The Yearbook of South Asian Languages and Linguistics: 2005 9783110186192, 9783110186185

South Asia is home to a large number of languages and dialects. Although linguists working on this region have made sign

226 83 3MB

English Pages 287 [280] Year 2005

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

The Yearbook of South Asian Languages and Linguistics: 2005
 9783110186192, 9783110186185

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

Introduction TANMOY BHATTACHARYA

1

Introduction

Some time at the beginning of the nine months that it took to produce this volume, one potential contributor mentioned that he could not work on a language other than his mother tongue. This ostensibly simple declaration quite apart from turning into a sociolinguist’s pedagogic fodder, at a personal level it brought back a range of academic memories for me. Not long ago, Neil Smith had inquired of me with uncharacteristic pointedness: Why Hindi and Bangla, why not Bangla and Welsh? in connection with a discussion we were engaged in, regarding adpositions, whereupon I did go and look at Welsh and that resulted in an interesting discovery and proposition.1 The lesson, however, brought home for me the stark reality of Universal Grammar (UG). Unquestionably, what we as syntacticians working within the context of South Asian languages have been participating in and responding to is the time-tested Boasian enterprise of the genealogical question. Over the intervening period, I have been intermittently revisiting the mildly disturbing thought that perhaps it is neither UG nor typology that can show us the path. Perhaps the mechanics of a particular language hinges upon quite something else.2 After every train journey that I take in India, I am convinced more and more that it is not the question of something being there or not in my or your language but rather the question of a system-specific heuristics within one particular language that we should be responding to.3 However, in order to pursue such a path, we need to first make those train journeys. This volume, to put a spin on the saying, takes you for such a ride. The destination, if I am right about a system-specific angle to language understanding, may not quite be accessible yet but the fact that this volume includes a paper on Sign Language, specifically on Indian Sign language (Aboh, Pfau and Zeshan, pp. 11–44), is an indication which unmistakably promises the possibility of such a future destination. If anything, it is the study of Sign language with a radical import for Interface designs

2

Tanmoy Bhattacharya

(Chomsky 1995) that most intimately aids our understanding of language as a system-internal, system-specific network arising out of Interface requirements. Hopefully, such examples will remedy the situation that most previous work on South Asian languages is beset with, relying for the most part, and for good reasons, on empirical findings and parametric conclusions. This volume is no exception in this regard; but then, the stops on the way bring into shape, into they not, our destination. Too often in the past, volumes on South Asian languages have either entirely bypassed or provided mere lip service to core syntactic phenomena, giving the implicit impression that those syntactic phenomena do not somehow matter for our languages. I am reminded of the absence of volumes especially on long-distance dependency structures in these languages. Ironically though, it was only such studies that provided the most significant breakthroughs in the progress of syntactic study of South Asian languages. The centre which led to the foundation that these breakthroughs are in some way or other based on is undoubtedly the Central Institute of English and Foreign Languages (CIEFL), Hyderabad. Many later dissertations on the syntax of a number of South Asian languages written within or outside South Asia drew inspiration from the early Ph.D. dissertations produced as a result of research conducted at CIEFL in the 1980s and the 1990s. The latter decade began with Mahajan’s MIT dissertation in 1990 on A/A 0 distinction and movement theory. A lot of syntactic research, and all the work on scrambling and binding, produced in that decade was a reaction to this work. To ignore this part of the history will be to belittle subsequent development in the field of syntactic studies on South Asian languages. Mahajan’s work and later work that it sparked o¤, in my view, are a culmination of the untiring work of the earlier generation of syntacticians. The tradition of typological, descriptive syntactic work on South Asian languages has had a healthy development and continues to play a significant part in our understanding of the empirical facts. The works of Alice Davison, James Gair, Peter Hook, Yamuna Kachru, Colin Masica, K.V. Subbarao and Boris Zakharin, among others, continue to inform more theoretical syntactic studies of the recent past and of today. Some of these scholars, in addition to unearthing rich empirical treasures, have contributed significantly to our understanding of basic workings of the Principles and Parameters model from the perspective of South Asian languages. It is no wonder then that the mixed bag that this volume deliberately represents, includes four scholars from the list above. I do not quite agree with the view that South Asian languages are not typically conducive to syntactic analyses and theories that have come out of the so-called West. In fact, I believe, some of us have been looking in the wrong place. Although there is a certain amount of legitimacy in terms of

Introduction

3

how much polemics can be found a rightful place within a supposedly scientific enterprise as ours, as well as how much of socio-political situation can be invoked to legitimize our (in)action, there has to be a modicum of accountability as far as taking the field forward is concerned. Any syntax writing, in this regard, must be useful for the reader. Those who view ‘usable’ as a dirty word and an excuse for reinstating a dominant canon of writing conventions in generative syntax need to outgrow such a morally stagnant viewpoint and participate in the conventions of writing in a way that is hopefully understandable for fellow linguists everywhere. The way syntactic theorization has progressed, I think, it is no longer advisable or even necessary to engage in polemics in the domain of syntax writing. The problem of absence of more studies on long-distance syntactic phenomena in the context of South Asian languages is not addressed in this volume either. Although the volume does not make the pretence of having resolved this problem, it nonetheless adds to the existing literature on South Asian linguistics in terms of how the students and researchers of this group of languages will be able to make use of the rich empirical reporting and theoretical demonstrations that are presented here. In fact, if I am not mistaken, this is the only volume of its kind which represents all the four language families of the region and a significant paper on Indian Sign language. It also has two papers on Sanskrit, one a diachronic/synchronic study on the occurrence of the injunctive in Vedic (Kiparsky) and the other a typological study of the strict ‘‘OV’’ character of Sanskrit with particular reference to Dravidian languages (Hock). 2

Summary of the Papers

In a way, the present volume is a first of its kind to present a serious syntactic study of clitics in the context of South Asian languages and Santali, an Austro-Asiatic/Munda language spoken mostly in central India. In Santali ‘Backernagel’ clitics: distributing clitic doubling (pp. 189–218), Kidwai examines Santali pronominal cliticization and claims that Santali represents a rare type of cliticizing language that imposes ‘‘a second to last’’ requirement on clitic placement. The incorporation of pronominal arguments and the distribution of a root complementizer, the author suggests and demonstrates, are inter-related syntactic phenomena. This attempt to connect two apparently unrelated syntactic events, as it were, to account for a particular phenomenon, is the essence of the new direction I mention at the beginning of this Introduction that South Asian syntax slowly but surely is taking and showing the way to the rest of the syntactic world. Kidwai locates the problem in clitic doubling rather than in clitic movement and argues for an adoption of Longobardi’s (1994) suggestion to

4

Tanmoy Bhattacharya

propose that the doubled clitic is the copy of the [number] feature of a pronominal in an N 0 position. More specifically, the subject or the EA clitic, as Kidwai calls it, attaches to the ‘‘wrong’’ host. Furthermore, she shows that the host and the clitic do not form a semantic unit. The proposal is couched within a Distributed Morphology (DM) analysis of ditropic cliticization. In DM, the Morphology module is located in the PF component that interprets the output of syntactic derivations. The paper concludes that Santali does not provide evidence that diatropic clitics exist, rather that these clitics are copies of number marking that fail to delete and have to be rescued by operations of a post-syntactic morphology component. In contrast, the paper by Hook and Zia (pp. 165–188) on a language of Gilgit in Northern Pakistan, Shina, presents an extremely rich descriptive account of a particular class of predicates (impersonal causative expressions) which show an alternation with respect to the verbal layer embedding them as well as with respect to the modifiers that they take. Furthermore, it presents the interesting phenomenon of the subject of these experiencer verbs being always null with a feminine gender feature. There are a number of such predicates of sensory perception in which an intransitive verb alternates with either a causative or a derived transitive: (1)

a.

baa´l taa´t-o bu´lu-n boy hot-M became-PresPerf ‘The boy is hot.’

b.

baa´l taa´t-i th-ar-ee´gi-n boy hot-F do-Caus-Pst3sgF-Pres3sg ‘The boy is (feeling) hot.’

The ending -ee´gi of the verb in the (b) variant is third person singular feminine regardless of the person, number, gender of the experiencer of the sensation. The presence of the causative morpheme in (1b) clearly provides surprising evidence for the presence of a verbal shell above the lexical verb, a possibility that Hale and Keyser (1993) convincingly argue for and the minimalist framework most famously adopts for the structure of the verb phrase (Chomsky 1995). Although Hook and Zia do not show it, an extension of the above findings can easily find a theoretical home, if required. Such a home is already found in Davison’s (pp. 83–116) analysis of phrasal predicates in Hindi/Urdu like the following: (2)

maiN-ne/ *mujhe un-kii yaad kii. I-erg *I-dat 3pl-gen-fs memory-fs do-pf-fs ‘I remembered, recalled them.’

Introduction (3)

5

mujhe/ *maiN-ne un-kii yaad aaii. I-dat *I-erg 3pl-gen-fs memory-fs come-pf-fs ‘I remembered/missed them.’

Davison shows how N and V contribute to the properties of the complex predicate and derives the right phrase structure, case and agreement properties from the analysis proposed. In particular, she compares two theoretical approaches, one based on incorporation, as in lexical syntax, the other based on predicate composition. The proposal to incorporate the nominal into the verb, whether null, overt or present as causative a‰xes, explains how N can be related in various ways to V. Already we see a theme emerging that brings these two papers (Hook and Zia and Davison) together; both are about the argument structure of the verb and the structure of the VP. The discussion of the causatives in both the papers is thus not accidental. On the absence of reflexive benefactives in Kannada (pp. 237–254), the representative paper on Dravidian languges by Lidz and Williams, is a fuller exploration of both the argument structure of the verb and a certain long-distance property, binding, that ditransitives in Kannada exhibit. In particular, ditransitives with benefactive morphology cannot have a reflexive indirect object, or have reflexive marking on the verb. The benefactive ditransitive is as follows: (4)

Hari Rashmi-ge pustaka-vannu kalisi-koTT-a Hari Rashmi-dat book-acc send-ben.pst-3sm ‘Hari sent a book to Rashmi.’

The paper thus gives a theoretical explanation for the fact of noncooccurrence of the reflexive and benefactive verb-auxiliaries in Kannada. The impossibility of cooccurrence is shown below: (5)

*Hari tann-age pustaka-vannu kalisi-(koNDu)-koTT-a Hari self-dat book-acc send-(vrm)-ben.pst-3sm

The authors argue that the benefactive structure, when its indirect object is a locally bound anaphor, is subject to two conflicting requirements. The anaphor must be local to its antecedent, and the direct object must be local to its Case licenser, but these conditions cannot be met in the same derivation. Thus reflexive benefactives are syntactically underivable. With respect to argument structure, the authors show that the Theme argument is generated higher than the Goal argument in the base structure generally, but not so in the case of the benefactives where the situation is just the reverse. Lidz and Williams thus touch upon a traditional problem that is gathering a renewed research interest in the syntactic studies on South Asian languages.4

6

Tanmoy Bhattacharya

Similarly, the paper on finiteness and control in Sinhala by Gair (pp. 117–144) not only raises important questions with regard to the traditional notion of finiteness, it also finds a way to relate a so-called head property (tense/finiteness) to long-distance dependency relations like control and binding. Gair is particularly reluctant to wholeheartedly endorse allegiance to a particular theoretical framework. However, given that control and binding facts form the core of the paper, it firmly ensconses his account in the principles and parameters tradition and at the same time presents a rich array of empirical findings. The interesting fact is the following. The distinction between finite and nonfinite verb forms does not seem to fall into neat compartments in Sinhala. Finite verb forms can occur with overt case marking; whereas non-finite (participial) forms can occur independently. Specifically, the paper deals with verbal inflection and features of phonologically unrealized subjects and control in non-finite structures in relation to infinitives. Further, it is shown that overt subjects surfacing in infinitival clauses seem to be an areal phenomenon: Malayalam seems to have it. Similarly, the Nominative versus Dative case alternation in pulawani and oona constructions is also interesting, exhibiting another parallel between Sinhala and Malayalam. Gair concludes that ‘‘finite’’ and ‘‘non-finite’’ do not isolate sets of forms in the language with sets of properties not otherwise accounted for. The paper by Dasgupta (pp. 45–82) takes us to the realm of ‘‘discourse’’ Interface from the observations of Bangla variable word order in negated sentences. The paper uncovers a set of data that crucially impinges upon our understanding of the system-internal network of forces I mention at the beginning. The data dealt with in the domain of past subjunctives (glossed as pa-sbj here) is as follows. First, note that an a‰rmative clause like (6a) or a simple negative clause like (6b) cannot support the particle ba (glossed here as mprt for mood particle): (6)

a.

*Dilip baadaam khacche ba Dilip nuts eat.pr-prog mprt

b.

*Dilip baadaam khaacche na ba Dilip nuts eat.pr-prog Neg mprt

The two ways to rectify them are to either alter the tense-aspect-mood information of the verb ‘eat’ by making it past subjunctive (7a) or by inserting an interrogative kæno (7b): (7)

a.

Dilip baadaam na-i ba khelo Dilip nuts Neg-emph mprt eat.pa-sbj ‘So what if Dilip doesn’t eat nuts’

Introduction b.

7

Dilip baadaam khaacche na-i ba kæno Dilip nuts eat.pr-prog Neg-emph mprt why ‘Why indeed is Dilip not eating nuts?’

The interaction of subjunctivity with the use of the modal particle ba forms the core of the paper. The theoretical manoeuvres assume the presence of a feature of Mood-Salience. Dasgupta takes (7) to mean that Past Subjunctive and Interrogatives share some property, called Mood-Salience. The Past Subjunctive is an M-Salient TAM, Interrogative K-words are MSalient Wh-words. Formally, the M-Salient feature covertly moves to a clausal functional head M, with interpretive consequences. One such consequence is that all and only M-Salient items can, via the M head of a finite clause, license the Modal Particle ba observed in (7) above as a companion of the Past Subjunctive. Thus the paper boldly asks the question: How are aspects of clause structure that lie at the interface of discourse integrated? Given all the current indications, not least among which is Platzack’s (1999) proposal of a Discourse Form in his list of multiple Interfaces, the paper throws open the issue of Interface designs and its e¤ects on our understanding of a systeminternal network. To return to the Sanskrit papers, Kiparsky’s paper entitled The Vedic Injunctive: Historical and Synchronic Implications (pp. 219–236) examines the status of a verb-form, the Injunctive, in early Vedic Sanskrit that poses a problem for the classification of Vedic verbs in terms of the inflectional categories of tense and mood, both diachronically and synchronically. Kiparsky demonstrates that this form is not unmarked for tense and mood but is underspecified, and therefore incompatible with a Paradigmbased theory of morphology but compatible with morpheme-based, nonrealizational morphological theories. In particular, the competition between unspecified and specified verb forms is modelled in OT implementations of such theories. As for the disappearance of the Injunctive in post-Vedic language, he suggests that immediately after the Rigvedic period, tense and mood were grammaticalized as obligatory inflectional categories which triggered the loss of the Injunctive. The story of the Injunctive in early Vedic in terms of the morphological/ syntactic headedness character of tense and mood overlaps with the story of finiteness in Sinhala (Gair, pp. 117–144) and Dasgupta’s conjectures about the mood particle ba in Bangla (pp. 45–82) as the use of the Injunctive alternates with moods and tenses. The issue of finiteness is also brought up in Hock’s paper entitled How strict is strict OV? A family of typological constraints with focus on South Asia (pp. 145–164). It focuses on the issue of ‘‘strict OV’’ typology

8

Tanmoy Bhattacharya

with particular reference to the Dravidian languages, with additional consideration of Indic and Tibeto-Burman word order typology. Hock questions various diagnostics to determine what is meant by ‘‘Strict OV’’ typology— specifically what is meant by finiteness—and appeals for a more careful comparative study of Dravidian languages including relatively neglected north Dravidian languages. He shows that the syntactic behavior of southern Dravidian languages has changed historically, which raises questions about what is meant by ‘‘Strict OV’’ typology. The paper brings out the relation between the various embedding strategies in the so-called OV languages and the place of either a quotative marker or a relative-clause embedding clitic which is a grammaticalized form meaning ‘or/ also/ and’ in typologically and geographically disparate languages such as Burushaski, Basque, and South Dravidian languages. Yet another typologically driven study is the paper entitled Internally Headed Relative Clauses in Sema (pp. 255–272) by Subbarao. It is also our only representative paper from the Tibeto-Burman group of languages. The paper provides typological evidence in support of the occurrence of an NP as the head of a head internal relative clause in Sema. The evidence is based on word order, case marking on the internal head, scope of adverbs, and absence of the nominative marker on the internal head. The example in (8) is of a head internal relative clause in Sema where the head occurs internal to the embedded relative clause and is a constituent of that relative clause just as in some other Tibeto-Burman languages such as Angami, Mizo and Hmar and unlike any Indo-Aryan language: (8)

[nO - nO timi lau yesI (pesI) tsI - ke- u] ye you [þTR] person to letter ACC give NOZ DEF [TR] khusuwo tall/long i. ii.

‘The person you gave the letter to is tall.’ (IO as head) ‘The letter you gave to the person is long.’ (DO as head)

Furthermore, the ambiguity concerning the relativizable head is accounted for in terms of the hierarchical/linear ordering of arguments in the clause. To come full circle, let me end the summary of papers included in the volume by looking briefly at Aboh, Pfau and Zeshan’s important contribution entitled When a wh-word is not a wh-word: The case of Indian Sign Language (pp. 11–44). The paper ably discusses Wh-questions in Indian Sign Language (ISL). I find their non-theoretical introduction to sign languages and deaf communities in general and in India in three separate sections especially important and useful. Sign languages (or at least some of them) challenge the universal position of the Wh element in the left periphery of the clause at some level

Introduction

9

of representation (that is, either overtly or covertly)5. Across languages, Wh-questions make use of a combination of a manual and a non-manual marker. Thus they di¤er from each other with respect to their question word paradigms, the syntactic position of Wh-signs, and the nature of nonmanual marking. In terms of the position of the Wh-sign, the most common are clause-initial, clause-final, or a combination of both, that is, a doubling of the Wh-sign. It is also common for a sign language to make use of more than one of these options. ISL is typologically unusual because of the strict syntactic constraints on the placement of the Wh-sign exclusively clausefinally. Thus ISL provides an argument for a rightward [Spec, CP] position. Aboh, Pfau and Zeshan also chalk out an alternative analysis of the Wh particle as a clause-typing particle that encodes the head of an interrogative phrase within the C-system and attracts the relevant Wh-sign within its specifier. However, the picture is complicated due to the availability of Whexpressions. 3

Finally

Many regular readers of this Yearbook will remember that in 2000, with much enthusiasm, I had embarked upon a project of getting a regular journal on South Asian Linguistics started. I still dream about it. For obvious reasons, therefore, I am grateful to Raj Singh for encouraging me to at least partially fulfil that dream by guest editing the current volume. I am also grateful to all of the contributing authors, who with their eagerness, made the editing of this volume more than a genial enterprise. The frequent power-cuts, the jammed printer, the dingy cyber cafe´, were all made a little more bearable by the arrival of an email from yet another co-operative author magically bringing about such conviviality that working became a pleasure. I was amazed and inspired by the promptness with which all the referees responded to the papers sent to them with very short deadlines. Finally, that wonderful bunch, my research students at the University of Delhi, sympathized with my state of mind towards the end of the editing period and greatly helped by keeping away and by waiting patiently with their dissertations, proposals and chapters. Haobam Basantarani, on the other hand, thankfully, helped with the cross-checking of references, by being at hand. Notes 1. Interested readers may wish to look at Bhattacharya (1999) for a syntactic account of the presence of prepositional agreement in Welsh and some other V-initial languages within the framework of Antisymmetry (Kayne 1994).

10

Tanmoy Bhattacharya

2. A fuller elaboration of these ideas and their implementation may be found in Bhattacharya (2005a, b). 3. See Bhattacharya and Thangjam (2004) for a system-specific rationale for cleft questions in Meiteilon. 4. See particularly Bhattacharya and Simpson (2004) and Malhotra (2005). 5. But see Simpson and Bhattacharya (2003) for a challenge to the strict typological divisioning of the position of the Wh expression.

References Bhattacharya, Tanmoy 1999 The structure of the Bangla DP. Ph.D. diss., Department of Phonetics and Linguistics, University College London. 2005a India not as a Linguistic Area: An insight from Minimalism. Talk delivered at Prof M.B. Emeneau Centenary International Conference on South Asian Linguistics, CIIL, Mysore. 2005b The Myth of Areal Linguistics: A view from the Interfaces. Invited talk at the Workshop on the syntax and acquisition of Asian languages, Nanzan University, Nagoya. Bhattacharya, Tanmoy and Andrew Simpson 2004 The position of the goal argument. Ms. School of Oriental and African Languages (SOAS), University of London. Bhattacharya, Tanmoy and Hindustani Devi Thangjam 2004 Why cleft? In Proceedings of the 23 rd SALA, Doug Bigham, Mark Brown, Q Wan Kim and Sadia Rahman (eds.). Stanford: CSLI Publications. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. Hale, Kenneth and Samuel Jay Keyser 1993 On argument structure and the expression of syntactic relations. In Kenneth Hale and Samuel Jay Keyser (eds.), The View from Building 20: Essays in Honor of Sylvain Bromberger, 53–109. Cambridge: MIT Press. Kayne, Richard 1994 The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Longobardi, Giuseppe 1994 Reference and proper names: A theory of N-movement in syntax and Logical Form. Linguistic Inquiry 25(4): 609–665. Mahajan, Anoop 1990 The A/A-bar Distinction and Movement Theory. Ph.D. Dissertation, MIT. Malhotra, Shiti 2005 Asymmetry of Objects and the Larsonian VP-shell for Hindi/Urdu. M.Phil. diss., Department of Linguistics, University of Delhi. Platzack, Christer 1999 Multiple interfaces. In Conceptual Structure and its Interfaces with other Modules of Representation, U. Nikanne and E. van der Zee (eds.). Oxford: The Oxford University Press. Simpson, Andrew and Tanmoy Bhattacharya 2003 Obligatory Wh-movement in a Wh-in-situ language. Linguistic Inquiry 34(1): 127– 142.

When a Wh-word is not a Wh-word: The Case of Indian Sign Language ENOCH O. ABOH, ROLAND PFAU and ULRIKE ZESHAN

Wh-questions in Indian Sign Language (IndSL) have a number of intriguing characteristics. First, IndSL allows only for the clause-final placement of a wh-sign. Second, sign languages typically have at their disposal a full paradigm of monomorphemic wh-words and make little use of complex whelements. In IndSL, however, the general wh-sign (G-WH) optionally combines with various phrases to express di¤erent meanings. Third, these complex wh-expressions allow for wh-split in the sense that the associate phrase can remain in situ. We first consider possible derivations—involving leftward or rightward movement of a wh-phrase—that are essentially similar to analyses brought forward for American Sign Language, another sign language that allows for sentence-final wh-elements. We show that these accounts are faced with serious problems when applied to IndSL and we propose an analysis that di¤ers in important respects from those proposed for ASL. Our analysis relies on the assumption that G-WH, unlike wh-signs in other sign languages, is not a wh-phrase but rather a wh-question particle that encodes the head of an interrogative phrase InterP. Depending on the presence or absence of an associate phrase, either the whole Focus Phrase dominated by Inter or the associate phrase are subsequently moved to SpecInterP. Support for this analysis comes from the distribution of non-manual question markers (facial expressions) and from the availability of other sentence-final clause-typing particles with which G-WH forms a paradigm. 1. Introduction1 Based on a wealth of typological data, researchers have come up with intriguing generalizations concerning the structure of wh-questions in spoken languages. In particular, it has been observed that cross-linguistically, whwords either remain in their base position (as, for instance, in Japanese) or appear in a sentence-initial position (as, for instance, in English). Based on this observation, scholars working within the Generative Grammar framework have proposed that SpecCP, the landing site of wh-movement

12

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

is located at the left edge of the clause universally. Wh-phrases must raise to this position, either overtly (English) or covertly (that is, at the level of Logical Form, as in Japanese) in order to check a [þwh]-feature in C. More recently, sign languages have entered the stage. Surprisingly, some sign languages seem not to obey the above generalization in that wh-elements may appear in a sentence-final position, even if this is not their underlying position. This is a striking finding, at least under the assumption that the representations and rules as made available by Universal Grammar are modality-independent. In this chapter, we will discuss wh-questions in Indian Sign Language and we will also compare the patterns to those described for American Sign Language. Before turning to the sign language data, however, a few words have to be said about sign languages and deaf communities in general (sections 1.1. and 1.2.) and about Indian Sign Language and the Indian deaf community in particular (section 1.3.). 1.1. Sign languages and other forms of manual communication Sign languages are visual-gestural languages that are used in communities of deaf people. They are often confused with other kinds of manual communication, and this confusion is responsible for some of the prejudices against and misconceptions about sign languages that were prevalent before the emergence of sign language linguistics and that are still not entirely eradicated. Unlike the co-speech gestures used by hearing people, sign languages are full-fledged, natural human languages with complex structures at all levels of linguistic organization.2 Just like spoken languages, sign languages di¤er from one another in their grammatical structures. The common misconception that sign language should be universal across the world is based on the confusion between signs and gestures. The grammatical structures of sign languages are entirely independent of and di¤erent from the spoken languages used in the same region, and there is no necessary one-to-one correspondence of sign languages and spoken languages in a particular region. For instance, Indian Sign Language is used all over India and even in neighbouring Pakistan. In the same region, a large number of di¤erent spoken languages are used that belong to di¤erent language families and sub-families. It is a common misunderstanding that sign languages are based on the spoken languages used in the same region and are nothing more than a representation of a spoken language ‘‘on the hands’’. Another form of manual communication that needs to be distinguished from sign language is home sign. If gesture is the most rudimentary form of manual communication, then home sign is somewhere in between gesture and a full-fledged sign language. Home sign develops, mostly at the level of

When a Wh-word is not a Wh-Word

13

the family, when a deaf person grows up in isolation from a deaf community. This often happens in villages in India because of a lack of access to the sign language communities in urban centres. An individual deaf person’s attempt at communicating on an ad hoc basis with family members and other immediate contacts gives rise to home sign, which uses gestures and idiosyncratic signs, but does not have the same linguistic status as a primary sign language used in a deaf community. The prejudice that sign languages are more ‘‘primitive’’ than spoken languages is sometimes based on a confusion between the sign languages of deaf communities and the home sign systems of isolated deaf people. 1.2. Deaf communities as linguistic and cultural minorities In recent decades, there have been two quite di¤erent views of deafness—a medical paradigm and a cultural paradigm. In the medical view, deafness is a disability that should be treated and, if possible, cured. The cultural paradigm, however, considers deaf communities to be linguistic and cultural minority groups that should have the same status as other minority groups (e.g. Padden and Humphries 1988, 2005; Erting et al. 1994; Ladd 2003). In support of the latter view, the following criteria are important: – The group has its own language, the regional or national sign language. – There is regular in-group interaction between the members of the group, for example during deaf sports competitions, religious services in sign language, and the like. – The group has its own institutions, such as deaf associations. – There are shared collective experiences and values within the group, such as experiences of linguistic oppression and positive attitudes towards sign language. – The group has its own norms of communication, its own history and cultural heritage, and/or its own art forms. All of these factors are true of Indian Sign Language and the Indian deaf community, and it is therefore clear that Indian Sign Language is a minority language used by a minority cultural group of deaf Indians. 1.3. Indian Sign Language Indian Sign Language (IndSL) has a unique linguistic status among the languages of India because it is the only indigenous pan-Indian language. IndSL is used in the deaf community all over India, and across the border in Pakistan. Although this is not clear at the present stage of research, it does seem as if the same sign language is also used in other parts of the Subcontinent, for instance, in Nepal, Bangladesh, or Sri Lanka (Woodward 1993).

