (INF) AUX to us ke liye aap ko apnaa taur tariiqaa badalnaa then that for you DAT self's strategy change (INF) hogaa... AUX (fut.) hamdardii haasilkarnaa hogii... sympathy get (INF) AUX (fut.) 'If you want to win her heart, you will have to change your approach... (You will) have to win (her) sympathy'. (Tanhaiyan) Sentences 43 and 44 are spoken by a household servant, one of whose distinctive characteristics is that he uses formal, high-flown speech. Here ko (DAT) is associated with formality, politeness, and deference. In the drama Tanhaiyan, the NP-ne + INF sentences are mostly uttered by young characters, and at 'upbeat' moments in the story. On the other hand, older, more serious characters use ko constructions; also, younger characters use ko to emphasize the seriousness of an occasion. For example, in 45 and 46, Sania, a light-hearted young girl, uses ko constructions when behaving in a serious, responsible mode. (45) Sania: Draevar se kaho gaaRii nikaale—mujhe jaanaa hai driver to say car take out I (DAT) go(INF) AUX Tell the driver to bring the car—I have to go'. Ani: kahaaN jaanaa where go (INF) 'Where are you going?'

The Urdu and Hindi Ergative Postposition ne · 27

Sania: faikTrii jaa rahii huuN factory going am Tm going to the factory (importantly)'. (Tanhaiyan) (46) Sania: biibii,... buqraat, naaslaa lagaaeN— Bibi Buqraat breakfast set out mujhe hauspiTal jaanaa hoi I (DAT) hospital go (INF) AUX 'Bibi, Buqraat, put breakfast (on the table)—I have to go to the hospital'. (Tanhaiyan) Here, the use of the dative stresses outside obligation, hence the importance of the action.

• 6. Possible Sources of Semantics of the NP-ne + INF Construction with 2nd Person Subjects It seems probable that multiple causes have contributed to the use and semantics of the present NP-ne + INF construction with 2nd person subjects. a. Influence from the 'future imperative' usage, which consists in this variety of Urdu of the infinitive with nominative turn or sometimes aap as optional subject but without AUX. This usage has morphological similarity (use of INF) and has the semantics of imposing weak obligation on an addressee. Sentence 47, without an expressed subject, and 48, with the subject aap, illustrate this construction. (See also example 31) (47) bahan sania ko taang nahiiN karnaa jaa-ke sister Sania to annoy not do (INF) go-CP 'Don't annoy your sister Sania over there'. (Tanhaiyan) (48) aap zaaraa se yeh baat kahnaa, you Zara to this word say (INF) to dekh yeh kyaa jawaab detii hai then see she what answer gives 'Tell Zara this and then see what she says'. (Tanhaiyan) b. Semanticization of ne as a marker of potential conscious choice in sentences with past tenses of some intransitive verbs (section 2). c. Phonological confusion. Consider examples 33 and 11, repeated here for convenience. (33) jin jin se turn ne baateN karnii thiiN whom whom with you ERG words do (INF f.pl.) AUX (f.pl.) karliiN have done (f.pl.) 'You have spoken to whomever you wanted to/were supposed to'. (Tanhaiyan) [+/- CONSCIOUS CHOICE]

28 · Elena Bashir

(11) ham ne Dalnaa hogaa pyaaz we ERG put (INF) AUX (fut.) onions 'We need to/will add onions'. (PTV program, Pot Luck, 16 September 1996) [+/- CONSCIOUS CHOICE] The semantics of these utterances is such that either ne or ko could appear. A simple mishearing of the dative forms twnheNorhameN, or simply a tendency to replace a VN (nasalized vowel) syllable with a stronger CV syllable could result in the transposition of the nasal element to the syllable-initial position, resulting in—eN > ne. It is worth noting in this connection that Urdu is a second language for most of its speakers in Pakistan, and the dative forms tumheN and hameN usually have to be carefully taught to new speakers, whose first tendency is to use the separate postposition fa? for the dative meaning.28 d. Possible continuing substratum influences from Panjabi and Siraikd. In Panjabi, there is a regular verbal formation, called the 'Naa form' (see Puar 1990, for example), which has the semantics of the Urdu NP-ne + INF form.29 A typical 3rd person sentence of this type is 49, in which ne appears in the type of construction which we are discussing for Urdu. Since ne regularly occurs with 1st and 2nd person finite subjects in Urdu, it could easily generalize from the 3rd person to 1st and 2nd persons in the ne + INF construction. (49) muNDe ne ciTThii UkhNu hE boy (OBL) NE letter (f.) write (INF f.) AUX 'The boy is going to write a letter'. (Puar 1990:16; his glossary) In Siraiki, the personal pronounsηκκΝ'Γ, tusaN'you', andosoWwe' are identical for nominative, oblique, and ergative uses. Example 50 shows these pronouns used as subjects with infinitives in an obligative reading, and as an ergative subject of a transitive verb. In 51 a [+ CHOICE] use of the INF construction is shown. (50) tusaN DakTar scehib ku milNce you (NOM/ERG) doctor sir DAT meet (INF + AUX) 'You must meet the doctor'. (Shackle 1976:139; his glossary)30 AsaN vAnNce we (NOM/ERG) go (INF -f AUX) 'We must go'. (Shackle 1976:149; gloss his) AsaN uku khedda diTha we (NOM/ERG) him (DAT) playing saw 'We saw her playing'. (Shackle 1976:149; his glossary) (51) par bal nA bcehir avNa ha, te nA aya but child not out come (INF) AUX (pst.) and not came 'But the child was not going to come out, and didn't'. (Shackle 1976: 139; his glossary) The presence in Siraiki (and probably varieties of Hindko) of a single form which has both standard ergative function (subject of transitives in past

The Urdu and Hindi Ergative Postposition ne · 29

tenses) and is used in NP + INF constructions in both volitional and obligative meanings may be exerting a substratum effect on the contemporary ne construction in Pakistan.31

• 7. The NP-ERG + INF 2nd Person Construction as a Conceptual Blend This construction can be interestingly considered as a conceptual blend (in the sense of Turner and Fauconnier [1995]) of semantic and morphological elements from several sources.

ι Semanticization of ne as a marker of potential conscious choice

Frequent use in Siraiki of 2nd person plural form which is same for NOM and OBL THE NEW CONCEPTUAL BLEND

aap ne INF AUX (pres.) ('instructional/ injunctive mode') Urdu/Hindi 'future imperative* construction NP (NOM) (INF). [- CONSCIOUS CHOICE] semantics

Presence of ne in Panjabi 3rd person sentences with [+ SOURCE SPECIFIED] semantics

Figure 2: Sources of the conceptual blend in NP - ne + INF (2nd person)

These four sources of influence contribute elements as follows: 1. Association of ne with the parameter of potential [CONSCIOUS CHOICE]; 2. Blurring of distinction between NOM and ERG forms; 3. Association of infinitive constructions with [-CONSCIOUS CHOICE]; impetus to apply construction to 2nd person assertions; and 4. Association of ne with infinitive constructions. The resulting blend puts ne in an infinitive construction which has the complex semantics resulting from combining a 'future imperative' [-CONSCIOUS CHOICE] with a [+ SOURCE SPECIFIED] form. In fact, in the context of 2nd person assertions (i.e. instructions or injunctions), the SPEAKER'S expectation that the subject will perform the action in question is also a manifestation of the [+ SOURCE SPECIFIED] semantics of ne.

30 · Elena Bashir

• 8. Possible Future Developments It appears that ne is in the process of being reanalyzed as a marker of the archicase category SOURCE, rather than specifically as a marker of [CONSCIOUS CHOICE], which is only one of the manifestations of the category. Three points along this continuum of regrammaticization are represented by the situations in Hindi, Delhi Urdu/Hindi, and Pakistani Urdu. In view of the developments observed so far, it seems that the future scenario for ne may involve the following elements: (a) Increasing contrast between ne and ko as subject markers. Under influence of the semantics of ne + INF constructions the range of ko will be increasingly restricted to situations where strong outside compulsion is overtly marked (e.g. withpaRnaa 'fall, befall'), or to indicate politeness (via diminution of volitionality/autonomy of speaker). (b) Continued expansion of the range ofne (i) To indicate all agential subjects. There are foreshadowings of this possibility for both past tense intransitive verbs and for future time reference. The sentences in (52) and (53) represent 'mistakes' and false starts which are commonly produced by second language speakers in Pakistan. Past intransitive (52) maiN ne baazaar gayaa I ERG bazaar went € I went to the bazaar'. Future (53) maiN ne... main jaauuNgaa I ERG will go 'I...will go', (false start, corrected with maiN) It is easy to see how the meaning of future time reference could become associated with ne from prototypical sentences like maiN-ne karaaciijaanaa Ί am going to Karachi'. The fact that NP-ne INF is not specified for either [+ CONSCIOUS CHOICE] or [- CONSCIOUS CHOICE] allows its meaning to generalize for 1st and 2nd person subjects to a general sense of intention/need to act with its associated implied future time reference, and in the case of 3rd person subjects, expected action. If speakers are, in fact, perceiving ne as the mark of a 'real' (volitional) subject, i.e. agent, since it marks the logical subject in so many agential (past tense transitive) sentences, then ne may be in the process of being transformed into a kind of marked nominative, as suggested in Bubenik (1989: 381). As a further development along these lines, it is possible that with intransitives the 3rd person sg. AUX, hai, will be replaced by an AUX form showing personal agreement, e.g. by a construction like *mai-ne jaanaa huuN. (ii) obligative meaning (< semantics of 2nd person ne + INF constructions and from future imperative). The development in 2nd person constructions

The Urdu and Hindi Ergative Postposition ne · 31 ('instructional mode') represents a tension between the semantics of the archirole, SOURCE (speaker expectations), and that of GOAL (i.e. addressee). An interesting question is whether the 2nd person NP-ne INF + AUX construction will (a) retain its role as marker of a subject that is ambiguously or simultaneously both volitional and non-volitional; (b) develop into a neutral future meaning; or (c) die out. • NOTES 1. The origin of ne as an ergative marker in Western Hindi is recent (around 200-300 years ago), and its source is uncertain. Several proposals have been advanced to explain its origin, (a) It developed from a phonologically reinforced and/or metathesized reflex of -ena, the ΟΙΑ instrumental ending of α-stem nouns (Masica 1991: 344-45). (b) That it is identical with the dative ne of Gujarati and nuuN of Panjabi. 'While Gujerati used this particle for both object and agent, Hindi having already an objective postposition ko, naturally restricted the use oine to the case of the agent' (Kellogg 1965:132). According to Kellogg, its ultimate source is the root lag 'to attach to', (c) It was originally a dative marker and came, on the loss of the old instrumental, to be extended to the subjects of past tense transitive verbs (Beames 1966:270, 272). 2. In the Urdu examples, vowel length is represented by doubling the vowel symbol and nasalization by following a vowel; for consonants, retroflexion is represented by capitalization, with the exception of , which represents the velar fricative /y/. In the Panjabi and Siraiki examples, represents retroflex/N/. represents the palatal affricate, and and the voiceless and voiced palatal sibilants, respectively. Abbreviations used are: ACC AUX CP

accusative auxiliary conjunctive participle CORREL correlative CS causative DAT dative

ERG EMPH f. fut. IMPER INF PFV

ergative emphatic feminine future imperative infinitive perfective

m. NOM p. pi. REL sg. TOP VOC

masculine nominative past plural relative singular topicalizer vocative word

3. Tuite et al. (1985: 264) give the following intransitives allowing optional ne: ciix- 'scream', chiiNk- 'sneeze', thuuk- 'spit', Sarmaa- 'be ashamed', khaaNs- 'cough', muui- 'urinate', muskwaa- 'smile', jhaaNk- 'peep', khusphusaa- Svhisper', khatkhataa- *knock', hichkichaa'hesitate', AwAiVwfl- tvihinny'1gudgudaa- 'tickle', bdbilaa- Vail', baRbaRaa- 'mutter', giRgiRaa'p\ead\gungunaa~ 'hum', tutlaa- 'lisp', lalkaar- 'utter a battle cry', cillaa- 'shout' also belongs in this category (Hook et al. 1987:150). Commenting on this, G.C. Narang says that only thuuk·, muut-,jhaaNk-, and khatkhataa- take optional ne in Delhi Urdu. Barker (1967:1: 329) mentions chiiNk-, muut-, hag- 'defecate', thuuk-, andjhaaNk- as taking optional ne, but classifies them as 'stems which are transitive but whose English equivalents are treated as intransitive'. Tuite et al. argue that the use of ne with this group of verbs is a result of the interaction of their semantics with their phonological shape and their status in the transitivity paradigm of the language. The verbs in question have a defective transitivity/causativity paradigm in that they are intransitive stems from which no causatives are derived, and they contain either a tense vowel or /aa/, the canonical phonological shape of transitive and first causative verbs, which normally take ne with perfective participle based tenses.

32 · Elena Bashir 4. T. Mohanan (1994: 72) makes the same argument for Hindi, giving the following forceful example. (a) roam ko acaanak Ser dilchaa. vah/ *us ne ciUaayaa Ram DAT suddenly lion appeared he (NOM) he ERG screamed 'Ram suddenly saw a lion; he screamed*. (b) us ne/ *vah jaan buujh-kar ci aaya he ERG/ he (NOM) deliberately shouted 'He shouted deliberately*. She further states that '...when a NOM subject co-occurs with verbs that, have an option between NOM and ERG subjects, the action must be nondeliberate' (Mohanan 1994: 73). 5. See Davison (1997 ms) and J. Singh (1993), for analysis of the assignment of case to ne in the Government-Binding framework. 6. Thanks to Peter Hook for focussing on this point. 7. Alternatively, Miriam Bun suggests that this example may simply represent influence from Panjabi, in which ne does not occur with 1st person subjects. 8. AsifAgha(1997)analyzeswhatIamcallingtheNP-A» + INF,NP-ne + INF,andNP0Ι-INF constructions as instantiations of a unified aspectual category, which he calls the 'telic' aspect. Thus, Agha would ca\\jaanaa in mujhejaanaa hai Ί have to go* the 'telic aspectual* form rather than the infinitive. In his analysis, these forms are distinct from the infinitive, and represent a reanalysis of the infinitive in progress: infinitive > 'telic' aspect. In favor of his analysis, Agha adduces the fact that subjects and objects of verbs in the -naa form follow the same case marking constraints and agreement patterns as for the language as a whole. Agha's analysis is similar to one presented in unpublished discussions by M.K. Verma in 1982; \fenna, however, called this third aspectual category the 'prospective*. Bubenik (1989), however, perceives the historical reanalysis from gerundive > infinitive (i.e. verbal adjective > verbal noun) as still in progress, in effect interpreting the process as moving in the opposite direction (in personal conversation—p.c. hereafter). The interesting questions raised here for Indo-Aryan language studies are: what, if any, is the 'prototypical* IA infinitive? In which languages is it found? In which direction is the reanalysis going (considered separately for each language)? I am in basic agreement with the analysis proposed by Verma and Agha, which is consistent with the behavior of ne which I am describing. Of the two terms proposed, I think 'prospective*, since it suggests that the common semantics of the NP-A» + INF, NP-ne + INF, and NP0 + INF constructions (including the future imperative*), as well as the 'infinitive proper* is the unrealized orpotential nature of the action, is preferable. In the interest of historical continuity of this discussion, however, I will continue to use the term lne + INFINITIVE* in this paper. 9. Although the commonly held view is that the ne + INF construction is a result of present-day Panjabi substratum (i.e. first-language) influence, I think the story of ne and its relation to Panjabi is more complex than this, and also involves interaction at a greater time depth. 10. The construction with ne + INF is not accepted .by many speakers of Urdu or Hindi, who correct it by replacing ne with ko. According to Anvita Abbi (p. c.), the construction is confined to the Hindi of Delhi, but even there it is not as widespread as the examples cited here indicate that it is in the Urdu of Pakistan. G.C. Narang has expressed the same opinion, commenting that examples 3b, 4a, 5a, 5b, and 7b of the present paper, for example, would not be acceptable in Delhi Urdu. The discussion in this paper applies to the Delhi Hindi with respect to 1st and 3rd person usages, and with the qualifications mentioned above. I have no information about 2nd person uses in the languages as spoken in Delhi. In order to further address the question of whether or not this usage is found in India to the same extent that it is in Pakistan, I listened to the full sound track of the Indian Urdu film Tiisrii Manzil for instances ofne + INF. Interestingly, in this film there were no examples of it.

The Urdu and Hindi Ergative Postposition ne · 33 Also, it is clear that even in contexts where the semantics is clearly [ - CONSCIOUS CHOICE] the characters use ko constructions. For example: (c) mujhe aap se koi boat nahiiN karnii I (DAT) you from any speech not do (INF-f.sg.) don't want to talk to you at all'. (d) maiN jaauuNgaa to sirf aap ke saath I will go if-TOP only you with 'If I go it will be only with you, wamaa mujhe jaanaa hό nahiiN—yeh samajh liijiye otherwise I (DAT) go (INF) EMPH not— this understand (IMPER) otherwise I won't go, be clear about this!' (e) Hero: suniitaa mujhe turn se kuch kahnaa hai Sunita I (DAT) you with something say (INF) AUX 'Sunita, I want to say something to you'. Sunita: aw ab mujhe jaanaa hai— baaibaai and now I (DAT) go (INF) AUX— bye bye 'And now I'm leaving—bye bye!' Contrast these examples with 16 from Tanhaiyan. 11. A fact stressed by Agha (1997) in arguing for a unified aspectual status for -naa constructions. 12. Tanhaiyan ('lonelinesses'), written by Hasina Moin, is a classic Pakistani television drama. The examples in the present paper have been transcribed from the video recorded version of the drama. I have not had access to a written version of the script An interesting question is whether all the examples of the NP-ERG + INF construction which I found in the videotaped version also occur in the written text. 13. The television drama Aanch ('heat, trial'), written by Naheed Sultana Akhtar, is set in Karachi. The parental home of the heroine is a traditional Urdu-speaking household; while the husband of the heroine is a modern, educated, affluent businessman. 14. The humorous, lightly satirical drama serial /utfiera Ujala ('darkness and light') is set in the Panjab; the chief characters are minor police officers. Their speech is Tanjabi-ized Urdu'. 15. The construction NP-0 INF, in which the subject is not expressed and the NP represents the direct object rather than the subject also exists. For example: (f) biibόjό akbar saahab aaye— Bibi Akbar Mr. came kah rahe haiN ki aap se boat karnii hai he says that you with speech (f.) do (INF f.sg.) AUX 'Bibi-ji, Mr. Akbar has come; he says he wants/has to talk to you'. (Tanhaiyan) 16. Note the retroflex intentionally used by the author to signal the Panjabi accent of this character. 17. Example 27 illustrates the tendency of urban Pakistani parents to use aap with their small children. Many parents do this to get their children in the habit of saying aap rather than turn or tu, so that they do not unwittingly give offense by using turn or tu to a person of higher status. 18. The context in which these examples occurred was entirely independent of the work on this paper. The family background of the speaker in 28 is Panjabi, but the speaker in 29 is from a Hyderabad (Deccan) family. 19. Demille Nelson (1990), The Gold Coast, p. 508. 20. Contrast with this the following comparable exchange, which is set by the author in a traditional Urdu-speaking Karachi household. (g) A: phir aap ne kyaa kahaa then you ERG what said 'Then what did you say?'

34 · Elena Bashir B: kahi kyaa—sun-kar cup horahii say (IMPFV) what—listen-CP quiet kept 'What was I to say; I just listened and kept quiet*. (Aanch) 21. Example 42 was noted in Oslo, where a sizable community of Pakistani (Panjabi) immigrants is settled. It occurred when the trains were all halted at a station because of a power outage ahead, and the train immediately in front of the one in which the speaker was seated was in a position to reverse direction and shift onto another track to return in the direction from which it had some. The speaker was guessing that the train would probably return. 22. This meaning is similar to the meaning of NP + INF sentences with inanimate subjects, which are not marked with either ko or ne. With inanimate subjects there is no mental state, and only speaker expectation is involved. (h) ab koi tufaan nahuN aaegaa jitne tufaan now any storm not will come how many storms (m.pi) aane the wo sab aa-ke guzar duke come (INF m.pl.) AUX (p.m.pl.) those all come-CP have passed 'Now no storms are going to come. All the storms that were to come have passed1. (Aanch) (i) aaj barf paRn hat today snow (f.) fall (INF f.sg.) AUX 'It's going to snow today*. (Agha 1997:38) 23. T. Mohanan (1994: 75) considered but rejected moving away from the notion of CONSCIOUS CHOICE towards a more generalized notion of SOURCE. The thinking in this paper is independent of Mohanan (1994), ultimately derived from analysis benefiting from DeLancey's cognitive model of event structure (DeLancey 1985). 24. The expression of the distinction between actions or events reported from the perspective of the 'prepared mind' versus the 'unprepared mind* in languages in which this distinction is morphologized is discussed in Bashir (1988) and (1995). I have also argued (Bashir 1995) that the compound versus simple verb choice available in Urdu and Hindi is another place in the grammar of this language where this distinction finds expression. If Agha's analysis of -naa forms as constituting a distinct aspectual category (regardless of what name is settled on for it) is correct, and I believe that it is, then the consolidation of a 'potential' aspect, with its implication of prospective (i.e. anticipated) action, may be participating in emergent grammaticization of the category 'expected* vs. 'non-expected' action in Urdu. 25. E.g. maiN karaadi jaanaa caahtaa huuN Ί want to go to Karachi'. 26. E.g. maiN karaadi jaauuNgaa Ί will go to Karachi' (future); maiN karaadi jaa rahaa huuN Ί am going to Karachi' (present continuous). 27. The notation [0 CONSCIOUS CHOICE] indicates 'unmarked for the parameter of CONSCIOUS CHOICE (CC)'. That is, the form can appear in situations where either + CC or - CC is involved, or where the parameter of CC is not relevant The intransitives allowing fluid subject marking all take animate subjects, so SOURCE is manifested as CONSCIOUS CHOICE. 28. Examples 28 and 29 also suggest that generalized second language learning factors may be at work in the development of this construction. 29. Importantly, ne occurs in Panjabi both as an ergative marker and in this construction only in the 3rd person. 30. In the Siraiki examples, implosive consonants are represented by underlining. 31. Similar situations are also found in other Indo-Aryan languages. Masica (1990:336) reports, following Wallace (1985) that in Nepali either the agentive postposition -le or the dative postposition -laii can be used with the infinitive for both abstract inner necessity and specific outer compulsion readings. Marathi and Gujarati also show distributions of volitional and obligational meanings over DATIVE and ERGATIVE cases that may have influenced the

The Urdu and Hindi Ergative Postposition ne · 35 development of this construction in Hindi and Urdu at an earlier stage. In addition, the phonological identity of Marwari ne (DAT) with Hindi ne (ERGATIVE) may have played some role in the early development of Hindi ne. For example, (j) mha ne jodhpur jaaNa he I DAT Jodhpur go (INF) is 'I have to go to Jodhpur*. (Magier 1990: 215)

• REFERENCES Agha, Asif. 1997. Form and function in Urdu-Hindi verb inflection. Department of Applied Linguistics, University of California, Los Angeles, ms. Barker, Muhammad Abd-al Rahman. 1967. A course in Urdu, 3 vols. Montreal: Institute of Islamic Studies, McGill University. Bashir, Elena. 1988. Topics in Kalasha syntax: an areal and typological perspective. Ann Arbor, Michigan: University of Michigan-Ann Arbor, Ph.D. dissertation. University Microfilms. . 1995. Causal chains and compound verbs. Proceedings of the Symposium on Complex Predicates in South Asian Languages (held in Madison, Wisconsin 1989), ed. by Manindra K. Verma and Usha Nilsson. New Delhi: Manohar. Beames, John. 1966. A comparative grammar of the modern Aryan languages of India (originally published 1872-79). Reprinted Delhi: Munshiram Manoharlal. Bubenik, Vit 1989. On the origins and elimination of ergativity in Indo-Aryan languages. Canadian Journal of Linguistics 34 (4). 377-98. Butt, Miriam. 1995. The structure of complex predicates in Urdu. Stanford, CA: Center for the Study of Language and Information. Butt, Miriam and Tracy Holloway King. 1991. Semantic case in Urdu. Chicago Linguistic Society 27. Chicago: Chicago Linguistic Society. Davison, Alice. 1997. Ergative case licensing in a split ergative language. University of Iowa, Department of Linguistics, ms. DcLancey, Scott 1985. Lhasa Tibetan evidentials and the semantics of causation. Berkeley Linguistics Society 11.65-72. Hook Peter, Omkar Nath Koul and Ashok Kumar Koul. 1987. Differential S-marking in Marathi, Hindi-Urdu and Kashmiri. Chicago Linguistic Society 23. Chicago: Chicago Linguistic Society. Jaiswal, M.P. 1962. A linguistic study of Bundeli. Series: Orientalia Rheno-Traiectina, Vol. 8. Leiden: E.J. Brill. Kellogg, S.H. 1965. A grammar of the Hindi language. 2nd ed. (originally published 1893). Reprinted London: Routledge, Kegan Paul Ltd. Magier, David. 1990. Dative/accusative subjects in Marwari. Expenencer subjects in South Asian languages, ed. Manindra K. Verma and K.P. Mohanan. Stanford: Center for the Study of Language and Information, Stanford University. Masica, Colin P. 1990. Varied case marking in obligational constructions. Experiencer subjects in South Asian languages, ed. Manindra K. Verma and K.P. Mohanan. Stanford: Center for the Study of Language and Information, Stanford University. . 1991. The Indo-Aryan languages. Cambridge: Cambridge University Press. Mohanan, Tara. 1994. Argument structure in Hindi. Stanford, CA: Center for the Study of Language and Information, Stanford University. Nairn, C.M. 1975. Introductory Urdu, Vol. II. Chicago: Committee on Southern Asian Studies, University of Chicago. Puar, Joginder Singh. 1990. The Panjabi verb: form and function. Patiala: Publication Bureau, Punjabi University. Shackle, C. 1976. The Siraiki language of central Pakistan. London: School of Oriental and African Studies.