14

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

The deaf community in India is mostly based in the urban centres, where deaf people come together in educational institutions, deaf clubs and associations, and social gatherings. IndSL has a substantial number of dialects that are based on regional di¤erences rather than on sociolinguistic or ethnic factors. Regional dialects of IndSL are mutually intelligible in face-toface communication. All varieties essentially share the same grammar, with very little variation across regions. By contrast, the vocabulary of IndSL is less stable, may change more rapidly over time, and varies to some extent across regions. On average, regional dialects of IndSL share about 75% of their vocabulary. Multidialectism and familiarity with several regional vocabulary variants is widespread in the Indian deaf community. No formal census data are available to account for either the total number of deaf and hard of hearing people in India or the total number of IndSL users. The number of sign language users has been estimated to be around 1.5 million people (Vasishta, Woodward and Wilson 1978). IndSL so far has no o‰cial status in India. Few people outside the deaf community even know about its existence. Unlike many minority spoken languages in India, IndSL has not been o‰cially recognized in any form. However, IndSL is by now a well-documented language. A sketch grammar (Zeshan 2000a, 2003a) and a reference grammar (Zeshan 2000b) describe the linguistic structures that make up the grammar of IndSL on the levels of phonology, morphology, syntax and discourse. IndSL vocabulary has been documented for several regions (e.g. Vasishta, Woodward and deSantis 1980; Ramakrishna Mission Vidyalaya 2001). Research articles have focussed on individual aspects of the structure of IndSL (Zeshan 2001, 2003b), and Miles (2001) has documented the history of deafness and gestural communication in South Asia (of which sign language is a small, but important part) from antiquity through modern times. Available language resources also include video-based course materials for teaching IndSL as a second language. IndSL is indigenous to the Indian Subcontinent and is not related to any other known sign language. A minor influence from British Sign Language can be seen in the use of a two-handed manual alphabet (fingerspelling) for representing English words. However, the lexicon and grammar of British Sign Language is very di¤erent from IndSL. No manual alphabet for representing indigenous Indian languages is widely in use. The sign language exists in a situation of language contact with the spoken languages of the region, but, since most deaf signers have limited competence in the spoken/written languages of the region, the influence on linguistic structures of IndSL is rather minimal. No contrived sign system is in use to represent a spoken language in the manual modality, such as the various systems of ‘‘Manual Coded English’’ do in some other countries. Very few schools for the deaf make systematic use of any form of manual communication in teaching situations.

When a Wh-word is not a Wh-Word

15

IndSL shares a number of structural properties with other sign languages, such as the use of the sign space for verb agreement, or the use of complex signs with numeral incorporation. Other structures, however, are quite different from what can be found in the better-known sign languages. The structure of interrogatives in IndSL is particularly interesting in this respect. This chapter is organized as follows: in section 2, we will familiarize the reader with properties of wh-questions in sign languages in general and in American Sign Language in particular. In section 3, we are going to introduce the relevant data from IndSL. We will first have a look at word order (section 3.1.) and then at the realization of wh-questions in this sign language (sections 3.2. and 3.3.). To anticipate the most important observation: IndSL has only one general wh-sign which always appears in sentencefinal position. We will discuss and compare di¤erent analyses for the observed patterns in section 4. The first line of reasoning assumes that the general wh-sign is a wh-phrase which is moved to a specifier position within the outer functional layer (section 4.1.). According to the second approach, however, the general wh-sign is a particle that is base generated in a functional head (section 4.2.). Section 5 summarizes our main findings. 2. Wh-questions across Sign Languages Before turning to the realization of wh-questions in IndSL, we shall first have a look at properties of wh-questions in sign languages in general (section 2.1) and in American Sign Language (ASL) in particular (section 2.2). It will turn out that across sign languages, wh-questions are only superficially similar in that they make use of a combination of a manual and a non-manual marker. On closer inspection it turns out that sign languages differ from each other with respect to their question word paradigms, the syntactic position of wh-signs, and the precise nature of non-manual marking. 2.1. Manual and non-manual marking of wh-questions All known sign languages mark wh-questions both manually and nonmanually. Manual marking of wh-questions involves question words, and sometimes question particles. Non-manual marking involves facial expressions, head movements and body positions. With respect to the structure of wh-questions, there are striking di¤erences between di¤erent sign languages. In a typological study across more than 30 sign languages around the world, Zeshan (2004) reports on question word paradigms, the syntactic position of question words, and non-manual marking of wh-questions. Firstly, cross-linguistic comparison shows that sign languages vary widely with respect to the size and internal organization of their question word

16

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

paradigms. Some sign languages, including the better-documented European and North American sign languages, but also East Asian sign languages, have fairly large question word paradigms. Large question word paradigms may have an internal organization with groups of question signs sharing a common phonological feature, such as the use of finger wiggling for various interrogatives of quantity in Chinese Sign Language. Others have fewer question words, which can sometimes be combined with other non-interrogative signs to express specific interrogative meanings. IndSL is particularly striking in that respect in that it has a minimal question word paradigm with, in some dialects, only a single question word. This pattern is quite exceptional across sign languages. Sign languages also di¤er from each other with respect to the syntactic position of the wh-sign, the most common positions being clauseinitial, clause-final, or a combination of both, that is, doubling of the whsign. It is common for a sign language to make use of more than one of these options. For some sign languages, it has been reported that the whsign may also remain in situ (e.g. ASL; cf. section 2.2.). Again, IndSL is typologically unusual because of the strict syntactic constraints on the placement of the wh-sign, which is exclusively clause-final (cf. section 3.2 for details). In all known sign languages, non-manual activities (facial expressions, head and body movements) represent an integral part of the grammar. At the level of syntax, they are particularly important for marking clause types, such as questions, negatives, topicalizations, and conditionals. While the function of such non-manual behaviors is similar across sign languages, their realizations and the particular constraints applying to their use can be quite di¤erent in di¤erent sign languages. IndSL, for instance, canonically marks wh-questions by raised eyebrows and a backward head tilt, but other sign languages use quite di¤erent patterns, such as furrowed brows and a forward head position (e.g. ASL), or even a tense headshake, as observed in Turkish Sign Language. It has been argued that non-manual markings in sign languages function similarly to intonation in spoken languages: both are suprasegmental and allow for spreading over a variable number of words/signs in the clause (Sandler 1999; Wilbur 2000; Pfau 2002). All manual signs that co-occur with a particular non-manual behavior are said to fall under its scope, and di¤erent sign languages have di¤erent restrictions with respect to the scope of non-manual markers. In addition to the simultaneous marking of wh-questions both manually and non-manually, many sign languages (amongst which IndSL) allow for an additional realization of wh-questions where no wh-sign is present and the wh-question is expressed through non-manual marking only. Such structures, however, are restricted to contexts where the target meaning of

When a Wh-word is not a Wh-Word

17

the wh-question is recoverable from the context, or is intended to remain vague. 2.2. Wh-movement in American Sign Language For ASL, wh-questions have been described and analyzed in some detail. In this section, we will summarize the main findings since this will provide a convenient starting point for our analysis of wh-questions in IndSL. In the literature, there has been some disagreement about wh-questions in ASL and the proper analysis thereof. While some scholars have argued that wh-movement in ASL proceeds leftwards to SpecCP, others have tried to show that this is not the case and that SpecCP is on the right in ASL, thereby making a strong argument against the claim that wh-movement proceeds leftwards universally. In this section, we will summarize the main arguments for the competing analyses. In section 4.1, we will present a rightward and a leftward analysis, respectively, for wh-questions in IndSL. Word order in ASL is underlyingly SVO. Other word orders which, in fact, are frequently observed have been argued to result from the presence of CP-external constituents that may adjoin to the left or right of CP, such as topics (Liddell 1980; Aarons 1994), pronominal right dislocation, possibly in combination with pro drop (Padden 1988; Neidle et al. 2000), and sentence-final tags (Neidle et al. 2000).3 Wh-signs, too, may occupy a non-canonical position. There is consensus in the ASL literature that wh-signs may appear either in situ or in a sentence-final position. In the examples in (1), this is only shown for a whobject (Petronio and Lillo-Martin 1997: 26¤ ) but wh-subjects may also appear in sentence-final position. In (1a), the underlying declarative sentence is given. In (1b) the wh-phrase WHAT is in its post-verbal base position. Note that the authors (as well as Neidle et al. 2000: 111) use a TP-final adverbial to demonstrate that the wh-object is not in situ in (1c). Crucially, a non-wh noun phrase cannot appear in a position following the adverbial (1d) (Petronio and Lillo-Martin 1997: 37).4 (1)

a.

JOHN BUY BOOK YESTERDAY ‘Yesterday John bought a book.’

b.

JOHN BUY WHAT YESTERDAY ‘What did John buy yesterday?’

c.

JOHN BUY YESTERDAY WHAT ‘What did John buy yesterday?’

wh

wh

d.

*JOHN BUY YESTERDAY COMPUTER

18

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

There is also consensus in the literature about the fact that wh-elements can be doubled, i.e. that they may occupy a sentence-initial and a sentence-final position. Actually, doubling is a very common strategy in ASL. Below, this is shown for a wh-object (2a) and a wh-adjunct (2b), respectively (Petronio and Lillo-Martin 1997: 32f ). wh

(2)

a.

WHAT JOHN BUY YESTERDAY WHAT ‘What did John buy yesterday?’ wh

b.

WHY STUDY LINGUISTICS WHY ‘Why do you study linguistics?’

Let us now have a look at the competing analyses that have been o¤ered to account for the above data. Proponents of the leftward analysis (LA: Lillo-Martin 1990; Petronio 1993; Petronio and Lillo-Martin 1997) claim that SpecCP and wh-movement is leftward universally. In ASL, however, wh-movement is optional which accounts for the grammaticality of (1b). Obviously, the LA is challenged by examples such as the one in (1c). For this and similar constructions, the LA assumes that a null whelement (e) has been moved to SpecCP to check the [þwh]-feature in C. The final wh-element in (1c) as well as the sentence-final doubles in (2) are claimed to be base-generated doubles that occupy the sentence-final head of CP and are licensed by a [þfocus]-feature in C (cf. the structure in (3a) below).5 (3)

a.

Leftward analysis (LA) CP SpecCP  WHAT/e

C’ IP

C° [+foc]/[+wh]  WHAT J. BUY t YESTERDAY base-generated double

When a Wh-word is not a Wh-Word b.

19

Rightward analysis (RA) XP WHAT

base-generated topic

CP C’

IP

SpecCP  WHAT C° [+wh]

J. BUY t YESTERDAY

In contrast to that, proponents of the rightward analysis (RA: Neidle et al. 1997, 1998, 2000; Neidle 2002) are convinced that wh-movement in ASL does not proceed leftwards, i.e. that SpecCP, the landing site for whmovement, cannot be assumed to be on the left universally. Similar to the LA, these authors assume that wh-signs may remain in situ in ASL; in contrast to the LA, however, they claim that the wh-sign in (1c) has been moved rightwards to a sentence-final SpecCP. According to the RA, constructions with doubled wh-elements involve a combination of rightward movement of a wh-constituent with a base-generated wh-topic (cf. the structure in (3b) above).6 An important argument for the RA comes from patterns of non-manual marking. Neidle et al. (2000) take the non-manual wh-marking to be an overt realization of the wh-feature in C. This non-manual marking needs manual material to be articulated with. In cases where wh-movement has resulted in the occurrence of a manual wh-sign in SpecCP, it is possible for the nonmanual marking to locally associate with the wh-sign (under Spec-head agreement), that is, to be co-articulated with the wh-sign only. In other words: with a sentence-final wh-element, spreading of the non-manual marking over the entire clause (as in (1c) above) is optional (4a) (Neidle et al. 2000: 113). The same option, i.e. wh-marking on the wh-sign only, is not available when the wh-sign remains in situ (4b). In this case, the non-manual marking has to spread over all manual material within its c-command domain (i.e. the entire CP, as in (1b) above) since there is no manual material available locally in SpecCP for the non-manual marking to be co-articulated with.

20

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan wh

(4)

a.

[TEACHER LIPREAD t i YESTERDAY]TP WHOi ‘Who did the teacher lipread yesterday?’ wh

b.

*TEACHER LIPREAD WHO YESTERDAY [þwh]C

While the (un)grammaticality of the ASL data in (1) and (2) is not disputed, there are also some constructions about which there is disagreement in the literature. First of all, according to the LA, a sentence with initial wh-object like the one in (5a) should be grammatical, while according to the RA this should be ungrammatical, since there is neither a wh-sign in situ nor in SpecCP (Lillo-Martin 1990: 214 versus Neidle et al. 2000: 110). Secondly, the RA predicts (5b) with a wh-object in situ and in sentence-final position to be ungrammatical, since actually, these two are one and the same element. Given that wh-movement is optional, the RA also predicts (5c) with a wh-object in situ and a sentence-initial wh-topic to be grammatical (Neidle et al. 1997: 261). The LA would make exactly the opposite predictions, but, to the best of our knowledge, the proponents do not discuss such examples. wh

(5)

a.

? WHO JOHN LOVE ‘Who does John love?’

b.

*JOHN BUY WHAT YESTERDAY WHAT ‘What did John buy yesterday?’

wh

wh

c.

WHO JOHN SEE WHO YESTERDAY ‘Who, who did John see yesterday?’ wh

d.

? BREAK-DOWN [WHO (POSS) CAR] ‘Whose car broke down?’

Finally, the RA explicitly claims that complex wh-phrases may appear sentence-finally, as in (5d), since SpecCP which is assumed to be on the right may host phrases. According to proponents of the LA, however, such structures are ruled out, since the final wh-element is analyzed as a double base-generated in C, a head position which may not host phrases (Neidle et al. 2000: 136 versus Petronio and Lillo-Martin 1997: 37).7 It is not our intention to take a strong position with respect to what is the correct or incorrect analysis for ASL. What is worth noting in conclusion of

When a Wh-word is not a Wh-Word

21

this section, however, is the fact that the competing accounts are motivated by the observation that wh-signs in ASL can in fact appear in various syntactic positions: in situ, sentence-finally, and sentence-initially (at least in the doubling constructions). In the next section, we are going to see that things are strikingly di¤erent in IndSL.8 3. Wh-questions in Indian Sign Language In this section, we are going to familiarize the reader with the relevant data from IndSL. Before turning to wh-questions, we will first have a look at constituent order in IndSL (section 3.1.). In section 3.2., we will show that the question word paradigm as well as the position of wh-signs in IndSL is much more constrained than what has been described for ASL (and other sign languages): there is only one general wh-sign which can only appear in sentence-final position. However, things get more intricate due to the availability of complex wh-expressions. The meaning and use of these expressions as well as the possibility of wh-split will be discussed in section 3.3. 3.1. Constituent order in Indian Sign Language IndSL is a verb-final language because the verb sign always appears at the right edge of the sentence (6a–c). Since IndSL, just like other sign languages, does not have a copula verb, sentence-final placement is also observed for adjectival (6d) and nominal (6e) predicates (Zeshan 2003a). (6)

a.

MAN INDEF WALK ‘Someone is walking.’

b.

APPLE CHILD EAT ‘A child eats an apple.’

c.

TOMORROW INDEX1 DELHI INDEX3 GO ‘I am going to Delhi tomorrow.’

d.

WOMAN SAD ‘The woman is sad.’

e.

WOMAN INDEX3 SERVANT ‘That woman is a servant.’

The order of arguments, however, is fairly free and is based on pragmatic factors. Just as in ASL, topicalization of constituents is a common strategy in IndSL. In contrast to ASL, however, topics are not obligatorily marked by a particular non-manual expression (6b). Frequently, topics are characterized by the very absence of a non-manual marker which accompanies the rest of the sentence (see example (15a) below). Zeshan (2003a) reports that sentences with more than one overt lexical argument are not very productive in IndSL. Rather, situations tend to be

22

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

split up into several propositions, each one containing one argument, or one of the participants is understood from the context. Sentences with one pronominal and one lexical argument are more frequently attested. This state of a¤airs suggests that IndSL includes null arguments (Rizzi 1986; LilloMartin 1986). There are very few signs that may follow the predicate in IndSL. All of these belong to the class of ‘‘functional particles’’ which assign a clause to a particular clause type and which have scope over the whole clause. Amongst these functional particles are the manual neutral negation marker NEG ((7a); cf. figure 1 below) and the completive aspect marker COMPL (7b). neg

(7)

a.

INDEX1 WORK NEG ‘As for me, I am not working.’

b.

YESTERDAY FATHER DIE COMPL ‘Yesterday (my) father died.’

Other sentence-final particles are a contrastive negative marker, two positive imperative markers, one negative imperative marker, and an existential sign (Zeshan 2003a: 164). See section 4.2.2. for further discussion of these particles.

3.2. Position of wh-signs in the clause As mentioned above, IndSL has a minimal wh-sign paradigm. In fact, there is only one non-compositional wh-sign, namely the general wh-sign G-WH (see figure 2).9 This sign is related to a co-speech gesture which commonly accompanies interrogatives in the hearing communities of India and Pakistan.

Figure 1. Negation sign NEG

Figure 2. General wh-sign G-WH

When a Wh-word is not a Wh-Word

23

As far as non-manuals are concerned, wh-questions in IndSL are marked by raised eyebrows and a backward head position with the chin raised (see figure 2). This is in contrast to ASL where wh-questions are marked by furrowed brows and squinted eyes. As in ASL, the eyebrow marking may spread, but the head/chin position is usually most pronounced on the manual wh-sign. As is shown by the examples in (8), the sign G-WH covers the whole range of question words in other languages. In this case, the interpretation of the wh-sign has to be inferred from the context. This kind of discourse-linked interpretation of signs (as well as of zero constituents) is frequently observed in IndSL. In order to express more specific meanings, IndSL has the option of combining G-WH with other non-interrogative signs (cf. section 3.3.). Note that in the examples below, we neglect the possibility of (partial) non-manual spreading (spreading patterns will be discussed in section 4). wh

(8)

a.

CHILD ANGRY G-WH ‘Why is the child angry?’

c.

INDEX3 COME G-WH ‘Who is coming?’

wh

b.

INDEX2 AGE G-WH ‘What’s your age?’

d.

INDEX2 GO G-WH ‘Where are you going?’

wh

wh

Interestingly, the placement of G-WH is much more constrained than the placement of wh-signs in ASL. In (9a), we give a base sentence with canonical S-O-V sign order. A striking contrast to the ASL data presented in (1) and (2) is that G-WH can only appear in sentence-final position (9b). That is, the wh-sign may neither remain in situ (9c), nor may it appear in sentence-initial position (9d) or be doubled (9e). Below this is only shown for a direct object but exactly the same distribution is observed for whsubjects and -adjuncts. (9)

a.

SEARCH FATHER INDEX3 BOOK ‘Father is/was searching for a book.’

b.

FATHER INDEX3 SEARCH G-WH ‘What is/was father searching?’

c.

*FATHER INDEX3 G-WH SEARCH

[in situ]

d.

*G-WH FATHER INDEX3 SEARCH

[sentence-initial]

e.

*G-WH FATHER IX3 SEARCH G-WH

[sentence-final]

[initial & final]

Remember that the ASL counterparts of (9c) and (9e) are claimed to be grammatical by proponents of the leftward and the rightward analysis.10

24

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

3.3. Complex wh-expressions and wh-split It may not always be the case that the context allows for an unambiguous interpretation of the general wh-sign G-WH. To express specific wh-words, IndSL signers may use composite expressions which consist of a combination of G-WH with an associate phrase. Common combinations are FACE G-WH (‘who’), PLACE G-WH (‘where’; cf. figure 3), TIME G-WH (‘when’; cf. figure 4), and NUMBER G-WH (‘how many’). However, this option is not available for ‘What’, ‘Why’, and ‘How’; these meanings can only be expressed by the general wh-sign G-WH alone.

Figure 3. PLACE G-WH ‘where’

Figure 4. TIME G-WH ‘when’

The examples in (10) illustrate the use of three of these composite whexpressions. Note that at present, the IndSL composite interrogatives do not show any sign of evolving into compounds. The two parts of the composite interrogatives are always clearly separate signs, that is, phonological reduction or assimilation phenomena which are typical for sign language compounds (Klima and Bellugi 1979) are not observed (Zeshan 2003a: 201). (10)

a.

INDEX2 FRIEND SLEEP PLACE G-WH ‘Where does your friend sleep?’

b.

INDEX3 ASK FACE G-WH ‘Who did s/he ask?’

c.

INDEX2 BOOK TAKE NUMBER G-WH ‘How many books will you take?’

Interestingly, while the wh-sign G-WH always appears sentence-finally, the associate phrase of the complex wh-expressions may remain in situ, i.e. we observe wh-split, as exemplified by the examples in (11).11 We take this to be further evidence for the assumption that composite interrogatives are phrasal and not compounds.12

When a Wh-word is not a Wh-Word (11)

a.

INDEX2 FRIEND PLACE SLEEP G-WH ‘Where does your friend sleep?’

b.

INDEX2 [BOOK NUMBER] TAKE G-WH ‘How many books will you take?’

25

In conclusion of this section, let us repeat the main characteristics that distinguish wh-questions in IndSL from those reported for ASL. First, we have seen that IndSL has a minimal question word paradigm: there is only one non-compositional question sign, the general wh-sign G-WH. This striking fact will be one of the motivations for the wh-particle analysis presented in section 4.2. Secondly, IndSL has a number of complex wh-expressions to express more specific meanings. Most of these complex wh-expressions allow for wh-split. Thirdly, the wh-sign—be it by itself or part of a complex wh-expression—always appears in sentence final-position and cannot be doubled. This observation makes a strong argument for the rightward movement analysis to be presented in section 4.1.1. 4. Accounting for the Data: wh-phrase vs. wh-particle In this section, we are going to present and compare three analyses for wh-questions in IndSL. The rightward and leftward movement analyses sketched in section 4.1. basically follow the reasoning outlined for ASL in section 2.2. In both these analyses, G-WH is treated as a wh-phrase which targets SpecCP. In contrast to what has been proposed for ASL, we don’t need to be concerned with the complex issue of wh-doubling since, as we have seen, this phenomenon is not attested in IndSL. The analysis given in section 4.2. di¤ers crucially from the above ASL proposals. Here we propose to analyze G-WH not as a wh-phrase but rather as a wh-particle which is base-generated in a functional head in the outer functional layer (Rizzi 1997, 2001). For all three proposals, it will be considered how they can account for sentences with the general wh-sign G-WH and for sentences involving wh-split. 4.1. G-WH as a wh-phrase 4.1.1. Rightward movement of G-WH Similar to the ASL data presented in 2.2., the IndSL data can be seen as a serious challenge to the assumption that SpecCP, the landing site of wh-movement, is universally on the left. In fact, the IndSL data make an even stronger argument for a rightward SpecCP since the wh-sign G-WH always and only appears sentence-finally. Crucially, the leftward analysis for ASL has been motivated and supported by the existence of sentence-initial wh-elements.

26

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

According to the rightward analysis, G-WH is moved to the sentencefinal specifier of CP. In sentences with a complex wh-expressions—glossed as [XXX G-WH] in (12) below—there are two options available. Either GWH is extracted from the complex expression, stranding the specifying element in its pre-verbal base position (option 1), or the whole complex moves to SpecCP (option 2). (12)

c-command domain of [+wh]

CP C’

IP SpecIP  subject DP  c [XXX ti] d ti

I’ VP

C  [+wh] I

V

SpecCP  c G-WHi d [XXX G-WH]i

non-manual marking under Spec-head agreement

Following Neidle et al. (2000), we assume that non-manual wh-marking is associated with a [þwh]-feature in C. Since in contrast to ASL, in IndSL, at least the wh-sign always moves to SpecCP, there is always manual material locally available for the wh-marking to be associated with. We therefore predict that it should be possible for the non-manual marking to be coarticulated with the wh-sign only. This prediction is borne out as has already been illustrated by the examples in (8) above. In the case of sentence-final complex wh-expressions, the pattern of nonmanual marking is somewhat less straightforward. Since the complex whexpression is hosted by SpecCP (option 2 in (12)), it is expected that the whole constituent is non-manually marked under Spec-head agreement. This, however, need not be the case, as is exemplified by the grammatical example in (13) (also compare figure 3 above). We therefore assume that it is su‰cient for the non-manual marking realizing the wh-feature to target part of the phrase in SpecCP only.13 wh

(13)

[INDEX2 FRIEND t i SLEEP [þwh]C [PLACE G-WH]i ]CP ‘Where does your friend sleep?’

When a Wh-word is not a Wh-Word

27

As in ASL, optional spreading of the wh-marking—at least the eyebrow marking—is possible. When spreading occurs it has to target the entire ccommand domain of C (14a), i.e. it may not spread over only part of the material contained under IP, as is illustrated by the ungrammatical example (14b). The same is true in ASL, as is shown in (14c–d). wh

(14)

a.

[[FATHER INDEX3 SEARCH]IP [þwh]C [G-WH]SpecCP ]CP ‘What is/was father searching?’ wh

b.

*[[FATHER INDEX3 SEARCH]IP [þwh]C [G-WH]SpecCP ]CP wh

c.

[[TEACHER LIPREAD t i YEST.]IP [þwh]C [WHOi ]SpecCP ]CP

d.

*[[TEACHER LIPREAD t i YEST.]IP [þwh]C [WHOi ]SpecCP ]CP

wh

The initial part of a clause, however, may be outside the scope of the nonmanual marking in case the respective constituent has been topicalized to a position above SpecCP, i.e. to a position outside of the the c-command domain of C. Zeshan (2003a: 199) points out that this is commonly the case in IndSL. In (15a), for instance, the subject occupies a position outside of CP. As mentioned before, topics in IndSL are not obligatorily marked nonmanually. This is in contrast to ASL, where a topicalized constituent—e.g. JOHN in (15b)—is always accompanied by a non-manual marker (Neidle et al. 1997: 268). wh

(15)

a.

[MANj ]TopP [[t j t i STAY]IP [þwh]C [G-WHi ]SpecCP ]CP ‘The man, why did (he) stay?’ t

b.

wh

[JOHNi ]TopP , [[YOU SEE t i ]IP [þwh]C [WHERE]SpecCP ]CP ‘John, where did you see (him)?’

In sum, the fact that wh-elements in IndSL always appear sentence-finally can be captured in a straightforward way by assuming rightward movement of such constituents. The analysis sketched above is quite similar to the account for ASL given in 2.2. In contrast to ASL, however, the IndSL account does not call for the assumption of base-generated sentence-initial wh-topics. Moreover, the fact that phrasal wh-constituents also commonly appear in final position supports the claim that the landing site is SpecCP

28

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

and not C, as has been claimed for ASL by proponents of the leftward analysis (Petronio and Lillo-Martin 1997). Finally, the distribution of nonmanual marking—i.e. optional spreading in the presence of a manual whsign versus obligatory spreading in its absence—can be seen as further evidence for the syntactic structure proposed in (12).

4.1.2. Leftward movement of G-WH Let us now consider how a leftward analysis could account for the IndSL data. The analysis given below di¤ers from the ASL account sketched in 2.2. in two important respects. First, in contrast to Petronio and Lillo-Martin (1997), we do not have to assume the presence of a base-generated double in C. Remember that this assumption was mainly motivated by the availability of wh-doubling constructions which, as we have seen, are not attested in IndSL. That the position of the wh-element in IndSL cannot be C is shown by the fact that complex whexpressions can appear in sentence-final position. The postulation of a sentence-final C head is somewhat problematic anyway: on the one hand, Petronio and Lillo-Martin try to account for the ASL data within a leftward movement scheme that has been claimed to be universal; on the other hand, however, the structure they propose does not follow Kayne’s (1994) antisymmetric Spec-head-complement scheme. Secondly, and related to the first point, we do not assume movement of an empty element to SpecCP— at least not when a manual wh-sign is present in the sentence.14 Obviously, the leftward analysis requires further movement operations in order to derive the surface sign order. To account for the sentence-final placement of G-WH, leftward movement of the wh-sign (or the complex wh-expression, respectively) to SpecCP has to be followed by remnant movement of the entire IP to a specifier position above CP, as illustrated in (16) for a sentence with a wh-object.

(16)

When a Wh-word is not a Wh-Word

29

While it is possible to derive the correct sign order based on this account— subject-verb-G-WH in (16)—the leftward analysis is faced with a number of problems. First of all, it is not clear what specifier position remnant IPmovement targets. Building on Petronio and Lillo-Martin (1997), Neidle (2002), and Nunes and de Quadros (2004), however, it could be suggested that remnant movement of IP past G-WH is triggered by focus operation. Secondly, the observed distribution of the non-manual marker cannot be accounted for in a straightforward way. Remember that frequently, the non-manual marker is articulated with the wh-sign G-WH only (8). This pattern can be explained when we assume that wh-marking is established at surface structure under Spec-head agreement, similar to what happens under the rightward analysis. However, the spreading facts—which are correctly predicted by the structure in (12)— pose a serious challenge for the leftward analysis. Crucially, in (16), the IP is outside of the c-command domain of the [þwh]-feature in C at surface structure. Therefore, examples such as the one in (14a) where the nonmanual marker accompanies the whole sentence can only be accounted for when we assume that non-manual marking is established before movement operations takes place.15 Such an account is confronted with serious problems. First, we are no longer able to account for the structures with non-manual marking on the wh-sign only. Once we accept that non-manual marking precedes movement operations, we predict that it is always all the manual material under IP that is non-manually marked. Secondly, as we have already seen, topicalized constituents may be outside the scope of the non-manual wh-marking (15a). Of course, in a structure like the one given in (16), following IP-movement, a constituent (e.g. the subject DP) could be moved to a specifier position above XP. We would then have to stipulate, however, that the non-manual marking on this constituent is deleted at surface structure. Thirdly, one looses the generally accepted generalization that nonmanual marking is a surface structure phenomenon, as has been argued for negative headshakes (Neidle et al. 2000; Pfau 2002; Pfau and Quer 2002), topic marking (Aarons 1994), and relative clause marking (Wilbur and Patschke 1999; Pfau and Steinbach, in press), amongst others. In sum, when comparing the rightward movement analysis to the leftward movement analysis, it appears that the former accounts for the IndSL data, in particular for the observed patterns of non-manual marking, in a more straightforward way. Again, we wish to stress that by reaching this conclusion, we do not imply that the same necessarily holds for ASL (or other sign languages with clause-final placement of wh-signs). After all, ASL, in contrast to IndSL, also makes use of sentence-initial wh-signs.

30

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

4.2. G-WH as a clause typing morpheme The previously discussed analyses presuppose that the IndSL general whelement G-WH is similar to English wh-phrases in nature in that it functions both as a wh-operator that ranges over a variable and as an interrogative clause typing element in the sense of Cheng (1997). However, this need not be the only logical or possible characterization. In this section, we provide cross-linguistic facts from spoken languages and evidence from IndSL which suggest that G-WH is better analyzed as a clause typing morpheme that encodes the interrogative feature in wh-questions. 4.2.1. Question particles in Lele and Indian Sign Language Comparative studies on spoken languages show that languages realize wh-questions in various ways. In Lele (an SVO East Chadic language), for instance, whquestions require a question particle that also surfaces in yes-no questions (see Frajzyngier (2001) for the discussion). Example (17) is a yes-no question that involves the question particle ga`. (17)

Kiya ha`b ku`lba´ ke-y ga`? Kiya find cow gen-3sg[m] inter ‘Did Kiya find his cow?’

The Lele wh-questions under (18) further indicate that those derive from a combination of the question particle and a wh-element that may either remain in situ or be fronted due to focusing. In example (18a) the object whelement we´y ‘who’ surfaces in situ and precedes the sentence-final question marker. In the examples (18b–c), however, the subject wh-element we´y and the object wh-element me ‘what’ appear in sentence-initial position where they are focus-marked by the focus particle ba, while the question particle ga` occurs sentence-finally. (18)

a.

Me` a`y we´y ga` 2sg[f] marry who inter ‘Who did you marry?’

b.

We´y ba e´ ga`? who fm go inter ‘Who went away?’

c.

Me ba gol dı´ ga`? what fm see 3sg[m] inter ‘What did he see?’

Setting aside the issue of the interaction between focusing and wh-fronting (Bosˇkovic´ 2002), what these Lele examples suggest is that languages

When a Wh-word is not a Wh-Word

31

may allow for a labor division between the clause typing morpheme (here the interrogative question particle) and the wh-element. More precisely, languages may allow a split between the interrogative feature and the whfeature proper. With regard to clause structure, this could be evidence that the wh-feature and the interrogative feature are not licensed by the same syntactic head, contrary to what is often suggested in the literature on the basis of English-type languages. For instance, the simultaneous realization of the focus marker and the question particle, in Lele, suggests that these two elements belong to di¤erent projections (i.e. FocP and InterP, see Rizzi 2001, Aboh 2004 for discussion). In this light, the discussed IndSL data suggest that this language uses a strategy that is parallel to the Lele wh-question strategy, and which can be represented as in (19) where YP is a wh-phrase that surfaces in situ or in the left periphery of the clause due to focusing. (19)

a. b.

. . . . . .YP. . . . . . . . . . . . . . . . .ga`. . . . . . . . .YP. . . . . . . . . . . . . . . . .G-WH. . .

[Lele] [IndSL]

Assuming this is the right characterization, an immediate conclusion that arises is that IndSL and Lele di¤er in that the former (not the latter) has null variants of wh-phrases. Put another way, we are now suggesting that in a sequence like (9b), repeated here as (20) for convenience, there is a silent hwhati in the object position (see below for a refinement of this analysis). (20)

FATHER INDEX3 hwhati SEARCH G-WH ‘What is/was father searching?’

This state of a¤air is not surprising within IndSL grammar since we have mentioned in section 3.1. that the language uses null arguments as long as they are unambiguously recoverable from the discourse. In this regard, it is interesting to note that in cases where the silent hwhi cannot be identified from discourse, IndSL resorts to wh-questions where the questioned argument or adjunct is overtly realized by the signs FACE, PLACE, NUMBER, or TIME, as has been illustrated in (10). Under this description, IndSL has a sentence-final particle G-WH (similar to Lele ga`) that is associated or combined with a null or overt wh-element to form the wh-question. We return to the analysis of these sequences later, but before getting on to that it is worth mentioning that IndSL manifests other clause typing morphemes together with which GWH forms a paradigm. These empirical facts provide the ground for the analysis of G-WH in terms of clause typing.

32

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

4.2.2. Clause typing in Indian Sign Language As mentioned in section 3.1., IndSL makes use of a cluster of particles including mood and negative markers that all occur in sentence-final position and clause-type the proposition. According to Zeshan (2000a: 97), these signs ‘‘have a relatively simple structure as compared to other signs’’, and form a closed class, two typical properties of functional items. The sentences under (21) illustrate those signs that are relevant for the current discussion (Zeshan 2000a: 95f ), but we refer the reader to Zeshan (2000a, 2003a) for a detailed description of the IPSL function signs. (21)

a.

YOU STUDY IMP ‘You have to study!’

[Imperative]

b.

INDEX1 TEA NEG ‘I haven’t had tea yet.’

[Neutral negation]

c.

INDEX1 TEA NEG2 ‘I don’t want any tea.’

[Contrastive negation]

d.

STUDY USEFUL EXIST ‘Education is really useful.’

[Existential]

Not all of these particles may co-occur, but when they do, they appear to follow a fixed order. For instance, the examples under (22) suggest that the negation sign NEG must precede the question particle G-WH. (22)

a. b.

MAN UNDERSTAND NEG G-WH ‘Why doesn’t/didn’t the man understand?’ *MAN UNDERSTAND G-WH NEG

These facts suggest that the IndSL clause typing morphemes manifest the same syntactic domain, which relates the proposition to the discourse and may encode interrogative, emphasis, mood, or negation. This, it seems to us, provides strong empirical grounds for treating these particles as functional heads. In addition, Zeshan (2000a, 2003a) observes that, to the extent that these function signs are associated with non-manual markers, they manifest similar scope properties. Their scope domain extends leftward from the right edge of the sentence. Put di¤erently, the non-manuals minimally scope over the sentence-final sign (quite often the sign with which they are associated), or on successively bigger constituents with the exclusion of topics. A consequence of this scope pattern is that the whole proposition (or clause) may be within the scope of the non-manual. As we suggest in the following paragraph, the scope properties of the non-manuals can be accounted for if we

When a Wh-word is not a Wh-Word

33

assume that the function signs are functional items that head distinct projections within the clausal left periphery and attract some material into their specifiers, with the non-manual signaling the Spec-head relation that is established between the attractee in the specifier position and the attracting head. This obviously raises the issue of why these clause-typing morphemes always occur sentence-finally. 4.2.3. Head final versus head initial approaches to Indian Sign Language In a traditional GB framework that adopts the directionality parameter, an apparently straightforward way to account for these facts would be to say that IndSL is a head-final language in the sense that heads take their complements to the right. Under Rizzi’s (1997) split-C hypothesis, this would mean that the wh-question particle realizes an interrogative projection (InterP) within the left periphery, which projects between ForceP and FinP and whose complement occurs to the left. If we further assume, for the sake of argument, that specifiers within the left periphery occur to the right, then we can account for wh-questions in IndSL by saying that SpecInterP hosts a null operator that is identified by the null wh/pronominal element that sits in situ as schematized in (23). (23)

Assuming that wh-elements are formally licensed under Spec-head configuration (Rizzi 1996, 1997), we may further conclude that the wh/pronominal element must move to SpecInterP at LF to be interpretable. This analysis can be further grounded on the fact that occasionally, G-WH co-occurs with a PLACE/TIME/NUMBER expression that surfaces in situ or immediately to the left of G-WH. With regard to the in situ cases (24a), one may propose that the wh/pronominal element moves at LF. As suggested before, the surface string can then be represented as in (24a 0 ). With regard to cases

34

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

where the PLACE, TIME, or NUMBER sign surfaces to the left of G-WH (24b), we argue that the displaced PLACE/TIME/NUMBER sign surfaces in the specifier of a focus projection that projects below InterP (see Petronio and Lillo-Martin 1997; Rizzi 2001; Bosˇkovic´ 2002; Nunes and de Quadros 2004; Aboh 2004) as illustrated in (24b 0 ). (24)

a.

INDEX2 FRIEND [PLACE] SLEEP G-WH ‘Where does your friend sleep?’

b.

INDEX2 FRIEND t i SLEEP [PLACE]i G-WH ‘Where does your friend sleep?’

a 0.

b 0.

This analysis apparently accounts for the IndSL data in a straightforward manner. However, it involves empirical and conceptual shortcomings that indicate that it might not be the right characterization. Setting aside the is-

When a Wh-word is not a Wh-Word

35

sue of the necessity of the directionality parameter and its relevance to core syntax, one such drawback is the apparent dichotomy between the specifier position within the complementizer system and the inflectional system. Assuming that IndSL is an SOV language, the position of arguments and adjuncts in IndSL suggests that specifiers within the inflectional domain occur to the left. On the other hand, the proposed analysis assumes that specifiers of the left periphery are to the right with the theory providing no explanation for such choice. The same argumentation could be made for functional heads, which occur to the left or to the right in IndSL (Kayne 1994; Bu¨ring and Hartmann 1997). Similarly, it is not obvious in this framework why the IndSL wh/pronominal elements never move overtly to SpecInterP, even though they can move to SpecFocP. Given these issues, we propose an alternative perspective that assumes, along the lines of Kayne (1994) that phrase structures are of the type specifier-head-complement (see Sˇarac et al., in press, for Austrian Sign Language). Under this approach, the IndSL clause structure may look like representation (25) if we assume, as previously, that the left periphery can be split into distinct functional projections. (25)

Assuming this is the right characterization, and keeping in mind the same ingredients as before, we can account for an IndSL wh-question like (24a), where the sign PLACE remains in situ, by saying that SpecFocP hosts a null operator that binds the null wh/pronominal element inside the proposition. We further argue that the interrogative clause-typing head Inter has scope over the focus phrase that is attracted into its specifier (Aboh 2004). More precisely, we are proposing that the wh-element is licensed in SpecFocP, while the proposition (here the focus phrase) as a whole moves to SpecInterP due to clause-typing. According to Chomsky (2001), this would mean that Inter has strong EPP features that are checked by the fronted

36

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

proposition (including the focus projection). The structure for sentence (24a) is represented in (26). (26)

We further conclude that the semantic correlate for this derivation is that wh in situ may be regarded to have focus neutral interpretation. In terms of this analysis, the spreading of the non-manual can be either on the clause typing head under Inter (i.e. G-WH), or the whole focus phrase (FocP) that has raised to SpecInterP. On the other hand, constructions where the PLACE/TIME/NUMBER expression appears left adjacent to G-WH are treated as focused whquestions where the associate phrase raises to SpecFocP to check its focus features and then to SpecInterP to check the EPP feature under Inter. We further speculate that movement through SpecFocP triggers movement of the remnant FocP past SpecInterP to a higher position, say SpecTopP (Rizzi 1997; Aboh 2004). At this stage of the discussion, it is not clear whether this last step of the movement is triggered by an attracting feature under a higher probe (e.g. Top) or whether it is contingent on movement of the PLACE/TIME/NUMBER expression through SpecFocP. Whichever the case, this analysis implies a distinction between neutral wh-questions (i.e. in situ wh-questions) and moved PLACE/TIME/NUMBER whquestions, the latter being more ‘emphatic’ than the former, as discussed by Bosˇkovic´ (2002) for Slavic languages. The derivation is represented under (27).

When a Wh-word is not a Wh-Word

37

(27)

The spreading pattern here is expected to be either on the interrogative head (G-WH) or on this head and on the phrase in its specifier (PLACE in (26)). Note, however, that this analysis predicts that the nonmanual cannot scope over the moved remnant FocP assuming the latter lands in the specifier of the topic phrase. Apparently, this is compatible with the data discussed previously which show that topics are outside of the scope of the clause typing non-manuals, but further study is needed to reach a final characterization of the spreading patterns in such a context. This aside, the proposed analysis has the advantage of capturing the apparent wh-splitting facts in a straightforward manner. The di¤erence between examples such as (24a) and (24b) reduces to the distinction between a neutral versus a focused wh-question. In addition, a general pattern that seems to emerge from this description is that sentence-final particles in this language have scope properties that force their complement to their specifiers, with this being reflected in the spreading behavior of the non-manuals. From a comparative and typological view, this an interesting result in that it shows that IndSL clusters with certain spoken languages (e.g., Lele, Gungbe) in this respect. 5. Conclusion The present paper compares two approaches to wh-questions in IndSL. Following the traditional approach to wh-signs, the first analysis proposes

38

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

that G-WH is a phrase that is licensed in a specifier position within the Csystem. However, the alternative approach that we propose indicates that G-WH is a wh-question particle that encodes the head of an interrogative phrase within the C-system and attracts the relevant wh-sign within its specifier. A consequence of this analysis is that IndSL allows for null wh-phrases, in neutral wh-questions, whenever these are contextually recoverable. In ambiguous contexts, however, an overt associate phrase (i.e. PLACE/NUMBER/TIME) is inserted that may further move to SpecInterP. The existence of a sentence-final clause typing wh-particle in IndSL opens the issue of the typology of wh-questions and of clause typing morphemes in signed and spoken languages. We hope to return to these issues in future work. Notational Conventions The sign language examples have been transcribed in a way that is widely used in sign language linguistics. The transcription consists of a line with glosses in capital letters representing the manual signs. For the IndSL examples, the glosses are given in English because the sign language is not limited to a particular spoken language region, for example the Hindispeaking area. A line on top of the glosses indicates non-manual features such as facial expressions, head movements, and the like, which simultaneously co-occur with manual signs; their co-occurrence with manual signs is indicated by the length of the line on top of the capital letter glosses. The following abbreviations and symbols have been used in the examples: COMPL EXIST G-WH IMP INDEF INDEXx

NEG NEG2 neg t wh

completive aspect marker existential marker general wh-question sign imperative marker indefiniteness marker (or IXx ) pointing sign used for localizing referents in the signing space, and for pronominalization of these referents (e.g. INDEX2 pointing towards the addressee for second person singular pronoun) neutral negation marker contrastive negation marker non-manual marker for negation (negative headshake) non-manual topic marking (raised eyebrows, head tilt) non-manual marker for wh-questions (see text for details)

When a Wh-word is not a Wh-Word

39

Notes 1. We are very much indebted to our informants Anjali Agrawal, Uday Bhaskar, Satya Sundare Das, Neil Fredrick, Sudip Ghosh, Rama Krishna, Dharmesh Kumar, Tushar Maganbhai, Gopal Motwani, Biswambhare Naik, Riju Sarma, and especially to our deaf research assistant Sibaji Panda; without their patient help this research would not have been possible. 2. For grammatical and psycholinguistic aspects see Klima and Bellugi (1979), for an overview of sign language phonology and syntax see Sandler (2000) and LilloMartin (2001) and references cited therein. 3. This is in contrast to a suggestion made by Bouchard and Dubuisson (1995) who claim that Quebec Sign Language and ASL—and possibly sign languages in general—exhibit free word order. In fact, they argue that sign languages are not characterized by an underlying hierarchical phrase structure but rather make use of entirely di¤erent organizing principles. See Kegl et al. (1996) for a thorough critique of their arguments. 4. The notational conventions for sign language examples are explained at the end of this chapter. 5. Petronio and Lillo-Martin (1997) propose that the final wh-elements are an instance of a more general double construction that is used for focus or emphasis in ASL. Sentence-final doubles are also observed with focused constituents, such as modals, verbs, negative signs, and quantifiers; cf. the examples in (i) (Petronio and Lillo-Martin 1997: 30f ). (i)

a.

NONE DEAF IN FAMILY BEFORE NONE ‘There have not been any deaf people in (my) family.’

b.

NANCY HATE ICE-CREAM HATE ‘Nancy hates ice cream.’

A similar argument is made for Brazilian Sign Language (LSB) by Nunes and de Quadros (2004). As in ASL, only heads allow duplication in LSB. 6. In order to strengthen this claim, Neidle et al. (2000) point out that wh-phrases in topic position are attested in other languages, such as Chinese and German. 7. More recently, Neidle (2002) has also argued that focus plays an important role in the analysis of ASL wh-questions. Her argument, however, is quite di¤erent from the one o¤ered by Petronio and Lillo-Martin (1997). Neidle (2002) only considers constructions with a single wh-phrase. She observes that a di¤erent interpretation is associated with the constructions involving wh-movement than with those in which the wh-phrase has remained in situ: moved wh-elements receive a focus reading. Consequently, she argues that there is an intermediate landing site to the left of TP (and to the right of the position in which topics occur) for focused whphrases, on their way to the CP-final specifier position. This intermediate position, she labels SpecFP. Wh-phrases that lack a focus feature are blocked from raising to SpecCP: due to the lack of the focus feature, they cannot raise to SpecFP but without this intermediate step, they cannot move further to SpecCP. 8. For further analyses of wh-questions within the generative framework see Nunes

40

9.

10.

11. 12.

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan and de Quadros (2004) for Brazilian SL, Cecchetto and Zucchi (2004) for Italian SL, and Sˇarac et al. (in press) for Croatian SL and Austrian SL. Some Indian dialects of IndSL make use of another non-compositional wh-sign to express the meaning ‘what day’. In the following discussion, this sign will not be considered. The other two conceivable options, namely combination of a wh-sign in situ with a sentence-initial or a sentence-final wh-sign, are also ungrammatical. The former option has been claimed to be possible in ASL by Neidle et al. (1997); cf. (5c) above. Remember that given an appropriate context, questions without wh-signs are also possible in IndSL (see Petronio and Lillo-Martin 1997: 35 for discussion of covert wh-words in ASL). Note that it is not entirely clear whether the complex expression FACE G-WH (‘who’) allows for wh-split. Further research is necessary to clarify this issue. In sentences containing the complex quantificational expression NUMBER GWH (‘how many’), even more options are available. Besides wh-split (11b), it is also possible to separate the quantifier from the noun phrase, as has been shown in (10c). Alternatively, the whole constituent—quantified noun phrase and complex wh-expression—may appear in sentence-final position (i). (i)

INDEX2 TAKE [BOOK NUMBER G-WH] ‘How many books will you take?’