36 · Elena Bashir Singh, Bahadur. 1966. Dialect of Delhi. New Delhi: Representative, South Asia Institute, University of Heidelberg. Singh, Joga. 1993. Case dependencies in Hindi. Papers from the Fifteenth South Asian Language Analysis RoundtaWe Conference, 1993. Iowa City: South Asian Studies Program, University of Iowa. luite, Kevin J., Asif Agha and Randolph Graczyk. 1985. Agentivity, transitivity and the question of active typology. Chicago Linguistic Society 21:2. Chicago: Chicago Linguistic Society. Turner, Mark and Gilles Fauconnier. 1995. Conceptual integration and formal expression, internet ms. Wallace, William D. 1985. Subjects and subjecthood in Nepali: an analysis of Nepali clause structure and its challenges to Relational Grammar and Government and Binding. University of Illinois-Urbana: Ph.D. dissertation.

What Has Bhartrhari Got to Say on Language?1 ι ASHOK R. KELKAR ι Bhartrhari has not received due attention for a variety of reasons. After a word about Bhartrhari and his works, a sketch of the Indian intellectual tradition in which to place him is presented in terms of its key questions, the resulting affiliations, and broad periodization. The difficulties in presenting Bhartjhari to the modern reader need to be overcome. Thus, no citations; Sanskrit terms are parenthesized; the risk of tidying up his thought is taken. A conspectus of Bhartrhari's thought on language is best presented under three headings in a certain order. Language as communication: Bhartrhari's conceptual framework and ideas of related causal dependencies are touched upon. Language as human practice: For Bhartrhari it is human practice that sustains speech power and its acquisition and the power of the speech bond. Language as cognition: All cognition, even seemingly non-linguistic cognition, is sustained by specific and generalized language competence. Bhartrhari's position on the interpretative element and on the presence of chains of signation was well-motivated. Bhartrhari's thought on language is considered in the perspective of Indian and Western thought on language.

•I Anybody who has heard of ancient Indian achievement in the field of grammar has heard of P nini and his grammar of the Sanskrit language. Unfortunately, the same cannot be said of Bhartrhari in spite of his crucial position in the development of the Indian grammatical tradition. The present study is a modest, if somewhat unconventional, attempt to set things right. I hope that those familiar with the tradition will not find it too distorting and those not so familiar will be induced to find out more about his thought.

38 · AshokR.Kelkar As it happens all too often in chronicling ancient and medieval India, not much is known about Bhartrhari. He lived probably in the 5th century CE, in any case not later. The once widely shared supposition that Bhartrhari (B from now on) the grammarian-philosopher is the same as Bhartrhari the highly popular legendary poet-king-renouncer has now been mostly given up by scholars. B is the author of Vakyapadiya (Of the sentence and the word') a probably incomplete verse treatise on the theory of grammar and language; Mahδbhδsyadipikδ ('a little light on the great commentary') a sub-commentary on Patanjali's commentary on Pδnini's grammar available in a short fragment and not too widely studied; Vrtti or Tika ('commentary') an incomplete commentary on Vakyapadiya (VP from now on) that is now believed to have been composed by B himself, and Sabdadhδtusamiksδ ('close search into speech as the basis [of what there is]') a treatise cited but now lost. We shall be concerned here with VP and its auto-commentary. VP has three books (kδndas), namely, Brahma-Kanda ('the book on brahman the Overwhelming Other'), also called Agama-samuccaya ('a collection of traditions') dealing chiefly with language as such and its relation to what there is; Vakya-kδnda ('the book on the sentence'); and the relatively longer and probably incomplete Prakirna-kanda ('the book of miscellaneous (topics)', also called Pada-kδnda ('the book of the wprd'), divided into many topics (samuddesas) some of which have either been lost or remained unwritten. Though the three books appear to constitute a cohesive whole without constituting a connected argument, it has been argued that the title Vakyapadtya properly applies only to the first two books, and the whole work should be called TrUcδndi ('[a work] in three books'). B's contribution to the theory of grammar and reality was recognized as a watershed by grammarians and as a significant contribution to an on-going debate by philosophers of language and reality. There was, however, a certain waning of interest after the 11th century CE, judging from the paucity and bad state of manuscripts, from the paucity of citations and commentaries, and, above all, from the relative neglect in traditional teaching and scholarly activity. Perhaps grammarians found B too philosophical, philosophers found him too philological and students found him too difficult. Among modern Western scholars and West-influenced scholars in India and elsewhere, B has come into his own only in recent decades, judging from editions, translations, citations, studies and from seminars and conferences. A proper understanding of B's thought calls for its proper placing within the Indian grammatical and philosophical traditions. A concern for language (for which the metonyms Sabda, vac, vam, bhδsδ, speech are commonly used and, as will be seen later, differentiated semantically in careful use), and therefore a concern for linguistic forms ranging from word elements to sentence sequences (for which, again, the metonyms sabda and vacana are commonly

What Has Bhartrhari Got to Say on Language? m 39 used with some differentiation), as well as a concern for grammar (yyakarana) was widely shared by all thinkers whether grammarians or not. Traditional Indian thought involves taking a position on certain key questions. It will be useful to set these along with alternate answers and names of groups or individuals to be associated with the answers. (1) What is the chief object of philosophizing? a. Understanding (jnδna) of what there is (yastu). al. Understanding how to understand (vastu-jfwna), according to grammar (including B), hermeneutics (mόnδmsδ), logic (nyaya). a2. Chiefly, understanding what there is to be understood (yastuartha), according to pluralist cosmology (vaisesika) and of course various positive sciences like health science (δywveda), astromathematics (jyotisa) and normative sciences (like ritual, law, poetics). b. Understanding of how to cope with one's life. bl. Chiefly, managing life with success (as befits a worldly one, sδmsδrika), according to the common sense school (lokayata), political economy (arthasastra, dandanόi). b2. Chiefly, changing oneself in such a way as to be free from the need to manage (as befits a liberation-seeking one, mumuksu), according to Samkhya-yoga, Upanishadic thought, Vedδnta, Pratyabhijnδ (the so.called Kashmir Shaivism), Buddhism, Jainism, various god-participation (bhakti) cults. Note: Quite often, thinkers under a take positions under b as well rather than remain neutral. Thus, B (grammar), later logic, later pluralist cosmology take up b2; later hermeneutics takes up position bl with the stipulation that rectitude (dharma) be adhered to. Again, thinkers under b often take up positions under a rather than accept some received account. Thus, Sδmkhya and Vedδnta propose a paucalist cosmology, Buddhism proposes its own pluralist cosmology and its own logic as does Jainism. (2) Understanding how language works is crucially helpful in any serious philosophizing. That is, understanding language vehicle (sabdalvac), language meaning (sabd-artha), and the bond (sambandha) between the two. What then is the chief object in understanding how language works? a. Chiefly, understanding language vehicle, according to phonology (siksδ), lexicon, grammar (Pδnini and others). b. Chiefly, understanding language meaning, according to etymology (nirukta), hermeneutics, logic, poetics (kδvya-sδstra). c. Chiefly, understanding language bond, according to grammar (Vyδdi, Vasurδta, B, Kaunda, Nδgesa), rhetoric (alamkδra-sδstra). (3) How does understanding or cognition come about (jMna-utpatti)! From what sources (jnapaka)?

40 · AshokR.Kelkar a. Chiefly by resolute search (anviksiki), according to logic, pluralist cosmology, grammar (including B), hermeneutics, positive sciences, common sense, political economy, Buddhism, Jainism. b. Chiefly by vision (darsana), according to Sδmkhya-yoga, Upanishadic thought, Vedδnta, Pratyabhijnδ. c. Chiefly by precedent (parampara) according to normative sciences. Note: Quite often, thinkers under each head resort to the remaining two activities too. (4) How can one make sure that understanding really, amounts to right understanding (jnapti)! Discovery (pratyaya) without scrutiny (pariksana) is lame, scrutiny without discovery is blind. a. Cognition is never free from interpretation (nirόpana), what there is being forever inaccessible, according to Buddhism (Nδgδrjuna), grammar (B). b. Cognition needs to reach out to what there is. bl. What there is being independent of the cognizer, (jnδfr), typically material (jada), according to common sense. b2. What there is being inherent in the cognizer, typically consciousness (cetana) or monitoring (vijfiδna), according to Advaita Vedanta, Pratyabhijnδ, Vijφδnavδdi Buddhism (otljer than Nδgδrjuna). b3. What there is being either of these, according to grammar (including B), hermeneutics, logic, pluralist cosmology, Sδmkhya-yoga, Dvaita Vedδnta, Sthaviravδdl and Bδhyδrthavδdi Buddhism, Jainism. (5) Transition between condition (sthiti) and process (gati) in either direction or between conditions or between processes indicates causal dependency (kδrya-tva) as distinct from causal non-dependence (nityatva). How to understand causal dependency? a. Chiefly, as genesis (utpatti) in the course of which inflow (karyiri) undergoes activity (kriyδ) yielding outcome (kδrya) thanks to effectants (kδranas) along with accessories (saha-kδrins) if any. Thus, uncooked rice undergoes cooking yielding cooked rice thanks to fire, water etc, along with the occasion for cooking it, etc. Genesis activity typically leads to a fresh episode in which the outcome now becomes the inflow, cooked rice getting digested or cooked, so to say, in the belly in the example at hand. al. Process alone exists, effectant and outcome merely subsist, according to Vijnδnavδdi Buddhism, Vedδnta (Gaudapada) Jainism. (This is the a-jati-vada position.) a2. Process and outcome exist, according to Sthaviravδdl and Bδhyδrthavδdi Buddhism. (This is the samutpδda-vδda position, that tends to highlight causal dependency chains \pratitya~samutpada].) a3. Process, outcome and condition exist, inflow being pervasive (vyδpaka) or inclusive of outcome, according to logic, pluralist cosmology. (This is theparinδma-vδda position.)

What Has Bhartrhari Got to Say on Language? · 41

b. Chiefly, manifestation (abhivyakti) in the course of which unmanifest (vyanjaka/a-vyakta) undergoes manifesting, (abhivyanjanal parindma) yielding manifest (vyangya/vyakla) thanks to power-transmission (sakti-sancara), along with manifesting-assistants (abhivyanjaka), if any. Thus, a seed undergoes sprouting and growing yielding a seedling thanks to life transmission along with soil, moisture etc. Manifesting typically metamorphoses the unmanifest into the manifest, seed into seeding in the example at hand. bl. Condition, unmanifest, manifest exist, manifest being no other than (tadatmd) the unmanifest, according to Sδmkhya, early Vedδnta. (This is the sat-kδrya-vδda position.) t>2. Condition, unmanifest exist, manifest being no other than a figuring out imposed (yivartd) on unmanifest, according to grammar (B), later Vedδnta (Samkara onwards). Note: (1) The terms under a are often substituted for the corresponding terms under b. B, in spite of his b2 position, not only does this but on occasion adopts a3 stance. (2) In certain cases, the inflow/unmanifest and the outcome/manifest are seen to share a site (δsraya/adhikarana). (3) Causal dependency may well be conducive, compulsive, obstructive, or preventive (anukόla, bandhaka,pratikula,pratibandhaka respectively). (4) While the distinction between the causally dependent (karya) or the causally non-dependent (nitya) is crucial in Indian thought, it is open to a thinker, given the occasion, to suspend consideration of the distinction and provisionally use the more fluid distinction between the achievable (sadkya) and the available (siddha). Thus, language analysis consistently ignores language history and treats everything, in the language being looked at, as available whether it is liable to be lost at a later time by virtue of causal dependency or not so liable. (6) When is man most himself when coming to terms with what there is? a. In understanding rightly what there is to be understood, according to Sδmkhya-yoga, Pratyabhijnδ, grammar (B). b. In accomplishing one's goal through action (karmari), according to common sense, hermeneutics. c. In accomplishing one's goal through action and at the same time understanding rightly what there is to be understood, according to Bhagavadglta, god-participation schools, Dvaita Vedδnta. d. In understanding rightly what there is to cope with, namely, suffering (duhkha/klesa), according to Upanishadic thought (at least some), Advaita Vedδnta, Buddhism, Jainism. Note: Thinkers under a, c, and (except Buddhism, Jainism) d recognize a non-interacting inner self (δtmari) beyond the interacting outer self (asmitδ/ahamkδra). The rest do not. Thinkers under d also take the

42 · AshokR.Kelkar position that there is nothing to be understood (Buddhism, Jainism) or that what there is cannot be understood, being the Overwhelming Other (Upanishadic thought, Advaita Vedδnta); B joins the second group though falling under a. (7) In what different ways does man interact with what there is? a. Being a man is no more than being a site for certain channels of consciousness (pudgala 'individual'), according to Buddhism, Jainism. b. Being a man is at least being a person (puru$a) with certain channels of interaction, according to the rest, including health science. Note: On either view, one does not think of man as body and soul put together or as body, mind, and spirit put together, but rather as a whole person to whom certain channels are available for interacting. There are certain broad correspondences between different schemes of the channels. Two such schemes widely accepted under a and b are presented here for comparison. al presentation (rόpa) a2 name-holding (nδma) a2a sensing (yedana) a2b mentation (samjna) a2c disposition (samskδra) a2d monitoring (vijnana)

bl b2 b2a b2b b2bl b2b2 b2b3 b2b4

gross (sthόla) subtile (sόksma) outer (bahya) inner (antar) minding (manas) disposing (citta) controlling (buddhi) referral (asmita/ahamkara)

Sδriikhya was probably the originator of the scheme. Each channel under b2 has a cognition aspect and an action aspect; for example, at b2a are recognized five sense modes (sight, hearing, touch, smell, taste) and five action modes (speech, handling, legging, defecation, urinationcoitus). Memory is the cognitive aspect of b2b2 and a2c. The scheme appears in Upanishadic thought as the doctrine of five sheaths covering the inner self: they roughly correspond to bl, b2a, b2bl-2, b2b3-4, the fifth sheath being the inmost state of the inner self. It will be noticed that there are certain broad correlations between the positions taken on each of the key questions. There are of course many complications. B appears to plough his own rather lonely furrow. The task of placing B's thought will not be complete without a rough indication of the periods in the unfolding of the Indian intellectual tradition. (This periodization does not wholly correspond to the periodization of political-social-economic history.) (1) Proto-historic period between the 20th and 8th century BC associated with Veda and brδhmana texts, with a poetic-mythic-ritual-magicalethnocentric-this-worldlyworldview.

What Has Bhartrhari Got to Say on Language? m 43 1st mythopoeia: the India pantheon out of the Varuna pantheon; asura demonology. 1st brahmanization: maintaining Vedic texts and practices and Sanskrit in a beleagured Δryδvarta (10th to 6th century BC). (2) Transition to historic period between 8th and 4th century BC associated with Upanishadic thought, the Epics, Buddhism and Jainism, gradual emergence of a prosaicKlialectic-eάόcal-technical-anthropocentric worldview admitting the other-worldly option, the cosmocentric option, and the genesis option (by the side of the accepted manifestation account of causal dependency), wide acceptance of the twin doctrines ofkarma-phala and rebirth. 1st spiritual ferment: Upanishadic thought, Buddhism, Jainism. 1st intellectual ferment: the emergence of disciplines and their gradual dissociation from the Vedas. 1st debrahmanization: Sδmkhya-yoga, Thntra. (3) Ancient period between 3rd century BC and 7th century CE associated with India as the home of a world-linked civilization. 2nd mythopoeia: the Trimόrti and Devi pantheon; rδksasa demonol2nd brahmanization between 3rd century BC and 4th century CE: emphasis on codification, hierarchy, constraints, anti-intellectualism. 2nd intellectual ferment: debates between and within Vedic, Buddhist, Jain, and Vedδnta camps between the 4th and 9th century CE. (4) Medieval period between the 8th and 18th century CE associated with bhakti cults, purδnas, insulation from non-Hindu contacts across and within the borders, the rise of Modern Indian languages and their poetry (little or no prose of intellectual discourse in these languages) 3rd intellectual ferment between the 8th and 18th century cfc. 2nd debrahmanization: Nδth and Kundalin'i Yoga, bhakti. 3rd brahmanization: intellectual complacency or intolerance. (5) Post-medieval period 19th century CE onwards associated with the challenge of modernity to the Indian tradition. B belongs to the 2nd intellectual ferment. He carries on from grammar (Vyδdi, Vajapyδyana, Audumbavδyana, Patanjali, his own teacher Vasurδta); borrows and reacts to Buddhist logic (Vasubandhu). He was supported by Mandanamisra, and grammar (Kaunda, NageSa^), influenced Pratyabhijnδ ( Abhinavagupta) and poetics ( Anandavardhana), and was opposed by hermeneutics (Kumδrila), logic (Vacaspati and Jayanta), Buddhist logic (Dinnδga, Dharmakirti), Vedδnta (Samkara). In a milieu of polemical dialogue or a sententious monologue, B's presentation is often in the likeness of an interior monologue in which he reviews various available positions and then offers a new position or selects an available position such that it will put the remaining positions in a new perspective that will make allowance for whatever

44 · AshokR.Kelkar plausibility they possess. His cognitive style is gentle rather than hard, depthseeking rather than tidiness-seeking. Once in a while, he treats us to charming cameos of ordinary life by way of illustration (a mother pacifying a crying child by saying, 'Be quiet or the tiger will cany you away') or metaphor (a birdcatcher plying his trade). Having placed B within the Indian tradition and having made the modern reader aware of the key questions that were being asked within that tradition, we shall now consider some of the other difficulties in understanding B and ways of mitigating them. Inadequate translations of key terms like Sabda (as word), artha (as sense) Jnδna (as knowledge), nitya (as constant or even eternal) in contexts in which they don't have these senses is another difficulty that we have sought to remove. His handling ofkδrikδs, which are supposed to be the verse counterparts of the prose aphorisms (sutras) of other thinkers, has not made for lucidity. His terminology is not stream-lined but fluid; this may be carelessness, but may also be due to his pioneer status. His doctrine otsphota, oi a-khanda-paksa, and ofsabda-tattva are often discussed in isolation from the rest of his thought, which make these seem stranger or more mystifying than they need to. His arrangement of topics does not always make for a clear grasp on our part of the overall structure of his thought or of its cross-disciplinary ramifications. Kant has said (Critique of Pure Reason B 370) that our job is to understand Plato better than he understood himself. This study is an attempt to present B's thought on language with a more careful arrangement of topics, a more careful use of terminology, but with a less careful separation between what B has taken over from the tradition and from other thinkers and what he has contributed himself and also a less careful separation between what he has to say tacitly or implicitly and what he has to say explicitly. We hope to bring in more tidiness without sacrificing either the gentleness or the depth. (What Kant said is not a piece of presumptiousness or arrogance but the recognition of one way of repaying our debt to a major thinker.)

• II The foregoing preliminary notes, sketchy as they are, should be helpful especially to a modern philosopher or scientist of language, for placing in perspective B's thought, which does not really recognize a split between the philosophy (mmδrhsδ) of language and the science (sδstra) of language or a split between a consideration of speech (vyδkarana), of interpretation (mlmδmsδ in the narrower sense of hermeneutics) and of cognition (nyaya that comprises logic, dialectic and epistemology as one sees them today). His thought on the narrowly grammatical questions (available in most of the 'topics' of the third book of VP) cannot be fully understood except in the

What Has Bhartrhari Got to Say on Language? · 45

context of certain prior steps (available in the first two books and some 'topics' of the third). The steps in their order are: Language as communication (pratipadana) Language as human practice (loka-vyavahδra) Language as cognition (jnana) The present account, as the title suggests, is confined to these three steps. No textual references will be offered. The Sanskrit terms are given in parentheses primarily on their first occurrence and should serve to offer bearings to the reader. (A newcomer should feel free to ignore them in the first reading.) Language as communication: Communication (pratipadana) presupposes comprehension (pratipatti). The first is an action episode (karman, kriya), but the second is a cognition episode (jnana andjnapti). In any communication episode, the communicator (pratipadayitr) brings about the comprehension of the meaning (pratipδdya/pratipanna) in the comprehendor (pratipaltr) with the help of some means of communication (pratipδdaka). (B's terminology is not all that stream-lined, it is more fluid and informal, as has already been indicated. This observation applies throughout.) All communication and comprehension is, in a deep sense, linguistic. This is true of seemingly non-linguistic communication as in animals, newborn humans, children, even humans when they use silent language as in pointing fingers, flying banners, or wooing the beloved. In what is normally called a linguistic communication-and-comprehensionepisode, namely, a speech transaction (yδg-vyavahδra), the speaker (vaktr) brings about speech reception (sabda-grahana) of the speech meaning (sabdδrtha/vδcya/ukla) in the listener (Srotr, Sabda-grahltr) being addressed (sambo· dhita) with the help of some speech act (vacana, ukti). One may present the causal dependency as genesis or as manifestation. (1) In the course of genesis, inflow speech act undergoes the speech transaction activity, yielding outcome speech reception thanks to effectant speech bond. In the site of the speaker, speech meaning undergoes speech performance (sabda-prayoga), yielding speech vehicle thanks to speech bond. In the site of the listener, speech vehicle undergoes speech performance yielding speech meaning thanks to speech bond. (2) In the course of manifestation, unmanifest speech (a-vyakta-Sabda) undergoes the speech transaction yielding manifest speech vehicle (yyaktasabda) thanks to speech power transmission (Sabda-sakti-sancara). In the site of the speaker, unmanifest speech undergoes speech power linkage (sabda-sakti-yoga), yielding figured-out (sa-vikalpa), qualityspecified (sa-guna), and accessibly formed (s-akara) speech meaning in successive increments, thanks respectively to inner, mediate, outer speech power transmission (pasyanti, madhyama, vaikharlvam). In the

46 · AshokR.Kelkar

site of the listener, manifest speech undergoes speech power linkage, yielding unmanifest speech in reversed stages. Thus, looking out of the window, one sees falling drops of water. One 'sees' that it is raining. One then recalls a dark shape seen on the horizon earlier. One 'sees' that one saw a dark cloud. One wonders about the connection between the ongoing experience and the recalled experience, and welcomes the rain on a stuffy afternoon. One 'sees' in a flash how the dark cloud has brought welcome rain. This inner seeing, this figuring out is inner speech. One silently says to oneself in English or Hindi or whatever and hears oneself saying this or that thing. This is figured-out inner speech getting translated, so to say, into figured-out mediate speech complete with classified segments in due sequence. One may then choose to speak out to another and choose what one speaks about and speaks out. The outcome is figured-out, quality-specified, and accessibly formed outer speech. One hopes that the listener will arrive at a seeing that is reasonably close to one's own seeing. Note: B has conveyed this account in terms of manifestation by using a popular metaphor already used by grammarians in other connections: the metaphor of bursting sound as on percussion (sphota) and resulting resonance (dhvani, nada) respectively for unmanifest speech and manifest speech. B does not turn down the genesis account, but finds the manifestation account much more illuminating. Language as human practice: A new term, 'speech power', has been introduced. Speech power is the power that sustains speech bond in a specific language (bhδsδ). Now what sustains the power of the speech bond? B places language squarely in human practice. (1) In the course of manifestation, unmanifest speaker-intention-claim (yivaksa) undergoes speech transaction yielding manifest listenerdiscovery (pratyaya) thanks to the intentionality-in-the-speech-bond (tatparya-sakti) in the site of the interlocutors (yak-pray oktrs). Note: The term tatparya-sakti came later, B usesyogya-bhδva. The term yogyatδ later came to mean the compatibility aspect of sentencecohesion, and tδtparya-sakti (literally 'that-for-ness') was assigned to intentionality-in-the-speech-bond. (2) In the site of a would-be interlocutor, typically a child, unmanifest specific language competence (sabda-sakti) undergoing language acquisition (sabda-sakti-graha) yields manifest specific language competence (Sabda-sakti), thanks to language faculty (sabda-bhavana), along with assistants like exposure to grown-up-interlocutor-transaction (vrddhavyavahδra) and to elucidation (vivarana) offered by grown-up interlocutors, grammarians or the like.

What Has Bhartrhari Got to Say on Language? m 47

Two more questions can be asked for which B has only partial answers. The question 'How do specific languages arise?' was traditionally answered: 'Through language degeneration' (apabhramsa). But B holds that language faculty also causes generalized language competence (Sabda-tattva) to be manifest. This in turn is unmanifest specific language competence. The question 'Where does language faculty come from?' was answered by B: 'Through traces (yasana) left from previous births' (which was the traditional Indian way of saying 'as innate capacity'). The traditional Indian way of asking the question 'What sustains the power of the speech bond?' was 'Is speech bond characterized by causal dependency (kδrya-tva) or causal non-dependence (nitya-tva)T Ordinarily, interlocutors and language specialists choose to regard speech bond simply as available (siddha). In a few cases (samjnδ-sabda), as given proper names and assigned technical terms, the speech meaning has been arbitrarily (yadrcchaya) assigned to the speech form within the span of memory (δdhunika). What about the majority of cases that remain? (3) Is the speech bond essentially causal-dependency-bound (karya) or causally independent (nitya)! a. It is causally dependent, being basically sustained by assignment (samketa) by convention (samaya). al. In accordance with God's will (isvar-eccha) which imparts power to it, according to early logic, pluralist cosmology. Note: This position has evoked the ironic query, 'God appears to have plenty of leisure to do it lexical item by lexical item?' This is actually a pious way of mitigating (a), except in the case ofsamjnδ. a2. In accordance with somebody's will, whether God's or (in the case ofsamjnδ) an ordinary person's, according to later logic. b. It is causally independent and the common property of people (loka). bl. Except when (in the case ofsamjnδ) effected by restrictive rule (niyama-krta), according to hermeneu tics, Vedδnta. b2. It is felt-to-be-appropriate (yogyatδ-rόpa) because people, conditioned by memory of constant association (nitya-sahacatya-smrti), actually impute (adhyasa) the meaning to the speech form; indeed they even identify the two (tδdδtmya), according to grammar. c. It is sustained by specific language competence which is manifest generalized language competence, according to B. Language as cognition: All cognition is, in a deep sense, linguistic; this is true of seemingly non-linguistic cognition, as in animals, newborn humans, children. Comprehension of what is communicated is of course only a special

48 · Ashok R. Kelkar

case of cognition. Generalized language competence covers both communication and comprehension. In taking this position, B does not receive much support. This position of B is his way of saying three things: (1) Some say that cognition needs to reach out to what there is, but what there is is inherent in the cognizer, typically the cognizer's consciousness. The impression that what there is can be independent of the cognizer is no more than the cognizer's imaginative figuring out. B argues that, while no cognition is free from interpretation (nirupana), this additive is not so much a matter wholly private to the cognizer as a matter publicly shared through generalized language competence if not through specific language competence. Even wholly silent cognition that does not get 'translated' into mediate or outer speech falls back on inner speech. How else to account space or time sense, for example? (2) Is there pure discovery (pratyaya) that is yet free from inner speech? Pure sensation, for example? B concedes that this can be. But such pure discovery is cognition in only a rudimentary sense: it cannot be retrieved, communicated, or manipulated in any way. Any cognition worth the name involves at least some figuring out (vikalpa) and any such figuring out attracts at least rudimentary mediate speech (such as 'this', 'that', 'something or other'). As we have already seen, inner speech is figured-out (sa-vikalpa). Figuring out, inner speech (pas· yanti), time-wise imaging or storying (kriya~vivarta)y space-wise imaging or picturing (mόrti-vivarta) are all channelled through disposing (citta). (B uses the cover termpratibha for all such capacities of citta.) Sooner or later, we have to recognize inmost recesses ofpafyanti, call itparapasyanti if you like. (Later, the terms parapasyanti or simply para came to be used.) (In any case, how else account for mystical experience except as inmost seeing, figured-out or non-figured-out as the case may be?) (3) Man is most himself in understanding rightly what there is to be understood. What there is to be understood cannot be understood, being the Overwhelming Other (brahman), the eminently unmanifest, any aspect of the manifest being no more than an imposed figuring out (yivarta). Considering that all that is manifest is open to interpretation through speech (sabda), what better way could there be of being most oneself in understanding rightly than through generalized speech competence (sabda-tattva). One has access to the Overwhelming Other through speech (sabda-brahma). Note: Compare similar claims about musical resonance (nδda-brahma). Another related position that B takes and that was widely resisted was his invoking unmanifest speech (a-vyakta-sabda or sphota in attempting certain

What Has Bhartrhari Got to Say on Language? m 49 problems in accounting for the operations of language recognized by Indian grammarians. (1) How is it that varying manifests are figured out as the same language unit, whether speech sound (varna), word (pada), or sentence (vδAya)? By virtue of the unmanifest kind (jδti-sphota). (2) How is it that fragmented manifests are figured out as the same language unit, whether speech sound, word, or sentence? By virtue of the unmanifest unity (akhanda-sphota). Actually, B proposed his version of the sphota doctrine in connection with varna-jati'Sphota, akhanda-pada-sphota, and akhanda-vδkya-sphota. Others applied it to the remaining cases. The crucial point that concerns B's version of the doctrine is the Janus-faced role played by unmanifest speech sound, unmanifest word, and unmanifest sentence. They are both signates (vacyas) and signants (vacakas). Thus, unmanifest word is the signate of speech sounds but the signant of speech meanings. B also takes up another related problem. (3) How is it that certain operator speech forms such as certain flexions or particles participate in the speech meaning formulation or comprehension without any speech meaning directly assignable to them? They are essentially signants that assist other signants in meaning formulation or comprehension. Thus, a negative operator within a word or a sentence has no signate of its own but merely bars the signate of the signant it accompanies. Note: Later, such operators were treated not as signants (vacakas) but as co-signants (dyotakas). B distinguishes between content speech forms (anvakhyeya-Sabda) and communicating speech forms (prδlipadaka-sabda); the terms chosen by him are not quite felicitous. This exposition of what B has got to say on language when read carefully with the preliminary notes should serve to reveal the wide-ranging scope and the deep penetration of this pioneering thinker on language. (Let it be noted that we have left out B's insights on the specific topics in language.) Those familiar with modern studies of B will appreciate that the tendency to mystify has been consistently resisted in the present account. It may be noted in passing how Saussure resembles B in adopting a semiotic framework in thinking on language and linguistic analysis.

• Ill How do we sum up the overall trend of B's thought on language as a philologist ('lover of speech') and as a philosopher ('lover of wisdom') in a terminology that is accessible to us?

50 · AshokR.Kelkar (1) Language (unmanifest speech power) is certainly a means (contributory cause) of communication (pratipadana). But it is much more. It is nothing less than a medium (manifestation) of human understanding or cognition (jnana) in that a generalized speech power manifests man's capacity for cognition and a specific language manifests that generalized speech power. a. So, as a philologist, he attends to the speech meaning rather than the speech vehicle. b. So, as a philosopher, he sees part meanings as no more than intersections of whole meanings (a-kfanda-sabd-artha), and sees universals (kinds, qualities, actions) rather than particulars (substances, individuals) as the basis of the unity of word meanings. (2) Language use (speech transaction, vδg-vyavahδra) is certainly the performing of a text calling for interpretation. But it is much more. It is nothing less than the performing of an act calling for a response. a. So, as a philologist, he attends to the actual specifics (speech transaction) rather than the abstract form (speech power, sabda-Sakti). b. As a philosopher, he sees configuration (kalpana) imposed by the minding faculty (manas) rather than any direct account of what there is out there (bδhyδrtha) as available to man. It should be interesting, and instructive, to compare B's thought on language with that of both early and later Wittgenstein. Earlier Wittgenstein, an associate of the Ideal language philosophers, appears to agree with B with respect to 2,2b but not so much with respect to 1, Ib. Later Wittgenstein, an associate of Ordinary language philosophers and a part-time philologist, appears to agree with B on both 1, Ib and 2,2b. One difficulty in comparing the two thinkers is that the problematics they are disposed to select for closer attention in applying their basic insights, present a somewhat limited overlap. This, of course, is also a measure of the difference between the traditions within which the two respectively operate. It will be interesting to make similar comparisons with other Indian and Western thinkers on language. The following schemata are only a tentative beginning, subject of course not only to elaboration but also to possible corrections. (1) A. Language is chiefly a means of human communication and its use is chiefly the making of a text. B. Language is chiefly a means of human communication and its use is chiefly the performing of an act. C. Language is chiefly a medium of human cognition and its use is chiefly the making of a text. D. Language is chiefly a medium of human cognition and its use is chiefly the performing of an act.

What Has Bhartrhari Got to Say on Language? m 51 Note: It will be seen that B fits best under (D), and so does later Wittgenstein but earlier Wittgenstein fits best under (B). (2) Philologists: A. Pδnini, the siksδ phonologists; Bloomfield. B. Theprakriyδ grammarians (like Bhattoji), the prδtiSδkhya phoneticians; Sδussure, the Prague School, Halliday. C. Nage§a; Chomsky. D. Bhartrhari; Sapir, the rebels against Chomsky. Note: The cleavage between (A, C) and (B, D) appears to be more salient than that between (A, B) and (C, D). (3) Philosophers of language A. Logic (nyδya school); Locke. B. Ideal language philosophy, earlier Wittgenstein. C. Hermeneutics (mimδmsδ school); Descartes, Frege. D. Bhartrhari; Kant, Humboldt, later Wittgenstein, Ordinary language philosophy. Note: The cleavages appear to go the same way as with philologists (under 2). There is of course more to B and other thinkers on language than such pigeonholing, but some diversion was needed after a strenuous journey! • NOTES 1. This is a thoroughly revised version of a paper presented at the seminar on 'Word and Sentence: Two Perspectives: Bhartrhari and Wittgenstein' at Sahitya Akademi (The Indian Academy of Letters), New Delhi, between 12 and 14 December 1994. The paper owes its existence to my senior friend Professor K.J. Shah who would not take no for an answer from an absolute novice in Bhartrhari studies. Unfortunately he did not live to attend the seminar that he had planned. To his fond memory I dedicate this essay. A novice badly needs both encouragement and nitpicking criticism. Both were presented in ample measure by my junior friend, Professor Ashok Aklujkar (Department of Sanskrit, University of British Columbia, Vancouver) who suffered both versions and my obstinateness. Comments will be welcome from the knowledgeable and the curious and may be sent to the following address: 7 Dhananjay, 759/83 off Bhandarkar Road, Pune 411004, India.

• REFERENCES Texts and translations Abhyankar, R.V. and V.P. Limaye (eds) 1965. Vakyapadiya ofBhartrhari. Pune: University of Pune.

(Bare text)

Rau and Wilhelm (eds) 1977. Bhartrharis Vakyapadya. Franz Steiner for Deutsche Morgenlδndische Gesellschaft. (Bare text.) Iyer and K.A. Subramania (eds and transl.) Vakyapadiya ofBhartrhari. Kanda 1, vrtti, Vrsabha's commentary. 1965. Pune: Deccan College, English translation 1966. Pune: Deccan College.

52 · AshokR.Kelkar Kδnda 2, vrtti, Punyarδja's commentary. 1983. Delhi: Motilal Banarsidass. English translation 1977. Delhi: Motilal Banarsidass. Kanda 3, Helaraja's commentary, 2 parts. 1963,1973. Pune: Deccan College, English translation part 1.1971. Pune: Deccan College, 1971 part 2.1974. Delhi: Motilal Banarsidass. (of course no vrtti for book 3 is available.) Abhyankar, K.V. and V.P. Limaye (eds) 1970. Mahδbhδsyadipikδ of Bhartrhan. Pune: Bhandarkar Oriental Research Institute. Critical ed. with English translation 1985-91. (δhnikas 1-7). 8 fascicules, by various hands. Pune: Bhandarkar Oriental Research Institute. Indian intellectual tradition Mohanty Jatindra Nath. 1974. Indian philosophy. Encyclopaedia Bntannica.Macropaedia Vol. 9, 313-34.15th ed. Chicago; London. Potter, Karl H. 1963. Presuppositions of India 's philosophies. Englewood Cliffs, NJ: Prentice-Hall. Bary, William Theodore de, et al. (compilers) 1958. Sources of Indian tradition. New York: Columbia University Press. Indian rpt: Delhi: Motilal Banarsidass, 1963. (Extracts in English translation.) Indian thought on grammar, language, logic, cognition and reality Biardeau, Madeleine. 1964. Theorie de la connaissance et philosophic de parole dans le brahmanisme dassique. Paris, The Hague: Mouton. Chakraborti, Prabhat Chandra. 1931. Thephilosophy of 'Sanskrit grammar. Calcutta: University of Calcutta. . 1933. Linguistic speculations of the Hindus. Calcutta: University of Calcutta. Kunjunni Raja, K. 1963. Indian theories of meaning. Chennai (Madras): Adyar Library; revised 1969. (Based on University of London Ph.D. dissertation, 1954.) Matilal, Bimal Krishna. 1971. Epistemology, logic, and grammar in Indian philosophical analysis. (The Hague: Mouton.) Sections 1:4,29-34; 3:5,109-13 on B's thought . 1985. Logic, language and reality: An introduction to Indian philosophical studies. Delhi: Motilal Banarsidass. -. 1990. The word and the world: India 's contribution to the study of language. Delhi: Oxford University Press. Ruegg, David Seyfort 1959. Contributions a I'histoire de la philosophic linguistique indienne. Paris: deBoccard. Coward, Harold G., Raja K. Kunjunni (ed.) 1990. The philosophy of the grammarians. Being: Encyclopaedia of Indian philosophies, ed. by Potter, Karl H., Vol. 5. Delhi: Motilal Banarsidass. Steal, John F. (ed.) 1985./I reader on the Sanskrit grammarians. Cambridge, MA: MIT Press, 1951; Delhi: Motilal Banarsidass. Bhartrhari's thought Aklujkar, Ashok Narhar. 1970. The philosophy of Bhartjhari's THkandi. Harvard University: Ph.D. dissertation, (unpublished). Gaurinatha Sastri. 1959 [1981]. The philosophy of word and meaning: Some Indian approaches with special reference to Bhartrhari. Calcutta: Government Sanskrit College. Rptd 1981. . 1980. A study of the dialectics ofsphota. Delhi: Motilal Banarsidass. . 1991. The philosophy of Bhartrhan. Delhi: Bharatiya Vidya Prakashan. Iyer, KA. Subramania. 1969. Bhartrhan: A study of the Vakyapadlya in the light of the ancient commentaries. Pune: Deccan College. Matilal, Bimal Krishna, P.K. Sen. The context principle and some Indian controversies over meaning. Mind 97:73-97, January 1988. (Frege, Mlmδmsδ, and Bhartjhari.) Bhate, Saroja, \ashodhara Kar. 1992. Word index to the Vakyapadtya of Bhartrhan (together with the complete text). Delhi: Eastern Book Linkers. Bhate, Saroja and Johannes Bronkhurst. (eds) Proceedings of the First International Conference on Bhartrhari, University of Poona, January 6-8,1992. Being: Asiatische Studien/Etudes asiatiques. Bern: Peter Lang. 47:1,1993. Including: Aklujkar, Ashok N.An Introduction the study of Bhartrhari, 7-36; Ramseier, Yves, Bibliography ofBhartrhari, 235-67.

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes* • MARIO PALASCHKE and WOLFGANG U. DRESSLER · Among the assimilation processes that Old IndoAryan consonant clusters underwent in Middle Indie, the change of plosive-sibilant clusters (-TS- and -ST-) to -TTh- has not yet found an adequate explanation. We will focus only on recent substantial attempts, i.e. syllabic analyses. Our counterproposal is based on two subtheories of Natural Phonology, those of universal preference and of system adequacy. Among universal natural phonological processes we will refer to the processes of sibilant aspiration and palatal affricate formation. Within the phonological system adequacy of Middle Indo-Aryan languages, the reduction of the phonemic and allophonic space of sibilants and the role of aspirate palatals will be particularly relevant.

• 1. Introduction • 1.1. Middle Indie in the History of Modern Linguistics In comparison with the Old Indo-Aryan (ΟΙΑ) languages, the Middle IndoAryan (MIA) languages or dialects have never aroused the same interest among grammarians and linguists. This is true of the tradition of Indian grammarians, where MIA was mainly dealt with as an appendix to Sanskrit grammars such as in the eighth chapter of Hemacandra's Sanskrit grammar, as well * This contribution represents an elaboration of parts of the first author's MA thesis, Vienna University 1997. We would like to thank Brett Kessler (Stanford University) and Julian Mondez Dosuna (University of Salamanca) for bibliographic help.

54 · Mario Palaschke and Wolfgang U. Dressler

as of modern Western linguistics. If MIA data is included in an analysis at all, it is prakritisms which found their way into the oldest stages of Indo-Aryan through borrowing (see, e.g. Wackernagel 1896: xviii) on the one hand. On the other hand, MIA provides evidence for the areal distribution of unclear phenomena (e.g. Emeneau 1988). In theoretical linguistics, researchers were mainly attracted by the accuracy of descriptions and formalisms adopted by Sanskrit grammarians. • 1.2. Syllable Structure vs. Natural Processes There is one chapter of MIA historical grammar which appears repeatedly in theoretical linguistic research: the 'Middle Indie Assimilation' (see Geiger 1916: §51). Grammont (1933), for example, cites MIA data in his chapter on assimilation, but in the most recent contributions dealing with MIA assimilations (Murray 1982; Vaux 1992; Kessler 1995), this data is used to show how a change in syllable structure triggers sound change. Moreover, they try to provide analyses, which should allow us to reduce the different kinds of assimilations between ΟΙΑ and MIA to the single phenomenon of syllable structure change. A major argument for these analyses is found in the ΟΙΑ gemination processes reported by Sanskrit grammarians and phoneticians (cf. Varma 1961) and first brought to the attention of Western scholars by Jacobi in 1881: We see that in most cases in which Prakrit shows gemination as a result of total assimilation of the consonants, these consonants were pronounced as doubled consonants in Sanskrit. If we assume a causal connection between these two phenomena, we obtain an explanation for the Prakrit gemination as well as an understanding of the mysterious doubling of consonants in Sanskrit (Jacobi 1881: 609, our translation). In this paper, we will argue that the reduction to one single principle is far from being adequate to account for the various changes within consonant clusters from ΟΙΑ to MIA. The example of those consonant clusters which consist of plosives and sibilants (-TS- and -ST-) shall provide arguments in favor of a dynamic phonological model of processes and rules (Natural Phonology, in our case) which are able to account for chronological steps as well as areal differences and intradialectal variation. • 1.3. The Middle Indie Assimilation As a rule—not without exception—ΟΙΑ consonant clusters are eliminated in MIA by either inserting a vowel or by assimilation (or fusion). In most MIA languages just -NT-, -TT- and -777i-, i.e. homorganic nasal-plosiv clusters,

Middle Indie Aspirate Formation: Syllable Structure OS. Natural Processes · 55 geminates and aspirated geminates, are allowed. In order to cover the Middle Indie Assimilation, Geiger (1916: §52) proposed an assimilation hierarchy, according to which 'stronger' consonants assimilate 'weaker' consonants. This hierarchy was discussed in several papers concerning sonority and consonantal strength scales in modern theoretical linguistics (e.g. Hankamer and Aissen 1974; Murray 1982): (1) Plosives > Sibilants > Nasals > / > v >y > r exemplified by: (2) ΟΙΑ agni-Tire' OIA Svapna- 'sleep' OIA aSva- 'horse' OIA Sukla- 'white'

> > > >

P. aggiP. soppaP. assaP. sukka-

Between consonants of the same level of the assimilation hierarchy, we find regressive assimilation: (3) ΟΙΑ rakta- 'read' > MIA rattaOIA utkanth 'desire' > MIA ukkanth Geiger himself was ready to admit some further exceptions: If the consonant cluster contains an aspirate, the aspiration appears at the end of the assimilated consonant group: kh + y becomes kkh, k + th becomes tth. Aspiration usually also appears, if the original consonant group contains a sibilant: st becomes tth (Geiger 1916: §51, our translation). Apart from Pali, which preserves si-clusters, due to its artificial, canonical and Sanskrit-oriented character, inscriptional Prakrits and G ndh ri, which preserve »^clusters in terms of an area laterale, all ΟΙΑ clusters containing sibilants appear as aspirated geminates in MIA: (4) a. Sibilant + Plosive: Skt. niska- 'gold ornament' > P. nikkhab. Plosive + Sibilant: Skt. aksi- 'eye' > P. akkhi- or acchiSkt. apsaras- 'nymph' > P.acchar Skt. kutsita- 'contemptuous' > P.kucchitaAs can already be noticed in these examples, there is a different outcome of the assimilation or fusion processes according to the sequencing of sibilant and plosive. Whereas Sr-clusters always change to 7TA, an aspirated geminate at the place of articulation of the plosive, rS-clusters nearly always result in the aspirated palatal geminate cch. The fact that the group -As-develops into -ccft- in the Western languages, but -kkh- in the Eastern languages is discussed in Berger (1955) and Tedesco (1956).