See Boster (1996) for the discussion of similar constructions in ASL in which the quantificational expression HOW-MANY appears in sentence-initial position without the NP it modifies. 13. Similar cases are known from spoken languages. See, for instance, Aboh (2004) for a similar conclusion on verb focusing in Gungbe. 14. If it was not for the availability of complex wh-expressions, it might, of course, be tempting to assume that in IndSL, it is always an empty element that is moved to SpecCP while the wh-sign occupies a [þfocus] C. This would imply, however, that IndSL wh-questions always have a focus reading (as claimed by Neidle 2002 for wh-questions with sentence-final wh-signs in ASL). Moreover, IndSL does not make use of other focus doubling constructions which are taken to be evidence for the focus analysis by Petronio and Lillo-Martin (1997) (cf. footnote 6). Note that in section 4.2., we will propose an analysis for IndSL which involves focus operations. These operations, however, are di¤erent from what has been proposed for ASL. 15. Alternatively, we might speculate that IP remnant movement is followed by feature movement of [þwh] to a functional head above XP, as also discussed (and rejected) for ASL by Neidle et al. (2000: 146).

References Aarons, Debra 1994 Aspects of the syntax of American Sign Language. Ph.D. diss., Department of Linguistics, Boston University.

When a Wh-word is not a Wh-Word

41

Aboh, Enoch O. 2004 The Morphosyntax of Complement-Head Sequences: Clause Structure and Word Order Patterns in Kwa. Oxford: Oxford University Press. Bosˇkovic´, Zˇeljko 2002 On multiple wh-fronting. Linguistic Inquiry 33: 351–383. Boster, Carole T. 1996 On the quantifier-noun phrase split in American Sign Language and the structure of quantified phrases. In International Review of Sign Linguistics, William H. Edmondson and Ronnie B. Wilbur (eds.), 159–208. Mahwah, NJ: Lawrence Erlbaum. Bu¨ring, Daniel and Katharina Hartmann 1997 Doing the right thing. The Linguistic Review 14: 1–42. Cecchetto, Carlo and Sandro Zucchi 2004 Why is SpecCP on the right in sign languages? Paper presented at 27 th Generative Linguistics in the Old World Colloquium (GLOW), Thessaloniki, April 2004. Cheng, Lisa 1997 On the Typology of Wh-questions. New York: Garland. Chomsky, Noam 2001 Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 1–52. Cambridge, MA: MIT Press. Erting, Carol J., Robert C. Johnson, Dorothy L. Smith and Bruce D. Snider (eds.) 1994 The Deaf Way: Perspectives from the International Conference on Deaf Culture. Washington, DC: Gallaudet University Press. Frajzyngier, Zygmunt 2001 A Grammar of Lele. Stanford, CA: CSLI. Kayne, Richard 1994 The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Klima, Edward and Ursula Bellugi 1979 The Signs of Language. Cambridge, MA: Harvard University Press. Ladd, Paddy 2003 Understanding Deaf Culture: In Search of Deafhood. Clevedon: Multilingual Matters Ltd. Liddell, Scott K. 1980 American Sign Language Syntax. The Hague: Mouton de Gruyter. Lillo-Martin, Diane 1986 Two kinds of null arguments in American Sign Language. Natural Language and Linguistic Theory 4: 415–444. 1990 Parameters for questions: Evidence from wh-movement in ASL. In Sign Language Research: Theoretical Issues, Ceil Lucas (ed.), 211–222. Washington, DC: Gallaudet University Press. 2001 One syntax or two? Sign language and syntactic theory. Glot International 5: 297–310. Miles, Mike 2001 Signs of development in Deaf South and South-West Asia: histories, cultural identities, resistance to cultural imperialism. Internet URL: http://www.independentliving. org/docs3/milesm2001a.pdf. Neidle, Carol 2002 Language across modalities: ASL focus and question constructions. Linguistic Variation Yearbook 2: 71–98. Neidle, Carol, Judy Kegl, Ben Bahan, Debra Aarons and Dawn MacLaughlin 1997 Rightward wh-movement in American Sign Language. In Rightward Movement, Dorothee Beerman, David Leblanc and Henk van Riemsdijk (eds.), 247–278. Amsterdam: Benjamins.

42

Enoch O. Aboh, Roland Pfau and Ulrike Zeshan

Neidle, Carol, Judy Kegl, Dawn MacLaughlin, Ben Bahan and Robert G. Lee 2000 The Syntax of American Sign Language. Functional Categories and Hierarchical Structure. Cambridge, MA: MIT Press. Neidle, Carol, Dawn MacLaughlin, Robert G. Lee, Ben Bahan and Judy Kegl 1998 The rightward analysis of wh-movement in ASL: A reply to Petronio and LilloMartin. Language 74: 819–831. Nunes, Jairo and Ronice Mu¨ller de Quadros 2004 Focus duplication of wh-elements in Brazilian Sign Language. Paper presented at 35th Conference of the North Eastern Linguistic Society (NELS 35), Storrs, CT, October 2004. Padden, Carol 1988 Interaction of Morphology and Syntax in American Sign Language. New York: Garland. Padden, Carol and Tom Humphries 1988 Deaf in America. Voices from a Culture. Cambridge, MA: Harvard University Press. 2005 Inside Deaf Culture. Cambridge, MA: Harvard University Press. Petronio, Karen 1993 Clause structure in American Sign Language. Ph.D. diss., University of Washington, Seattle. Petronio, Karen and Diane Lillo-Martin 1997 WH-movement and the position of Spec-CP: Evidence from American Sign Language. Language 73: 18–57. Pfau, Roland 2002 Applying morphosyntactic and phonological readjustment rules in natural language negation. In Modality and Structure in Signed and Spoken Languages, Richard P. Meier, Kearsy A. Cormier and David G. Quinto-Pozos (eds.), 263–295. Cambridge: Cambridge University Press. Pfau, Roland and Josep Quer 2002 V-to-Neg raising and negative concord in three sign languages. Rivista di Grammatica Generativa 27: 73–86. Pfau, Roland and Markus Steinbach in press Relative clauses in German Sign Language: Extraposition and reconstruction. To appear in Proceedings of the 35th Conference of the North Eastern Linguistic Society (NELS 35). Ramakrishna Mission Vidyalaya 2001 Indian Sign Language Dictionary. Coimbatore: Sri Ramakrishna Mission Vidyalaya Printing Press. Rizzi, Luigi 1986 Null objects in Italian and the theory of pro. Linguistic Inquiry 17: 501–557. 1996 Residual verb second and the wh-criterion. In Parameters and Functional Heads. Essays in Comparative Syntax, Adriana Belletti and Luigi Rizzi (eds.), 63–90. Oxford: Oxford University Press. 1997 The fine structure of the left periphery. In Elements of Grammar. Handbook in Generative Syntax, Liliane Haegeman (ed.), 281–337. Dordrecht: Kluwer. 2001 On the position ‘‘Int(errogative)’’ in the left periphery of the clause. In Current Studies in Italian Syntax: Essays O¤ered to Lorenzo Renzi, Guglielmo Cinque and Giampaolo Salvi (eds.), 287–296. New York: Elsevier. Sandler, Wendy 1999 The medium and the message: Prosodic interpretation of linguistic content in Israeli Sign Language. Sign Language & Linguistics 2: 187–215. 2000 One phonology or two? Sign language and phonological theory. In The First Glot

When a Wh-word is not a Wh-Word

43

International State-of-the-article Book. The Latest in Linguistics, Lisa Cheng and Rint Sybesma (eds.), 349–383. Berlin: Mouton de Gruyter. Sˇarac, Nina, Katharina Schalber, Tamara Alibasˇic´ and Ronnie Wilbur in press Crosslinguistic comparison of sign language interrogatives. In Sign languages: a cross-linguistic perspective, Pamela Perniss, Roland Pfau and Markus Steinbach (eds.). Berlin: Mouton de Gruyter. Vasishta, Madan, James Woodward and Susan deSantis 1980 An Introduction to Indian Sign Language (Focus on Delhi). New Delhi: All India Federation of the Deaf. Vasishta, Madan, James C. Woodward and Kirk L. Wilson 1978 Sign language in India: regional variation within the deaf population. Indian Journal of Applied Linguistics 4: 66–74. Wilbur, Ronnie B. 2000 Phonological and prosodic layering of nonmanuals in American Sign Language. In The Signs of Language Revisited, Karen Emmorey and Harlan Lane (eds.), 213– 244. Mahwah, NJ: Lawrence Erlbaum. Wilbur, Ronnie B. and Cynthia Patschke 1999 Syntactic correlates of brow raise in ASL. Sign Language & Linguistics 2: 3–41. Woodward, James C. 1993 The relationship of sign language varieties in India, Pakistan, and Nepal. Sign Language Studies 78: 15–22. Zeshan, Ulrike 2000a Sign Language in Indo-Pakistan. A Description of a Signed Language. Amsterdam: Benjamins. 2000b Geba¨rdensprachen des indischen Subkontinents [Sign Languages of the Indian Subcontinent]. Munich: LINCOM Europa. 2001 Mouthing in Indopakistani Sign Language: Regularities and Variation. In The Hand is the Head of the Mouth: The Mouth as Articulator in Sign Languages, Penny Boyes Braem and Rachel Sutton-Spence (eds.), 247–271. Hamburg: Signum. 2003a Indo-Pakistani Sign Language grammar: a typological outline. Sign Language Studies 3: 157–212. 2003b ‘Classificatory’ constructions in Indo-Pakistani Sign Language: grammaticalization and lexicalization processes. In Perspectives on Classifier Constructions in Sign Languages, Karen Emmorey (ed.), 113–141. Mahwah, NJ: Lawrence Erlbaum. 2004 Interrogative constructions in sign languages: Cross-linguistic perspectives. Language 80: 7–39.

Q-baa and Bangla Clause Structure PROBAL DASGUPTA

The Past Subjunctive form in Bangla shares the morphology of the simple (Indicative) Past but di¤ers syntactically. One di¤erence is that it can directly license a modal particle /baa/, a particle here called Q-baa as it occurs with other Tense-Moods only if licensed by a Q, an interrogative. The analysis of this particle’s distribution in various clause types helps probe some aspects of clause structure. The proposal that each /baa/ needs both Local Focus licensing from an Emph(asizer)-bearing constituent that preposes and Mood Salience licensing from an M-Salient item that need not move turns out to provide an adequate account of the distribution of /baa/, assuming that interrogatives and the Past Subjunctive count as bearing M-Salience. Introduction The present methodological exercise follows the descriptive procedures of one formal syntactic system in order to show that certain phenomena that currently prevalent assumptions place within syntax cannot be handled without redrawing the boundaries of this domain. To the extent that this demonstration adheres as closely as necessary to familiar formal syntactic canons, it follows that what requires revision is the delineation of syntax as a domain. Unless this apparent anomaly in the formal syntax paradigm can be shown to disappear under the assumptions of some model claiming to supersede the one employed in this exemplification, model choice issues within that paradigm as it stands have little bearing on the discussion. At the empirical level, the present study, on the basis of some syntactic diagnostics for Bangla (a.k.a. Bengali) provided by Dasgupta (1996) that help detect the Present Subjunctive, continues the investigation of the subjunctive mood in Bangla, a language that does not morphologically distinguish subjunctives from indicatives. Using one such diagnostic—the position of Neg—to identify a hitherto unreported Past Subjunctive, and focusing on the fact that Past Subjunctive occurs only in root sentences, the present study addresses the interaction between verbs bearing this

46

Probal Dasgupta

Tense-Mood and their frequent companion, the modal particle /baa/. This particle is called Q-baa in this paper since it occurs with Tense-Moods other than the Past Subjunctive only if licensed by an interrogative. It is argued here that the task of adequately characterizing within the syntax the double licensing of the root sentence particle Q-baa makes it appropriate to postulate a natural class comprising interrogatives, the modally salient Past Subjunctive and nothing else. The first author to observe the existence of the Tense-Mood exemplified in (1), Roy (1996), calls it the Quirky Past; her project there does not involve placing it in any overall grid of Tense-Moods. Her quirky past is morphologically identical to the simple (Indicative) Past given in (2) (the glosses use abbreviations listed at the end of the paper): (1)

Past Subjunctive a. Dilip baadaam khelo-i baa Dilip nuts eat.PaSbj-Emph MPrt ‘So what if Dilip eats nuts(/Let Dilip eat nuts)’ b. Dilip baadaam naa-i baa khelo Dilip nuts Neg-Emph MPrt eat.PaSbj ‘So what if Dilip doesn’t eat nuts’ c. ? Dilip baadaam naa khelo-i baa Dilip nuts Neg eat.PaSbj-Emph MPrt, ¼ (1b) d. *Dilip baadaam khelo naa-i baa Dilip nuts eat.PaSbj Neg-Emph MPrt

(2)

Past (Indicative) a. Dilip baadaam khelo Dilip nuts ate ‘Dilip ate nuts’ b.

Dilip baadaam khelo naa Dilip nuts ate Neg ‘Dilip didn’t eat nuts’

c.

*Dilip baadaam naa khelo Dilip nuts Neg ate

But the pattern of well-formed and ungrammatical cases makes a syntactic contrast evident. The type of Past found in (2), whose indicative character is independently obvious, elicits postverbal negation. But the type of Past observed in (1) requires preverbal negation. The diagnostics that look serviceable for the Present Subjunctive in Dasgupta (1996) invite the conclusion here that the preverbally negated Past in (1b, c) and its a‰rmative counterpart in (1a) are Subjunctive in character. It is thus their Subjunctive feature that triggers the placement of Neg to the left of the verb in (1b, c). I

Q-baa and Bangla Clause Structure

47

shall use the negation diagnostic throughout to bring out the Subjunctive/ Indicative contrast. I do not discuss here the fact that another diagnostic highlighted in Dasgupta (1996), the contrast between the copula /(aa)ch/ and its main verb counterpart /thaak/, also supports the diagnosis of the forms given in (1) and in (2) as Subjunctives and Indicatives, respectively. Crucial to the present study is one distinctive property of the Past Subjunctive: it consistently carries an intonation peak, contrasting in this respect both with the indicative Past and with the Present Subjunctive studied in Dasgupta (1996). Now, this parallels a clear contrast that obtains quite generally between interrogatives, associated with an intonation peak, and relatives, which keep a low profile. Bangla’s overt morphological distinction between the K-words of the interrogative and the J-words of the relative series might seem to make a suprasegmental contrast functionally unnecessary. Nonetheless, even this language associates interrogatives, not relatives, with consistent prominent intonation. Later in the study, we shall see that interrogation and the Past Subjunctive form an intriguing natural class. Examples (3)–(5) present the two kinds of wh-words in Bangla and at the same time illustrate the basic traits of clause structure in the language: headfinal constituent order; the way headed relatives—usually extraposed—as in (4) contrast with sequential relatives as in (5); the unavailability, in interrogatives like (3) and sequentials such as (5), of the wh-fronting process restricted in Indo-Aryan to the headed relatives of (4); and other details— see Dasgupta (1987, 1997) for a larger range of examples relevant to these concerns: (3)

Dilip kon netaake helikaptxaar upohaar diyeche? Dilip which leader.Dat helicopter gift has-given ‘To which leader has Dilip made a gift of a helicopter?’

(4)

Prodip eak netaar kathaa bollo jaake Dilip helikaptxaar upohaar diyeche Pradip a leader.Gen about spoke who.Dat Dilip helicopter gift has-given ‘Pradip mentioned a leader to whom Dilip has made a gift of a helicopter’

(5)

Nirbaaconer aage je netaa je loker ghus election before which leader which man’s bribe kheyeche, jetaar par se to taar bhaaipoke caakri has-eaten, winning after he Prt his nephew.Dat job debe-i will-give-Emph (roughly): ‘A leader who has taken a man’s money before the election will of course after winning give that man’s nephew a job.’

48

Probal Dasgupta

The present study formalizes such intrinsic prominence, in an ad hoc fashion for reasons clarified above, as a diacritic M-Salience (Mood-Salience) property, specifiable in terms of phonological and interpretive consequences. Phonologically an M-Salient item bears prominent intonation, as we have seen. The formal proposal, in the system adopted here, is that the M-Salient feature covertly moves to a clausal functional head M, with interpretive consequences. One such consequence, of empirical interest in the present study, is that all and only M-Salient items can, via the M head of a finite clause, license the Modal Particle /baa/ observed in (1) above as a frequent companion of the Past Subjunctive. Both the M-Salient Tense-Mood of (1) above and the M-Salient interrogative features of (6) below license /baa/, which we shall call Q-baa because of cases like (6). In contrast, the clauses in (7) crash because Q-baa receives no M-Salience Licensing: (6)

Prodip kon netaake-i baa ghus debe? Pradip which leader.Dat-Emph MPrt bribe will-give ‘Which leader, indeed, will Pradip bribe?’

(7)

a.

*Je netaa je loker-i baa ghus khaak, se which leader whichever man’s MPrt bribe may-eat he take tobu kichu debe naa him still anything will-give Neg ‘Whichever man a leader may accept bribes from, he still won’t do a thing for him’

b.

*Prodip cestxaa korche jaate Dilip baadaam Pradip e¤ort is-making so-that Dilip nuts naa-i baa khaae Neg-Emph MPrt eat.PrSbj ‘Pradip is trying (to ensure) that Dilip does not, indeed, eat nuts’

The present study o¤ers an account of the licensing of this strategically interesting Modal Particle Q-baa in several kinds of clauses and, en route to this goal, brings out the crucial properties of several Tense-Mood forms and their interaction with aspects of clause structure in Bangla, especially certain phenomena concerning the M shell. Section 1 of the study, Assumptions about clause structure, lays out the syntactic framework adopted and empirical claims about Bangla that are taken for granted. Section 2, Morphological assumptions, places the study in the context of the substantivist paradigm. Section 3, The modal particle Q-baa, addresses the specific issues by carefully describing Q-baa; this particle requires Mood Salience licensing from some M-Salient item within its

Q-baa and Bangla Clause Structure

49

domain as well as Local Focus licensing from an item that must bear a Dissociative Emphatic feature and must prepose. Section 4, Negation and the pasts, uses these results to investigate the interaction between Neg Placement and the Tense-Moods highlighted earlier, further clarifying the role of M in clause structure. 1. Assumptions about Clause Structure In this study, the operative syntactic theory in the background is Chomsky (1995), plus the assumption (which later models in that tradition adopt in various forms) that Attractors house an uninterpretable formal feature exhibiting Suicidal Greed. Accordingly, I presuppose underlying universal head-mediality and typologically specific preposing in Bangla to yield the invariant surface head-final order (apart from Scrambling) exemplified below. (8)

Toruner posaa baaxdor baaraandaae kalaa khaacche Torun.Gen pet monkey veranda.Loc banana/s eat.PrProg ‘Torun’s pet monkey is eating bananas on the veranda’

(9)

darjaar pichone /pichon theke door.Gen behind /behind from ‘behind/from behind the door’

(10)

Sobitaar ceye lambaa Savita.Gen than tall ‘taller than Savita’

How is the transition from underlying head-medial to surface head-final order e¤ected? The answer assumed here rests on small p. This functional node, which is to big P what small v is to big V, encodes some observations (unpacked in Dasgupta 2002) about ‘‘light postpositions’’ such as /theke/ in (2) and ‘‘Case marked nominals’’ like the Genitive /N(e)r/ formats in (8) or (10) and the Locative /Ne/ format in (8). The small p element is central to the analysis of clause structure in this study. For there are morphological reasons—widespread in the Indo-Aryan language family, as illustrated below—for associating small p not just with nouns, as in the (a)-examples below, but also with verbs, as in the (b) and (c) cases: (11)

Marathi a. pustak madh-uun book inside-from ‘from inside the book’ b.

tithe jaa-uun there go-ConjP ‘having gone there’

50 (12)

Probal Dasgupta Hindi-Urdu a. kitaab ke (andar) book Gen(m) (inside) ‘(inside) of the book’ b.

(13)

kaam puuraa kar ke work complete do ConjP ‘having completed the work’

Bangla a. kutxir-e; maatxi-te cottage-Loc; ground-Loc ‘in the cottage; on the ground’ b.

Paakistaan theke phir-e Pakistan from return-ConjP ‘having returned from Pakistan’

c.

Paakistaanke kichu di-te Pakistan.Dat something give-Inf ‘to give Pakistan something/anything’

In the morphologies of the Indo-Aryan family in general, as (11)–(13) illustrate, nonfinite inflectional formats and particles as in the (b) and (c) examples are akin to (a)-type Case formats and particles as well as other light postpositions. Syntactically, a nonfinite verb has the diagnostically useful property of not permitting Rightward Argument Scrambling. It is as invariantly clause-final as an adposition is phrase-final. So I base my account of surface head-finality in Bangla on the following package of assumptions: (14)

Bangla nominals base their Case, and nonfinite clauses their corresponding subordination features, at small p; verbal complexes base their Acc Case features at small v and their Nom Case features at T.

These features interact with those of other shells and drive the preposing that e¤ects head-finality. Within a nominal, this means DP preposes to the Spec of the p where Case features are based. In a clause, nominal arguments prepose to the Specs of T and v. If a p that heads a clause bears subordination features, this p selects a dependent M, making MP prepose to the Spec of that p. On the grossest assumptions, then, ignoring split VP and other internal details, the derivation of the simple finite clause (15) proceeds from (16a) to (16b): (15)

Torun kalaa khaacche Torun banana/s eat.PrProg ‘Torun is eating bananas’

Q-baa and Bangla Clause Structure (16)

a.

functional heads þ[VP [ pP Torun][V 0 [V khaacche][pP kalaa]]] ! V to v to T, Object Preposing, Spec VP to Spec TP:

b.

[TP [ pP Torun][T 0 [ pP kalaa][T [V khaacche][T 0]][VP traces]]]

51

The independent MP and CP outside the TP are trivial in (16b). If one switches over to an embedded clause with a phonologically overt C /je/, one finds that this C is clause-initial, as exemplified in (17), where a zero M encodes the assumption that independent finite M precedes its TP complement: (17)

Baaxdor jaane naa [ je [M 0] Torun kalaa khaacche] monkey knows Neg that M Torun banana/s eat.PrProg ‘The monkey doesn’t know that Torun is eating bananas’

The clausal head p mentioned above and presupposed in (14) is instantiated as the clause-final Quotative particle /bole/ in (18), where I withhold judgment on where the M is, an issue addressed in the derivation of (18) given later: (18)

[[Torun kalaa khaacche] bole] baaxdor jaane naa Torun banana/s eat.PrProg Quot monkey knows Neg ‘The monkey doesn’t know that Torun is eating bananas’

I thus propose that M is selected in Bangla by either clause-initial C as in (17) or clause-final p as in (18). The background assumption is that, throughout Indo-Aryan, every C is finite and the head of a nonfinite clause is always a very light p. (18) illustrates the heavier Quotative p of Dravidian-influenced Indo-Aryan languages (Bangla /bole/, Marathi /mhanxuun/), which, unlike the standard Indo-Aryan very light p, selects finite M. Being a subordination-bearing particle, Quot ends up clause-final. This suggests that its features make the MP prepose to Spec of pP, the step from (19b) to (c) in derivation (19a–d) for example (19): (19)

a.

b. c. d.

Baaxdor jaane naa [ p bole] [M 0] [TP Torun kalaa monkey knows Neg Quot Torun banana/s khaacche] eat.PrProg ! T to M, Spec TP to Spec MP: Baaxdor jaane naa [ p bole] [MP Torun kalaa khaacche] ! MP to Spec pP: Baaxdor jaane naa [MP Torun kalaa khaacche] [ p bole] ! functional projection processes upstairs: khaacche] [ p bole] baaxdor jaane [MP Torun kalaa Torun banana/s eat.PrProg Quot monkey knows naa Neg

52

Probal Dasgupta

I presume that nonfinite clauses bearing subordination features as in (11b)–(13b,c) above with their nonfinite verb mimic opaquely what (18) with its finite verb plus subordination functor unpacks transparently. In other words, clausal p uniformly selects M, finite in (18) and nonfinite in (11b)–(13b,c). These assumptions and claims do not perhaps provide a su‰ciently helpful picture of clause types in Bangla. The morphosyntactic mechanisms assumed here are elaborated in Dasgupta, Ford and Singh (2000). Our goal in the present study is to understand options and impossibilities that correlate with degrees of clausal independence. More needs to be said, on the basis of a slightly closer look at C in (17) and p in (18). First, the option of Scrambling an argument like /cheletxaake/ ‘the boy’ rightward past the canonically clause-final finite verb available with C in (20)—see (20 0 ) for the canonical order—is excluded if the selector of T is either the heavy p Quot as in (21a) or an a‰xal p as in (21b) (again, see (21 0 a, b) for the canonical order): (20)

(with Rightward Argument Scrambling) je aami fon korbo cheletxaake that I phone do-Fut the boy ‘that I will call the boy’

(20 0 )

(canonical order) je aami cheletxaake fon korbo that I the boy phone do-Fut ‘that I will call the boy’

(21)

(rightward scrambled, unacceptable) a. *aami fon korbo cheletxaake bole I phone do-Fut the boy Quot b.