56 · Mario Pcdaschke and Wolfgang U. Dressler

m 2. Syllabic Analyses • 2.1. Murray (1982) Murray tries to get rid of the 'shortcomings' in Geiger's analysis (such as having both progressive and regressive assimilation) by suggesting an analysis based on the principles of Natural Generative Phonology (cf..Hooper 1976). In this framework, marked syllable contacts have to be improved according to a scale of consonantal strength and to 'preference laws' for syllable structure (see Vennemann 1988). In Murray's analysis, the first step is an improvement of the syllable contact by the gemination process in ΟΙΑ already mentioned in 1.2, followed by the reduction within the syllable onset according to the 'Law of Initials': 'Medial syllable-initial clusters should be possible word initial clusters' (Vennemann 1972:11). (5) word-initial: Skt. krayavikraya- 'commerce' > P.kayavikkayasyllable-initial: Skt: *tak.ra- Vhe/ > *tak.kra- > P.takkaFor ~ST- and -TS-clusters, Murray admits that these cases remain problematic, but he assumes the following chronological steps (1982: 181 n. 13): (6) a. 1. 2. 3. b. 1. 2. 3.

nis.kanisk.kanik.khaak.siak.k§iak.khi-

gemination: 'Law of Finals' aspiration and metathesis gemination: violation of Hooper's principle1 aspiration

This analysis does not explain the special status of these clusters, since the aspiration of s remains unexplained and has to be dealt with as an exception to the usual total assimilations. In addition, it does not cope with the fact that 73clusters usually result in -cch~. m 2.2. Vaux (1992) Vaux's analysis also assumes two steps in the change sT > TTh: We know that all [sic!] sibilants are debuccalized in Middle Indie, merging with /h/; I propose that the sibilants have already begun this process by the time gemination develops, so that the gemination after sibilants is actually the same process we see after /h/ (Vaux 1992: 287). Since /h/ lacks all supralaryngeal features, it copies the supralaryngeal features from the consonant. The resulting pre-aspirated geminate ATT ends up as the post-aspirated geminate TTh according to Bartholomae's Law.

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes · 57 The problem for this analysis is quite evident: there are no clusters like -hTin ΟΙΑ, but only 'clusters' like -ΑΓ-, i.e. visarga + plosive, which always include a morphological boundary. Pischel (1981[1900]: §329) and Schmidt (1924:16) cite examples of'clusters' with intermediate morphological boundaries (which differ from clusters without them): (7) Skt. antahkarana- Internal organ' > P. antakkaranaSkt. antahp ta- 'post fixed in the middle of the place of sacrifice' > P. antapp a• 2.3. Conclusion about Syllable-based Explanations for the Middle Indie Assimilation Murray (1982) and Vaux (1992) represent ambitious attempts towards generalizations for explaining the various ways of cluster reduction from ΟΙΑ to MIA. Nevertheless, both analyses fail to explain the exceptions to total assimilation, which already Geiger had to admit in 1916. The question why sibilants end up as aspiration on the assimilating consonant remains open, as well as, for example, thetrivial fact that the palatal glide palatalizes dental plosives as in the following example (Murray 1982:180): (8) *sat.ya > *sat.tya > *sat.cya > *sac.cya > sacca 'true5 In fact, all dental plosives are palatalized by a following palatal glide, no matter whether they are aspirated or voiced or not: (9) Skt. kathya 'steady' > P. kaccha M. and S. nevaccha Ornament', incorrectly Sanskritized to nevathya2 Skt. adya 'today' > MIA. ajja Skt. dhy na 'meditation' > MIA. jh na The exceptional status of this development becomes evident when looking at other palatal-glide clusters: (10) Skt. vikhy ta- 'famous' > JM. vikkh yaSkt labhya- 'to take' > P. labbhaSkt. Gun dhya 'rich in virtues' (a name) > JM. Gunaddha These examples show that the fusion of a palatal glide with a preceding plosive is restricted to dental plosives. Since no syllable-based explanation for this restriction is in sight, we must look elsewhere for explanation. One possibility is offered by process phonologies which have established that palatalization of dentals must be distinguished from palatalization of other consonants (cf. Dressier 1984b: 47-48; Lahiri and Evers 1991). One type of process phonology is Natural Phonology.

58 · Mario Palaschke and Wolfgang U. Dressler

• 3. Natural Phonology Natural Phonology, not to be confused with superficially similar 'concrete trends' in process phonologies (cf. Brόck et al. 1974), has been founded by Stampe (1969,1980; Donegan and Stampe 1979) as a pioneering endeavor to model what is natural (or unmarked) in one and only one component of language. Later on, the investigation of linguistic naturalness has been extended to morphology, syntax and textlinguistics (cf. Hurch and Rhodes 1996; Dziubalska-Kolaczyk 1996: 225-332), with a metatheoretical basis in semiotics (cf. Dressier 1984a, b). According to Natural Phonology, both phonemes and their phonetic realizations, within any language, are the outputs of phonological processes, which are the remnants of universal natural processes inhibited in language acquisition. Processes 1) merge conceivable human sounds into the phoneme inventory of each language (paradigmatic prelexical processes), 2) others govern the phonotactics of phonemes (syntagmatic prelexical processes), 3) still others derive phonetic outputs from phonemes. Within the third type, prototypically assimilatory backgrounding (lenition, weakening) processes are opposed to prototypically dissimilatory foregrounding (fortition, strengthening) processes. The same process, e.g. devoicing of obstruents, may have just one or more of these three basic functions of phonological processes. Thus, small children have been observed to devoice obstruents in general. Children acquiring, for example, Ojibwa, Delaware, Tonkawa, Wiyot, Yukagir as their first language, can maintain this universal natural process,'because these languages disallow voiced obstruents. Thus the universal process of obstruent devoicing explains both the implication that the presence of voiced obstruents in the phonemic inventory implies the co-presence of unvoiced obstruents in the same inventory (cf. Maddieson 1984:27) and that small children apply context-free devoicing to obstruents. However, children acquiring, for example, Russian, Polish have to restrict this devoicing to word-final position, children acquiring Northern Standard German to syllable-final position—and in Sanskrit, final devoicing is restricted to pausa position. Children acquiring English have to learn to suppress this process completely, but it may re-emerge in English aphasics, i.e. aphasia may impair such language-specific inhibitions. If native speakers of a language (e.g. Yoruba) that has no final obstruents learn a second language that has final obstruents (e.g. French), then they often apply final devoicing to final voiced obstruents. This is another evidence for the universality and naturalness of this process, which has not been suppressed by children when acquiring their native language Yoruba, because they had no occasion to do it. Diachronie change is reanalyzed as change in inhibitions of universal natural processes. Thus if diachronic change introduces obstruent devoicing, then this is due to failure in suppression of this process in language acquisition.

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes · 59

Natural phonological processes may be antagonistic. Thus, final obstruent devoicing may co-occur with intervocalic obstruent voicing within productions of the same child and within the same language (as in Sanskrit sandhi). Phonological naturalness has either phonetic bases, as in the case of obstruent devoicing and obstruent voicing (i.e. other things being equal, it is easier to produce unvoiced than voiced obstruents, but it is easier to produce voiced obstruents in intervocalic position) or cognitive, psychological or sociopragmatic bases. For example, rhythmic organization of sequential activity is not only a property of speech production and perception, but also of music and of any motor activity. Natural Phonology has been a pioneer in systematically using 'external' or substantive evidence (from phonetics, first and second language acquisition, language impairments, sociolinguistic and phonostylistic variation, poetic use of language, diachrony, etc.) on par with internal evidence from the investigation of corpora (including those elicited in field-work). Therefore, diachronic change is of great concern to Natural Phonology. Universal natural processes as well as other phonological tendencies are understood as universal preferences based on extralinguistic factors. In this way, natural processes are first of all universal preferences with universal preferential hierarchies of application: for example, obstruent devoicing occurs preferentially in final position, and even there with hierarchical grading: prepausal < word-final < syllable-final. Diachronic change basically consists in changes according to such hierarchies, e.g. in regard to obstruent devoicing: context-free to context-sensitive (or vice versa) or pre-pausal to word-final (extension) or vice-versa (reduction) or suppression of the process altogether, i.e. loss of obstruent devoicing (cf. Luschόtzky 1997). This universal level must be distinguished from the typological level which refers to how universal preferences favor or disfavor one another, and how preferred and dispreferred options may combine into phonological types, and how such phonological types are related to types in morphology and syntax. Typological properties may also be areally restricted and constitute typological areas (cf. Ramat 1988). Appurtenance of a language to a language type or to a typological area (such as the area of the Indian subcontinent) is gradual, in the sense of Wittgensteinian family resemblances. The third level of naturalness theory is language-specific system adequacy. On this level, Natural Phonology studies principles of basis of articulation, constraints of phonemic space or what has been studied in Generative Phonology under the headings of maintenance of underlying contrasts via counterfeeding and of rule conspiracies (i.e. several phonological processes have coherent results). Another area of system adequacy is represented by the interplay of prosodic and segmental phonology, e.g. in compensatory lengthening processes, which help to maintain syllabic, or moraic quantity relations.

60 · Mario Palaschke and Wolfgang U. Dressler

M 4. Aspiration and s m 4.1. Sibilants and Aspiration in Ancient Indian Phonetics The close similarity between sibilants and aspiration was already recognized by ancient Indian grammarians and phoneticians, as summarized by Allen (1953: 26): The general term for the fricatives is üsman, literally 'hot, steaming', perhaps because of their resemblance to the hiss as escaping steam: it is glossed by Uvata as vayu, 'wind'.... The term is applied not only to the letters J§s, but also to - -x -h and h,... and the breathy release of the aspirated stops (sosman). There is no special term corresponding to 'sibilant'.... Furthermore, we assume with Allen (1953:35), also in light of the modern Indo-Aryan languages, that both the aspiration of the voiced aspirates/gh, jh, dh .../ and Sanskrit has been voiced [fi]. Visarga, on the other hand, seems to be voiceless, which is shown by the complementary distribution of /h/ in sandhi-rules. In voiced contexts /s/ undergoes rhotacism, whereas in voiceless contexts it becomes visarga: (11) rsibhis idyas -> rsibhir idyah agnis hotδ -» agnir hotδ paribhόs asi -»paribhur asi • 4.2, s -> h as a Natural Process Within the framework of Natural Phonology, the process s -> h has been described as a natural process of lenition/backgrounding/weakening (e.g. in Hurch [1988], who cites examples from various areas such as second language acquisition and diachrony). Phonetic investigations by Mindez Dosuna (1984, 1985) have shown that the distribution of this process in consonant clusters is influenced by articulation: 'The duration of/s/ in /sKAclusters varies in accordance with the degree of homorganity between the sibilant and the stop' (Mendez Dosuna 1984:152). His measurements showed that the dental sibilant is always articulated longer before a dental plosive than before a nonhomorganic plosive. From this he concludes that in historical change the natural processs ->A, if applied in consonant clusters, should start in /sp/-clusters, proceed to /sk/-clusters, and should involve /st/-clusters at the end. In sandhi, according to this hierarchy, this process has been generalized to labials and velars, but not to dentals: (12) janah pibδti 'the man drinks' janah karoti 'the man does' janas taviti 'the man is strong'

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes · 61

In MIA and Modern Indo-Aryan languages, we find evidence that /st/-clusters are retained longer than other possible sibilant-plosive clusters: (a) In the Western ASoka-inscriptions of Girnδr /st/ is often retained: asti, hastiy etc. (Hinόber 1986:113). Graphic variants make us assume that these examples are occurrences of historical orthography, since the language of Girnar with its conservative character tends to maintain archaisms^Nevertheless the fact that the inscriptions show /st/, but not /sp/ and /sk/ is taken by Hinόber (ibid.) as strong evidence for the process -st- > -tth- to be the latest, particularly since Pali also retains or restores the cluster /st/. (b) In the north-western languages both homorganic clusters /st(h)/ and / st(h)/ remain unchanged up to the MIA languages: (13) AS. (Shδhbδzgarhi, Mδnsehrδ): asti—AS. (Kalsi, Jaugada): athi Skt. hasta- > G. (Niya): hasta, PS. hast Skt. astau > G. (Niya): /*astha/, PS. asta (Hinόber 1986: §231 with further references) This also seems to be plausible in the light of the fact that in Gδndhδri—in contrast to other MIA languages—the three sibilants are not merged to one sibilant, but remain three different phonemes. Sibilants in general appear to be better preserved in this area. M6ndez Dosuna (1987) further shows that aspiration of s is not to be understood as being triggered by syllable structure conditions. But once the process s -> h appears in a language, it can be extended from its original context to other contexts. In this light, we may assume that aspiration of s extended from sandhi-contexts to word-internal contexts, i.e. hierarchically from a more probable to a less probable position. • 4.3. -TS- > -cch-

All plosive-sibilant clusters—with the Eastern dialectal exception of ks > kkh—develop into an aspirated palatal geminate in MIA. In this case, not the entire oral gesture of the sibilant is lost, i.e. the oral gesture of friction is preserved in affrication. This is why in these cases we find another exception to total assimilation, namely reciprocal, partial assimilation: (14) a. ks > Western cch, Eastern kkh3 Skt. ksudraka- 'small' > AS. (Girnδr) chudaka-, AS. (Kalsi) khudakab. ts > cch Skt. matsara- 'envy' > MIA maccharac. ps > cch Skt. jugupsati 'takes care' > P. jugucchediFor this kind of sound change we find evidence, in terms of prakritisms, in the oldest documents of Old Indian literature, from the Rigvedic hymns onwards. Such forms cited in Wackernagel (1896: §135) are:

62 · Mario Palaschke and Wolfgang U. Dressler

(15) a. cch < ps krcchra-'painful' < *krpsraguccha 'bush' < *grpsab. cch < ks rcchar -'ankle' < rksal chava-, paricchava- 'sneezing9: ksava-, pariksava-

c. cch < ts

ucchanna-'destroyed' < utsannaucchalati 'flies upwards' < *ut-salati ucch dana- 'rubbing the body with oil': uts danamaccha- 'fish': matsya-

Wackernagel (1896) provides many examples of the type of 13a-c, but we could not find any example of a prakritism which would have showed the NilA result of the change ΟΙΑ -ST- > ΜΙΑ -77Λ-. Thus not only the results of the cluster-reduction from OLA to MIA are different according to the sequencing within the cluster, but they also seem to have happened at different times.

• 5. Intermediate Steps • 5.1. Pie-aspiration In his discussion of-Sr-clusters, Murray (1982:181) suggests the intermediate step 'aspiration and metathesis9, but he is not explicit about how these two processes should be interrelated. When Vaux (1992:288) arrives at a 'virtual' intermediate step of -ACC-, he 'simply invoke(s) the mechanism known as Bartholomae's Law4, which moves the laryngeal feature specifications of a consonant cluster to it's final segment...'. Since Bartholomae's Law does not only apply to geminates, but with consonant clusters containing an aspirated element in general, we assume that it would be better to invoke a mechanism like Bartholomae's Law or a restricted version of this law (provided that geminates are to be treated as consonant clusters). In addition, Bartholomae's Law does not mention pre-aspiration, due to the simple fact that there is no pre-aspiration in ΟΙΑ. Thus we should assume a paradigmatic prelexical process, within the framework of Natural Phonology (see Lusch tzky 1988:8Φ85), which prevents any kind of pre-aspiration in terms of system adequacy. In addition it fits the universal scale of naturalness presented inter alia in Hurch (1988:61-63):

(16) Ο > *C 'Postaspirated consonants are more likely to be expected than preaspirated ones.'5 Hurch (ibid.) cites examples from the Spanish dialect of Sevilla, which shows a similar development to MIA. In this dialect, the natural process s ->A is active.

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes · 63

Spanish ST-clusters appear as aspirated plosives as in estos casos 'these cases' ['ethokhaso]. Hurch assumes that a process of metathesis converts preaspiration to post-aspiration. In this context, it is important to emphasize that this process is paradigmatic, i.e. a process restricting the aspirated consonants of a language to post-aspirated consonants. Thus as soon as the sibilant is debuccalized, this process automatically moves the aspiration to the plosive release of the following consonant. This is a matter of system adequacy. In this light we agree with Grammont (1933:188): [Skt. asli] developed into *athti in Pali, but since Pali, such as Sanskrit, does not accept any aspirated plosive before another plosive, the aspiration is immediately moved to the second plosive, which gives atthi\.... (our translation and emphasis). This does not justify an intermediate diachronic step of pre-aspiration as assumed in Vaux (1992). • 5.2. Affrication In -rS-clusters the development appears to be more complex. Murray (see 2.1) just provides an explanation for the example aksi- 'eye' > akkhi-, but none for the development -TS- > -cch-. From the fact that these clusters change their place of articulation to a palatal, we conclude that these clusters were aspirated and affricated simultaneously. First, we assume an application of the natural process of affrication of palatals as in Stampe (1973: 52): (17) [+ sonor, - continuous, + palatal] -»[+ affricate], (antagonistic to the universal processes suppressing affrication: Luschόtzky 1992:143). This process covers the well-known tendency for palatal stops to change into affricates (cf. Ladefoged 1982 [1975]:148).6 The combination of the two natural processes of affrication and aspiration cannot, however, explain the changes /ks, ts, ps/ > /cch/, because there is no universal natural process which would change /ks, ts, ps/ into palatals. An explanation, in our opinion, must be sought in terms of phonological system adequacy. The historical preludes for our particular problem area of system adequacy are: (a) The Indo-Aryan languages maintained and expanded post-aspirated obstruent phonemes. (b) The genesis of the palatal affricate phonemes /c/ and /')/ through palatalization of /k/ and /g/ before Proto-Indo-European /e/ and later phonemization (and morphologization of the natural process of palatalization). These changes left two open slots within the phoneme inventory of aspirated/unaspirated voiced/voiceless obstruents (and nasals): aspirated palatal affricates.

64 · Mario Palaschke and Wolf gang U. Dressler (c) There are very few ΟΙΑ words with ch which have a direct Proto-IndoEuropean origin: one important source of ΟΙΑ ch (intervocalically -ccA-) is the PIE cluster sk' (consisting of sibilant plus palatalovelar plosive), as the following comparison demonstrates: (18) OLA. ch y - 'shadow' - Gr. σκτύί ΟΙΑ. prcchati 'asks': Gr. (έπτ) doKoLat. posc , OHG forsc n Another source is the following sandhi-rule merging a dental plosive and a following palatal sibilant: (19) pad-'foot' + suffix -Sas > pacchas 'foot by foot' Finally, as a very frequent origin of ch in ΟΙΑ Wackernagel (1986: §135) mentions prakritisms such as krcchra- 'painful' < *krpsra-. This looks like a conspiracy which has filled the open slot of ch.1 (d) The open slot of jh was Filled by onomatopoetic words such asjalajjhala 'the sound of falling drops', loan words such asjh vu- "Ikmarix indica', and finally—as in the case of ch—by prakriticisms such as Ved.jaj(h)jhatir 'splashing waters': jaksati 'laughs' (Dressier 1969:169), which represents a dialectal variation (cf. Hin ber 1986: §236). This conspiracy (generative concept) in filling this open slot (structuralist concept) clearly belongs to the area of system adequacy in assuring optimal efficiency in combining distinctive features within the phonemic inventory. (e) It belongs to system adequacy of ΟΙΑ and MIA languages that, of all conceivable (i.e. universally pronounceable) affricates, they allow only the universally most favored one, i.e. the series of palatal affricates. In other words, a paradigmatic, prelexical process merges all affricates into palatal affricates (cf. Lusch tzky 1992 chapter 4): (20) affricates -» palatal Second, we assume the natural process of aspiration (see 4 and 5.3 c, d). • 5.3. System Adequacy With these prerequisites, we can offer a more satisfactory explanation of the genesis of MIA cch. (a) The palatal place of articulation of cch in the results of /ks, ts, ps/, as in 14 is due to the system adequacy of palatal affricates only (see 5.2 [e]), guaranteed by a prelexical paradigmatic process. (b) The phonemic and allophonic space (another concept of system adequacy) of ΟΙΑ sibilants was greatly reduced in MIA languages. As mentioned above, most MIA languages merged the three ΟΙΑ sibilants to just one (most languages to s, M gadhi and Ardham gadhi to/). The natural phonological process of aspiration already applied to sibilants in ΟΙΑ sandhi and was extended to other contexts in MIA.

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes · 65

In the north-western languages, represented in MIA by Gδndhδri, all three sibilants were retained: (21) Skt. sapta- 'seven' > G. (Niya) sataSkt. Sata- 'hundred' > G. (Niya) Sada, (GDhp.) Sata Skt. dosa- 'fault' > G. (Niya, GDhp.) dosa As already shown in 13, in the north-western languages more sibilant-plosive clusters resisted cluster reduction. The reduction of the distinctive-phonemic role of sibilants and of their phonotactic role goes hand in hand, another case of system adequacy. (c) In the outcomes of /ks, ts, ps/, aspiration was preserved, presumably for maintaining lexical contrast to words with c, as the following minimal pairs demonstrate: (22) a. kacca- < kδca- via quantity metathesis kaccha- < kaksab. chitta- < ksetra- 'field' citta- < citra- 'strange' This may explain why /s/ in these clusters both contributes to the creation of affricates and is still preserved by the feature of aspiration, i.e. it gets lost as a segment, due to fusion with the preceding segment in the creation of affricates, but maintains its distinctive force. (d) The non-existence of pre-aspirate consonants (replaced by postaspirate ones) is another aspect of system adequacy (cf. 5.1).

• 6. Conclusion We hope to have shown that no unique principle can explain Middle Indie aspirate formation. Specifically, we have insisted on the inadequacies of syllabic explanations. In general, we may expect only rarely to be able to subsume several diachronic changes, which occurred during the same time-span, under one general principle, for otherwise we should be able to predict future changes. In our case, the dialect differences mentioned in 14a already prove that competing sound changes are possible. In this sense, no complete explanation for diachronic change is ever available, but only the (difficult!) definition of possible and of probable changes and the identification of scenarios of probable changes (cf. Dressier 1997). Our explanation has followed the three subtheories of Natural Linguistics. Most important are, on the one hand, universally available natural phonological processes, which have the status of universal preferences (partially comparable with the 'preference laws' of the competing syllabic analyses). The typology of phonological processes (cf. Dressier 1984a, b; Back 1991; Luschόtzky 1997) helps to distinguish probable, possible and unexpected changes. On the

66 · Mario Palaschke and Wolfgang U. Dressler other hand, there are the either hard or soft constraints of language-specific system adequacy which either restrict sound change (hard constraints) or increase/decrease the probability of changes (soft constraints), i.e. system adequacy filters the universal possibilities of change. What we have not done yet, is to study the intermediate filter of typological adequacy, i.e. the influence of prosodic phonological types on segmental phonological types8 and the typological area of the Indian subcontinent. • APPENDIX List of Abbreviations Ai G. G.Dhp Gr. JM. Lat M. MIA £. OHG P.

Asoka Gδndhδri Gδndhδn-Dharmapada Greek Jaina-Mδhδrδstri Latin Mδhδrδstri Middle IndoAryan Sauraseni Old High German Old Indo-Aryan Pali

PS. Skt.

Pasai Sanskrit

• NOTES 1. In order to include gemination, Murray reformulates the principle as follows: Tor a sequence Q$Qn within a simplex (uncompounded) word, the preferred syllable structure is such that m £ r'. 2. Appearing in an Indie play (see Hinόber 1986: §247 with further references). 3. For further, detailed discussion see Hinόber (1986: §232-36). 4. This law, established by Christian Bartholomae in 1882 for Vedic, e.g. badh- + -to ->· baddha 'bound', probably goes back to Indo-European (see Mayrhofer 1986:115). 5. In this notation the sign' > ' means A is more natural than B'. It does not express A changes to B'. 6. For the distinctive-feature identification of affricates, cf. Dasgupta (1972:120-21). 7. Many instances of the geminate cch in Prakrits and prakritisms are due to the morale lengthening of ch after short vowels, dealt with in terms of Natural Phonology by Back (1991:161-165). This is another instance of system adequacy, i.e. of the interplay between prosodic and segmental phonology. 8. For an audacious attempt in terms of Natural Phonology, which crucially involves the Indian subcontinent, see Donegan and Stampe (1983).

• REFERENCES Allen, W. Sidney. 1953. Phonetics in Ancient India. London: Oxford University Press. Back, Michael. 1991. Die natürliche Prozessbasis des natürlichen Lautwandels. Stuttgart: Steiner.

Middle Indie Aspirate Formation: Syllable Structure vs. Natural Processes · 67 Berger, Hermann. 1955. Zwei Probleme der mutelindischen Lautlehre. Mόnchen: Kitzinger. Brόck, Antony et al. (eds) 1974. Papers from the parasession on natural phonology. Chicago: Chicago Linguistic Society. Dasgupta, Probal. 1972. Coronality, old Indo-Aryan palatals, and natva. Indian Linguistics 33.99122. Donegan, Patricia and David Stampe. 1979. The study of natural phonology. Current approaches to phonological theory, ed. by Daniel Dinnsen, 126-73. Bloomington: Indiana University Press. . 1983. Rhythm and the holistic organization of language structure. Papers from the 19th regional meeting of the Chicago Linguistic Society, 337-53. Chicago: Chicago Linguistics Society. Dressier, Wolfgang U. 1969. Altindisch /jh/. Die Sprache 15.168-70. . 1984a. Explaining natural phonology. Phonology Yearbook 1. 29-51. . 19845. Morphonology. Ann Arbor: Karoma. . 1997. 'Scenario* as a concept for the functional explanation of language change. Language change and functional explanations, ed. by Jadranka Gvozdanovi, 109-42. Berlin: Mouton de Gruyter. Dziubalska-Kolaczyk, Katarzyna (ed.). 1996. Naturalness within and across components. Sprachtypologie und Universalien-Forschung 49/3. Emeneau, Murray B. 1988. The dialects of old Indo-Aryan. Sanskrit studies ofM.B. Emeneau. Selected Papers, ed. by B.A. van Nooten, 107-17. Berkeley: Center of South and Southeast Asian Studies, University of California. Geiger, Wilhelm. 1916. Pali Literatur und Sprache. Strassburg: Trόbner. Grammont, Maurice. 1933. Traite de phonetique. Paris: Delagrave. Hankamer, Jφrge and Judith Aissen. 1974. The sonority hierarchy. Papers from the Parasession on Natural Phonology, ed. by Anthony Brόck et al., 131-45. Chicago: Chicago Linguistic Society. Hinόber, Oskar von. 1986. Das ältere Mittelindisch im Oberblick. Wien: Verlag der Φsterreichischen Akademie der Wissenschaften. Hooper, Joan B. 1976. Introduction to natural generative phonology. New York: Academic Press. Hurch, Bernhard. 1988. άber Aspiration. Ein Kapitel aus der Natürlichen Phänologie. Tόbingen: Narr. Hurch, Bernhard and Richard Rhodes (eds) 1996. Natural phonology: The state of the art. Berlin: Mouton de Gruyter. Jocobi, Hermann. 1881. Das quantitδtsgesetz in den prδkritsprachen. Zeitschrift für vergleichende Sprachforschung auf dem Gebiet der indogermanischen Sprachen 25. 292-98. Kessler, Brett 1993. Syllables and Sandhi in classical Sanskrit. Paper delivered at the West Coast Conference on Formal Linguistics XII, Santa Cruz, CA. . 1995. Ambisyllabicity in the language of the Rigveda. Stanford, CA: Stanford University, ms. Ladefoged, Peter. 1982 [1975]./I course in phonetics. New York: Harcourt Brace Jovanovich. Lahiri, Aditi and Vincent Evers. 1991. Palatalization and coronality. The special status of coronals: Internal and external evidence, ed. by Carole Paradis and Jean-Francpis Prunet, 79-100. San Diego: Academic Press. Luschόtzky, Hans-Christian. 1988. Sixteen possible types of natural phonological processes. Wiener Linguistische Gazette 43. 79-103. . 1992. Zur Phonologic der Affrikaten. Frankfurt am Main: Hector. . 1997. Synchrone und diachrone Aspekte der phonologischen Prozesstypologie. Wien: Universitδtsverlag. Maddieson, Ian. 1984. Patterns of sounds. Cambridge: Cambridge University Press. Mayrhofer, Manfred. 1986. Lautlehre (Segmentale Phonologic des Indogermanischen). Indogermanische Grammatik, Vol. 1/2. Heidelberg: Winter. Mendez Dosuna, Julian. 1984. Homorganity and phonetic length: On some regular diffusion patterns in sound changes involving/sp/, /st/, /sk/ clusters. Wiener Linguistische Gazette, Supplement 3: Discussion Papers for the Fifth International Phonology Meeting, 152-56. . 1985. La duration de s en los grupos sp, st, sk. A proposito del orden regular de algunos cambios fondticos. Symbolae Ludovico Mitxelena septuagenario oblatae, ed. by J.L. Melena. Vol. 1,647-55. Vitoria Gasteiz: Institute de Ciencias de la Antigόedad, Universidad del pais \fesco.

68 · Mario Palaschke and Wolfgang U. Dressler Mlndez Dosuna, Julian. 1987. La aspiracion de 5 como proceso condicionado por el contacto de sflabas. Revista de la Sociedad Espanola de linguistica 17.15-35. Murray, Robert 1982. Consonant cluster developments in Pali. Folia Linguistica Historica 3.16384. Palaschke, Mario. 1997. -ST- > -TTh- im Mittelindoarischen oder akhkhatöcrtyä 'Silben bildend*. Wien: MA Thesis, University of Vienna. Pichel, Richard. 1981 [1900]. A grammar of the Prakrit languages. 2nd revised edition of the English translation by Subhadra Jha. Delhi: MotilaJ Banarsidass. Ramat, Paolo. 1988. Macroaree tipologiche.Xnmi/i dell'Istituto Orientale di Napoli 10.81-95. Schmidt, Richard. 1924. Elementarbuch der Sauraseni mit Vergleichung der Mähärästri und der Mägadhi. Hannover: Lafaire. Stampe, David. 1969. The acquisition of phonetic representation. Papers from the 5th regional meeting of the Chicago Linguistic Society. 443-54. . 1973. On Chapter Nine, issues in phonological theory, ed. by Michael Kenstowicz and Charles Kisseberth, 44-52. The Hague: Mouton. -. 1980. dissertation on natural phonology. New York: Garland. Tfedesco, Paul. 1956. Review of Berger 1955. Zwei Probleme der mittelindischen Lautlehre. Language 32.498-504. Varma, Siddheshwar. 1961 [1929]. Critical studies in the phonetic observations of Indian grammarians. Delhi: Munshi Ram Manohar Lai. Vaux, Bert 1992. Gemination and syllabic integrity in Sanskrit. Journal of Indo-European Studies 20.283-303. Vennemann, Theo. 1972. On the theory of syllabic phonology. Linguistische Berichte 18.1-18. . 1988. Preference laws for syllable structure and the explanation of sound change. Berlin: Mouton de Gruyter. Wackernagel, Jacob. 1896. Altindische Grammatik. Vol. 1: Introduction generate und Lautlehre. 2nd edition 1957. Gφttingen: Vandenhoeck and Ruprecht

B

Open Submissions

Specificity in the Bangla DP • TANMOY BHATTACHARYA · In this paper* I offer an analysis of the DP structure in Bangla (Bengali) based on specificity effects obtained within the DP. I propose that the Bangla DP has a three-layered structure, the layer intermediate between the DP and NP being a QP, based on the position of the Q/ Num + Classifier complex in the DP. The specifier of the QP acts as the landing site for specific NPs. This leftward movement, I suggest, is due to a [specificity] feature of the Q head. KINSHIP INVERSION is another instance of DP-internaJ NP movement which I claim to be driven by the same [specificity] feature of the Q head. These two types of movement out of an nP-sheU are instances of overt NP movement inside the DP in Bangla.

• 1. Introduction The paper is organized as follows. In this section, I suggest a three-layered structure of Bangla DP. In section 2,1 suggest that the XP intermediate between the DP and the NP is a QP. In sections 3 and 4,1 investigate the position of the Demonstrative (Dem) and the nature of the complex head Q, which, I argue, contains Q/Numerical (Num) and Classifiers (Cla). In section 5,1 briefly look at the position of Adjectives (Adj) in Bangla and suggest that they may be generated as NP-specifiers. In section 6, the main section of the paper, I examine the specificity effects obtained inside the Bangla DP. The last section provides the final argument in favor of equating clausal and phrasal structure through an investigation of the base position of the Possessive (Poss).

* I am indebted to Rita Manzini, Neil Smith, Michael Brody and ProbaJ Dasgupta for comments, criticism and suggestions on earlier drafts of the paper; all remaining errors are mine.

72 · Tanmoy Bhattacharya

• 1.1. The Layered DP Most of the research in the syntax of DPs has concerned the similarity between clausal and phrasal structure. A plausible hypothesis is that these approaches can be subsumed under a common structure like the following: (1)

Layer 1

The most influential approach is due to Abney (1987) who argues that Noun Phrases are headed by the functional category Determiner (D). D is known to be similar to the Infl in accommodating agreement features. In the following structure (2), for example, John gets Case in [Spec, DP] from the 's morpheme in D, similar to clausal subjects getting Case from Infl by specifier-head agreement. (2) John book

Szabolsci (1983) had earlier argued in favor of an Infl head in Noun Phrases in her study of Hungarian possessor constructions. She further argues (Szabolsci 1995) that Noun Phrases also contain a pre-determiner, COMP-like, A' position. Later research concentrated on the region between the DP and the NP (Ritter 1988; Valois 1991). These studies, more or less, proposed the following structure for DPs:

(3)

Specificity in the Bangla DP · 73

What is the X in 3? There have been several ideas: X = NumP (Ritter [1991] for Hebrew, Valois [1991] and Bernstein [1993] for French) X = QP (Giusti [1991] for Romanian, Löbel [1989] for German) X = KP (Sigurdsson [1993] for Icelandic; Tang [1990] for Chinese for whom it is a 'Klassifier' Phrase) X = ArtP (Santelmann [1993] for Swedish) X = BP (Dasgupta and Bhattacharya [1993], Bhattacharya [1995], Bhattacharya and Dasgupta [1996], for Bangla where B = 'Badge') In fact, there has been a general proliferation which means that there is more than one XP between DP and NP of functional projections within the DP structure—for all we know, there could be several XPs between DP and NP, or so the trend indicates. I would like to propose, against this trend, that perhaps the DP structure should be really seen as in 3 in line with the classical sentential structure. It is interesting to note, in this connection, that independent of the sentential structure, research in nominals in general (that is, irrespective of the framework) has tended to report such tripartite partitioning in nominal phrases.1 In this connection, let us look at Bangla. Notice that Num-Cla and Adj-N in 4 behave like independent units; the word order is relatively free. To keep the account easy to follow, we will assume, at this point, that all the phrases below have similar truth conditions. (4) a. ei tin-Te Sobuj boi2 this 3-CLA green books 'these three green books' b. ei Sobuj boi tin-Te c. tin-Te Sobuj boi ei d.? tin-Te ei Sobuj boi e.* tin-ei-Ta Sobuj boi f.* tin-Te Sobuj ei boi

Dem Num-Cla Adj N Dem Adj N Num-Cla Num-Cla Adj N Dem Num-Cla Dem Adj N Num-Dem-Cla Adj N Num-Cla Adj Dem N

There are probably more (im)possible orders but this is enough to show that there is enough freedom of movement as long as Dem, Num-Cla, Adj N form three separate units. There is, therefore, reason to believe that Bangla Noun Phrases may perhaps fit into a tripartite structure of the Noun Phrase. A comparison between 3 and the classical clause structure (CP-IP-VP) would lead us to think of XP to be similar to IP in nature. The Infl being a functional element, it takes predicates as arguments, for example, it quantifies over predicates of events provided by the VP. It makes sense therefore to think of the XP to be predicative in nature. However, in proposing the classical DP structure, Abney (1987: 76) used similar arguments to equate D with Infl: The function of the Det is to specify the reference of the NP. The N provides a predicate; and the Det picks out a particular number of the predicate's

74 · Tanntoy Bhattacfanrya extension. The same function is performed in the verbal system by tense, or Inflection. In a clause, VP provides a predicate, i.e., a class of events and the T locates a particular event in time. In this system, D seems to be doing two things at the same time: fixing up the reference of the phrase as well as quantifying over the event variable (or its nominal equivalent) of the NP. I suggest that these two functions be separated. Such a division of labor will need to make reference to another functional position between D and N which is predicative in nature. Based on recent research on the UP positions of quantificational elements in a clause (Beghelli and Stowell 1997) and given the maximal identity between clausal and phrasal syntax, I claim that the highest functional position of a phrase is the position for referential elements. The quantificational/ predicative function of Abne/s D, I claim, is performed by X. One evidence in Support of this theory (Zamparelli 1996) actually instantiates this XP as a 'Predicative Phrase*. He proposes the following structure: (5)

SDP

(Referential) (Predicative) (Kind-denoting) (adjectives) NP-

In 5 SD is the STRONG Determiner head and PD is the WEAK Determiner head. SDP in this system is the only 'referential' part of the DP and is the locus of pronouns, demonstratives, proper names and strong determiners, as well as numerals and (in)definites in their strong/referential sense. PDF denotes the predicative part of the DP. It is the locus of weak determiners: indefinites and numerals in their nonspecific reading. It denotes a property which is predicated of the SD head. KDP is the kind-denoting part of the DP, containing the NP proper.3 Zamparelli's arguments for noun phrases containing predicative material is based on the evidence that they can be negated and modalized.4 All that 5 shows, for us, is that there is indeed predication involved between D and NP. Without going into further detail, we therefore note that the 'space' between DP and NP in a configuration like [DP... NP] is predicative in nature.

• 2. The Quantifier Phrase My proposal will be to suggest that this 'space' is uniquely occupied by a Quantifier Phrase (QP). I will present the specificity argument for QP in section 6

Specificity in the Bangla DP · 75

which will further verify the need for the intermediate XP to be QP. Löbel (1989) also proposes a similar structure for German. She observes that the relation between the Q and the N is that of'countability' or rather the function of the category Q is to ensure the countability of the NP. For a [+ Count] N, Q may be morphologically realized as tf plural suffix in English and German: (6) a. drei [Q0] Baum-e COUNT three tree-s b. drei [Q stück] Wild-0 MASS three head game 'three head of game' Now let us look at some more data from Bangla before we proceed further to propose any structure. (7) tomar your (8) tomar your (9) tomar your (10) tomar your

ei this ei this ei this ei this

notun new SOb all kOek-Ta some-CLA tin-Te three-CLA

SaRi sari notun new notun new notun new

SaRi sari SaRi sari SaRi sari

So a maximally occupied phrase may have the following order of constituents: (11) Poss D Num-Cla Adj NP Q-(Cla) If we accept that numericals or quantifiers are quantifying expressions, then their occurrence at the same position is not surprising. Moreover, these two different kinds of quantifying expressions can never co-occur: (12)* tomar ei kichu tin-Te notun SaRi your this some 3-CLA new sari Furthermore, if we assume that the possessive phrase tomar 'your' occupies the [Spec, DP] position at some stage of the derivation, then we have a structure like the figure on the following page. What is immediately visible about this structure is that a head-initial word order is presumed in 13. Bangla is typically an SOV language. How do we explain this? I assume with Kayne (1994) that the universal underlying word order is Specifier-Head-Complement (S-H-C). Languages which display a different word order on the surface must employ movement to arrive at some non-universal outcome. Any movement in Kayne's model is leftward by default since asymmetric C-command will imply precedence.

76 · Tanmoy Bhattacharya

(13)

I adopt the S-H-C order or the LCA based on the following: (14) Dasgupta (1996) has provided an account of Oblique Case in Hindi/ Urdu and the phenomena of agreement in the postpositional phrase which crucially relies on the conjecture that Kayne's assumptions hold. In that account, the complement of the adposition moves to the specifier of a P-related functional head, yielding argument-adposition order and the agreement details. Kayne (1994) reported that agreement between an adposition and its complement is observed only in case of postpositional languages. Hindi/Urdu, which is closely related to Bangla, shows a confirmation of this prediction. To be more precise, Kayne (1994) (based on p.c. with Hale) reports that there are postpositional languages which show agreement between the adposition (postposition) and its complement, while prepositional phrases never show such agreement. Mar cz (1989) (cited in Kayne) reports that P-DP order is possible in Hungarian only when the adposition is of the class that never shows agreement. If we compare this with Dasgupta's analysis of Hindi/ Urdu Oblique Case as a case of DP-P agreement then we get confirmation of Kayne's prediction. ºï see the issues more clearly, consider the following Hindi data: (15) a. [makaan house

ke]\ poos t| GEN near

MASC (OBL) MASC

Specificity in the Bangla DP m 77

b. [makaan kii]i or ti house GEN towards FEM (OBL) FEM

The facts of agreement (oblique) and Case choice are explained in terms of complement-to-specifier movement of the agreement nominal by feature checking needs. If we adopt an underlying SVO order for Bangla then we would expect the order D-NP within the DP. On the default assumption that the Dem is a D, DNP is in fact the surface order within a Bangla DP. With the S-H-C order, the phrase-initial position of the Dem in Bangla is explained.5 Having given some justification of the order adopted for 13, let us investigate the following properties in connection with 13 in turn: • • • • •

Dem as D° (Section 3) Num/Q -I- Cla as fused Q° heads (Section 4) Adj-Noun order in Bangla (Section 5) Specificity in the DP (Section 6) Base position of Poss (Section 7)

• 3. Dem and the Three-layered DP Consider the following evidence which puts into question the headedness of Dems: (16) a. ei du-To boi this two-CLA book 'these two books' b.* boii ei du-To ti c. ei boii du-To t\ (17) a. ei du-To lal boi this two-CLA red book 'these two red books' b.* [lalboi]i ei du-To ti c. ei [lal boi]i du-To ti The starred b phrases show that leftward movement of the NP across ei 'this' is barred. The crucial barrier here seems to be the Dem. If the Dem is a head, then it is difficult to see how it can act as a barrier to XP movement. If the Dem is not a head then it cannot occupy either D (our initial conclusion as in 13) or any other head we may now be forced to place between D and Q. Suppose there is one, does it make our task easier? Consider the structure for the DP in the following page. That is, I have generated the Dem as a specifier of the intermediate XP projection. The fact that the Dem may not be equated with D° is well established

78 -· Tanmoy Bhattacharya

(18)

(Giusti [1997]; Bernstein [1993]; Brugo [1996] etc). They argue that the definite article (at D°) and the Dem can co-occur in many languages: (19) a. el libro este/ese/aquel (Spanish) Brugi (19%) the book this/that/that 'this book' b. bäiat-ul acesta (frumos) (Rumanian) Giusti (1997) boy-the this nice 'this nice boy* For 19b, Giusti suggests that N to D movement ofbäiat 'boy9 takes place across the Dem as well as the Adj, if present. This shows that the Dem is neither at D, which has the article-u/, nor in an intermediate head, otherwise the Head Movement Constraint (HMC) would be violated for this local-N movement. Additionally, 19c shows that the Dem also blocks the AP movement to [Spec, DP] which is otherwise allowed in Rumanian. c. frumosul (*acesta) bäiat nice this boy The intermediate head position at whose Spec the Dem is located is needed as an escape hatch for the N to D movement to proceed. However, in Bangla, as I will show later (section 7), there is no N to D movement in general. There is no convincing evidence, therefore, to posit a head X° between D° and Q° although the Dem behaves like an XP. One alternative is to generate the Dem as an adjunct to QP. • 3.1. Dem as QP Adjunct As it will be argued in section 6, specific NPs move to [Spec, QP]. Both this movement and our decision to postulate an XP between DP and NP are

Specificity in the Bangla DP · 79

driven by specificity. It is therefore important to ask if there is any specificityDem interaction. Consider the following: (20) a. ei du-To boi (deictic) this two-CLA book 'these two books (here)' b. ei boii du-To tj (specific) 'these two books' In other words, in 20a the deictic meaning is more important (shown in the translation by here), whereas in 20b the specificity of the books is more important. We will look at the specificity effect obtained in 20b in greater detail later. The deictic effect in 20a is less easily stated in syntactic terms. Bernstein (1997) analyses the following contrast in terms of the idea that the deictic effect in 21 a reflects a Dem to D° that does not take place in the syntax for 21b. (21) a. this woman (right here) = this woman (deictic) b. this woman (from Paris) = a woman (indefinite specific) That is, for Bernstein (1997), deixis is obtained through a movement of the Dem to D°, whereas in case of the indefinite there is no movement of the Dem. Imagine, however, a system where the Dem is ambiguous, not between a deictic and an indefinite interpretation, but rather between a deictic and a definite interpretation. Applying Bernstein's system to such a language would involve viewing both deixis and definiteness (which most standard analyses relate to D) as due to some interaction with D°. In other words, Bernstein's system is unable to distinguish between the two different effects associated with Dem in terms of the two types of movement. Bernstein's account of deixis is unsatisfactory for another reason. She proposes movement of an XP (Dem) to an X (D°), a conceptually undesirable move for which she provides no specific motivation. I suggest that deixis is made possible by the maximal Dem merging at [Spec, QP]. This is then the mechanism responsible for the deictic effect in 20a, for example. Let us now see if there is any evidence for the existence of a head between D and Q. In Bernstein (1997) an FP is proposed where the functional head F is the 'Demonstrative reinforcer' like here/there in colloquial Scandinavian varieties and non-standard English: (22) a. den här mannen (Swedish) (Bernstein 1997) the here man-the 'this man' b. den dar bilen the here car-the 'that car'

80 · Tanmoy Bhattacharya

(23) a. den herre klokka (Norwegian) the here watch-the 'this watch' b. del derre huset the there house-the 'that house' (24) a. this here guy (English) b. this there car The Dem in her system is the specifier of this FP. Whether such a head is wellmotivated for Bangla is unclear.61 will, therefore, settle for the solution that Dems are generated as QP adjuncts, noting perhaps the difference between two specifiers as sites for two different syntactic effects. The revised structure, with Dem as adjunct to QP, is presented below:

(25)

What we have learned from this section is as follows: Dems behave as XPs rather than as heads and that since there is no independent motivation in Bangla to posit a head between D and Q, the only alternative is to generate the Dem as an adjunct to QP.