(21 0 )

*aami fon korle cheletxaake I phone do-CondP the boy

(canonical, well-formed variants of [21]) a. aami cheletxaake fon korbo bole I the boy phone do-Fut Quot ‘that I will call the boy’ b.

aami cheletxaake fon korle I the boy phone do-CondP ‘if/when I call the boy’

To present the second di¤erence between (17) and (18), it is necessary to consider the matrix as well:

Q-baa and Bangla Clause Structure

53

(22)

Tomraa dhore nite paaro je aami (to) you-Pl for-granted take-Inf can that I (of-course) aasbo come-Fut ‘You people can take it for granted that I will (of course) come’

(23)

[Aami (*to) aasbo bole] tomraa dhore [I (*of-course) come-Fut Quot] you-Pl for-granted nite paaro take-Inf can ‘You people can take it for granted that I will (*of course) come’

The (17)-type embedded clause in (22) is well-formed both with and without the topicalizing particle /to/, here glossed as ‘of course’. But (23), with an (18)-type downstairs clause, is well-formed without and unacceptable with this particle—abstracting away from the marginal improvement of (23) on successive-cyclic movement of /aami to/ ‘I, of course’ to the matrix CP, not the structure of concern here. These two di¤erences between types (17) and (18) are handled here in terms of the following moves. C and p select independent and subordinate M respectively. C has no features that might trigger the preposing of MP to Spec of CP. An MP selected by p is subordinate and must prepose to Spec of pP. Its head, the subordinate M, is a‰xal and, once the chips of Checking fall where they must, ends up sandwiched between T to its left and p to its right. (For details concerning how M can be an ‘‘a‰xal’’ head in a substantivist theory, see section 2.) Independent M, at a first approximation that I hold constant throughout this paper, is a null but non-a‰xal head, its light feature matrix awaiting supplementation by any modal Prt (Particle) present in the clause covertly moving its relevant features—presumably P-features, P for ‘‘peripheral’’—to M. Scholars often use the Prt /to/ as a clause type diagnostic. In the present account, the operational flexibility of independent M is what allows an indicative finite clause to support a Prt like /to/ in (22). In contrast, /to/ in (23) leads to crash if it covertly moves its independent P-features to a subordination-committed a‰xal M. This study regards Prt (exemplified also by the Emphasizers /i/ and /o/) as a category that does not project. Prt always unifies into the feature structure of its sister, which projects. Independent M’s flexibility also enables the finite clause in (12) to scramble an argument rightward, by mechanisms associable with the larger operating area of an independent finite clause. This is not a matter confined to the indicative; Rightward Argument Scrambling also occurs in clauses bearing the Present Subjunctive of Dasgupta (1996), such as the subordinate clause in (24):

54 (24)

Probal Dasgupta Tumi caao naa je oraa fon kare aamaake you want Neg that they phone do.PrSbj me ‘You don’t want them to call (lit. that they call) me’

Like a regular indicative, a subjunctive too is a finite MP selected by C and containing a null but independent (non-a‰xal) M. It too has the ‘‘larger operating area’’ for which I take Rightward Argument Scrambling to be an observational diagnostic. Subjunctives are on a par with indicatives as independent finite clauses. An attempt to work out how Rightward Argument Scrambling operates would disorganize this study beyond repair if o¤ered with any seriousness. Just for the record, here is a stop-gap account. Consider the embedded CP of (24): (25)

kare]]] [CP [C je][MP [M 0][TP oraa aamaake fon that M they me phone do.PrSbj

For concreteness, I shall arbitrarily take the position that Leftward Argument Scrambling preposes not a focus, but an antifocus. (In derivations exemplifying only this movement, the scrambled item’s exit leftwards leaves a true focus, if any, adjacent to the verb’s final landing site and thus eligible for focal interpretation.) Let /aamaake/ ‘me’ in the TP in (25) be the antifocus and undergo Leftward Scrambling, adjoining to TP (or to MP, if one prefers a multiple Spec of MP variant—not adopted here—of the analysis), in a first manoeuvre. After this the derivation can tamely terminate with null M moving to C if M has no relevant P-features, though the case of current interest does not in fact halt here since M is not so empty: (26)

[CP [C je][MP [M 0][TP aamaake [TP oraa t fon kare]]]] that M me they t phone do.PrSbj

My proposal is that what one calls Rightward Argument Scrambling actually involves subsequent stranding of this Leftward-Scrambled argument. In other words, the TP-body deserts it and preposes to Spec of MP, and the argument gets left behind on the right: (27)

[CP [C je][MP [TP oraa t fon kare] [M 0] [TP aamaake [TP that they t phone do.PrSbj M me t]]]] t

Why should a TP move to Spec of M at all? One of the P-features in the phonologically null but (in this case) syntactically active M’s feature matrix may serve as its attractor, a matter left open here.

Q-baa and Bangla Clause Structure

55

I reiterate that this unpacking of the observation about the relatively large operating area of an independent finite clause merely indicates what kind of account serious research would have to build. The phenomenon of Rightward Argument Scrambling is used here only as a diagnostic for the independent finite clause status shared by subjunctives and finite indicatives. One serious claim in this analysis is that non-a‰xal M occurs in every independent finite MP, indicative or subjunctive, selected by C. Equally serious is the contrast with the a‰xal M of either a nonfinite clause headed by an a‰xal small p (where the p-M interaction is managed at the verb inflection level, as in (28) ¼ (13c)) or a clause headed by the Quotative p /bole/ which selects a finite MP headed by a‰xal null M forcing the TP to prepose to Spec of MP and the MP to prepose to the Spec of p, as in (29) ¼ (18): (28)

[ pP [MP PRO Paakistaanke kichu traces] [ p 0 [ p [M [T [V PRO Pakistan.Dat something traces dite]]]]]] give-Inf ‘to give Pakistan something’

(29)

[ pP [MP [TP Torun kalaa khaacche] [M 0]][ p 0 [ p bole] Torun banana/s eat.PrProg M Quot tMP ]] baaxdor jaane naa MP-trace monkey knows Neg ‘The monkey doesn’t know that Torun is eating bananas’

This study focuses on some ways in which aggressively independent clauses, always selected by C, di¤er from clauses that exhibit various degrees of subordination, all the way down to standard p-selected nonfinite MPs with an a‰xal T-M-P sequence encapsulated in the verb’s inflection. Alternative formalisms of clause cartography that articulate the fine structure of the left periphery seem to have been fashioned entirely without reference to the Indo-Aryan material. It remains to be seen if such formalisms can sidestep the di‰culties highlighted in the present study that beset all formal syntaxes. Unless they can, the choice between the present T-M-P system and an alternative cartography that segments the Comp family (and perhaps also segments a corresponding family in the periphery of the small v projection) into Topic-zero, Focus-zero and other projections has little pertinence to empirical research. 2. Morphological Assumptions The syntax of the degrees of clausal independence, outlined in section 1, closely interacts with the morphology of TAM, Tense/Aspect/Mood. I con-

56

Probal Dasgupta

tinue to elaborate here the approach to Bangla TAM phenomena developed in texts where TAM-Neg interaction is used as a crucial probe (Dasgupta 1994, 1996). One subgoal of this section is to provide some relevant highlights of this approach. Another subgoal is to place this study in the context of the substantivist paradigm built around Whole Word Morphology (Ford, Singh and Martohardjono 1997; Dasgupta, Ford and Singh 2000; Singh and Starosta 2003). One major di¤erence between the more commonly known formalist paradigm in linguistics and the substantivist alternative within which the present study is located appears at the level of their respective approaches to the characterization of formal objects. The formalist approach adopted by many generative grammarians uses the machinery of specific descriptive levels to represent formal objects. A combinatorics particularizes each level, which derivations map on to neighbouring levels (some variants of the theory substitute linking by correspondences for derivational mapping). In contrast, the substantivist approach operates on the assumption that levels of linguistic characterization must converge on formal objects. Morphology and syntax co-specify a word. Syntax and pragmatics cocharacterize a sentence. The implementations in Dasgupta, Ford and Singh (2000) show that the minimalist programme can serve as a meeting-ground where substantivism recuperates and extends results obtained under formalist assumptions. The model developed there (and assumed here) inherits Ford, Singh and Martohardjono’s (1997) axiom that words are always seamless, integral objects, specified as such by the morphology. At the same time, the feature composition of these seamless, integral words drives syntactic movement for feature checking, a matter of the syntax module co-specifying what words are. In the present study, these matters become relevant because the status of Neg in Bangla requires discussion. Asamiya or Kannada present the Neg feature as part of the verbal word, while English allows negation to occur freely either at a nominal site (Not a single woman voted for her) or in the verb inflection complex (They don’t have to not show up to make their point). Bangla has a negative particle that looks independent, as in English, but is tied to the verb inflection system, as in Asamiya. The substantivist method of having morphology and syntax co-specify words gives us a grip on this phenomenon. Later in this section, I shall develop a substantivist formal device, Word Extension Strategies, in order to handle the details of Bangla negation. Substantivism is built around the devices of Whole Word Morphology (WWM) but is not confined to this core. WWM uses bidirectional Word Formation Strategies (WFSs; see Singh, Ford and Martohardjono 1997

Q-baa and Bangla Clause Structure

57

for basic clarifications). WFSs (30) and (31) show the formalism at (a), the associated interpretive correspondence at (b), and examples of pairs linked by the WFS at (c), by way of indicating briefly how the formalism works. Note the alternation between /a/ and /o/ at /bhadro/ vs /abhodro/; this is handled by the automatic phonology, which WWM prescinds from. (30)

(31)

a.

/VX/Adj $ /anVX/Adj

b.

WFS (30a) links a‰rmative on the left to negative on the right

c.

examples: ukto ‘explicit’ abohito ‘informed’ onutapto ‘repentant’

anukto ‘inexplicit’ anobohito ‘uninformed’ anonutapto ‘unrepentant’

a.

/CX/Adj $ /aCX/Adj

b.

WFS (31a) links a‰rmative on the left to negative on the right

c.

examples: sambhab ‘possible’ asombhab ‘impossible’ bibecok ‘considerate’ abibecok ‘inconsiderate’ bhadro ‘polite’ abhodro ‘impolite’

In the present study, I find it necessary to postulate unidirectional WESs, Word Extension Strategies, which map words onto extended words by attaching a clitic such as the Dissociative Emphasizer /i/: (32)

a.

/X/Word ! [/X/Word /i/Emph ]Word

b.

WES (32a) maps word on to dissociative-emphasized word

c.

examples: sab ‘everything’ boi ‘books’

sab#i ‘absolutely everything’ boi#i ‘precisely books’

I take the position that the phonology and syntax modules can look into an extended word of the form Word-Emph, which they co-specify with the morphology. This move gets the phonology right, letting /sabi/,/boii/ violate what are otherwise word level phonotactic constraints with impunity. The move also expresses some aspects of the syntax; if we want to build an account of Wackernagel phenomena, more needs to be said in the syntactic feature composition of the WES input to get the placement right. Lexically speaking, the move we have made amounts to claiming that clitics have no distinct entries, but exist in the form of WESs, subject to principles constraining the class of possible clitics.

58

Probal Dasgupta

Before I approach topics that WES can handle well, I begin with cases of negation that it is not supposed to touch! Readers familiar with substantivism will recall that WWM constitutively abhors zero units. Consequently, the overall architecture of substantivism prevents a syntax postulating a zero T as the head of TP from the fatal move of taking a ‘‘zero word’’ and attaching a clitic to it. In other words, if we propose that Bangla verbs are normally negated by Neg cliticization WESs, then substantivism predicts that the zero T cannot do this, and that the negation of verbless clauses has to work di¤erently. Bangla does exhibit verbless a‰rmative clauses. Suppose we do describe their finiteness in terms of a phonologically null T in the syntax: (33)

Modhu lokkhi chele [T 0] Madhu well-behaved boy T ‘Madhu is a good boy’

(34)

Tumi dustxu meye [T 0] you naughty girl T ‘You are a naughty girl’

It turns out that they negate either with an invariable /naa/, as in— (35)

Modhu lokkhi chele naa Madhu well-behaved boy Neg ‘Madhu is not a good boy’

(36)

Tumi dustxu meye naa you naughty girl Neg ‘You are not a naughty girl’

—or with an overt agreement-bearing negation in the default present tense (no other tense is available for this form), as in: (37)

Modhu lokkhi chele nae Madhu well-behaved boy NegAgr ‘Madhu is not a good boy’

(38)

Tumi dustxu meye nao you naughty girl NegAgr ‘You are not a naughty girl’

What is the right syntactic representation for these strings? It is important to note that I am doing one module at a time and will return to morphology much later. Syntactically speaking, I take the position that such a Neg, with Agr as in (37) and (38) or without Agr as in (35) or (36), bears T features

Q-baa and Bangla Clause Structure

59

and is Merged as a negative T. In (35) and (36) /naa/, on its way from the lexicon to the numeration, has picked up an optional default Present value for its T feature. The agreement-inflected negation in (37) and (38) has picked up both this value and whatever specification enables it to manifest phi features. The task of expanding this account so that it describes just how the inflectionally invariant items /nei/ ‘is not there, does not exist’ and Negative Perfect /ni/ encapsulate all pertinent features is not addressed here. The features of T and Asp[ect] turn out to constitute the site at which Bangla manages the formal diversification of its negatives. Hindi-Urdu, in contrast, appears to do it at a modal site. The general non-indicative Neg /na/ in Hindi-Urdu and the imperative Neg /mat/ contrast with the regular finite indicative Neg /nahiix/, with no Agr and not a whisper about T or Asp. I conclude that negation is T-oriented in Bangla and M-oriented in Hindi-Urdu. This point is unpacked more fully in Dasgupta (2005). The surface distribution of negation in Bangla is as follows. An invariant and preverbal /naa/ characterizes the subjunctive and all nonfinites. In the imperative, Neg retains this invariant /naa/ form but occupies the postverbal position characteristic of the indicative. See chart (39) and the examples that follow it: (39)

(40)

(41)

Form and Position Chart for Bangla Neg Postverbal Neg

Preverbal Neg

Inflectable for T/Agr

Some Indicatives: (V) Neg-Agr



Invariant /naa/

Other Indicatives, Imperatives: (V) /naa/

Subjunctives or Nonfinites: /naa/ V

i.

Inflectable Postverbal Neg: Dilip baadaam khaae ni Dilip nuts eat NegPerf ‘Dilip hasn’t/hadn’t eaten/didn’t eat nuts’

ii.

Inflectable Neg with V absent: Dilip rogaa nae Dilip thin NegAgr ‘Dilip isn’t thin’

i.

Invariant Postverbal Neg with V Indicative: Dilip rogaa chilo naa Dilip thin was Neg ‘Dilip was not thin’

60

Probal Dasgupta ii.

Invariant Postverbal Neg with V imperative: Kaal Diliper kaache jeyo naa tomorrow Dilip.Gen at go-FutImp Neg ‘Do not visit Dilip tomorrow’

iii. Invariant Neg with V absent: Dilip rogaa naa Dilip thin Neg ‘Dilip isn’t thin’ (42)

i.

Invariant Preverbal Neg with V subjunctive: Jodi Dilip kalaa naa khaae, taar baaxdorer if Dilip bananas Neg eat.PrSbj, his monkey.Gen bhaari majaa great fun ‘If Dilip does not eat bananas, it is great fun for his monkey’

ii.

Invariant Preverbal Neg: Prodip Dilipke caar din kono phal naa khete bollo Pradip Dilip.Dat four days any fruit Neg eat-Inf told ‘Pradip told Dilip not to eat any fruit for four days’

The basic observation that Bangla Neg marking is postverbal with typical finites and preverbal with nonfinites and dependent finites goes back to Trivedi (1905), who furthermore proposes that preverbal and postverbal /naa/ are one and the same element. Implementing Trivedi’s proposal today means setting aside the possibility—worth pursuing elsewhere—that Hany Babu’s (1997: 94) decision based on Zanuttini’s (1991) distinction between a typically inflected Neg head and a normally invariant Neg specifier might provide the right account for Bangla, where postverbal negation sometimes varies and preverbal negation is always invariant. My quest here is for a formal implementation of the unity of Neg that squares with the Bangla facts. I opt for an analysis that makes [þNeg] a feature that occurs only at TAM heads in Bangla. This account derives the obvious di¤erences among the negations from TAM properties of these heads themselves. The following initial exposition reflects only the actions of the syntax module. Later, when the morphology comes in at the discussion surrounding (45) and (46), details of the account emerge more clearly. The syntactic part of this account rests on the idea that an independent TAM head need not transmit any crucial features up the tree. Neg can thus directly appear in the feature composition of such a head. V can be merged as an adjunct to that nucleus, yielding a V Neg head like (40i)’s

Q-baa and Bangla Clause Structure

61

/khaae ni/ ‘hasn’t eaten’, whose negative nucleus /ni/ follows its verb margin /khaae/; or, to take a more apposite example (since we are not providing an account of /ni/), /khaae naa/ ‘doesn’t eat’. In contrast, a dependent TAM head anaphoric to a matrix tense must contain a verb whose features are transmitted up the line of relevant heads and co-refer to the antecedent referential T. Such anaphora is inconsistent with the occurrence of Neg features in the dependent TAM head proper. A Neg can only occur adjoined to such a head, as in (42ii)’s head sequence /naa khete/ ‘not to-eat’ with the negative margin /naa/ adjunct-merged to the verb nucleus /khete/. One advantage of this account is that it makes sense of the facts in the Chittagong dialect of Eastern Bangla studied in Roy (1996). This dialect, unlike standard Bangla (called ‘‘Bangla’’ throughout this paper, and elsewhere in the literature, for ease of exposition) lacks Tensed or phifeatured Neg and employs only dependent /na/ quite generally. My proposals here correctly predict the preverbal placement of Chittagong Bangla /na/. To return to standard Bangla, its postverbal Neg is handled here by first merging it as an independent tensed head on its own. The V later headmoves to it, adjoining to its left to yield V Neg order both for overtly inflected Neg as in (40) and for the invariant Neg as in (43). Here is an abbreviated sample derivation: (43)

Derivation (ignoring small v, split VP, etc.): a. [TP [T naa][VP Dilip [V 0 [V chilo] rogaa]]] Neg Dilip was thin ! VP-internal processes; V to T; Spec VP to Spec TP ! b.

[TP Dilip rogaa [T [V chilo][T naa]][VP traces]] Dilip thin was not traces

As for the preverbal type of Neg, its features make it depend on a subordinate verb, and every subordinate verb in Bangla requires preverbal Neg. The factor responsible must be the tense features anaphoric to the matrix tense, which for the sake of smooth handling of this anaphora needs to be visible at the T head and therefore must occupy prime space at the T site— even in the case of as independent a subordinate verb as a subjunctive. Consequently a Neg that serves a verb bearing anaphoric T features occurs not at the nucleus of the head, but in a head-adjunct position. Obvious questions arise about how it gets merged. The answer I adopt is that V movement to T lexicalizes this node, at which Neg then merges as a headadjunct, syntactically speaking (for the lexical component of the description, see below):

62 (44)

Probal Dasgupta if-clause: derivation (ignoring internal details): a. [CP [C jodi ][MP [M 0][TP [T 0[þNeg]][VP Dilip khaae if M [þNeg] Dilip eat.PrSbj kalaa]]]] bananas ! VP to TP processes as in (43); realization of þNeg as adjunct /naa/ ! b.

[CP [C jodi ][MP [M 0][TP Dilip kalaa [T [Neg naa][T [V if M Dilip bananas Neg khaae]]][VP traces]]]] eat.PrSbj traces

Does this syntactic description contradict theories in which Neg heads its own projection? So it seems at first sight. But no naive notion of a head Neg survives Laka’s (1990) familiar recasting of it in terms of a Sigma head at which a language exercises its choice among negation, emphatic a‰rmation, and, in Spanish, Imperative features. If a Spanish Sigma can harbour such TAM features, evidently Laka can be read as pioneering the idea that the Neg merger site counts as a TAM head. Thus the present account implements one of the earliest rigorous proposals in this domain. Another aspect of the present account that might seem unusual is the proposal that the composite head containing Neg and T in (44b) arises by merging Neg as an adjunct to T. Again, such a claim receives support from data presented in the standard literature. Thus, Laka (1990: 29–30) derives the uninterruptible Basque sequence /ez da/ ‘no has ¼ has not’ by moving the lower T head /da/ to the higher Neg head /ez/. But the general theory of adjunction leads us to expect the adjunction of a lower T to a higher Neg head to yield T Neg, not Neg T. At this juncture we note that Laka does make the relevant point (1990: 30, fn 12) that ‘‘Although it is orthographically separated from the inflected verb, the negative element is a clitic on the auxiliary.’’ Given that Neg is a proclitic, and given that Sigma must be higher than T for Basque (as argued by Laka 1990: 27–38), one sound option would involve first moving T to an abstract [þNeg]-specified Sigma head and then realizing this feature by merging /ez/ as a head-adjunct to the /da/-lexicalized Sigma head to yield /ez da/. In other words, even though some details of the present account of Neg are designed to ensure descriptive continuity with earlier work on the interaction between negation and the Bangla subjunctive system, nonetheless the moves made in this account simultaneously preserve theoretical continuity with empirical and conceptual antecedents in the careful literature on the relevant functional heads.

Q-baa and Bangla Clause Structure

63

What I have provided so far are the syntactic coordinates. Morphologically, for the forms built around invariant /naa/ that the discussion focuses on, I postulate the following WESs: (45)

(46)

a.

/X/T; Indic=Imp ! [/X/T; Indic=Imp /naa/T; Neg ]T; Indic=Imp; Neg

b.

WES (45a) maps tensed verb on to negated tensed verb

c.

examples: chilo ‘was’ jeyo ‘go’

chilo naa ‘was not’ jeyo naa ‘don’t go’

a.

/X/V; Sbj=Nonfin ! [/naa/Neg /X/V; Sbj=Nonfin ]V; Sbj=Nonfin

b.

WES (46a) maps verb on to negated verb

c.

examples: khaae ‘eat.Sbj’ khete ‘eat.Inf ’

naa khaae ‘not eat.Sbj’ naa khete ‘not eat.Inf ’

The collaboration between lexicon and syntax envisaged here organizes the implementation as follows. The WESs endowing these negated forms with morphological validity (a type of lexical recognition) only specify the lexical form of the output. The syntax, merging feature structures and generating a formally valid network of nodes, provides sites where lexical output can be niched without violating syntactic principles. The syntactic account above describes the sequences of syntactic operations validly constructing particular subtrees. The WESs provide words that fit into them. Let us take a closer look now at the TAM feature specifications with which Neg interacts. The central goal of this study is to examine the relation between TAM diversity and degrees of clausal independence. This makes it useful to look at the contrasting properties of the Past (Indicative) /khelo/ ‘s/he ate’ and the Past Subjunctive /khelo/ ‘s/he eat.PaSbj’. Though morphologically identical (apart from certain copula shape alternation phenomena not discussed here, as noted in the introduction), these two Pasts di¤er syntactically, for instance, in relation to Neg. The examples of these di¤erences provided below introduce one additional common factor to maximize comparability—the Modal Particle Q-baa: (47)

Past (Indicative) Dilip baadaam khelo naa-i baa keano? Dilip nuts ate Neg-Emph Prt why ‘Why, indeed, did Dilip not eat nuts?’

(48)

Past Subjunctive Dilip baadaam naa-i baa khelo Dilip nuts Neg-Emph Prt eat.PaSbj ‘Dilip had better not eat nuts’

64

Probal Dasgupta

To complete this section’s task of clearing the ground for an analysis of these contrasts, I need to address two issues, one general and one particular. The general issue runs like this. Given that I subscribe to Whole Word Morphology, a strong integralist decision to express inter-word relations in terms of Word Formation Strategies and not to recognize a‰xes as separate entities in a synchronic grammar (Singh, Ford, and Martohardjono 1997; Singh and Dasgupta 1999; Dasgupta, Ford, and Singh 2000; Singh and Starosta 2003), can I extract TAM commonalities and yet keep faith with my strong integralist morphological doctrines? Despite appearances, this is not a conceptual but a terminological mater. For Whole Word Morphology does, of course, recognize shared properties. On this basis I can and do note that two TAM formats in Bangla—the Past Indicative and the Past Subjunctive—share phonological and semantic properties. While WWM does not posit and cannot refer to a‰xes, it can and does refer to formats that in certain cases correspond to what morphemic theories describe as a‰xation. The particular issue that needs to be addressed has to do with a‰xal heads of syntactic projections, such as the a‰xal M head discussed in section 1. Does that notion clash with substantivist doctrine? It does not. Substantivism is happy to postulate syntactic entities that are a‰xal and correspond closely to conventional inflectional a‰xes. Content words merge at lexical heads and climb until their features have been checked. Some functional heads are independent and have words, function words, merged at them. Other functional heads are dependent and serve as sites where an appropriate feature structure gets checked by virtue of inflectional features; these are a‰xal heads. 3. The Modal Particle Q-baa I turn now to the task of describing carefully the interaction between the properties of the Past Subjunctive Tense-Mood and those of Negation and the Modal Particle Q-baa. To provide a context for this description I return to examples (47) and (48). They show how clauses featuring the Past TenseMoods place Neg and Q-baa di¤erently. Neg and the relevant Tense-Moods are familiar by now. The present section focuses on the distribution of the Bangla Modal Particle Q-baa, which seems to have received no specific attention in the literature as yet. The analysis o¤ered here focuses on some distinctive properties of Q-baa that associate it with a neighbour that bears emphatic or focal features and a neighbour exhibiting Mood-Salient features. I begin by illustrating the centrality of these properties at the observational level. A simple a‰rmative clause like (49) or a simple negative clause like (50) cannot support Q-baa:

Q-baa and Bangla Clause Structure (49)

*Dilip baadaam khaacche baa (permutations of this also *) Dilip nuts eat.PrProg MPrt

(50)

*Dilip baadaam khaacche naa baa (permutations also *) Dilip nuts eat.PrProg Neg MPrt

65

What if the Dissociative Emphasizer /i/ that precedes Q-baa in the wellformed cases (47)–(48) is the crucial factor? One tries adding /i/ to (49) and (50)—choosing its apparently preferred location to the immediate left of Q-baa as in (47)–(48)—and still fails to achieve well-formedness, as the Emph is necessary but not su‰cient: (51)

i.