• 4· Q + Cla as a Fused Head In this section I will discuss the content of the Q head, as in 25, and claim that Num-Cla (and Q-Cla) in Bangla is part of the QP domain. I will show that both a Q + Cla and a Num + Cla quantifies a following nominal argument (N or Adj + N). A quantifier in Bangla followed by a clitidzed Ta appears to modify a noun. Consider the following cases of Q-Cla sequences: (26) a. kichu-Ta doi some-CLA curd

Specificity in the Bangla DP m 81

b. SOb-Ta doi all-CLA curd c. khanik-Ta doi some-CLA curd Let us now see how a Q + Cla sequence combines with verbs: (27) a. kichu-Ta dekhechi some-CLA seen-1 '(I) have seen some' b. SOb-Ta dekhechi all-CLA seen-1 '(I) have seen all' c. khanik-Ta dekhechi some-CLA seen-1 '(I) have seen some' This is similar to the set in 26 if it implies a zero N; 27a, for example, means / have seen some of it. In both these sets I take it that the Q quantifies over Ns. Consider also the following set of data where the Q-Cla sequence seem to quantify over an adjective: (28) a. dilli-r cee kichu-Ta bORo Delhi-Gen than somewhat-CLA big 'somewhat bigger than Delhi' b. ag-er cee khanik-Ta bhalo before-GEN than a little-CLA good 'a little better than before' c. ekhan theke Onek-Ta dur here from alot-CLA far 'quite far from here' The data above shows that a Num/Q-Qa sequence is followed either by N (26), zero N (27), Adj (28), or Adj-N (9, 10). In other words, the maximal sequence noted in 13 and revised in 25 for the Bangla DP can account for the data above. Consider the following additional data involving a Q-Cla sequence (the data can be replicated for other Qs and Numerals). (29) a. ami boi du-To dekhechi I book two-CLA seen-lP º have seen the two books' b. ï phOlkO-Ta kheyeche s/he fruit some-CLA eaten-has-3P 's/he has eaten some (specified) fruits' 29b may have a more clearly definite reading in contexts that invite it, a matter I do not investigate here. The specific reading shown in the gloss provided

82 · Tanmoy Bhattacharya

suffices for our purpose. Notice that 29 must be syntactically related to the following example where the Q precedes the nominal, and where the DP has an indefinite (and default non-specific) reading: (30) a. ami du-To boi dekhechi I two-CLA book seen-lP º have seen two books' b. o kO-Ta phOl kheyeche s/he some-CLA fruit eaten-has-3P 's/he has eaten some fruits' Comparing 29 and 30, we see that 29 gives the effect of a floating quantifier.7 In other words, the quantifier is stranded. ºï see this more clearly, recall a floating quantifier example from a better-studied language, French: (31) a. tons les enfants ont vu all the children have seen b. les enfants ont tons vu the children have all seen

ce film this movie ce film this movie

In 31b the quantifier lous appears dislocated from its position in 31a. These sentences are identical at some level of representation since the universal force of the quantifier tous is identical in both sentences, tons is, however, crucially a subject-oriented quantifier. The predominant view in the literature is that floated quantifiers mark the position of the subject traces. Bobaljik (199S: 131) argues that in object-shift languages, object-oriented floated quantifiers appear on the left edge of the VP. If NP is the phrasal equivalent of VP then we can assume that in the context of a Noun Phrase, the left edge of the VP is where the QP is in 25. Sportiche's (1988) analysis of the structure of this construction shows that there is no actual 'floating' of the quantifier involved. Instead, generalization 32 provides a better key to the distribution of floating Qs: (32) (Floating) Qs may appear in the NP-initial position (Sportiche, 1988: 427) This generalization, coupled with the VP-internal subject hypothesis proposed by Sportiche, leads to the following possibility. The subject originates within VP. When subject moves out to the [Spec, IP] position, it strands the Q, which thus remains at the left edge of the VP. We return now to 29 on the basis of this understanding of the phenomenon of Q stranding which I take 29 to exemplify. Our account of the relation between 30 and 29 is that the nominal moves out of its base-generated position to a higher position. Based on Sportiche's analysis and incorporating Bobaljik's (1995) claim that object-oriented FQs are possible in object-shift languages, I conclude that 29 involves phrasal equivalent of object-shift that leaves the Q stranded.

Specificity in the Bangla DP · 83

The movement of the NP across Q-Cla (or Num-Cla as a special case) in 29 is acceptable, which means it does not count as a violation of any locality constraint on movement, which I take to show that the Q-Cla must be a head. I will thus interpret the floating Q behavior of the Q-Cla sequence as evidence that this sequence as a whole counts as a Q head In the next section I will find additional evidence for treating Q-Cla as the head of QP.

• 5. Adjective-Noun in Bangla Before we proceed to look at specificity facts within the Bangla DP, let us briefly discuss the placement of the Adj in the Bangla DP. If we look at our canonical DP structure in 25 again, we notice that the adjective phrase is in the Spec of NP. The status of attributive adjectives has been controversial. I will adopt an NP-shell structure with the adjective as the specifier of NP. We will see (in 35) that an Adj-N unit may not be broken in Bangla. One way of dealing with such a descriptive fact would be to propose that Adj and N in Bangla form a constituent AP as a Spec of NP achieves this effect. If adjectives were to head an AP with an NP complement, then Object'-shift (suggested for 29 and discussed in detail later) would require whole APs to move leftward to [Spec, QP] in our structure; there is no strong motivation for such a movement, for example, adjectives are not ± definite/specific in Bangla. Bangla does not (while the closely related language Hindi/Urdu does) show any agreement of the Adj and the head N. The attachment site of the A (or AP) in Bangla, therefore, should not affect the analysis of DPs that I am proposing here. However, as I have noted earlier in 4, Adj-N nevertheless forms a close unit in terms of distribution in the clause. Keeping the possibility of adjectival agreement in Hindi/Urdu in mind, a structure which proposes the placement of APs on the specifiers of NPs seems optimal. Notice that in such a structure, it is possible to check agreement features in a Spec-head configuration to account for the Hindi/Urdu facts.8 Extending the move made by Chomsky (1995) for the verbal projection, I am postulating nP as the (maximal) projection of the N system, in other words, as the outermost nP-shell. Notice that enlarging the nP-shell in our canonical DP structure in 25 thus would still maintain the three-layered structure that I intended to preserve as far as possible. Conceptually motivated along the lines of Chomsky's enlargement of the VP projection, this nP proposal receives initial empirical support from Adj-N agreement in Hindi. Additional evidence for the nP-shell will be presented in section 7 where I discuss the base position of the Poss. Consider now a base-generated Dem Q Adj-N sequence in Bangla: (33) ei duTo lal boi this two-CLA red book 'these two red books'

84 · Tanmoy Bhattacharya

The preferred order is Dem QP Adj-N as in 33. Notice that the NumeralClassifier sequence may occur right after the Dem as in 33 or postnominally as in 34, but never in an intermediate position as in 35. (34) ei lal boi duTo Dem Adj-N QP this red book two-CLA 'these two red books' (35)* ei lal duTo boi *Dem Adj QP N this red two-CLA book With regard to 35,1 noted earlier (in 4) that the Adj-Noun unit may not be broken. This descriptive fact falls out of the DP structure that I have proposed in 25 above since AP-N forms a constituent NP in 25. Bangla also disallows any leftward movement of the adjective in the overt syntax. In other words, extraction of the specifier of NP in general is not allowed. In order to understand how 34 is derived, let us look at the structure for 33.1 propose that 33 is derived from 36, ignoring the fine-grained nP-shell structure for the the moment.

(36)

The derivation of 34 from 36 proceeds by moving the whole phrase NP to [Spec, QP] to yield the following:

(37)

boi

Specificity in the Bangla DP · 85

Analysis 37 gives rise to an observation and a query: (38) (i) NP movement leaves the Q stranded as in cases of Q-float (ii) What drives the leftward NP-movement? Regarding 38(i) we note that it provides additional justification for treating a Q-Cla structure as a complex Q head. If Q and Cla were two different heads, we would need additional head-to-head movement of the Q to Cla, and two leftward movements of the NP (first to [Spec, QP] and then to [Spec, ClaP]) to derive the right order for 34. A fused Q head, therefore, not only maintains the three-layered DP structure but supports a more economical derivation of specific DPs. With regard to 38(ii), I propose 39 for which section 6 provides independent language-specific evidence. (39) A presuppositional/specific feature of the Q head drives leftward movement.

• 6. Specificity in Bangla Diesing (1992) equates specificity with presupposition. For most authors, specificity essentially presumes an identified discourse referent. Mahajan (1990) proposes to treat nominal specificity in Hindi/Urdu as a syntactic property. This move, as Kidwai (1995) shows, does not work since it is not the case that all DPs that trigger verb agreement or are Case-marked are necessarily and unambiguously definite/specific in Hindi/Urdu. Since Bangla does not show number/gender agreement, there is in any case no obvious way of implementing Mahajan's proposals for Bangla. At least for Bangla, then, we accept by default the idea that Case and agreement facts alone cannot be used to decide whether a particular nominal is specific. We may note that the core of the specificity effect (movement of the object NP out of the VP) that Mahajan discusses can be imported into the type of theory that Diesing discusses, which, in essence, forces all presuppositional material out of the VP in LF. Note that in Bangla sentences like in 40 below, the'moved nominal induces a presuppositional/specific reading in 40a. In other words, 40a presupposes a prior discourse referent for books. It seems unnecessary to appeal to any syntactic defmiteness feature in the Bangla DP (located, say, at D) to explain the fact that the DP in 40a is specific. We may also note here that according to Groenendijk and Stokhof (1981) specificity ranges over not only indefinites but definites, numerical expressions, singulars and plurals. What matters in 40a is, therefore, specificity, the in the English gloss notwithstanding: (40) a. ami boi du-To dekhechi I book two-CLA seen-lP º have seen the two books'

86 · Tanmoy Bhattacharya

b. ami du-To boi dekhechi I two-CLA book seen-lP º have seen two books' I shall then presume that the phenomenon connected with the leftward movement of the NP in Bangla DP is specificity rather than definiteness. I adopt Diesing's suggestion that specificity can be equated with presuppositionality. Her theory represents, at LF, the syntactic reflex of the semantic/pragmatic notion of presupposition. Indefinite subjects in Dutch, and objects in Turkish allow a specific interpretation of indefinite NPs in certain syntactic contexts. Diesing notes that this can be explained by considering the essential semantic content of specificity to be presuppositionality. This proposal carries over directly to the following Bangla examples. (41) a. kal Ek-Ta chele eSechilo (non-presuppositional) yesterday one-CLA boy came 'a boy came yesterday* b. kal chele-Ta eSechilo (presuppositional) 'the boy came yesterday* The nominal in example 41b presupposes a prior discourse mention of its reference. It shows that the shifted nominal chele Ipoy* can receive a specific reading for the Noun Phrase. The presuppositional nature of the nominal in 41b is dear if we consider the fact that 41b can only be an answer to a which NP question 42b but not a what question 42a. (42) a. kal ke eSechilol yesterday who came *who came yesterday?' b. kal kon chele-Ta eSechilo? yesterday which boy-CLA came ^hich boy came yesterday?' Presupposing the existence of some boy or boys, the question in 42b targets the identity of the boy involved. The data in 41 and 42 constitute the evidence for a presuppositional analysis of DP-internal specificity in Bangla. I now turn to the task of working out, within the syntax of the Bangla DP, the mechanism responsible for the NP preposing that gives rise to such specific interpretations. • 6.1. Leftward NP-movement in Bangla Let us now look at the distinction between the following. (43) a. oi duTo lal boi those two-CLA red book 'those two red books'

Specificity in the Bangla DP · 87

b. oi lal boi duTo (specific) those red book two-CLA 'those two red books' Note, that the specificity obtained is not clear from the English gloss in 43. However, in the case of 43b where the nominal boi has moved out of its base position, the phrase is felicitous only if the nominal has a prior discourse reference. I extend Diesing's analysis here to NPs and suggest that a specific NP moves out of its immediate nP-shell to a higher position. Notice that 43b indicates a specific reading of the NP lal boi 'red books'. The N is specific or presuppositional in 43b and therefore it must move up.9 This leftward movement of the NP as shown in 37 is repeated here:

(44) Spec

boi

An important question to ask at this point is: What drives this movement? Since movement in the framework adopted for this purpose (Chomsky 1995) is feature-driven, the default option would be to formulate a mechanism for the movement observed in 44 in terms of a feature. Let us return to 39, where I suggested a feature of specificity (or presuppositionality). Let us now assume that a filled Q comes with an optional feature of specificity in the numeration. Certain nonsubstantive heads (like C, T or D in English) can be assigned a particular feature when they are chosen for the numeration. I propose that this option is exercized by the Q in a specific DP in Bangla. In Chomsky (1995), it is the feature of the ATTRACTOR (in this case Q) that forces movement. I propose that specificity is a non-interpretable feature and therefore, according to Chomsky (1995), must be checked either in overt syntax or at LF. To ensure that the checking occurs in overt syntax, I propose further that this optional feature picked up by a Q as it enters the numeration is STRONG. I constrain the proposal by adding that only a Q head that contains a Classifier element has the option of picking up this strong, non-interpretable feature [specific].

88 · Tanmoy Bhattocharya

In support of this aspect of the proposal, consider the following evidence. So far we have observed numeral expressions in Bangla carrying a Classifier Ta. There are, however, certain classifierless Num-N sequences in the language. The following examples are from Dasgupta (1983): (45) a. du deS-er moitri two country-GEN friendship 'friendship between two countries' b. tin caka-r gaRi three wheel-GEN vehicle 'three-wheeled vehicle' c. tin bOchor three year 'three years' d. car paS four sides Tour sides' e. tin dik three direction 'three directions' f. ora car ban tin bhai they four sister three brother 'they are four sisters, three brothers' A similar classifierless Num-N sequence appears in the following measure expressions:

(46) a. du gOj two yard b. tin hat three hand/arm 'three cubits' c. car miTar four meter d. paMc peala ca five cup tea (Dasgupta 1983:12-13) Note that, crucially, in none of these phrases can the NP move leftward to give a specific reading. The following, therefore, are not possible: (47) a.* deS diaer moitri 'friendship between the two countries' b.* caka tiner gaRi 'the three-wheeled car' c.* bOchor tin 'the three years'10

Specificity in the Bangla DP m 89

d.* paS car side four 'the four sides' e.* diktin direction three 'the three directions' f.* bon car bhai tin 'the four sisters, three brothers'11 g.* gOj dui 'the two yards'12 h.* hat tin arm three 'the three cubits' i.* miTar car metre four 'the four meters' j.* ca paMc peala 'the five cups of tea'13

A typical structure for the grammatical examples will be as follows:

car 'four'

I take this data to mean that the Q head in these cases never exercises the option of picking up a specificity feature when it enters the numeration. We notice two things about these examples: that Q lacks a classifier element, and that the NP cannot prepose across the Q. These two facts—the absence of specificity and the absence of leftward NP movement—must be correlated. The account that I propose, based on 39, explains this correlation by giving the Qhead the option of choosing a non-interpretable formal feature of specificity. Now consider the nature of this feature. Since I presume this feature to be -Interpretable, it must be checked in a Spec-Head configuration either in the covert or the overt component. In the immediately preceding discussion, I have shown the need to endow the Q head of QP with a specificity feature as a lexical option made available if the Cla features are present. Does this mean

90 · Ttmrnoy Bhattacharya

that the post-Q NP must move leftward whenever Qa occurs? No, for in the çïç-spe fic DP examples 40b, 41a and 43a, the post-Q NP remains in situ despite the presence of Cla in the Q. Choosing among the formal possibilities, we propose that when a Q bears a strong specificity feature, it selects an NP complement with a similar (specificity) feature, this one Interpretable. In a given derivation, the option of assigning the strong specificity feature to the Q and concomitantly selecting the interpretive specificity feature for its NP complement may or may not be exercised. But once such feature assignment has taken place, there is no further choice. The complement NP must prepose overtly to check (delete) this strong feature. This account provides a standard mechanism to drive movement of NP to the Spec of QP. If, however, the numeration contains a non-specific Q, as in 40b, 41a and 43a, then there is no need for feature checking and hence no overt preposing. Now consider the cases in 45 and 46. These DPs are without a classifier. According to the analysis presented above, the Q head in these DPs cannot carry any feature of specificity. The impossibility of using classifiers with these expressions is a morphological reflection of this fact. Why? Because the absence of classifiers precludes the choice of the strong specificity feature for Q. It may be of interest to point out that the Ns in these expressions seem to form a class of their own. These are similar to the bare adverbs discussed in Larson (1985).14 Tfemporal NPs that Larson discusses include 'calendricaT Ns where particular intervals of calendar years function as proper Ns for temporal periods. This property appears in our data set at 45c. Larson's discussion of NPs denoting location, direction and manner seems relevant also to our 45d,e. For the rest of the NPs in 45, one could imagine a function of 'relation' to be included in this special class of Ns.15 As for 46, measure phrases form a class of their own.16 Larson's specific proposal offers an account of Case marking for NPs headed by Ns of this special class without appealing to mechanisms outside the DP. While this idea is compatible with the framework I am using, I do not pursue it here. In the given context, the point that 45 and 46 help establish is that, in the absence of classifiers, the Q bears no attractor feature that could trigger complement NP preposing. Larson is relevant only to the extent that his idea might help future work trying to relate the unusual (classifierless) type of Q in such DPs with the apparently unusual N type that seems to co-occur with it. For the purpose of this paper, the task of explaining why the language tolerates bare Qs and why these Qs seem to select special Ns is not urgent and can be postponed. In conclusion I may note that one conceptually attractive property of our account of NP preposing in specific DPs in Bangla is the fact it provides yet another empirical argument for the Specifier-Head-Complement universal order that I have adopted on the basis of Kayne's work. To return to the basic pattern, compare the following: (49) a. [VNP]=>[vpNPi[Vti]] b. [QNP]=>[QpNPi[Qti]]

Specificity in the Bangla DP · 91

Our analysis of the Bangla DP along with the obligatory movement (either in covert/overt syntax) of the NP to the left of the Q mirrors the situation in a clause. Notice that QP in our formulation is predicative in nature (see 5 in this connection) like the vP-shell in a clausal structure.

• 7. Position of Poss In this section I will provide further evidence of DP-internal NP movement. I will claim that this new type of NP-movement—based on data previously unnoticed in this or a related language—nevertheless, is triggered by the same [specificity] feature responsible for clause-like 'object' shift inside the DP. For a detailed analysis of the data reported here, see Bhattacharya (forthcoming). Examples 7-10 provided earlier show that in the default order within the DP the Poss occurs highest in the tree. This involves either base-generating Poss in this position or setting up a derivation that takes it there. If we were to base-generate Poss at [Spec, DP], we would have to recast the analysis of deixis in section 3 where I proposed to move Dem to [Spec, DP] to obtain a deictic reading. Before I address this issue, consider the following data: (50) a. baba amar khub gorib father mine very poor b. chele amar khub duSTu son mine very naughty c. ma amar SOt manuS mother mine honest human These example indicates that the Poss need not occupy the highest available specifier slot in the tree. At first sight, this order looks restricted to kinship terms.171 will describe this phenomenon as 'Kinship Inversion'. There is evidence from other languages that this type of inversion is not entirely unexpected. For example, in Longobardi (1994), it is reported that in Italian, kinship terms have a cluster of properties not shared by other common nouns. He suggests that kinship terms, in fact, in these uses, behave like proper nouns. Proper nouns in Longobardi's theory obligatorily move to D. This seems too good to be true—I can simply say that kinship Ns in Bangla are like proper names (as in Italian) and therefore they move to D to give the order we notice above. Bangla, however, differs from Italian in one crucial respect; in the former, it is the whole NP which moves up. Consider the following: (51) a. amar buRo baba khub bhalo my old father very good b. *baba amar buRo khub bhalo c. buRo baba amar khub bhalo

92 · Tanmoy Bhattacliarya

(52) a. amar lokkhi ma ebar kheye nao my obedient mother now eat-CONJ take 'my obedient daughter, please eat now' b.* ma amar lokkhi... c. lokkhi ma amar... 5 Ib and 52b clearly indicate that the N moves up along with all its modifiers. In Bangla, therefore, it is a case of NP movement and not N (to D) movement as noticed in Italian and other languages. The data above has shown us that Possessives are not the highest Spec in the tree. Therefore I conclude that Poss is base-generated at a specifier that is not the highest in the DP. Could we merge it at [Spec, QP], then? But placing it there creates more problems than it solves. I have earlier proposed that a [specificity] feature of the Q head moves the NP leftward to [Spec, QP]. More importantly, the following is also possible: (53) amar boi\ du-To t, my book two-CLA 'My two books' That is, we can have a Poss (like amar 'my') preceding the moved specific noun. Now if I claim that the Poss is generated (or even, ends up) at [Spec, QP] then the Object* shift story is problematic unless I resort to multiple Specs. There is nothing in principle to avoid generating multiple Specs for QP, but crucially, empirically and conceptually, a possessive nominal has nothing to do with a Quantifier Phrase. There seems to be no reason, therefore, to generate it at [Spec, QP]. Where does it originate, then? • 7.1. Possession and Specificity If we ignore the cases of kinship inversion, the derived position of the Poss is [Spec, DP]. Does the raising of amar 'my' proceed via [Spec, QP]? This should be easy to investigate since we have seen that only those NPs are attracted to this Spec position which can check the [specificity] feature of the Q head. The following data shows that a Poss always permits—if the intonation is suitably modified—a contrastive reading: (54) amar chele khub bhalo my son very good a. 'my son is very good' b. 'MY son is very good' (55) robin-er gaRi-Ta gEche Robin's car-CLA has gone a. 'Robin's car is gone!' b. 'ROBIN'S car is gone!'

Specificity in the Bangla DP · 93

Although this reading can be forced upon any noun when emphasized, the Poss by its very function restricts the set of possible 'sons' or 'cars' in the above examples. That is, the Poss always picks out a member from a particular set of nouns. 'My son' or 'Robin's car' are identifiable, specific son or car. Let us investigate this position further. Amar chele 'my son' as in 54, contrasts chele 'son' with other members in the set of relations/things/objects belonging to me. So the very use of amar reduces the set of objects that belong to everybody to objects that belong to me. Consider the following sentence: (56) amar CHELE khub bhalo, meye-Ta-i bOjjat my son very good daughter-CLA-EMP nasty 'my SON is very good, it's only the daughter who's nasty!' The focus on chele 'son' now picks out and contrasts chele as opposed to other objects that may belong to the narrow set already created by amar. The other members of the MY-set each denote alternative sets in the sense of Rooth (1985). However, at the moment of calculating the focus, only one of the alternative sets is picked out by the denotation of the NP. This state of affairs is represented as follows:

(57)

Stage I

Stage II

The diagram in 57 seems to imply that both Poss and Focus act as restrictive modifiers on a set. However, although it is arguable whether Poss can indeed be seen as a restrictive operation on a set, the 'reduction' in Stage II only indicates the set (out of other alternatives) finally chosen to receive the focus intonation. Let us now consider the following: (58) AMAR chele khub bhalo, (lomar-Ta bojjat) my son very good your-CLA nasty 'MY son is very good, (it's yours who is nasty!)' In 58 Stage I has taken place but not Stage II, as focus on amar sets it up against a similar Poss-set outside the my-set:

(59)

Stage I

Stage II

94 · Tanmoy Bhattacharya

Stage I in 59 creates the A/y-set. Now, when the my of the A/y-set is focused, the alternative sets created must be from among members outside the domain of the A/y-set. That is, the alternative sets may be made up of objects bearing the relation of possession to you, him/her, or X. So, Stage II cannot take place inside the Afy-set. The diagram in 59 shows this state of affairs where Stage II depicts only one of the many possible alternative sets. Let us see if this explanation holds water for the other order we have been looking at, that is, the marked order of NP-Poss. This order, I suggest, breaks up the thematic relation between the possessor and the possessed. I will assume that the thematic properties of Poss in 60 are satisfied covertly in the case of KI. (60) bhai amar ar chaRbe na brother mine any more leave-will not 'Brother mine will not leave me any more!' As per the set-theoretic account offered so far, I take this to mean that the Afyset that is created is in some sense 'diffused'. I represent this state of affairs as follows:

(61)

Stage I

Stage II

The shaded portion in 61 represents an underspecified area. I believe that a general theory of underspecified semantics as in Reyle (1993) can be implemented for such underspecified sets. However, such an exercise is beyond the scope of the present paper. For the present, note that in 61, it is still possible to perform a Stage II operation of the type shown in 57, but not a Stage Ð operation of the type shown in 59. This is due to underspecification. That is, 62a is a possible derivation from 61, but 62b is not.

(62)

Given what I have said, it would seem that in the marked order (NP-Poss), focusing the NP would still be acceptable but focusing of the Poss would not be allowed, since the only meaningful function of the latter would be to set it up

Specificity in the Bangla DP · 95

against another Poss-set, which, due to underspecification, cannot take place in 62b. Not surprisingly, the data corroborates this prediction: (63) a. CHELEi amar tt khub bhalo,... son mine very good,... 'SON mine is very good,...' b.* cheld AMAR f, khub bhalo,... In a nutshell, we can say that, a priori, there is no semantic reason to prevent the Poss (in syntax) from passing through the [Spec, QP] position. This prediction supports syntactic constraints on local movements as in a phrase like (64). (64) a. amar-r ei du-To boi my-GEN this two-CLA book Apart from such obvious syntactic advantages as in 64,1 will claim that an analysis of possession in terms of specificity has at least two other distinct advantages. First, it has been noticed in some languages, the presence of a Poss makes the NP definite or presupposed. Ghomeshi (1997) reports this fact for Persian. Object nouns in Persian may occur with the definite marker-ra, the indefinite enclitic-/, or without any marker as shown in 65a. However, whenever a Poss is present, the object NP must appear with the definitive marker 65b. (65) a. ketab-o/ketab-i/ketab xund-am book-ra/book-iNDEF/book read-Is º read the book/a book/books' b. ketab-ejian-o/*jian-i/*jian xund-am book-EZ Jian-ra/Jian-lNDEF/Jian read-Is º read Jian's book' Given the conclusion in section 6, presuppositionality is a LF reflection of syntactic specificity. I will therefore consider 65 as evidence in favor of a specificity analysis of Poss. Second, such an analysis allows us to distinguish between two types of spedficty—strong fcnd weak—which is well established in the literature on specificity (Groenendijk and Stokhoff, [1981]; and Ludlow and Neale, [1991] among others). I will claim that specificity due to Poss moving to [Spec, QP] is weak specificity which does not require the identification of the referent, whereas specificity due to NP moving to [Spec, QP] is strong specificity which strongly requires such identification. The final movement of the Poss to its derived position—that is, to [Spec, DP]—is due to a feature like [Poss] in D which attracts the Poss to its Spec. A piece of direct evidence that the theory outlined above is on the right track is the observation that the Poss does not move up to [Spec, DP] in case of kinship inversion. This is because, as I have mentioned earlier, inversion breaks up the thematic relation of, say, Possession. It is now straightforward to see that in

96 · Tanmoy Bhattacharya

these cases there is no [Poss] feature to check at [Spec, DP], with the result that the Poss remains in its base-generated position in kinship-inversion cases. What I am suggesting is this: the Poss can indeed raise up to [Spec, QP] to check the [specificity] feature of the head, but cannot stay there. This is not only because the surface order of Dem-Poss-NP is not allowed in Bangla, but because the Poss in question has another Interpretable feature which checks a similar feature of the D at [Spec, DP] and we end up getting the order PossDem-NP as in 64. It is not a coincidence that Poss are marked with the Genitive. This marking reflects their derived position, which is [Spec, DP]. That [Spec, DP] is a site for GEN case checking is independently suggested by various scholars (Ritter [1988]; Miyagawa [1993], and others). There are these two facts to consider. We have seen that the Poss can pass through [Spec, QP] but cannot remain there. We have also seen that the Poss is not the highest Spec in the DP tree since kinship inversion leaves the Poss stranded. These pieces of evidence indirectly point to the conclusion that the Poss is base-generated at a position lower than [Spec, QP]. According to some authors (Giorgi and Longobardi [1991]); Malten [1992] and others) Possessives are like adjectives (for some languages) and therefore must be generated within the NR If that is the case then in the current framework, we can generate the Poss within the nP-shell.18

(66)

With this I have now accounted for all the positions in the structure of the Bangla DP that I proposed in 25. In this section I have provided an additional motivation for the [Spec, QP] position (and therefore, QP), a position where the kinship noun moves to and through which the Poss moves up to [Spec, DP] in nonkinship DPs. The presence of the Poss in the nP-shell in 66 brings about complete isomorphism with the vP-shell structure of a clause. This is a final piece of evidence for the detailed comparability between clausal and phrasal structures. • NOTES 1. Rijkhoff (1990:24), working within the Functional Grammar framework, suggests the following to be the canonical structure of the NP:

Stem Asp Suff be Asp Suff Aux Tns Agr This structure is illustrated in the following sentence: (2) hari-ji daur-ait rah-ait cha-1-aah Hari-3h run-Imp be-Imp Aux-PT-3h 'Hari had been running'. Maithili, however, differs from most of the Indo-Aryan languages with respect to verb agreement, in that a finite verb can register the grammatical features of one to three referents.2 Examples in 3a-c illustrate single, double and triple agreement, respectively: (3) a. ham sut-l-au(h) IN sleep-PT-lN º slept'. b. ham hun-kaa madat kar-1-i-ainh IN 3h-ACC help do-PT-lN-3hACC º helped him'. c. ham to-raa hun-ak kitaab de-1-i-au-nh IN 2mh-ACC 3h-GEN book give-PT-lN-2mhACC-3hGEN *I gave his book to you'. The controllers of verb agreement in all the three types of verb agreement include not only the arguments of a predicate and the possessors (as shown in 3a-c) but also non-arguments like nominals in postpositional phrases and possessors therein as well as deictic referents in a discourse. Examples are given below: (4) a. to hun-kaa-lel kaaj kae-1-ah-unh (agr with non-argument) 2mhN 3h-OBL-for work do-PT-2mhN-3hOBL 'You worked for him'. b. ham toh-ar ghar-par ge-1 ch-al-i-ah (agrwith the possessor in a non-argument) IN 2mh-GEN house-at go-pel Aux-PT-lN-2mhGEN º had been to your house'. c. ham o-kraa maar-1-i-ah (agrwith a deictic referent) IN 3nh-ACC beat-PT-lN-2mh º beat him (who is related to you, etc.)'. Table 1 presents the three types of agreement paradigms in Maithili.3 (See Appendices 1 and 2.1-3 for the illustration of intransitive paradigm with daur'run' and the transitive paradigms with the verb dekh- 'see', respectively.4) As we see in Table 1, the agreement paradigms in an intransitive clause show a binary contrast of nominative and non-nominative cases. That is to say, nominative referents have one set (listed vertically on the leftmost side) and

The Complexity of Maithili Verb Agreement · 141

f

V IJr 5s

l IT

s

5 g

fff

11 i X·^

X-N

^^

JS

43

A

X-N

X-N

^^

s^

—''^~*'^ 5

>

C B 3

1 11

— ^£ ^ * 3 •S -S ? ? ?r

B

•ô

â

.S

II

f ·§

II

!! l X

JZ

A* Έ

e

'?

«|*

g

fff

f

l

ill

•f

|f f I

l

'^

*^

g i

·&

å Ñ

cd

'a

J^

s. CO

1

*T

3

J^

c' -9 *?

«Ë iS, 5 -

X—S

SS f

*?

CO

•f -ji

•s -s ^ s' '? '? 4 ? ? 4

*->

Q+ O« 3h > 2nh > 3nh > 2h/l

(ii) The controller has received prominence in discourse. Such an agreement comes exclusively from non-nominative set. The non-argument in this case can be either a nominal in a postpositional phrase or deictic referent. The agreement with the first type of non-argument is exemplified in 16: (16) a. ham tohar ghar-par ge-1 ch-al-i IN 2mh GEN house-at go-Pel aux-PT-lN º had been to your house', b. ham tor-e ghar -par ge-1 ch-al-i-a(h) FOC !N-2mhNN º had been to YOUR house (not somebody else's)'. 16a shows agreement only with the argument ham, while the verb in 16b is marked for both the argument ham and the focused non-argument tor-ey which is 2mh—a high—ranking referent. But the agreement with a focussed non-argument is precluded if it is lower than the argument in the honorific hierarchy. This is shown in 17b: (17) a. to okar ghar-par ge-1 ch-al-a(h)? 2mhN SnhGEN 2mh b. * to okar-e ghar-par ge-1 ch-al-ah-uk? 2mh-3nh 'Had you been to his house?' Another type of non-argument which may optionally trigger verb agreement in Maithili is the deictic referent in a discourse. It generally refers to the person

The Complexity of Maithili Verb Agreement · 149

addressed or 3 h referent present in the discourse, even though no such participant appears in the clause. Consider the following examples in this regard: (18) a. ham ohi aadami-sa bhet-al-i INom that man-to meet-PT-lN º met that man'. b. ham ohi aadami-sa bhet-al-i-ah lN-2mh º met that man (whom you had referred to)'. c. to dhyaan lagaa-ke padh-1-ah? 2mhN attentively read-PT-2mhN 'Did you read attentively?' d. to dhyaan lagaa-ke padh-1-ah-unh 2mh-3h 'Did you read attentively (for your father's sake)?' Both of these non-argument controllers are determined by salience; that is to say, they are allowed to trigger verb agreement only if they need to be focused in discourse. 3.2.2. Triple Agreement Lastly, there is a triple agreement paradigm. The verb agrees with three referents, as shown in 19: (19) a. ham tohar baabu-ji-ke dekh-al-i-au-nh INom 2mhGEN father-3hACC see-PT-l-2mh-3h º saw your father'. b. ham tora hunak kitab de-1-i-au-nh INom 2mhAcc 3hGen book givePT-l-2mh 'I gave his book to you'. This triple agreement is, however, heavily constrained, in that it permits only the combination of IN and 2mh and 3h NN referents. The reason for the constraint seems quite logical in the light of the facts about the Maithili agreement system. First, 1 and 2h referents trigger agreement only if they are in nominative case; and of the two, only if 2h is selected as a controller then it will have to be combined with another controller in the second person, viz. 2mh, which is logically impossible; so, we are left only with 1 as a logically possible controller in this case. Second, the choice of 2mh/nh and 3h as non-nominative controllers in the triple agreement can be justified in terms of the personhonorific hierarchy in Maithili, which licences only the higher referents, namely, 2mh/nh and 3h, to control the verb agreement.

• 4. Conclusion In the preceding sections we have analyzed the verb agreement system in Maithili. This analysis reveals that in Maithili the grammatical features (person,

150 · Yogendra P. Yadava

honorificity, gender and even case) of one to three referents may yield some composite function on verbal inflection. This composite verb agreement comprises the features of the controlling referents not only in the theta-grid but also outside it (e.g. non-arguments in PPs, the matrix subjects in a raising constructions or missing deictic controllers). Besides, in a modification structure, the verb agreement may encode the features of the head NP or its modifier or both, depending on their status in the honorific hierarchy. To sum up, this paper is an attempt to explicate the complexity of the verb agreement system in Maithili. There remain, however, a number of issues which need to be further investigated and spelt out, especially in terms of its dialectal variations, discourse strategies8 and historical development. Though this paper is just descriptive in its approach and no attempt has been made to analyze the data provided here within the framework of any linguistic theory, it is hoped that these data would be of immense interest to any linguistic model. I APPENDICES Appendix 1: Intransitive Paradigm (daw 'run') Persons/Tenses Present 1 2nh 2mh 2h 2hh 3nh 3h 3hh

Past

daur-ait ch-i daur-ait ch-ae daur-ait ch-a(h) daur-ait ch-i daui-aljaa-ii ch-ai daur-ait ai-ch daur-ait ch-aith daur-ait ch-alhin(h)

daur-\-au(h) daur-1-áé daur-l-effc) daur-l- f/i) dauT-al ge-iai daur-Ë-É daur-1-áû/é dauT-\-athin(h)

Future daur-a-b daur-b-iie daur-b-a(h) daur-a-b daur-al je-t-ai daur-u-l daur-i-aa(h) dauT-t-athln(h)

Appendix 2.1: Present Transitive Paradigm (dekh-'s Subject

1 2nh 2mh 2h 2hh

Nonsubject

1 dekh-ait ch-ae dekh-ait ch-a(h) dekh-ait ch-/ dekh-al jaa-it

2nh

2mh

dekh-ait ch-iau(k)

dekh-ait ch-ia(h)

-

-

-

-

-

ch-ai(k)

2h dekh-ait ch-/

3nh dekh-ait ch-iai(k) dekh-ait ch-ahi(k) dekh-ait ch- indirect object > direct object, the interpretation is, as it were, context free. The scrambling operation presumably alters the interpretation of focus and scope. Finally—and this is another weakness of the questionnaire-dependent grammatical description—the authors provide no data on multiple questions with both what AND why questions. Since both of these question words compete for the same position, according to the authors, the natural question to ask is: what happens when both are present in the clause? How does the grammar of Kashmiri resolve this conflict? These are interesting questions that should have been addressed in a 'cognitive-descriptive' grammar of a language. The section on clause-types is followed by a discussion on the 'internal structure' of different clauses and phrases. This is followed by the presentation of coordination facts: the various ways in which Kashmiri grammar makes coordination (of sentences, phrases and words) possible. Coordination is followed by a discussion on negation, both sentence and constituent. The discussion of anaphora comes next, followed by a discussion of different nieans of expressing equatives. The last couple of sections in this chapter discuss issues of focus and topic in this language. The chapter on morphology presents the richness of the morphological system of Kashmiri. This chapter has two main sections: Inflectional morphology and derivational morphology. The richness of the inflectional system is faithfully recorded in more than 100 pages. The remainder of the 30-odd pages discuss mainly derivational morphology and, briefly, other word-formation operations like compounding. The chapter on phonology is weak, packing the entire chapter into 14 pages. Very little effort is made to go beyond Kachru (Braj, B. Kachru. 1969. A Reference Grammar ofKashmin. (Mimeo) Urbana: Department of Linguistics, University of Illinois.) and Grierson (Grierson, George. 1911.Ë Standard Manual of the Kashmiri Language, 2 vol. Oxford: Clarendon Press). The

250 · The Yearbook (1999)

authors might have spent some time discussing the stress pattern of Kashmiri, which is typologically similar to Khalkha Mangolian and other such languages. The generalization is that the stress falls on the leftmost heaviest syllable of the word; the last syllable is extrametrical. The discussion is present in Bhatt (Bhatt, Rakesh M. 1994. Word Order and Case in Kashmiri. University of Illinois, Urbana, USA: Ph.D. dissertation), which the authors have consulted. Clearly, in the next project we'd like the authors to undertake a detailed investigation of the phonology of Kashmiri. The two pages of the chapter on ideophones and interjections are followed by a lexicon of kinship terms, body parts, color, cooking and agricultural terminology, and a list of basic vocabulary. These chapters are followed by a comprehensive bibliography and a useful index. I applaud Kashi Wali, who, as the Acknowledgments' show, has invested a good part of the late 1980s and early 1990s researching the information for this book. It is not easy to record the grammatical properties of a language about which very little is known and/or available. And, unfortunately, it probably does not help to be constrained by the questionnaire format of descriptivegrammar writing. According to the series editor, the grammars in this (Routledge Descriptive Grammar) series bridge the gap between theory and (Description. Further, the editor writes that the questionnaire-framework provides a description of language that is both sufficiently comprehensive to cover the major structures of any language that are likely to be of theoretical interest and sufficiently explicit to make cross-language comparisons a feasible undertaking. However, these goals serve only the purpose of theory construction and thereby undermine the creative potential of natural languages. To conclude: this book mainly suffers from its own structural (questionnaire) design. There are numerous places, too many to mention, where it appears that the data were grafted upon the dotted line. I am afraid the book will not be of much use to those who want to learn and teach the language, or who wish to consult it for clarification of a grammatical point. It is designed for the consumption of those linguists who need to verify the reliability of a theoretical account—and even there it may fall short, as I have tried to show using the data in (1) and (2). What we now need is a grammar of Kashmiri, fully nuanced, without any responsibility to theory construction: a grammar ä la Quirk et al.—*4 Comprehensive Grammar of the English Language (Quirk, Randolp, Ian Svartvik and Geoffrey Leact. 1985. A Comprehensive Grammar of the English Language. London: Longman). Good descriptive grammars have, and will, outlast current trends in linguistic theory: they will stay with us as a useful resource for linguistic-conceptual thinking, language learning and teaching, and for language maintenance and survival.

D

Dialogue

Linguistic Revolutions: A Reply to Tanmoy Bhattacharya JOSEF BAYER Bhattacharya's review of my 1996 book reveals some attitudes and convictions concerning linguistic theory that I find quite astonishing, if we agree that linguistics has developed into a mature field of scientific inquiry. 'Mainstream' linguists are obviously struck by a disease which forces them to rephrase their work with every scientific revolution that changes some of thfc axioms of syntactic theory. The revolutions come in short intervals, and they come from the desks of Noam Chomsky and Richard Kayne. The rest of the 'mainstream' community has the job of deriving earlier results under the new assumptions and affirming the new theory by supplying more data and filling in the details. Of central interest seems not so much how we can explain linguistic data in general, but rather to be sure about what we can't do under the narrowest interpretation of the most recent proposal. As a consequence, various results of linguistic research then become untenable or in the worst case uninteresting once we have again been hit by a new revolution. Those who want to publish a larger piece of work run the risk of facing, while going to press, a revolution that affects basic assumptions. Very obviously, I had the bad luck of becoming such a victim. Years of research had convinced me that languages differ by their basic word order, and that these differences yield extremely interesting interpretive effects which somehow should be explained by linguistic theory. I showed in my book what would be required within the framework of Barriers enriched by Pollock-type articulation of functional projections. What I have seen of the Minimalist and of the Antisymmetry revolutions, that took place shortly before the final version of my book had to go to press, turned out to be quite incapable of dealing with the challenges of my data. My evaluation of the Mainstream' theories up to Chomsky's Bare Phrase Structure

254 · The Yearbook (1999)

and the manuscript of Kayne's The Antisymmetry of Syntax appears in my book, and it should be clear to the reader what prevented me from fully embracing the new theory. In his critique the reviewer obviously does not have minimalist explanations in mind which would cope with my data in a superior way. His point is rather that earlier assumptions have to be given up no matter what they bought us, and no matter what the new assumptions can buy us. When it comes down to the real problems, I find some of the theoretical assumptions in my book far more parsimonious than those to which Bhattacharya seems to submit. For instance, he finds the idea of rightward movement with subsequent deletion of the trace 'strange within Chomskyan linguistics'.11 derive from this that he finds nothing strange in the assumption of large numoers of unmotivated features that are postulated for one and only one reason: to cope with elementary word order variations. Bhattacharya, for example, seems to be convinced that current generative syntax offers insights into the extraposition phenomenon that would automatically shrink alternatives down to 'speculations'. Obviously he is unaware of the classical and more recent literature that tries to divorce this phenomenon from core syntax by seeking the relevant constraints in the human processing system or in prosodic phonology. A glance at the works of Hawkins, Truckenbrodt or various contributions to the recent volume edited by Beerman, LeBlanc and van Riemsdijk should suffice to get an impression. Interestingly, not even Chojnsky seems to be convinced that the full range of extraposition phenomena can be captured by a 'proliferation of [Larsonian, JB] shells' which would be needed to strand material in basic positions.2 Here the reviewer is clearly more Catholic than the Pope. The conclusion that the extraposition problem has been solved by the antisymmetry hypothesis appears to be unwarranted; and as far as I know, the number of syntactidans who remain open to unconventional alternatives is large. Since Bhattacharya finished his review, another revolution has taken place: Kayne (1998) indicates how the scope effects, which are treated at length in my monograph, can be derived without making reference to covert movement. Extensive movement of VP-remnants and, of course, still more features are the price that has to be paid for this simplification. If Kayne's proposal is to be generalized, 20 years of research on LF have to be done over. Let us hope that there will be enough 'mainstreamers' who find the time to tidy up the post-revolutionary battlefields. • NOTES 1. As I pointed out in section 7.4.5 of my book, the proposal is not strange at all from a diachronic perspective. Extraposed complements seem to have emerged from constructions in which the complement was related to an overt nominal placeholder which could then disappear under certain circumstances. I hope everybody agrees with me in the assumption that syntactic analyses will ultimately have to stand firm against this kind of evidence too. 2. cf. Chomsky (1995: 333).

Dialogue m 255 REFERENCES Beerman, D., D. LcBlanc and H. van Riemsdyk (eds). 1997. Rigfitward Movement. Amsterdam: Benjamins. Chomsky N. 1995. The minimalist program. Cambridge, Massachusetts: MIT Press. Hawkins, J. 1990. A parsing theory of word order universals. Linguistic Inquiry 21,223-61. Kayne, R.S. 1998. Overt vs. covert movement. Syntax 1,128-91. TVuckenbrodt, H. 1994. Towards a prosodic theory of extraposition. MIT, ms.