*Dilip baadaam khaacche-i Dilip nuts eat.PrProg-Emph baa (permutations also *) MPrt

ii.

*Dilip baadaam khaacche naa-i Dilip nuts eat.PrProg Neg-Emph baa (permutations also *) MPrt

I know of only two ways to turn such strings into well-formed sentences. One is to change the TAM and choose the Past Subjunctive, as in (48) above. The other way is to hold the TAM constant but insert an interrogative constituent, as in (52): (52)

Dilip baadaam khaacche naa-i baa keano? Dilip nuts eat.PrProg Neg-Emph MPrt why ‘Why indeed is Dilip not eating nuts?’

I take this to mean that the Past Subjunctive and Interrogatives share some property, which I will arbitrarily call M[ood]-Salience. The Past Subjunctive is an M-Salient TAM, as distinct from other TAMs. Interrogative K-words, unlike Relative J-words, are M-Salient wh-words. There is a diagnostic for M-Salience independent of the Modal Q-baa. An M-Salient item is strongly associated with phonological prominence, which it normally carries, as in (53). It may delegate this burden to certain neighbours, as in (54), but it cannot occur in consistently low-profile strings like (55), the way other TAMs or relative wh-words can in the equally lowprofile (56): (53)

a.

Past Subjunctive prominent: Naahae Dilip aaro kayektxaa baadaam KHELO suppose Dilip yet a-few nuts eat.PaSbj ‘Suppose Dilip DOES eat a few more nuts’

66

(54)

(55)

(56)

Probal Dasgupta b.

Interrogative prominent: Dilip KOTHAAE eato derite baadaam kinbe? Dilip where so late nuts buy-Fut ‘WHERE will Dilip buy nuts so late?’

a.

E baadaamgulo naahae NAA-I khelo Dilip these nuts suppose Neg-Emph eat.PaSbj Dilip ‘Suppose Dilip DOESN’T eat these nuts’

b.

Dilip eato derite baadaam KINBE kothaae? Dilip so late nuts buy-Fut where ‘Where will Dilip BUY nuts so late?’

a.

*Naahae Dilip aaro kayektxaa baadaam suppose Dilip yet a-few nuts khelo, colourless intonation eat.PaSbj

b.

*Dilip kothaae eato derite baadaam Dilip where so late nuts kinbe?, colourless intonation buy-Fut

a.

Dilip aaro kayektxaa baadaam khaabe Dilip yet a-few nuts eat-Fut ‘Dilip will eat a few more nuts’

b.

Dilip jekhaane eato derite baadaam kinche se Dilip where so late nuts buy.PrProg that dokaantxaa notun shop new ‘The place where Dilip is buying nuts so late is a new shop’

Returning to Q-baa, I can propose that it needs licensing from a constituent bearing Dissociative Emphasis (thus marked by the Emphasizer /i/) and from an M-Salient item—a Past Subjunctive verb or an interrogative constituent. As one step in fleshing out this proposal, I make the claim that Q-baa carries P-features that covertly move to M. This claim is going to help make sense of the following data set, independent of what has been discussed so far: (57)

Standard word order: a. Dilip tomaae baadaam naa-i baa dilo Dilip you nuts Neg-Emph MPrt give.PaSbj ‘What if Dilip doesn’t give you nuts?’

Q-baa and Bangla Clause Structure b.

(58)

67

Dilip dokaan theke baadaam naa-i baa kinlo Dilip shop from nuts Neg-Emph MPrt buy.PaSbj ‘What if Dilip doesn’t buy nuts from the shop?’

Displaced Neg-Emph Prt sequence: a. Dilip tomaae naa-i baa baadaam dilo Dilip you Neg-Emph MPrt nuts give.PaSbj ‘What if Dilip doesn’t give you nuts?’ a 0 . Dilip naa-i baa tomaae baadaam dilo Dilip Neg-Emph MPrt you nuts give.PaSbj ‘What if Dilip doesn’t give you nuts?’ b.

Dilip dokaan theke naa-i baa baadaam kinlo Dilip shop from Neg-Emph MPrt nuts buy.PaSbj ‘What if Dilip doesn’t buy nuts from the shop?’

b 0 . Dilip naa-i baa dokaan theke baadaam kinlo Dilip Neg-Emph MPrt shop from nuts buy.PaSbj ‘What if Dilip doesn’t buy nuts from the shop?’ In the ‘‘standard’’ word order considered earlier and repeated in (57), the Neg-Emph-supported Modal Particle immediately precedes the verb. Unlike a simple /naa/ that must stick to that position, however, a /naa-i baa/ sequence—Neg-Emph plus Q-baa—can appear displaced to the left in examples (58a, b) and further to the left in the primed examples (58a 0 , b 0 ), provided that one does not place any interruptor between Neg-Emph and Q-baa. I tentatively note that Neg, despite its distance from the verb, can keep its features connected to the TAM head system if the Q-baa it is attached to counts as keyed in to that system by virtue of features covertly moved to M. A formal account is provided in section 4. My more specific proposal that helps make sense of the distribution of Qbaa is that it needs two kinds of licensing: Mood-Salience Licensing from some Salient item and Local Focus Licensing from one bearing Dissociative Emph features. On this basis, I begin to build a formal account, considering first the M-Salience Licensing of Q-baa. In this section I shall take up cases without Neg, leaving negative examples for section 4 to handle. Consider the contrast between the interrogative /ki/ ‘what?’ in (59), which licenses Q-baa, and its relative counterpart /jaa/ ‘what’ in (60), which does not. Thus (60) is well-formed only if Q-baa is omitted. The sequential relative construction of (5) and the interrogative in (59) share many properties. But they di¤er in P-features. Unlike relatives, interrogatives are MSalient. The task, then, is to handle the derivation of (59), bearing in mind that the version of (60) with Q-baa is ill-formed:

68

Probal Dasgupta

(59)

Prodip tomaake-i baa ki khaawaae? Pradip you.Dat-Emph MPrt what feeds ‘And what does Pradip serve YOU?’

(60)

Prodip tomaake-i (*baa) jaa khaawaae, setxaa Pradip you.Dat-Emph (*MPrt) what feeds, that dekhe Dilip to abaak on-seeing Dilip Prt surprised ‘As for what Pradip serves YOU of all people, Dilip was really astonished when he saw that stu¤ ’

My first move is to claim that Q-baa merges as a non-projecting (particle-type) head with a sister neutrally called a ‘‘TamP’’ here. Let the /tomaake-i ki khaawaae/ ‘feeds you what’ segment of (59) allow Q-baa to be merged with it, yielding (61), a stage of the derivation to be followed by the preposing of the emphasized constituent ‘you.Dat-Emph’ through mechanisms explained below: (61)

[TamP [Prt baa][TamP tomaake-i ki khaawaae]] MPrt you.Dat-Emph what feeds

I make the general assumption that Q-baa is a functional head of the Particle type and therefore lets its substantive sister continue to project. (This assumption is as optional as TamP; systems that need to treat Prt as a projecting head will of course do so.) One specific assumption, stated in section 1 (discussion surrounding example (60)) for the Particle /to/ and also relevant for Q-baa, since both these Prts bear P-features, is that such a Prt may occur at various niches within the clause body but must covertly move its Pfeatures to the M head of the clause. As becomes apparent in the discussion below, many other items within the TamP sister of Prt engage in feature tra‰c with Prt, which thus serves as a conduit to the M head. My second move is to formalize one of these pieces of tra‰c, M-Salience Licensing: an uninterpretable M-Salience-seeking contextual feature in Prt’s feature bundle picks out, within its TamP sister, a matching M-Salient feature, which covertly moves to Prt and deletes that contextual feature. This process does not change the constituent sequence of string (61)—which will be a¤ected by another operation yet to be discussed—but it does create a bond between the Q-baa and the Interrogative /ki/ there. Summarizing so far: (62)

Some Properties of the Modal Particle Q-baa a. It belongs to the category Prt (Particle) but bears special contextual features.

Q-baa and Bangla Clause Structure

69

b.

When it Merges, /baa/ takes a TAM-type sister, forming [TamP Prt TamP], with Prt’s features not halting the projection from sister TamP to mother TamP, though there may be percolation or other mechanisms enabling the mother to inherit unsatisfied requirements from Prt.

c.

M-Salience Licensing: /baa/’s uninterpretable (contextual) formal feature [[þM-Salient]] picks out, within its TamP sister, a substantive feature [þM-Salient]. This matching feature covertly moves to Prt and deletes the contextual formal feature there.

My third move is to formalize Local Focus Licensing by adding a fourth clause (d) to the analysis (62). In formulating this clause, I am assuming that a contextual feature that triggers overt movement exhibits Suicidal Greed, one of the very few proposals subsequent to Chomsky 1995 that I adopt in this study, as mentioned at the outset (later versions formalize it under the Probe-Goal reworking of the system): (62)

d.

Local Focus Licensing: A strong ‘‘Local Focus’’ contextual feature of /baa/’s matrix Attracts, and is checked by, an item LI within TamP bearing [þEmph(atic), þDissoc(iative)], features whose e¤ect is to endow LI with the Emph /i/ and which LI picks up optionally at entry into the numeration.

I will not labour the point that (62d) needs to be read with the discussion of Word Extension Strategies in section 2. Syntactically speaking, the e¤ect of proposal (62d) is as follows. The derivation, after the stage shown at (61), has already string-vacuously complied with (62c) and now, in conformity with (62d), preposes the Emphasizer-laden constituent /tomaake-i/ ‘you.Dat-Emph’: (63)

[TamP tomaake-i [TamP [Prt baa][TamP ki khaawaae]]] you.Dat-Emph MPrt what feeds

I leave a couple of questions open: whether ‘you.Dat-Emph’ gets substituted or adjoined; whether the attractor feature that makes it prepose is Prt’s own contextual feature as stated in (62d) or some copy of the feature foisted on to a piece of TAM machinery to ensure trouble-free preposing; and whether further processes responsible for the proper placement of the agent /Prodip/—deliberately not shown in the derivation so far, but see (64) below—result in a multiple specifier construction and, if so, at what functional shell. With these bets hedged, the tree for (59) comes out as (64), with the ubiquitous hedge constituent ‘‘TamP’’ leaving open some

70

Probal Dasgupta

details concerning VP, vP, split-VP, TP, MP. That this leaving open does not a¤ect the points made here is of course an empirical claim: (64)

[TamP Prodip [TamP tomaake-i [TamP [Prt baa][TamP ki Pradip you.Dat-Emph MPrt what khaawaae]]] feeds ‘And what does Pradip serve YOU?’

If this is how the account handles (59), recast below as (65i) with its sequential format made explicit, one needs to confirm that it can handle sequences such as (65ii, iii): (65)

i.

sequence: XP-Emph MPrt WH V: Prodip tomaake-i baa ki khaawaae? Pradip you.Dat-Emph MPrt what feeds ‘And what does Pradip serve YOU?’

ii.

sequence: V-Emph MPrt WH: Dilip esab baadaam khaawaae-i baa kaake? Dilip these nuts feeds-Emph MPrt who.Dat ‘And who does Pradip SERVE these nuts to?’

iii. sequence: WH-Emph MPrt V: Dilip kaake-i baa esab baadaam khaawaae? Dilip who.Dat-Emph MPrt these nuts feeds ‘And WHO does Dilip serve these nuts to?’ I try (65ii) first. The relevant stage of the derivation is (66): (66)

[TamP [Prt baa][TamP kaake esab baadaam khaawaae-i ]] MPrt who.Dat these nuts feeds-Emph

It is now possible to prepose any constituent that counts as emphasized. The verbal head itself counts and if (optional) percolation applies, so do larger phrases including it. This e¤ectively allows movement of the minimal ‘feeds-Emph’ (67a) or of the larger ‘these nuts feeds-Emph’, which is the case one is considering at (65ii)/(67b) of the even larger ‘who.Dat these nuts feeds-Emph’, though this leads to a slight drop in acceptability shown at (67c): (67)

a. [TamP khaawaae-i [TamP baa[TamP kaake esab feeds-Emph MPrt who.Dat these baadaam t]]] nuts t

Q-baa and Bangla Clause Structure

71

b. [TamP esab baadaam khaawaae-i [TamP baa [TamP these nuts feeds-Emph MPrt kaake t]]] who.Dat t ?c. [TamP kaake esab baadaam khaawaae-i [TamP who.Dat these nuts feeds-Emph baa [TamP t]]] MPrt t I will ignore the acceptability drop at (67c) as possibly an e¤ect of other factors, and assume that the Emph feature on a head freely percolates up to its mother and beyond in the manner of the wh feature in overt whmovement languages. This entails that a particular derivation where percolation has proceeded up to some node will show as much pied-piped material (in addition to the obviously Emph-bearing head) as that node contains. The study of Emphasis, like that of Scrambling, is an entire subenterprise best left to its practitioners. Beyond making my provisional assumptions explicit, which I have, I do not address such issues as how Emph percolation and the determination of the scope of a focused item really work. In set (67), I examine only example (67b), the case of concern at (65ii). What happens after Q-baa is merged at (66) on the way to (67b)? Recall that the first thing that happens is string-vacuous. M-Salience Licensing applies, with Q-baa seeking out the M-Salient feature at the interrogative /kaake/ ‘who.Dat’ and having this feature covertly move to Prt and check its contextual feature. No constituent moves overtly. The second thing that happens is the preposing noticed at (67b), a derivation where the Dissociative Emph feature has percolated up to the projection containing just ‘these nuts feeds-Emph’ and thus it is this node that preposes, attracted by the Local Focus Licensing contextual feature at Prt. After stage (67b), the processes I have been systematically ignoring e¤ect the proper placement of the agent to yield (68), as desired: (68)

[TamP Dilip [TamP esab baadaam khaawaae-i [TamP baa [TamP Dilip these nuts feeds-Emph MPrt kaake t]]] who.Dat t ‘And who does Dilip SERVE these nuts to?’

I now turn to the final case, (65iii), repeated below: (65)

iii. sequence: WH-Emph MPrt V: Dilip kaake-i baa esab baadaam khaawaae? Dilip who.Dat-Emph MPrt these nuts feeds ‘And WHO does Dilip serve these nuts to?’

72

Probal Dasgupta

This derivation proceeds virtually like that of (65ii) just considered. Q-baa merges to derive ‘MPrt who.Dat-Emph these nuts feeds’. The M-Salience feature in Q-baa exhibits Suicidal Greed, driving covert feature movement from an interrogative and thus M-Salient constituent, ‘who.Dat-Emph’. Then the Local Focus feature acts, forcing the Emphatic item, here unambiguously ‘who.Dat-Emph’, to prepose and yield the order observed. Surely it is for Emphasis studies to say just why an emphatic argument like ‘who.Dat-Emph’ in (65iii) cannot have its Emph percolate upwards into TamP projections, whereas an emphatic verbal head like ‘feeds-Emph’ in (65ii) can. The examples considered so far feature interrogatives as the items satisfying the Modal Particle’s M-Salience requirement. I turn now to cases where this job is done by the Past Subjunctive. Consider: (69)

Dilip baadaam khelo-i baa Dilip nuts eat.PaSbj-Emph MPrt ‘It’s fine if Dilip eats/has eaten nuts’

Analysis (62) applies here as follows. The derivation builds a TamP ‘nuts eat.PaSbj-Emph’ and merges Q-baa: (70)

[TamP [Prt baa][TamP baadaam khelo-i ]] MPrt nuts eat.PaSbj-Emph

First, the Modal Particle’s M-Salience feature seeks out, triggers covert movement of, and gets checked by the M-Salient feature/s of ‘eat.PaSbjEmph’. The Local Focus feature in Q-baa’s TAM bundle further needs an Emph constituent preposed. By processes that are familiar by now, the Emphatic constituent ‘nuts eat.PaSbj-Emph’ moves and e¤ects the Checking required, yielding the surface order of (69). In other words, analysis (62) can also handle examples where TAM itself contains an M-Salient feature. I will thus assume that (62) adequately characterizes the relevant properties of Q-baa, which I can now use in order to study the three Liquid Tense-Moods in interaction with Negation—the business of the next section.

4. Negation and the Pasts On the basis of the results arrived at in section 3, the present section investigates the interaction between Neg Placement and the Past Tense-Moods highlighted earlier. The purpose is to further clarify the behaviour of clauses that exhibit various levels of in/dependence, using the tools developed so far.

Q-baa and Bangla Clause Structure

73

Some crucial examples of the Past Tense-Moods are given below in juxtaposition with ill-formed variants to facilitate analysis: (71)

(72)

Past (Indicative) a. Dilip baadaam khelo naa-i baa keano? Dilip nuts ate Neg-Emph MPrt why ‘Why, indeed, did Dilip not eat nuts?’ b.

*Dilip baadaam naa khelo-i baa keano? Dilip nuts Neg ate-Emph MPrt why

c.

*Dilip baadaam naa-i baa khelo keano? Dilip nuts Neg-Emph MPrt ate why

d.

*Dilip baadaam naa-i baa keano khelo? Dilip nuts Neg-Emph MPrt why ate

Past Subjunctive a. Dilip baadaam naa-i baa khelo Dilip nuts Neg-Emph MPrt eat.PaSbj ‘Dilip had better not eat nuts; it’s fine if Dilip doesn’t eat nuts’ b.

?Dilip baadaam naa khelo-i baa Dilip nuts Neg eat.PaSbj-Emph MPrt

c.

*Dilip baadaam khelo naa-i baa Dilip nuts eat.SubjPret Neg-Emph MPrt

I shall consider the well-formed (a)-cases first and turn to the deviant strings later. The examples under examination in this section involve not just the Modal Particle Q-baa and the factors from section 3 that facilitate or prevent its licensing, but also Neg, whose variable placement can now be handled in detail on the basis of the analysis given in section 2. Recall that that account is built around a basic contrast between the postverbal independent TAM-featured Neg with a regular finite verb— such as ‘ate’ in (71a)—and the dependent and thus preverbal Neg, instantiated here at (72a). The account provided in section 2 first Merges an independent Neg and then head-moves (left-adjoins) a finite verb to it, while it Merges dependent Neg as an X-zero level adjunct extending an already Merged head. Only TAM-bearing heads in this account are eligible hosts for dependent Neg. We can use these ideas directly to understand how (71a) first Merges an independent Neg as itself a TAM bearer to which ‘ate’ subsequently headmoves. But more needs to be said for (72a), where the Modal Particle Qbaa interrupts the sequence of preverbal Neg and verbal host. This is where

74

Probal Dasgupta

the actions of the syntax module show that they are independent of the elementary morphology formulated earlier. This independence raises new questions about the proper co-characterization of our phenomena across modules. Proper answers to those questions, not all of them formulable on the basis of data at our disposal, will have to allow reasonable amounts of independent action by each module, despite some overall imperative of cross-module cooperation to the extent possible. The scale of the problem posed by (72a) becomes obvious when we recall data showing that the Neg-Emph plus Modal Particle sequence can occur further to the left, away from the verbal complex. Thus consider sentence (72d), where the Neg-Emph Prt sequence is immediately preverbal, and variants (72d 0 , d 00 ), with this sequence displaced leftward: (72)

d.

Dilip dokaan theke baadaam naa-i baa kinlo Dilip shop from nuts Neg-Emph MPrt buy.PaSbj ‘What if Dilip doesn’t buy nuts from the shop?’

d 0 . Dilip dokaan theke naa-i baa baadaam kinlo Dilip shop from Neg-Emph MPrt nuts buy.PaSbj ‘What if Dilip doesn’t buy nuts from the shop?’ d 00 . Dilip naa-i baa dokaan theke baadaam kinlo Dilip Neg-Emph MPrt shop from nuts buy.PaSbj ‘What if Dilip doesn’t buy nuts from the shop?’ I propose the following treatment. In set (72d–d 00 ), the Merger of Q-baa occurs at point (72e)—suppressing the Agent, which is treated along the lines mentioned in the fuller presentation of the analysis of the Modal Particle in section 3: (72)

e.

[TamP baa [TamP dokaan theke baadaam naa-i MPrt shop from nuts Neg-Emph kinlo]] buy.PaSbj

The simplest case involves preposing of the emphasized constituent NegEmph from (72e) to yield (72f ), modulo proper Agent placement to derive (72d 00 ): (72)

f.

[TamP naa-i [TamP baa [TamP dokaan theke baadaam Neg-Emph MPrt shop from nuts t kinlo]]] t buy.PaSbj

Q-baa and Bangla Clause Structure

75

This treatment regards cases (72d, d 0 ) as more complex. ‘From the shop’ moves to the left in (72d 0 ), and both ‘from the shop’ and ‘nuts’ in (72d), perhaps into Spec of MP positions, if one chooses such an implementation of the TamP format employed here as a theory-neutral expository device. (72a) is like (72d 0 ), involving step (72f ) and additional preposing of one argument—‘nuts’, in (72a). Such an analysis needs to, and can, face some construction-specific questions. Why does Neg-Emph never pied-pipe any material? Presumably the reason is that the features of a functional adjunct to a lexical head cannot percolate, which prevents Neg-Emph from percolating its Emph feature upwards the way Verb-Emph could in section 3. Second question: how does one know that Neg-Emph can be Merged to the left of a subjunctive verb? This one receives a direct factual reply—other instances of the phenomenon exist: (73)

Oraa jodi tomaake naa-i dxaake, takhon ki they if you.Dat Neg-Emph invite.PrSbj, then what korbe? do-Fut ‘If they end up NOT inviting you, then what will you do?’

(74)

Aamraa baronco oder naa-i dxaaklaam we better they.Dat Neg-Emph invite.PaSbj ‘We had better NOT invite them’

Third question: what is the basis for the claim that the landing site of preposed Neg-Emph in such sentences has something to do with the Prt and does not just happen to seem connected to it? Again, this claim has an empirical basis: (75)

a.

aamraa jodi esab baadaam naa-i khaawaai we if these nuts Neg-Emph feed.PrSbj ‘if we do NOT serve these nuts’

b.

result of preposing Neg-Emph to left of Prt in (75a): aamraa naa-i jodi esab baadaam khaawaai we Neg-Emph if these nuts feed.PrSbj ‘if we do NOT serve these nuts’

c.

result of preposing Neg-Emph to arbitrary intermediate niche: *aamraa jodi naa-i esab baadaam khaawaai we if Neg-Emph these nuts feed.PrSbj

76

Probal Dasgupta

Data set (75) shows that another Prt with P-features, /jodi/ ‘if ’, attracts Neg-Emph to its left, optionally—the process applies in (75b) and does not in (75a); and that Neg-Emph cannot land at an intermediate site that the Prt does not attract it to, as in (75c). I take it, then, that the account of negation and the modal particle outlined in sections 2 and 3 can handle the well-formed cases (71a) and (72a). I now address the ill-formed variants, beginning with the regular Indicative set (71b–d), repeated for convenience: (71)

b.

*Dilip baadaam naa khelo-i baa keano? Dilip nuts Neg ate-Emph MPrt why

c.

*Dilip baadaam naa-i baa khelo keano? Dilip nuts Neg-Emph MPrt ate why

d.

*Dilip baadaam naa-i baa keano khelo? Dilip nuts Neg-Emph MPrt why ate

Ill-formed (71b)’s preverbal Neg minimally contrasts with acceptable (71a)’s postverbal Neg and with the lack of Neg in (76): (76)

Dilip baadaam khelo-i baa keano? Dilip nuts eat.PaSbj-Emph MPrt why ‘Why DID Dilip eat nuts?’

What makes (71b) unacceptable must then be its preverbal Neg clashing with its indicative verb, which calls for postverbal Neg for reasons reviewed above. But implementing this idea poses a formal challenge. For the analysis of negation in section 2, unaided, stops short of ensuring that (71a) blocks (71b). By hypothesis, (71a)’s independent Neg and (71b)’s dependent Neg di¤er featurally, distinguishing the numerations. This precludes Blocking. One must find some other way consistent with an otherwise wellsupported account to make (71b) come out ill-formed in Bangla despite the existence of dependent Neg in the Bangla lexicon. To this end, I propose that Bangla parametrically allows a dependent Neg extension of a head only if the head bears P-features that check at M— call them ‘‘M-features’’. This auxiliary hypothesis, based on subjunctiveindicative contrasts presented in detail in Dasgupta (1996), completes the account of (71a) vs. (71b). (71c, d) are tricky cases that help bring out some aspects of the analysis that might otherwise have remained inexplicit. In these sentences NegEmph is Merged as an independent TAM-bearing head to which the regular tensed verb later head-adjoins. At the juncture in the derivation at which the Modal Particle Q-baa comes in, the string looks like this (with suppression of Agent standard by now, and position of ‘why’ ignored):

Q-baa and Bangla Clause Structure (71)

e.

77

[TamP baa [TamP baadaam [khelo [naa-i ]]]] MPrt nuts [ate [Neg-Emph]]

If the functional head ‘Neg-Emph’ is to prepose, it must carry its lexical head-extension ‘ate’ along, as in the well-formed example (71a) considered earlier. Cases (71c, d) are unacceptable on this analysis because they would involve ‘Neg-Emph’ performing an illicit solo movement, stranding the lexical verb ‘ate’ in head-adjoined position. (71c) di¤ers from (71d) only with respect to the position of ‘why’, a detail ignored in representation (71e); thus, the reasoning addresses both the cases. That concludes the discussion of the default or indicative Past set. I turn to the ungrammatical cases in the Past Subjunctive set (72) repeated below for expository convenience: (72)

Past Subjunctive a. Dilip baadaam naa-i baa khelo Dilip nuts Neg-Emph MPrt eat.PaSbj ‘Dilip had better not eat nuts; it’s fine if Dilip doesn’t eat nuts’ b.

? Dilip baadaam naa khelo-i baa Dilip nuts Neg eat.PaSbj-Emph MPrt

c.

*Dilip baadaam khelo naa-i baa Dilip nuts eat.PaSbj Neg-Emph MPrt

The well-formed case (72a) has been done above. The Modal Particle Qbaa Merges. Its M-Salience is satisfied by covert movement of the MSalient feature of the Past Subjunctive verb. Its Local Focus contextual feature summons the Emph-bearing Neg. This Neg-Emph, as a functional head in a head-adjunct position, has not percolated its feature upwards, and therefore must prepose alone, yielding (72a). One is now looking at the other cases. The clearly ungrammatical case (72c) is straightforward. It tries to place a dependent Neg to the right of a subjunctive verb. But the present analysis makes it impossible for dependent Neg to end up in such a position. Such a Neg merges as a head-extension to the left of a subjunctive verb. This Neg, being emphatic, must move to Q-baa’s left all by itself, yielding (72a). A derivation from this to (72c) would have to prepose the subjunctive verb /khelo/ to the left of the Neg-Emph Prt sequence—and there is no landing site for a head there. With case (72b), the problem is not so much its ill-formedness as the fact that it sounds as good as it does. Elsewhere in Bangla, the sequence Neg plus Subjunctive Verb-Emph is generally excluded. Compare the well-

78

Probal Dasgupta

formed examples (73) and (74) above with the unacceptable result of placing their Emph on V instead of on Neg: (77)

*Oraa jodi tomaake naa dxaake-i, takhon ki they if you.Dat Neg invite.PrSbj-Emph, then what korbe? do-Fut ‘If they end up not inviting you, then what will you do?’

(78)

*Aamraa baronco oder naa dxaaklaam-i we better they.Dat Neg invite.PaSbj-Emph ‘We had better not invite them’

In contrast, (72b) is only marginally unacceptable. Why should this be so? I suspect that it has something to do with the following independently confirmable fact. Q-baa, which so compulsively calls for a Dissociative Emph bearing item to its immediate left, is able to inflict this feature on an item that otherwise could not bear it, provided that the item satisfies Qbaa’s need for M-Salience licensing. The principle seems, descriptively, to be that a Q-baa that can get M-Salience licensing in sentence S is not going to let S get away with withholding Local Focus licensing; if need be, Q-baa will simply impose Dissociative Emph features on the M-Salience licensor. Thus, interrogative constituents in Bangla normally cannot bear Emph features of any kind, witness the Dissociative Emph case in (79b) and the Associative Emph case in (79c); but the presence of Q-baa saves the interrogative bearing Dissociative Emph in (79a): (79)

a.

Tomraa kaake-i baa dxaakbe? you who.Dat-Emph MPrt invite-Fut ‘Who, indeed, will you invite?’

b.

*Tomraa kaake-i dxaakbe? you who.Dat-Emph invite-Fut

c.

*Tomraa kaake-o dxaakbe? you who.Dat-Emph invite-Fut

I have no explanatory account of this, but the descriptive package perhaps su‰ces to render the relative well-formedness of (72b) intelligible. This concludes the discussion of all the obvious examples of the way Negation and the two Past Tense-Moods interact with the phenomena surrounding the Modal Particle Q-baa. One finds a reasonable fit between theory and data.

Q-baa and Bangla Clause Structure

79

As one concludes, it is useful to recall the problematic phenomenon that led to the entire exercise—the fact that the Past Subjunctive form in Bangla shares the morphology of the simple (indicative) Past but di¤ers syntactically in its special ability, not shared by the simple Past, to directly license the Particle Q-baa—and the more general concern with some behavioural di¤erences between very dependent clauses and the very independent clauses that can harbour Particles bearing such P-features that link it to an independent (non-a‰xal) M head. The initial question seems to have been addressed. To that extent, one is ready to celebrate the discovery of a syntactic Past Subjunctive in a language where it shares all its morphology with the simple Past and has therefore been di‰cult to detect. The Past Subjunctive form in Bangla, unlike other Pasts and unlike other Subjunctives, has turned out to share an MSalience property with interrogatives. Future work will no doubt dissolve the descriptive notion of Mood Salience into more generally usable Pfeatures; but the present package su‰ces to turn the initial pretheoretical problem into one that is amenable to formal investigation. The general concern has been addressed to the extent that one now has a more detailed map of the independent MP in Bangla. My cartography has skirted a variety of boundary disputes, hence the ‘‘TamP’’ device and many hedged bets. But the account given above does make it clear, for instance, that the Past Subjunctive clause is consistently a root sentence. Its Pfeatures and the behaviour of its M do not allow it to depend on any superordinate clause. But it is subjunctive in that it requires pre-verbal or dependent Neg. The interaction between Neg and the TAM system (in relation to parametric properties of T-centred Bangla as distinct from, say, M-centred Hindi-Urdu), and the interaction between Emph features and other tra‰c in a clause that harbours as aggressive a marker of independent clause status as the Modal Particle Q-baa, are now better understood. This is an empirical step forward, to the extent that one can continue this e¤ort with other particles that show interesting properties. There is little doubt, however, that such facts as the distribution of Q-baa will need an adequate account in any serious descriptive framework, and that explanatory theories cannot a¤ord to neglect descriptive problems of this magnitude. Abbreviations and Transcription Acc: Accusative. Agr: Agreement. CondP: Conditional Participle. ConjP: Conjunctive Participle. Dat: Dative. Emph: Emphasizer. Fut: Future. Gen: Genitive. Imp: Imperative. Indic: Indicative. Inf: Infinitive. Loc: Locative. Lower-case m: masculine. MPrt: Modal Particle. M-Salience: Modal Salience. Neg: Negative. Nom: Nominative. Nonfin: Nonfinite. P: lexical Ad-

80

Probal Dasgupta

position. Lower-case p: functional Adposition. Pa: Past. Pl: Plural. Pr: Present. Prog: Progressive. Prt: Particle. Quot: Quotative. Sbj: Subjunctive. The retroflex stops tx txh dx dxh and flap rx contrast with dental t th d dh and alveolar r. The nasalized vowels ix ux etc. contrast with the oral vowels i u etc. The digraphs ea and aa symbolize low front and low back unrounded vowels respectively. The vowel symbol a in Bangla stands for a low back rounded vowel.

Acknowledgements For comments on earlier presentations of this material, I wish to thank Indrani Roy, Gautam Sengupta, Noam Chomsky, Ayesha Kidwai, Veneeta S. Dayal, Anoop K. Mahajan, participants at M.T. Hany Babu’s defence, the guest editor of this issue, the chief editor of the Yearbook, and the anonymous referee. The usual disclaimers apply.

References Chomsky, Noam 1995 The Minimalist Program. Cambridge: MIT Press. Dasgupta, Probal 1987 Sentence particles in Bangla. In Elena Bashir, Madhav M. Deshpande, Peter Edwin Hook (eds.) Select Papers from SALA7. Bloomington: IULC. 49–75. 1994 Verbless and nonfinite clauses in Bangla. PILC Journal of Dravidic Studies 4:1.57– 67. 1996 Remarks on subjunctivity. In Shivendra Kumar Verma, Dilip Singh (eds.) Perspectives on language in society: Papers in memory of Prof. Srivastava. Vol. 1. Delhi: Kalinga. 72–89. 1997 Postlexical morphology in parametric grammar. In Kumaralingam Gopalakrishnan Vijayakrishnan and Punnappurath Madhavan, (eds.) Phases and Interfaces. Hyderabad: CIEFL. 2002 Postpositions and parameters. In Vadlakonda Swarajya Lakshmi (ed.) Case for Language Studies: Papers in Honour of Prof. B. Lakshmi Bai. Hyderabad: Centre for Advanced Study in Linguistics (Osmania University) and Book Links. 123–140. 2005 Mejaajer herpher aar Baanglaa kriyaapad [Mood alternation and the Bangla verb phrase]. Onustxup 39:4.105–115. Dasgupta, Probal, Ford, Alan, and Singh, Rajendra 2000 After Etymology: Towards a Substantivist Linguistics. Muenchen: Lincom Europa. Ford , Alan, Singh, Rajendra, and Martohardjono, Gita 1997 Pace Panini. New York: Peter Lang. Hany Babu, Musaliyarveettil Tharayil 1997 The syntax of functional categories. Central Institute of English and Foreign Languages (Hyderabad) Ph.D. dissertation. Laka, Itziar 1990 Negation in syntax: on the nature of functional heads and projections. Massachusetts Institute of Technology Ph.D. dissertation.

Q-baa and Bangla Clause Structure

81

Roy, Indrani 1996 The expression of Tense and the placement of Neg in the Chittagong Bangla functional head sequence. University of Hyderabad M.Phil. dissertation. Singh, Rajendra and Probal Dasgupta 1999 On so-called compounds. The Yearbook of South Asian Languages and Linguistics 1999: 265–275. Singh, Rajendra and Starosta, Stanley (eds.) 2003 Explorations in Seamless Morphology. New Delhi: Sage. Trivedi, Ramendrasundar 1905 Naa [Not]. In his (1950) Raamendroraconaaboli, vol. 3. Kolkata: Bangiya Sahitya Parisat. 9196. Zanuttini, Ra¤aella 1991 Syntactic properties of sentential negation: a comparative study in Romance languages. University of Pennsylvania Ph.D. dissertation.

Phrasal Predicates: How N Combines with V in Hindi/Urdu ALICE DAVISON

Transitive complex predicates are productively formed in Hindi/Urdu by combining a complex event nominal with a variety of verbs. The thematic object of N has genitive, locative or other object case. I propose that this nominal is the syntactic object of the V, but instead of discharging an argument position of V, as a normal object would, instead the argument structures of N and V are merged. The combined argument structure forms a saturated phrase which meets LF conditions on full interpretation, with no open argument positions at the end of the derivation. This account derives the full range of case marking as well as the correct semantic roles of the syntactic subject, goals which are not met by argument transfer to empty light verbs or semantic incorporation, proposed as ways of deriving complex predicates. 1. Introduction Many languages form a large part of their verbal lexicon by combining a verb such as ‘do/make’ with an adjective or complex event noun.1 The combination of N and V constitutes a syntactic phrase with a single argument structure. Complex predicates of this sort are found in Japanese, Turkish, and Persian, among many others. Since complex predicates are so generally available, they must be the result of very general linguistic processes, which I will propose are part of syntax and the interpretative interface. In this paper I consider two kinds of syntactic derivations, incorporation and argument merger, or formation of a composite argument structure. The goal of this analysis is to show how N and V contribute to the properties of the complex predicate, and to derive the right phrase structure, case and agreement properties. Complex predicates form a large and productive part of the verbal lexicon of Hindi/Urdu (1)–(2): (1)

maiN-nee/ *mujhee -[[un-kii yaad ] kii]. I-erg / *I-dat 3pl-gen-fs memory-fs do-pf-fs ‘I remembered, recalled them.’

84

Alice Davison

(2)

mujhee/ *maiN-nee [[un-kii yaad ] aaii]. I-dat / *I-erg 3pl-gen memory-fs come-pf-fs ‘I remembered/missed them.’

(3)

[un-kii yaad ]. 3pl memory-fs ‘a memory, recollection about them.’2

(4)

maiN-nee kyaa ki-yaa? I -erg what? do-pf-ms ‘What did I do?’

The N yaad ‘memory’ takes a thematic object, the thing remembered (3), and implies an experiencer, the one who remembers. In combination with a V (4), this external argument is syntactically projected with nominative, ergative (1) or dative (2) case. While sentences like (1)–(2) are common and unremarkable in Hindi/ Urdu, their structure poses both general and language-specific problems for deriving the right syntax, morphology and interpretation. The N component of N-V is the syntactic object of V, with the null nominative object case and verbal agreement. (fs in this case) required when the subject is marked by a postposition, blocking agreement with the subject. The N yaad ‘memory’ has a syntactic object of its own which requires case licensing (3). The N selects the case for it, in this instance genitive (1)–(3). The V determines the case of the external argument of the sentence, ergative when the V is a verb like kar-naa ‘do’ (1), dative with V aa-naa ‘come’. The combination forms a syntactic phrase, which can be interrupted by negation and particles, modified by an adjective on N or reordered. There are also interpretive/semantic problems associated with this structure. N is a complex event nominal (Grimshaw 1990), which like V has an argument structure, as independently occurring words (3)–(4). But the semantic roles in these argument structures may not be the same. The external argument of yaad ‘memory’ has primarily an experiencer role (1)–(3), while the external argument of kar-naa ‘do/make’ is typically agentive (4) and the subject of aa-naa ‘come’ refers to an entity with or without internal motivation which moves to a destination. The argument structures of both N and V have to be reconciled as one argument structure, preserving the right semantic relations for the arguments. The subjects in both (1)–(2) have experiencer interpretation, so that these sentences may be synonymous in specific contexts. For arguments that ergative case is a structural case not linked to semantic agency, see Davison (2004a, b). Many but not all of these problems have been articulated from di¤erent descriptive or theoretical viewpoints, from earlier writers such as Bailey

Phrasal Predicates

85

(1963), Kachru (1982), Bahl (1974), Hook (1979), to more recent work by Montaut (1991), Verma (1993b), Mohanan (1994), Butt (1995), and Butt and Ramchand (2003). The goals of this paper are (a) to form a comprehensive picture of the facts about N-V complex predicates in Hindi/Urdu, (b) to propose an account which preserves the productive syntactic nature of complex predicates, showing their relation to related simplex predicates in the lexicon of the language, (c) to account for case and syntactic structure, and (d) to explain how the properties of N and V are combined. I will relate N-V predicates to other classes in the lexicon of Hindi/Urdu, drawing some conclusions about the di¤erences in their derivation and syntactic projection from N-V predicates. I compare two possible theoretical approaches, one based on incorporation, as in Hale and Keyser’s conflation process in lexical syntax (1993), the other based on Higginbotham’s predicate composition (1985, 1999). 2. The Phrasal Syntax of N-V Combinations The N of the N-V combination may be independently modified by an adjective or adverb (5)–(6). Negation or emphatic particles may intervene between the N and the V (6)–(7): (5)

us-nee moohan-kii bahut yaad kii 3s-erg Mohan-gen much memory-fs do-pf-fs ‘He/she remembered Mohan very much.’ (Bahl 1974: xxix)

(6)

coor-nee [apnee dimaag-kaa sahii isteemal ] nahiiN ki-yaa] thief-erg self ’s mind-m-gen right use-m not do-pf-ms ‘The thief did not make right use of his mind.’ (Hook 1979: 159)

(7)

mujhee pataa hii/bhii nahiiN thaa I-dat knowledge-m only/also not was-ms ‘I did not even know.’ (Montaut to appear: 296)

The adjective/adverb bahut ‘much’ in (5) intervenes between the genitive object phrase and the N. Negation in (6) and (7) intervenes between the N and the V, while N has an adjective modifier sahii ‘correct, right’ in (6). The N itself behaves like a syntactic direct object. It has gender, person and number agreement features reflected on the verbal complex under just the same circumstances that thematic object DPs require agreement. Object agreement is required when the subject has non-nominative case, and the direct object also is not marked with a postpositional case. Compare the object agreement in (1) above with the subject agreement in (8). The simplex verbs in (9a,b) show the same contrast of subject and object agreement. Verbal agreement is determined by the nominative subjects in (8) and (9a),

86

Alice Davison

but when the subject is marked by postpositional case, the verb agrees with a nominative object. The verb agrees with yaad ‘memory’ in (1)–(2), and the nominative syntactic direct object in (9b). (8)

pulis (waalaa) coor-kii talaash kar-eegaa police (man) thief-gen-fs search-fs do-fut-3ms ‘The policeman will search for the thief.’

(9)

a.

baccee billii deekh-eeNgee child-mpl cat-fs see-fut-3pl-m ‘The children will see a cat.’

b.

baccooN-nee billii deekh-ii child-mpl-erg cat-fs see-pf-fs ‘The children saw a cat.’

In passive sentences, the constituent singled out as the direct object is the N component of N-V in (10), while it is the thematic object of the simplex verb in (11). (10)

pulis-kee dwaaraa coor -kii talaash kii ga-ii police-gen means thief-ms-gen-fs search-fs do-pf-fs go-pf-fs ‘The thief was not searched for by the police.’

(11)

baccooN-see billii deekh-ii nahiiN gaii child-mpl-from cat-fs see-pf-fs not go-pf-fs ‘The children couldn’t bear to see the cat; lit. the cat was not seen by the children.’3

The N component of N-V complex predicates has these two properties in common with referential direct object DPs: it triggers agreement when the subject is not nominative, and it is the object promoted to subject in passive sentences.4 The N component also behaves like an independent N linked with a DP complement by the genitive postposition, which is su‰xed to DP, and agrees with N. (12)

a.

raam -kii gaaRii Ram-ms gen-f car-fs ‘Ram’s car’

b.

raam-kii yaad Ram-gen-f memory-fs ‘the memory of/about Ram’

I will argue below that complex predicates are syntactically derived; they are not lexical idioms. 3. Other Lexical Classes of Predicates in Hindi/Urdu In this section I compare N-V predicates like (5)–(7) with other kinds of predicates related to nouns, especially complex event nominals. The Hindi

Phrasal Predicates

87

lexicon includes simplex verbs related to the same kind of N found in N-V combinations, and in many cases coexisting with N-V combinations. There are also N-V combinations in which the N and V form a closer unit than in the preceding examples, in some sense intermediate between a phrasal verb and a verb stem. In section 4, I use the discussion of these predicate classes to ask whether there is a common derivation for all types of predicates, by incorporation. 3.1. Denominal verbs There are many denominal verbs in Hindi/Urdu, such as those in (13)–(15). The formation of denominal verbs is no longer freely productive, but it remained productive after the massive of borrowing of Perso-Arabic vocabulary, such as Persian daaG, talaash, etc. (13)

a. b. c.

daaG ‘mark of burning’ N daaG-naa ‘burn, brand’ V daG-naa ‘be branded’5 V

(14)

a. b. c.

khooj ‘search’ N bahas ‘argument, dispute’ N talaash ‘search, investigation’ N

(15)

a. b. c.

khooj kar-naa ‘search, discover’ bahas kar-naa ‘argue, debate (with)’ talaash kar-naa ‘search for a concrete object’

khooj-naa ‘to search for’ V bahas-naa ‘to argue’ V talaash-naa ‘search according to abstract criteria’ V

The simplex denominal verbs in (14) coexist with the corresponding N-V complex predicates (15), often with a di¤erent range of meaning (Gambhir 1993). In particular, the verb khooj-naa ‘search’ contrasts with khooj karnaa ‘search, find’. The N-V predicate as an accomplishment meaning is not found with the simplex verb (16).6 (16)

ganapat singh-nee [eek naii bimaarii-kii khooj] Ganpat Singh-erg one new illness-gen search.fs.nom kii hai do.pf.fs is ‘Ganpat Singh has discovered a new disease.’ (Bahl 1974: 222)

The meaning of the complex predicate is not always completely predicted from the meaning of N and the simplex denominal verb. The selection of lexical case properties, however, often relates denominal verbs to complex predicates (17)–(18) via the shared N stems, suggesting that lexical case is associated with N.

88 (17)

(18)

Alice Davison a.

baccee billii-see Dar-tee haiN child.mpl cat-from fear-impf.mpl are ‘The children are afraid of the cat.’

b.

baccooN-koo billii-see Dar lag-taa hai child.mpl-dat cat-from fear.ms strike-impf is ‘The children are afraid of the cat.’

a.

coor apnii karnii-par pachtaa-yaa thief self ’s misdeed-on repent-pf ‘The thief repented his misdeeds.’

b.

coor-koo apnii karnii-par pachtaavaa hai thief-dat self ’s misdeed-on repentance is ‘The thief repented his misdeeds.’

3.2. Diatheses, verb alternations Complex predicates in Hindi/Urdu fill in paradigmatic gaps with respect to denominal verbs. There are a large number of related verbs of the kind in (19). Related Ns include khool ‘hollow’, pheer ‘turn’ and kaaT ‘cutting’. The verbs in each set share a basic root, but are distinguished by stem alternations and su‰xes, as well as by the number of arguments projected syntactically. The transitive stem (19b) takes both subject and object arguments. A related detransitive verb (19a) has a phonologically reduced stem, and projects the thematic object as syntactic subject. The same reduced stem combines with the causative a‰xes -aa and -vaa, increasing the valence of the verb by the addition of a causer. (19)

a.

detransitive khulnaa ‘to be open’ phir-naa ‘to turn’ kaT-naa ‘to be cut’

b.

transitive khol-naa ‘to open’ pheer-naa ‘to turn’ kaaT-naa ‘to cut’

c.

causative1 khul-aa-naa ‘to cause to be open’ phir-aa-naa ‘to cause to be turned’ kaT-aa-naa ‘to cause to be cut’

d.

causative2 khul-vaa-naa ‘to get opened by another’ phir-vaa-naa ‘to cause to be turned’ kaT-vaa-naa ‘to cause to be cut’

Phrasal Predicates

89

As Gambhir (1993) noted, some denominal verbs do not participate in this paradigm (20), and instead complex predicates are formed with the same valence properties as the missing derivatives. (20)

a.

intransitive talaash hoo-naa ‘be searched’ *talash-naa ‘be searched’

b.

transitive talaash kar-naa ‘search for’ talaash-naa ‘search for’

c.

causative talaash kar-vaa-naa ‘have someone search for’ *talash-vaa-naa ‘have someone search’ (Gambhir 1993: 89)

There is a category of complex predicates in which the N and V, while remaining separate words, form a closer association than the N and V which I have called phrasal (1), (2), and (6). The di¤erence from the truly phrasal N-V combinations is that the thematic object is syntactically the direct object. It can trigger verb agreement (20), (21), (23): (20)

bhaiyaa-nee apnii taaqat isteemaal kii brother-erg self ’s strength.nom.fs use.ms do.pf.fs ‘Brother used his strength.’ (Hook 1979: 158)

(21)

coor-nee apnaa dimaag (*-kaa) (*sahii) isteemal nahiiN thief-erg self ’s mind.m (*gen) right use.m not ki-yaa do.pf.ms ‘The thief did not make right use of his mind.’ (Hook 1979: 159)

(22)

us-nee [moohan-koo] bahut yaad ki-yaa 3s-erg Mohan-dat much memory.f do.pf.ms ‘He/she remembered Mohan very much.’ (Bahl 1974: xxix)

(23)

siitaa-koo [apnee maaN-baap] yaad aa-ee Sita-dat self ’s mother-father.mpl memory.fs come.pf.mpl ‘Sita remembered/missed her parents.’ (Saxena 1985)

The thematic object can have nominative case (20), (21), (23), or dative direct object case (22). Unlike N in examples such as (2), N is not linked by genitive case or locative case to its thematic object. Rather, the V itself determines the direct object case. This option is possible only for certain Ns, and for adjectives like mazbuur ‘forced’ which assigns no case (see section 5). Both denominal verbs and N-V complex predicates are transparently related to N. But the facts of the language suggest that denominal verbs N-V are not optional variants of one another. A simplex verb and a complex

90

Alice Davison

predicate related to the same N coexists in the language, but often with different nuances of meaning or belonging to di¤erent aspectual classes, such as khooj-naa ‘search’ and khooj kar-naa ‘search, discover’ (14)–(15). Many denominal verbs participate in transitivity alternations (19), but not all do, and the gaps in the paradigm are filled by N-V predicates (20), suggesting again that V and N-V have semantic commonalities but are not automatically available variants of one another. Finally, N-V combinations are for the most part phrasal, linked to a genitive or locative-marked thematic object, the choice determined by N. But a subset of these N-V complex predicates also form a more closely associated combination distinct from a simplex verb. 4. Incorporation as Model of Derivation and Interpretation In this section I consider proposals for incorporating nominal information into verbs, whether null, overt or present as causative a‰xes, so as to explain how N can be related in various ways to V. I will first consider the derivation of denominal verbs and the transitivity alternation by means of Hale and Keyser’s conflation, or movement of N into a verbal head in lexical syntax (1993). I then explore the extension of conflation to analytic or overt phrasal causatives, asking if this kind of derivation can account for the di¤erence between clearly phrasal complex predicates (1)–(2) and the more closely associated N-V counterparts (20)–(23). I next consider whether the lexical syntax of the conflation account captures the full range of event structures in the verbs discussed here, particularly those with experiencer subjects, and particularly when the experiencer interpretation is dictated by N and not V. Finally, I consider semantic incorporation, which preserves phrasal structure but combines information in Logical Form; I compare the interpretation of incorporation of indefinite NP objects with the incorporation of complex event nominals with their own full argument structure. This discussion serves as a background to the alternative I will propose in section 5, based on Higginbotham’s predicate composition by argument identification (1985, 1999). 4.1. Denominal verbs and transitivity alternations Hale and Keyser (1993) draw comparisons between the overt phrasal syntax which licenses objects and the object relations implicit in derived verbs, especially verbs of change, such as clear. This deadjectival verb is derived by their process of conflation by incorporating the Adjective clear. In l(exical)-syntax, the verb clear is equivalent to ‘V-cause the screen Vbecome clear’, with the Adjective clear incorporated into the abstract null V of causation and the inchoative V-become. The theta roles are directly read o¤ the l-syntax: the subject of V-cause has the agent interpretation, while ‘the screen’ gets the role of subject of change or theme. Intransitive

Phrasal Predicates

91

activity verbs which are ‘unergatives’, like dance are derived in l-syntax by incorporating the N dance which is the l-syntactic object of V do. We must extrapolate a little from Hale and Keyser’s account of English to derive the Hindi transitive denominal verbs such as those in (14), (17a) and (18a). An activity verb like talaash-naa ‘search for’ would have the lsyntactic structure (24): (24)

The N talaash ‘search’ undergoes head movement to the null V and v in l-syntax. The N takes on the category V and the syntactic DPs which get interpreted as semantic arguments of the verb. The same structure is the basis for verbs of change and causation, such as kaaT-naa ‘to cut’, and kaTnaa ‘to be cut’, related to N kaat ‘cutting’: (25)

a.

b.