On Purism and Idealizations in Linguistic Theory: A Response ANJUM P. SALEEMI To the extent that it is possible to summarize the intriguingly complex and provocative arguments in the paper under discussion (i.e. Dasgupta 1997), what Probal Dasgupta (henceforth PD) says may be outlined as follows. His concern seems to be with two sets of issues. First, there is a set of technical issues pertaining to the purported main intent of the paper, which is to initiate a debate between two particular viewpoints of semantic interpretation attributed to Alec Marantz (AM) and Noam Chomsky (NC). One of these, namely the one attributed to AJM, is presumed to consider interpretive mechanisms to have access to syntactic derivations in PF (i.e. Phonetic Form: the interface with articulatory-perceptual mechanisms) and at various pre-PF stages as well as in LF (Logical Form: the interface with the conceptual-intentional system of cognition), whereas the other, claimed to be characteristic of NC's thinking, is supposed to strive to confine interpretation largely to LF, with PF having some role in the process. Thus, the key issue may be somewhat loosely described in the following terms (where N = numeration: the lexical reference set for a derivation). (1) Is interpretation carried out exclusively at the interface levels, particularly at LF, or can it be read off at any point in the course of a derivation from N to LF and PF? Second, there is a set of larger, conceptual issues related to the nature of human language in general that gets reflected in the paper because adopting one of the two derivational-interpretive views, presumably, has consequences

258 · The Yearbook (1999)

for one's position on these issues. In what follows, I shall consider each of the two kinds of issues, starting with the technical ones and then moving on to the conceptual ones. As the former are relatively concrete and therefore easier to deal with, I shall probably have a lot more to say about the latter, as their significance goes well beyond that of mere technicalities prevalent at any given time in a rapidly evolving field. Regarding the technical questions, I shall start by saying how much they remind one of the controversy surrounding generative semantics (GS), which (as is well known) revolved around the role of interpretive mechanisms at deep and surface structure. PD's line of reasoning, though couched in terms of LF and PF rather than deep and surface structure, appears to have some of the flavor of that controversy. If that is indeed the case, then I would like to point out that, at least to my mind, NC's position now is essentially the same as it was then, although quite obviously considerable technical and terminological differences have since emerged due to further developments in the theoretical apparatus. For example, deep structure, D-structure and LF are similar in some respects but crucially different in others. At any rate, that deep structure or LF (or any other comparable single level) can uniquely or predominantly bear the burden of semantic interpretation is not a claim that, as far as I know, NC has ever embraced; in fact as we know it was exactly such a claim that characterized the GS approach. The temptation of suggesting a conversion on the part of NC to a GS-like viewpoint notwithstanding, it seems to me that this is not a correct analysis of the situation. Perhaps there are ways in which the latest reincarnation of the transformational theory, as presented in Chomsky (1995), does indeed resemble GS. In respect of semantic interpretation, however, the Minimalist Program (MP) is certainly not any closer to GS than were its predecessors. It is true that LF, carrying as it does both the entities originating in the lexicon and those resulting from displacement before and after Spell-Out to PF, is the level that contains a good deal of meaning. However, such meaning is largely of the conceptual and logical nature, and is thus appropriately interpretable at LF. However, several remaining aspects of meaning, for instance those related to given-new information, topic-comment and focus-presupposition (see, e.g. Kidwai 1997, for a treatment of focus placement resulting from scrambling) fall within the domain of PF interpretation. Assuming that this conceptualization of interpretation as occurring at both LF and PF is correct, I do not see any real conflict between AM's and NC's viewpoints insofar as the interpretive mechanisms, as contrastively presented in (1), are concerned. It should be mentioned that the late insertion of phonological content of words as proposed in Distributed Morphology (Halle and Marantz 1993) does not impinge on this issue, since the assumption in DM is not that the semantic content of lexical items is not available before PF, but only that words construed as forms in morphophonology and morphosyntax are usually not crucial to pre-PF operations. There may well be some problems involved in adopting MP or DM together, or perhaps certain subtle

Dialogue m 259 differences in regard to their conceptions of word meanings. For instance, AM's view is possibly more Saussurian than NCs in the way for AM words as PF forms relate to the corresponding meanings. But the two approaches do not appear to me to be necessarily incompatible. It is hard to imagine that PD is truly convinced that AM and NC actually literally espouse the views attributed to them by him, nor does it appear even remotely plausible that he is engaged in deliberate misinterpretation. I suspect his motive, partly obscured by the immense complexity of the issues involved, is to exploit the two conceptually distinct approaches to the nature of interpretive mechanisms in order to raise certain issues of wide-ranging significance. In his view, MP, perceived as a theoretical retreat from many previous positions, can act as a springboard for further potentially beneficial revisionism. Let me now turn to this strong revisionist/programmatic undercurrent in PD's paper, which I think deserves some serious consideration. Needless to say, at least to some extent my interpretation of the programmatic subtext of PD's paper is based on a reconstruction of his argumentation in my own words, as there are several points which are not spelled out explicitly enough by him. For one thing, I think PD wants the computational system handling linguistic derivations to do more than just 'pure' syntax and a little beyond. He wants it to handle many more aspects of meaning than it is currently supposed to, and is somewhat suspicious of a narrowly modular view of the language faculty wherein it is primarily conceived as a syntactic system, with the consequence that several semantic and pragmatic aspects of natural language are simply left out of the picture, to be dealt with by some other presumably non-linguistic module(s) (cf . Saleerai, forthcoming, for a discussion of a range of parallel concerns). So for him the promise of MP is partly contained in the way it could accommodate many facets of language by reducing syntax to a minimum: the latter part of this perception is echoed by AM when he states that under the minimalist assumptions 'The syntactic engine itself—the autonomous principles of composition and manipulation Chomsky now labels "the computational system"—has begun to fade into the background' (Marantz 1995:380). According to PD, this change could lead to a view of linguistic derivations whereby syntactic combinatorics, meanings and pragmatics can operate in step with each other. He is particularly intrigued by the introduction in the system of lexical choice (i.e. numeration) as the reference set for a derivation. He suggests that the computational-derivational system should not be just about picking out the most optimal/admissible derivations from among the many convergent ones: rather, one of his suggestions is that, given a particular thought, the derivational apparatus should also strive to determine how many syntactically distinct ways there can be of expressing it, and which one of these is the least and which the most marked. Notice that such an expanded linguistic theory will still be concerned with the inescapably narrow formal issues, and therefore cannot be accused of sociologism. It, however, will be

260 · The Yearbook (1999)

more sensitive to issues of lexical choice and speaker intentions. Such a view will naturally force a redivision of labor between syntax on the one hand and semantics and pragmatics on the other, primarily by incorporating into the derivational machinery a wider range of semantic options. Here is one way of 'unpacking' (to use one of PD's expressions) the foregoing proposal in some more detail. Take the examples (2a), (2b) taken from PD. These are based on two different numerations, which appear respectively as (3a) and (3b). (2) a. John expected that he would win. b. John expected to win. (3) a. Numeration A = {John, expected, that, he, would, win} b. Numeration B = {John, expected, to, win} Simply put, in MP the main task will be to find out the most optimal/economical convergent derivation for each of the two numerations independently of each other. Thus, if derivations Dl, D2, D3, ...Dn are all convergent derivations of A, then some derivation Di is the most economical of these. PD, however, would also want linguistic theory to consider the fact that the strings (2a) and (2b), in spite of being the output of distinct numerations and the related derivations, are nearly synonymous, as they can be said to express the same proposition P, with the difference that we might wish to consider (2b) to be (in some sense) a less marked way of saying P. A theory of language that does not capture the similarity (and of course the differences) between (2a) and (2b), or for that matter between an active sentence and its passive counterpart, is missing something crucial. As is ^cell-known, under the earlier versions of generative grammar, such similarities were understood to follow from the assumptions, now abandoned, that there were transformational relationships between such related strings. In the absence of such relationships, clearly some other way must be found to accomplish the job of relating. The force of PD's argument that syntax in particular, and linguistic form in general, is not a pure, unified entity is enhanced by the suggestion that functional categories, though supposedly narrowly syntactic, are typically rooted in reference. Thus: Why should tense, determiners, etc. be indexical? And why are at least some grammatical features interpretable? Further, is it the case that even the uninterpretable features (e.g. gender) help fix the reference of a DP, even though the relationship between a particular grammatical feature and the non-linguistic entity it is associated with is more often than not arbitrary, to some extent like the connection between a word and its referent? PD also notes how the 'deconstruction' of nominal expressions into chains finds a sort of parallel in what he refers to as paradigmatic 'ladders', and how neither sentences nor words can have the identity they do without appealing to their particular lexical, morphological, syntactic, informational and referential contexts. In this context it should be interesting to point out the obvious fact of

Dialogue m 261 how inflections mediate between the lexicon and syntax, with words providing the proverbial meeting ground for syntax and meaning, form and content. This continuous mixing of one entity with another PD considers to be a design feature of language. His conclusion: 'The nature of language makes impurity necessary'. A major corollary of the foregoing points is that PD's programmatic agenda is less than comfortable with some of the methodological idealizations that have held sway in the field for a long time. As we know, one of PD's concerns is that linguistics remains too purist in its approach to its subject matter, namely human language. The purism gets reflected in several ways: in the methodology, in the prevalent conceptualization of the subject matter, and in the theoretical reasoning employed. It is.primarily embodied in the set of idealizations that modern linguistic theory has adhered to since its inception some decades ago. Clearly, any theory of a set of related empirical phenomena has to make some preliminary, a priori assumptions about its subject matter, the relevant data and what is likely to be the most effective methodology to describe and explain these data. As Chomsky says, Any serious approach to complex phenomena involves innumerable idealizations' (1995: 19). Over the last few decades, some idealizations have played an important role in the development of contemporary linguistic theory. The more important ones among these, in my view, are restated in (4) for the sake of easy reference. (4) a. Abstract away from language in actual use, as what is manifested is often considerably distorted by non-linguistic factors and performance limitations. b. Idealize away (to the necessary extent) from semantic complexity, especially if its reflexes in the formal structure of language are weak or non-existent. c. Idealize away (to a desirable extent) from the diversity exhibited within a language as well as across many languages, the idea being that deep generalizations based on a relatively small amount of data are better than superficial ones relying on a much wider database. d. Idealize away from the diversity created by late (i.e. adult) acquisition of languages, under the assumption that the problem posed by the (monolingual, normal) child's successful learning of the language of his immediate speech community is of a more fundamental character. e. Idealize away from diversity due to historical factors. This idealization will have the effect of excluding language change and contact phenomena (such as code-switching/mixing, pidginization and creolization) from mainstream theoretical considerations.

262 · The Yearbook (1999)

Among these idealizations, the one identified as (a), which says that structural mechanisms can (in fact must) be studied without investigating the uses they can be put to has a particularly distinguished track record, its dominance spanning the greater part of the era of modern linguistics. It is crucial to the reasoning that underpins the distinction between competence and performance and to the hypothesis that language is an autonomous module of human cognition. Further, one of its notable consequence is that the explanation of linguistic behavior in its entirety, warts and all, is usually not considered to be a worthy goal for linguistic theory to pursue. More narrowly construed, it is also central to the thesis of the autonomy of syntax from meaning, of both linguistic and pragmatic kinds. Combined together, the idealizations described in (4) impose severe a priori limits on the character of the core database of linguistic theory, which is constrained to operate 'under the idealized conditions of a homogeneous speech community* (Chomsky 1995:19). At the same time, they admittedly enable it to study the relatively reduced and uniform database in considerably greater depth and with sharper focus than would otherwise be possible. Since 'We do not expect to find "pure instantiations" of the initial state of the language faculty', and there are reasons to suspect that 'even if a homogeneous speech community existed, we would not expect its linguistic system to be a "pure case" ... [as] all sorts of accidents of history would have contaminated the system' (Chomskyl995), it seems paramount to abstract away from such accidents so that the initial state of the human language faculty can be isolated from the non-linguistic phenomena for the purpose of objective inquiry. Thus, in general, the generative approach fervently assumes that inadequate idealization can be dangerous, and insists on the virtual independence of form from content, of underlying enabling knowledge from its use, as without separating one from the other the empirical domain under investigation will appear to be too uncomfortably messy (or, to use PD's term, 'impure') to allow the researcher to tease apart different varieties of facts. Comparing language with the visual system and the system of motor coordination, Chomsky (1995) argues why 'a theory of the mechanisms of vision or motor coordination' should not try to 'explain why someone chooses to look at a sunset or reach for a banana. The problem of choice of action is real, and largely mysterious, but does not arise within the narrow study of mechanisms' (p. 227). Indeed, it does appear to be true that formal units and mechanisms (the syllable, the metrical foot, uninterpretable features, Merge, etc.) are just that: formal constructs. To understand exactly what they are one needs only to confront them on their own as structural entities, and to additionally consider the function that they perform in a given context would merely confound our understanding of their natural properties. Why should, then, one worry about the baffling impurities of superficial reality, if the general scientific belief, warranted by plenty of evidence, is that at deeper levels of reality lie a set of simpler mechanisms whose complex interaction gives rise to its messy appearance?

Dialogue m 263 Part of PD's argument to the contrary rests on a curious twist of the methodological tale that suggests that not all idealizations are desirable, and that some in fact could be detrimental to the cause of objective understanding. He points out that religious and historical purisms hampered our understanding of language as a natural phenomenon, and that what are perceived to be scientific trends could actually lead to similar negative consequences. If natural languages are in matter of fact considerably more complex than can be inferred from our idealizations, and if these later are likely to distort and oversimplify rather than clarify, then perhaps it would be quite reasonable to want to revise our methodology accordingly. PD seems to be convinced that such sanitized and idealized practice of linguistics manages to evade the systematic avoidance of several issues of extreme importance in the study of human language, erecting barriers to the proper understanding of the full complexity of its diverse phenomena. What I have done in the preceding pages is probably not much by way of analytical appraisal. My commentary is expository rather than critical, in part because I think the jury is still out on the issues under discussion. It is very hard to strike the right balance between the wealth and diversity of facts and the power and depth of an austerely conceived theory. In the end, it does not seem clear whether PD wants linguistic theory to adopt just some new methodological strategies and theoretical mechanisms, or a whole new and more complex view of language which is different from the narrower conception it currently prefers to maintain. Though his invitation to controversy appears to be ambivalent between a narrow and a broad debate, with neither lacking in substance or conceptual appeal, the latter certainly could contain many interesting hints for what the ultimate goals of linguistic theory might begin to look like under a somewhat expanded conception of language. • REFERENCES Chomsky, Noam. 1995. The minimalist program. Cambridge: MIT Press. Dasgupta, Probal. 1997. Knowing the word Trikke: Against purism in the study of language. Yearbook of South Asian languages and linguistics 1998, ed. by Rajendra Singh, 247-66. New Delhi: Sage Publications. Halle, Morris and Alec Marantz. 1993. Distributed morphology. The view from building 20, ed by Ken Hale and Sam Jay. Cambridge, MA: MIT Press. Kidwai, Ayesha. 1997. Word order and focus positions in universal Grammar. Jawaharlal Nehru University, ms. Marantz, Alec. 1995. The minimalist program. Government and binding theory and the Minimalist Program, ed. by Gert Webelhuth. Oxford: Basil Blackwell. SaJeemi, Anjum. Forthcoming. Skinner's razor and the varieties of mentalism. Proceedings of the Conference on Language and Knowledge, ed. by Bao Zhiming and Thiru Kandiah, Singapore: National University of Singapore Press.

On So-called Compounds* RAJENDRA SINGH and PROBAL DASGUPTA The purpose of this note is (1) to flesh out, with the help of some facts pertaining mostly to Bangla and Hindi, the somewhat cryptic remarks Ford and Singh (1996a: 168) and Ford, Singh, and Martohardjono (1997: 58ff) make regarding what are normally called compound words and (2) to invite our readers to tell us why we should not reject, lock stock and barrel, the standard characterization of compounds as a subclass of words made of two or more words (cf. Anderson [1992]; Bloomfield [1933]; Bloch and Trager [1942]; Lieber [1981]; Selkirk [1982]; Spencer [1991] and Williams [1981], amongst several others, virtually everyone, even Jespersen [1949]; for a notable exception, see Starosta [1997] and Starosta et al. [1997]). The uncritical acceptance of the characterization in question, unfortunately, seems to lead researchers not to even notice the sort of contradiction one finds in, for example, Vijaykrishnan (1994): although he notes, for example, that in the 'compounds' he is concerned with 'the constituent (inner) V of the verb compound is not accessible in the syntax' (p.266), he sees no difficulty in postulating or operating with structures like [ [X]n [Y]v]v. It seems necessary to issue such an invitation because those who reject word-internal syntactic compositionality are often asked 'But what about compounds?' as if they had never heard of them. The appropriate response to the question 'What about compounds,' we'd like to suggest, is: 'What about them?' We choose Bangla and Hindi endocentric 'noun-noun compounds' to make our case because 'compounds' like these are often presented as providing particularly telling evidence for 'the internal structure of words' (cf. Anderson).1 * This note is a revised version of a paper presented at the 1998 GLOW colloquium (20-22 January), held at CIEFL, Hyderabad. We are grateful to Alan Ford, Yves-Charles Morin, K.P. Mohanan, and Stanley Starosta for numerous discussions regarding 'compounds'.

266 · The Yearbook (1999)

Ford and Singh reject that characterization because their theory of morphology, which we shall, following a suggestion of Starosta's (p.c.), refer to as Seamless Morphology, does not allow them to entertain different kinds of morphologies for what are seen as different subclasses of words. They insist on a unified morphology without any internal structure beyond word internal phonological and semantic structure and claim that words standardly seen as morphologically complex are simply words that can be de-concatenated into two substrings, one representing a variable and the other representing a constant, with the help of some Word Formation Strategy (WFS), though the representations of these words do not contain any indication of the parts they are sometimes said to be made up of. The English word banality can, thus, be said to contain two substrings, banal and ity, and the word doghouse can be said to contain the substrings dog and house. In the former case, the substring banal takes, to anticipate a bit (cf. [3]), the place of the variable X in the WFS: 1X1M\. /Xiti/Noun and the substring ity is the constant specified in the relevant WFS. There is, they claim, no significant difference between the ity in banality and the house in doghouse: they are both constants. The P ninian's attempt to situate his various constructs—root, stem, affix, augment, etc.—within a structured hierarchy, as we shall see, seems unable to cope with the facts of the matter. The fact that the house in doghouse looks like the word house is only an appearance, as Bhartrhari (cf. Iyer 1977) would have said. It certainly falls short of the commutative possibilities of dog, which, as a moment's reflection will show, can be replaced by any number of words. Grammaticalization flies in the face of the common belief, not to say illusion, that in expressions such as doghouse, either of the two elements can be freely replaced. When this belief is turned into a hypothesis, as in Dowty (1979), it turns out to put too heavy a burden on pragmatics: what, for example, is the PRAGMATIC reason which allows terminvorschlag and terminvereinbarung in German but blocks *date proposal and *date agreement in English? Facts of'compounding' from various languages indicate quite clearly that 'compounds' come in sets, each set anchored in some specific constant. There are, in other words, exceptions to all of Levi's (1977) nine classes, but only because she has only nine. What is needed is the identification of as many sets as there are constants. The rather large number of rules Fanselow (1981) feels obliged to postulate for German compounds should, therefore, come as no surprise. If, as Ford and Singh maintain, 'compounds' are locally anchored in specific constants, representational analogues of the popular bi-motic view of them must be rejected. They cannot, in other words, be assigned structures such as (1), where ÷ and y represent categories:

(i)

÷

y

Ë Ë

Dialogue m 267 The problem is that there is no justification for assigning any category to any of the terminal nodes: a two element 'compound' is ONE word and NOT three words. Consider the following paradigm from Modern Hindi: (2) ghODAgADi UnTgADi bElgADi bhEnsAgADi

'horse-carriage' 'camel-carriage' Ox-carriage' 'water-buffalo carriage'

There is a rather large number of nouns, and it can be extended, that can occupy the place of the first element, the variable in the examples in (2), not surprisingly because that is what 'the place of the variable' means; equally unsurprisingly, there is not much that can take the place ofgADi, the constant. Notice that facts of this kind find a straightforward expression in the sort of morphological theory we have alluded to. Briefly, it expresses all word-formation processes with Word Formation Strategies that have the following shape: (3) /×/á 4+ /×/â

Where X and X are words, á and â are morphological categories (feature bundles), indicates a bidirectional relationship, and X is a semantic function of X, and ' represents the form-related difference(s) between X and X. The ' represents the constant involved in the morphological operation and everything else in the string is referred to as a variable. Given the 'minimalist' apparatus above, the facts of (2) are expressed as follows:

(4) Notice that (4) is not, as it indeed cannot be, different from (5) below, which expresses the generalization involved in what P ninians would call the suffixation of -ity (in words like banality): (5) /X/Adj./Xlti/N The idea is not to claim that -ity is a word of English, something morphologists like Lieber (1981 and 1992) come dangerously close to claiming, or thatgADi is not a word of Hindi but to claim that neither ghODA norgADi in the examples in (2) is a word of Hindi: neither can be accessed for any grammatical purpose, andgADi is in addition no longer commutable. By virtue of the fact that it is now a part ofXgADi, where it is no longer commutable, gADi has not only lost its syntactic quality of 'wordhood' but may also be poised at the brink of losing its semantic independence. Notice that whilst the classification of bothghODA andgADi as stems may appear to cope with the 'compounding facts' of languages that have rich morphologies and must strip their words of what some call 'inflection' before

268 · The Yearbook (1999)

putting them in their 'compounds', it does little to answer the question: 'Why are there these commutability differences between the two stems?', a strong indication, we believe, of the dispensability of constructs such as 'stem'. Only two sorts of things are needed for complex words: variables and constants, as we said before.3 The fact that both 'roots' and 'stems' can function both as variables (cf. nomin· in nominate and dog in doghouse) and as constants (-ceive as in English receive and gADi in Hindi ghODAgADi) shows that the Päninian atoms of morphology don't quite stack up the way they should. The pasting can't work because the cutting seems fundamentally flawed. Be that as it may, we now turn to the fact that the semantics of 'compounds' requires what, following Quine (1961), could be calledpada-categorematicity: the 'meanings' of the elements that go into the making of 'compounds' are word-held. Used as a word in a Bangla sentence the sound sequence /pOtro/ means 'letter or document' in some contexts and 'leaf in others. However, when the sequence /pOtro/ occurs in the compounds (6a)-(6e), it contributes to these words only the 'letter' meaning: (6) a. b. c. d. e.

prempOtro proSnopOtro niyogpOtro uttorpOtro pOdottEgpOtro

'love letter' 'question paper' 'appointment letter' 'answer script' 'resignation letter'

In contrast, when it occurs in the compounds (7a)-(7e), /pOtro/ contributes only the 'leaf' meaning to these words: (7) a. b. c. d. e.

billopOtro nimbopOtro amropOtro SOptopornipOtro pOddopOtro

'wood-apple leaf' 'margosa leaf' 'mango leaf 'chatimleaf 'lotus leaf

Even under conditions that should enable a pragmatic override, a boy giving a girl a leaf to express unusual love simply will not call it a /prempOtro/ 'love leaf, and a document eulogizing a lotus-flagged political party simply cannot be called a /pOddopOtro/ 'lotus document'. No such pragmatic overrides exist. The full word /pOtro/ does retain its full range of meaning options when used in a true syntactic construction. And the word-fragment /pOtro/ in a 'compound' does get semantically constricted. The language compels it to contribute the 'letter' meaning to the 'compounds' (6a)-(6e) and the 'leaf meaning to (7a)-(7e). This compulsion can be expressed formally. We might propose the following WFSs, details negotiable, for these consistent patterns: (8) WFS for (6a)-(6e): /X/N,act/abstr /XpOtrO/N, act/abstr

Dialogue m 269 (9) WFS for (7a)-(7e): /X/N, botanical