The syntactic object (25a) is the subject of V, the verb of change (25b), while the syntactic subject is the specifier of v or the overt causative a‰x (25a). The various verb stems associate with V and v. The association is not free however, as the full form of the stem kaaT- is associated only with the transitive structure. The reduced form kaT associates in the intransitive

92

Alice Davison

and causative versions. Assuming there is some way to guarantee that the right form of the stem is associated with the right projection in l-syntax, the conflation approach gives a reasonably good result.7 4.2. Can incorporation be extended to ‘analytic’ complex predicates? Here I consider whether complex predicates can be given a similar account, substituting lexical verbs for the null V and v. The presence of lexical verbs blocks head movement in l-syntax. A proposal of this sort is made for Persian complex predicates by Folli et al. (to appear). Generalizing this proposal to Hindi/Urdu confronts problems in object case marking, and subject interpretation. There are di¤erences of case marking between Persian and Hindi/Urdu. Complex predicates in Persian mark the thematic object for case in the same way as simplex verbs, with null nominative or the di¤erential object marker -ra. Complex predicates in Hindi/Urdu have three di¤erent options for marking the thematic object of complex predicates. First, the thematic object may be linked by genitive case to N (agreeing with N in number-gender features) (1–2). Second, some Ns select locative lexical case in combination with V (17b), (18b). These options are the most common ones. Finally, the N-V forms a close combination giving the thematic object the same case as a simplex verb, either null nominative case or the di¤erential object marker -koo (20–23). This third option is the least common in Hindi, although it seems to be the most the preferred marking in Persian. This third possibility might be captured using a structure similar to (25a) for a sentence like (22), repeated here as (26a): (26)

a.

b.

us-nee moohan-koo bahut yaad ki-yaa 3s-erg Mohan-dat much memory-f do-pf-ms ‘He/she remembered Mohan very much.’ (Bahl 1974: xxix)

Phrasal Predicates (27)

a.

93

maiN-nee [[un-kii yaad ] kii] I-erg 3pl-gen-fs memory-fs do-pf-fs ‘I remembered, recalled them’

b.

The more usual expression of complex predicates involves a NP object containing the predicate N and its thematic object, which gets a case determined by N, genitive or locative (27a). The syntactic structure required for such sentences would seem to be (27b), which represents the thematic object as a constituent of NP, rather than a specifier of VP. It is hard to see how (27b) could be derived from (26b) without stipulation of object lowering or V 0 raising, and subsequent restructuring of the thematic object DP as a constituent of NP. A further problem is the semantic role assigned to the subject, which does not follow from these structures. Neither (26b) nor (27b) explains how the N determines the semantic role of experiencer for the syntactic subject, as the subject is not a syntactic argument of N. Both structures predict that only V assigns a theta role to the subject (as is the case for Persian, according to Folli et al. (to appear). In section 5, I will propose an alternative account which directly generates (27b) and the less usual (26a), and allows the N to determine the semantic roles of the subject.

4.3. The semantic role of the subject While many N-V complex predicates like talaash kar-naa ‘search’ have agentive subjects, not all such predicates do. Complex predicates such as yaad kar-naa ‘remember, recall, miss’, afsoos kar-naa ‘regret’ and mahsuus

94

Alice Davison

kar-naa ‘experience, feel’ have non-agentive experiencer interpretations for the subject: (28)

ab diwaalii paas aa rahii hai, is-liyee baccee now Diwali near come prog is so maaN-koo bahut yaad kar-tee haiN children mother-dat much memory do-impf are ‘Now Diwali is approaching, and for this reason the children miss their mother very much.’ (Bahl 1974: 73)

(29)

yee gumraah aurateeN peeshtar nahiiN too sharaab-kaa these wayward women earlier not then liquor-gen nashaa utar-nee-kee baad [apnii haalat]-par intoxication come down-inf-gen after self ’s condition-on afsoos kar-tii haiN. regret do-impf are ‘These wayward women, not at first, but after coming down from the intoxication of liquor, feel regret about their condition.’ (Bahl 1974: 98)

(30)

kanyaa-nee mahsuus ki-yaa [ki woo jiivan-meeN Kanyaa-erg feeling do-pf that 3s life-in pahlii baar kisii puruS-kee aakarSaNpaash-meeN puurii first time some man-gen attraction-snare-in full taur-par bandh ga-ii hai.] state-in tied go-PF is ‘Kanya felt [that for the first time in her life that she was completely snared in the trap of attraction to some man].’ (Bahl 1974: 253)

Since the event N is the syntactic object of V, and is not raised to vP, it is not associated syntactically with the subject of vP in the syntactic structures (26a)–(27a). This representation predicts that the subject of N-V should have the subject semantic role assigned by V, which should be agent if the V is a transitive activity or accomplishment verb like kar-naa ‘do, make’. But in (28)–(30) the syntactic subjects of these sentences do not refer to causative agents or events which bring about a change of state. One of the advantages of the Hale-Keyser representation of the lexical syntax of verbs is that the syntactic projections are at the same time a representation of the event structure of the verb, at least for verbs of causation and change. There is an instigating entity (or event) which causes a change, which implicates a state (clear), or a relation (shelve/on the shelf ). Activities like laugh implicate an entity which is produced by the event. Depending on

Phrasal Predicates

95

the event complexity of the verb, the event structure can be represented schematically as (31): (31)

a.

n > (e ! r (relation)/s(state))

b.

(e ! s)

c.

e!n

transitive verbs of causation and change) inchoative verbs, change of state, unaccusatives activity verbs, unergative (Hale and Keyser 1993: 69¤ )

The subevents correspond to the verbal phrases within the l-syntactic structure underlying a verb. The semantic roles of the verb’s arguments follow from the VP/event with which they are associated, making it unnecessary to stipulate a traditional ordered array of theta roles. Verbs of perception, mental or psychological state would seem not to conform to this otherwise perspicuous description.8 The direct objects of such verbs are not ‘subjects of change’, undergoing a transition to a resulting state. Rather they could be interpreted as the locus of a feeling or perception. The instigating event or cause of the feeling or perception can be present in the context without being projected as a sentence constituent. In (28), for example, the approach of the Diwali festival arouses in the children the memory of their mother. The stimulus can be expressed as an optional adverbial clause rather than a syntactic subject or locative source, with both simplex and complex predicates (32)–(33): (32)

apnee saamnee doo banduuk-dhaarii DaakuuooN-koo self ’s front two gun-carrying robber-dat deekh-kar woo Dar-aa thaa see-prt 3s fear-pf was ‘Seeing two gun-carrying robbers in front of him, he became afraid.’ (Nespital 1997: 624)

(33)

kaaminii-kee pahlee din-kii baat sun-nee-kee baad maiN Kamini-gen first day-gen matter hear-inf-gen after I apnee-meeN ajiib hiintaa-kaa anubhav kar-nee self-in strange deficiency-gen experience do-inf lag-aa thaa begin-pf was ‘After hearing what Kamini said the first day, I began to feel a strange inferiority.’ (Bahl 1974: 248)

The same surface syntax can encode di¤erent event structures, either the causative/agentive schema for (20)–(21). The same event structure for

96

Alice Davison

psychological predicates can be expressed syntactically in two ways, one with a nominative or ergative subject (28)–(30), or with a dative experiencer subject (34). In both structures the syntactic subject refers to the experiencer of a mental or physical state, which may come about as a process. (34)

siitaa-koo apnee maaN-baap yaad aa-ee Sita-dat self ’s mother-father-mpl memory-fs come-pf-mpl ‘Sita remembered/missed her parents.’ (Saxena 1985)

The causative structure (26b) which has an overt V kar-naa ‘do’ accounts for only a subset of complex predicates such as (20), but it fails to derive the right case, constituent structure and subject semantic role in many other instances. One of the problems with (26b)–(27b) is that N and not the V provides the experiencer interpretation of the syntactic subject. When verbs like kar-naa ‘do, make’ and aa-naa ‘come’ are used in isolation, it is not the case that their subjects are interpreted as experiencers, but this interpretation is associated with N-V combinations. But the V does add to the aspectual interpretation of N-V. The sentence (35) contrasts the two expressions for ‘remember’. (35)

aur us-see mujhee bahut-sii baateeN yaad aa and 3s-from I-dat many-rather matter-pl memory come gayii thiiN jinheeN maiN yaad nahiiN kar-tii go-pf were which-dat I memory not do-impf ‘From that I remember many things which I do not want to recall/think about.’ (Bahl 1974: 73)

The dative subject version yaad aa-naa ‘memory come’ can only refer to involuntary memory, but some agency as well as perception may be conveyed by the nominative subject version yaad kar-naa ‘do memory’. 4.4. The semantics of incorporation and the interpretation of complex predicates Accounts of complex predicates have proposed many ways for the argument structure of N to be combined with V. For example, Grimshaw and Mester’s process of argument transfer (1988) assumes that the ‘light’ verb is denuded of theta roles, and the arguments of N are transferred into V. Their version of argument transfer requires a number of stipulations about the lexical representation of N and V. But denominal verbs of the kind discussed above are formed by syntactic incorporation of N into a null verb head, and so necessarily the argument structure of the resulting verb is

Phrasal Predicates

97

determined by N. The result is a syntactic unit which assigns case to its object.9 Phrasal combinations have been analyzed as the result of semantic incorporation, which leaves the phrasal surface syntax intact, but combines N and V at LF. These combinations are typically made up of an indefinite object and an activity verbs (Mohanan 1995; Dayal 2003; Chung and Ladusaw 2003). The N is interpreted not as a referential expression but as an intransitive predicate which restricts the meaning of the verb in more or less the way an adverb does. Dayal (2003) argues persuasively for an analysis of this kind for Hindi/Urdu, refining the proposal made by Mohanan (1995). This indefinite reading is associated with (36) with a nominative singular object, which can have a plural meaning absent when the case of the object is dative: (36)

anu baccaa/baccee-koo samhaal rahii hai Anu child-nom/child-dat look-after prog is ‘Anu is looking after children (one or more)/the child.’ (Dayal 2003)

The subject’s acts of ‘looking after’ are restricted to children as the direct object.10 Suppose that N-V predicates were accounted for in the same way. The event noun yaad ‘memory’ with its transitive argument structure ‘x remembering y’ restricts the object of kar-naa ‘do’. The result can be paraphrased more or less as ‘z’s acts of doing are restricted to x remembering y’. If this meaning is possible at all, it seems to be a heavily literal meaning with separate subjects of ‘do’ and ‘memory’, quite far from the actual meaning of the complex predicate, which has a single argument structure.11 So I will conclude that Hindi/Urdu has both syntactic and semantic incorporation, but for very specific classes of predicate. Denominal verbs and the intransitive-transitive-causative diathesis are derived by syntactic incorporation in l-syntax, while semantic incorporation involves indefinite objects without complex argument structure. Even if complex predicates were to be derived by some sort of semantic incorporation, it is not enough in itself to say that N is linked with V at LF; what is needed is an explanation of how the properties of N including its argument structure are linked to V, yielding a single argument-licensing unit. 5. Derivation of Complex Predicates In this section I will propose a alternative to incorporation for deriving N-V combinations with the right surface syntax, case marking and semantic roles for the subject and thematic object. The focus here is on syntactic derivation and constraints on derivation. The derivation I propose for N-V

98

Alice Davison

predicate formation adopts the bottom to top successive phrase formation of the Minimalist Program (Chomsky 1995 and subsequent work). Syntactic combinations are formed by MERGE, linking heads and phrases and projecting the head syntactically. The arguments of the head correspond to successively merged phrases licensed by a process of Theta Discharge.12 An alternative process to Discharge is Theta Merger, combining the argument structures of two syntactic elements in combination. 5.1. Case, agreement and semantic roles in Hindi/Urdu complex predicates The goal is to explain and di¤erentiate the three di¤erent syntactic outcomes for N-V, summed up in (37)–(39). The case of the thematic object classifies the three types: (37)

Type I—genitive thematic object, N-V agreement ganapat singh-nee [eek naii bimaarii-kii khooj ] Ganpat Singh-erg one new illness-gen search.fs-nom kii hai do-pf-fs is ‘Ganpat Singh has discovered a new disease.’ (Bahl 1974: 222)

(38)

Type II—locative thematic object, N-V agreement yee gumraah aurateeN . . . [apnii haalat-par] afsoos kar-tii these wayward women self ’s condition-on regret do-impf haiN are ‘These wayward women . . . feel regret about their condition.’ (Bahl 1974: 98)

(39)

Type III—dative or nominative thematic object, N does not agree with V a. siitaa-koo apnee maaN-baap yaad Sita-dat self ’s mother-father-mpl memory-fs aa-ee come-pf-mpl ‘Sita remembered/missed her parents.’ (Saxena 1985) b.

us-nee moohan-koo bahut yaad ki-yaa 3s-erg Mohan-dat much memory-f do-pf-ms ‘He/she remembered Mohan very much.’ (Bahl 1974: xxix)

The N is morphologically and syntactically the object of V in Types I and II, but not in Type III. N assigns object case in Types I and II, but not Type III. The N assigns semantic roles to both the subject and thematic object,

Phrasal Predicates

99

while V is responsible for the aspectual class of the predicate. V may contribute semantic roles in all cases. The V projection assigns the case of the thematic object. An account of complex predicates in Hindi/Urdu has to allow for many di¤erent kinds of argument structures, including those with experiencer predicates (37)–(39), as well as those with agentive subjects (20)–(21). It must derive correctly the respective contributions of N and V; N provides the semantic roles of the arguments, including the syntactic subject, and in most instances, the case-marking of the thematic object, while V accounts for the case of the subject, the event structure and aspectual class of the verb, and in some instances, the case of the thematic object. 5.2. Theta-related processes To simplify the exposition and to focus on the crucial features of the syntactic derivation, I will assume a lexical representation of the argument structure of verbs and event nouns as in (40). It identifies an ordered list of arguments, and contains a Davidsonian event argument (Higginbotham 1985; Speas 1990): (40)

h1, 2, 3, ei

The arguments in this lexical structure have semantic roles which are linked to specific phrases by the process of Theta discharge process. I assume this association is a consequence of MERGE and PROJECT (Chomsky 1995). Discharged theta roles are distinguished by 2* indicating that the position of arguments so marked is saturated and does not license further combination. By Theta projection, the argument structure is projected upward to the newly formed phrase. Theta/argument identification identifies one argument position in an argument structure with an argument position in another argument structure; theta or argument merger is the result of linking or identifying corresponding argument positions in two separate argument structures to form a new composite argument structure (Higginbotham 1999; Folli 2001). 5.3. An example of a well-formed derivation I begin by demonstrating how the operations of MERGE and Theta Discharge apply in a sample derivation of a sentence (41), in which the arguments are saturated constituents, not complex event nominals.13 A sentence like (41a) is built from the merger of the transitive verb khooj-naa ‘search for’ (41b), with its arguments. In addition to internal and external arguments, the argument structure will include the Davidsonian event argument e.

100 (41)

Alice Davison a.

b.

ham tumheeN khooj rahee thee we you-dat search prog were ‘We were searching for you.’

(Bahri 1992: 138)

khooj-naa ‘search (V)’ h1, 2, ei

The verb khooj ‘search’ merges with the DP tumheeN ‘you-dat’, discharging the 2 argument position and the V head PROJECTS upward, with its argument structure. This DP saturates the internal argument position. The result is (42). For concreteness, I use indices to represent the theta-discharge relation between the saturated argument position and the phrase receiving the semantic role. The combination in (42) MERGES with ham ‘we-nom’, forming (43): (42)

(43)

Finally, the event argument is THETA-BOUND by sentence tense and aspect, anchoring it directly or indirectly within a temporal representation (cf. Smith 1997; Giorgi and Pianesi 1997). The final result of the derivation of the sentence, which goes to the interpretative interface, is the well-formed LF representation in (44):

(44)

Phrasal Predicates

101

All positions in the argument structure are saturated; all arguments are linked to a position in the argument structure, including the event variable. The event argument is bound by the TENSE. This sentence derivation and the resulting LF representation conform to Higginbotham’s Saturation Principle, which can be adopted within the assumptions of the Minimalist Program as an interface condition, relating syntactic structures and LF interpretation. (45)

Saturation Principle: A constituent which has open argument position(s) may not itself discharge a theta role assigned by another head. (Higginbotham 1985: 561–2)

However, (45) would be violated if the nominal object is a complex event nominal, with aspectual structure and multiple argument positions to be saturated (Grimshaw 1990). The derivation of a sentence with a complex predicate of Type I or II works like the formation of a simple sentence in successive stages shown in (46), (48), and (49), with one important di¤erence stemming from the Saturation Principle. The N first merges with a DP, which gets the internal argument theta role in the N’s argument structure. The genitive or locative case is checked within the nominal projection which is formed by MERGE.14 The theta grid with one discharged role is projected up to the phrasal node (46): (46)

I will assume that the N khooj ‘search’ has no lexical case specified in its lexical entry, but afsoos ‘regret’ as in (38), does assign a lexical case (47a): (47)

a.

afsoos h1, 2, ei ‘regret’

b.

khooj ‘search’ h1, 2, ei

-par ‘on’ If no lexical case is supplied by the lexical entry of N, then structural genitive is the default.15 At this point, the derivation can take two distinct paths, one of which is shown in (48). Here the NP with an open theta position undergoes theta identification with V, and does not discharge the internal theta role of V. The discharged 2 position of N is identified with the 2 open position of V. The open external argument position of N is identified with the open 1 position of V, as are the event arguments of N and V.

102

Alice Davison

(48)

Finally, in (49), the DP external argument merges with the V projection, discharging the external argument theta role. The theta role of the external argument specified by N (bold) is available in the composite N-V structure. The LF (50) contains the composite argument structure (bold italics): (49)

(50)

The main feature of this account is that complex event NP itself does not discharge a theta role of V. It does not saturate a position in the argument structure of V. Higginbotham’s Saturation Principle (45) would be violated if the N, which has an open theta position in its argument structure, did discharge a theta role of V. Rather, information from the argument structure of NP at that point in the derivation is transferred to V and combined with the specification of the argument structure in V. The theta merger of N and V means that the external position of N is projected as non-distinct from the external argument position of V, and TENSE binds the common event variable. Type III complex predicates, which are all variations of Type I predicates, di¤er in two important ways from Type I. The N does not trigger verb agreement in Type III, and the thematic object has the nominative or dative case of an ordinary direct object. This structure is illustrated by (51).

Phrasal Predicates (51)

maiN-nee raam-koo yaad ki-yaa I-erg Ram-dat memory-f do-pf-ms ‘I remembered Ram’

103

(Type III)

The di¤erence is that Argument Identification/Merger is the first operation, taking place before the DP thematic object is merged (52)–(53). Finally, the subject is merged, as in the other derivations (54): (52)

(53)

(54)

Argument identification takes place at the X 0 level, before N has combined with a thematic DP object, as it would in Types I and II. The distinction between Types I and II predicates follows from the lexical properties of N. The distinction between Types I/II and Type III seems to stem from a lexical property of a fairly small number of Ns. They assign either genitive or structural object case. I propose that the Ns which assign nominative or dative structural object case may have a second lexical representation which blocks the assignment of genitive case, based on what Dubinsky (1997) proposes for some complex predicates Japanese. Dubinsky proposes that N

104

Alice Davison

loses its full categorial distinction, and is specified only as þN, non-distinct from an adjective. Adjectives may also have argument structures and assign case, but not genitive or lexical case. In fact, the structural case-marking of objects in the Type III N-V predicates is exactly like what we would expect from adjective-verb combinations (McGregor 1995). Hindi/Urdu has an AþV complex predicates, such as mazbuur kar-naa ‘make forced, force, insist upon’, which assign structural object case, as in (55a). This dative object case is structural, as it disappears in the passive (55b): (55)

a.

unhooN-nee DaakTar-koo [PRO fiis lee-nee] par 3pl-erg doctor-dat fees take-inf-on mazbuur ki-yaa forced do-pf ‘They forced the doctor [PRO to take his fees].’ (Bahl 1979: 37)

b.

us-kaa jiivan phir [PRO caTTanooN-par 3s-gen life again rocks-on Takaraa-nee]-kee liyee mazbuur ki-yaa ga-yaa hai dash-against-inf-gen-for forced do-pf go-pf is ‘Her life again was forced to dash against the rocks.’ (Bahl 1979: 37)

These case-marking facts follow from the inability of adjectives to assign object case, so that Argument Identification is forced to form an A-V unit which assigns structural case. Early argument merger is therefore forced for adjectives in general and for Ns which have a defective category specification. Early merger is a consequence of category specification in the lexicon. Some Ns, like yaad ‘memory’ and isteemaal ‘use’, have dual lexical representations, allowing them to combine in both Type I and Type III combinations. Optionality is a consequence of dual lexical representation, in this view. This dual representation of N as [þN-V] (Types I, II) and [þN] (Type III) provides an explanation for the di¤erences of case and phrase structure. The object merges with N low in the structure for Types I/II, while it is higher in the structure for Type III. The object position follows from the point in the derivation at which argument merger takes place. Early argument merger is forced when N or A cannot assign case, deriving type III structures in which the DP object merges higher up with N-V, while type I/II structures are derived when N does assign case to its object, allowing the DP with N, so that in the resulting clause structure it is quite low. It is not necessary to stipulate the high versus low position of the object, the criticism leveled at the structures (26b)–(27b). I have argued above that such

Phrasal Predicates

105

stipulations are required if complex predicates have the same structure as incorporation/conflation structures. 6. How N and V Combine to Form a Single Predicate I return here to the question of how V and N contribute to the meaning and aspectual properties of the N-V combination formed by argument merger motivated by the saturation Principle (45), focusing in greater detail here about lexical information and how it is combined when argument structures are identified. A somewhat similar proposal based on argument merger has been made for Persian (Karimi-Doostan 1997), though the structural di¤erences of case and lexicon between Hindi/Urdu and Persian entail some formal di¤erences from the account given here, which derives the variety of complex predicates in Hindi, counterparts of which are apparently not found in Persian. The argument merger approach was explored in earlier work by Jackendo¤ (1974) and Cattell (1984). I assume that V in N-V combinations is as fully specified as in independent occurrence without the event N, in contrast to earlier accounts, Grimshaw and Mester (1988) in particular, who represented V as ‘light’, devoid of thematic properties. But more recent discussions such as DiSciullo and Rosen (1990) point out that V in N-V is far from empty of specification. My position is that the verbs which enter into N-V combinations (Appendix III) have thematic properties and aspectual properties, but are inherently less specified than other verbs, such as those which incorporate indefinite objects like samhaal-naa ‘look after’, discussed above. The complex event nouns, on the other hand, are inherently more fully specified, not only for event structure and arguments but also for various specific default properties for the arguments. This asymmetry of lexical content accounts for the generally greater contribution of N’s properties when the argument structures are merged. The V and complex event N have similar lexical structures, so that corresponding information can be matched. I will use the lexical structures of Pustejovsky (1995), somewhat simplified, to focus on a representation of the lexical content of N and V. For a concrete example of the same N combined with di¤erent Vs, compare the complex predicates in (56): (56)

aur us-see mujhee bahut-sii baateeN yaad aa and 3s-from I-dat many-such matter-pl memory come gayii thiiN jinheeN maiN yaad nahiiN kar-tii go-pf were which-dat I memory not do-impf ‘From that I remember many things which I do not want to recall/think about.’ (Bahl 1974: 73)

106

Alice Davison

This sentence contrasts two versions of ‘remember’, the dative subject version with aa-naa ‘come’, and the nominative/ergative subject with kar-naa ‘do’. This contrast illustrates a subtle aspectual di¤erence. The dative subject version refers to an involuntary process, instigated by another event us-see ‘from that’, and leading immediately to a resulting state of remembering. This version has the properties of an achievement. The nominative/ ergative version is more an accomplishment, focusing on a process which takes some time. Some volition may be conveyed. The aspectual property of the N-V follows from the specification of the V involved. Following Pustejovsky (1995), I represent the event denoted by V in terms of subevents, which are in a precedence relation. A subevents may be the head of the whole event (distinguished by *), representing the focus of the event. Accomplishments focus on the initiating process, achievements on the resulting state (57a, b). (57)

(58)

a.

achievement: E ¼ e1: process, e2: state precedence: e1 < e2*(