Abstract
The goal of this chapter is first to give an outline of Bengali clause structure and clausal complementation in particular, and then to relate the findings of that investigation to the theory of operator movement that has been developed in the previous chapters.1 Bengali is a strict head final language, but it has one type of finite complement that can only appear in extraposed position. On the other hand, Bengali is a language which, like many head final languages, does not exhibit overt Wh-movement. As such it is similar to Chinese, Japanese and Korean all of which have received much attention in generative linguistics in recent years. Bengali is interesting in its own right, however, because it does not have, in the normal case, a scope marker that indicates the scope of the Wh-element; furthermore it can make use of other strategies for question formation which do look like scope marking and overt movement. Since Bengali allows for clausal complements to appear on either side of the verb, it can serve as an interesting testing ground for our theory. Assuming that a Wh-phrase unmoved in the syntax has to undergo movement to a scope position at the level of LF, our theory predicts that extraposed clausal complements only allow narrow scope readings, while complements in canonically selected position allow wide scope readings as well. Evidence that supports this prediction does not only come from Bengali, but also from various related Indo-Aryan languages. Crucially, we will show that the scope of the Wh-operator is constrained by the orientation of the verb that selects the complement containing the Wh-phrase. This will provide additional and hitherto unexplored evidence in favor of LF as a genuinely syntactic level of representation. The pattern found in Bengali will be shown to be very similar to the one found in German.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
Notes
Following suggestions by Probal Dasgupta, I will adopt the following conventions for the transcription of Bengali: T, D for retroflex stops, c, j for voiceless and voiced palatal stops, S for a palato-alveolar fricative, N for a velar nasal, Y, W for mid and y, w for high semi-vowels, E, O for low vowels; M indicates the nasal quality of a preceding vowel; after a consonant, h indicates aspiration. I am indebted to Krishna Bhattacharya and Probal Dasgupta for checking my transcription of the Bengali examples. For the representation of examples from other Indian languages, I follow the sources from which I drew them. The English glosses added to the Bengali examples reflect but do not always respect morpheme boundaries.
Recall that the same was found to be true for German finite complements.
For relevant discussion see Singh (1980), Dasgupta (1980; 1984; 1987), as well as Bal (1990) on Oriya, a closely related language spoken in the state of Orissa i.e., in the immediate vicinity of West Bengal. What I consider here is mainly based on these works.
Indo-Aryan languages such as Assamese, Bengali, Dakkhini-Hindi (spoken in and around Hyderabad), Marathi and Oriya are likely to have borrowed these from Dravidian; see Telugu -ani, Tamil endru,Kannada anta, Malayalam -enna,all of which are clause-final morphemes which derive from verba dicendi and function as complementizers.
Historically, je derives from the Indo-European pronoun *io. See, for example, Brugmann (1904) and Hettrich (1988); for Bengali, in particular, see Chatterjee (1926: 839ff.).
According to my informants, extraposition of bole-clauses with argument status is either highly marked or completely ungrammatical. The same seems to be true for Oriya, which has the same kinds of complementizers. In his study of complementizers in Oriya, Bal (1990) gives examples of extraposed CPs which are headed by the final complementizer boli. He seems to take these as grammatical, but adds then that they are “not typical” and that they “usually occur in situ,in the expected pre-head complement position,…” (pp.6f.). Alice Davison (p.c.) informs me that C-final clauses may occasionally be extraposed in Dravidian. For my argumentation this is not relevant as long as these extrapositions have a marked character. Notice that je-clauses and German daß-clauses in extraposed position never have a marked character.
See, for example, Koopman (1984) about the complementizer like morphemes na (for CPs in the context _ V) and ka (for CPs in the context V_) in Vata. Koopman claims that only the latter is a complementizer. Similarly, Bal (1990) claims that only boli but not je is a complementizer in Oriya. I will come back to his analysis.
The same is true for the Vata case mentioned in note 7 and for Turkish where the agglutinative morpheme -dik (or -dig) has been argued to be a complementizer for IP-internal clauses, while the clause initial ki seems to be a complementizer for extra-posed clauses. See Kural (1992) and Veld (1993) for recent discussion.
See Bal (1990) for Oriya and Wali (1988) for Marathi.
While Mahajan (1990) and Srivastav (1989) take Hindi ki to be a complementizer, Alice Davison (p.c.) disagrees with this analysis. Bal (1990) gives arguments against je as a complementizer to which we will turn immediately.
See Chomsky (1982) and Koopman and Sportiche (1982)
See Lenerz (1984) and literature quoted therein.
See Chomsky and Lasnik (1977). Various German and Dutch dialects do not observe the DFCF. See Bayer (1984) and Mayerthaler and Mayerthaler (1990) for Bavarian.
As I have already indicated, Bengali permits the occurrence of DPs to the right of the verb. Probal Dasgupta (p.c.) observes that such DP-shift is possible from CPs to the right of the verb, but not from those which stay in the canonical object position. Consider: (i) ami Sune-chilam [(je) babul e1 kha-be ruTii I hear -PTS1 COMP Babul eat-FUT3 bread I heard that Babul will eat bread (ii) *ami [babul e1 kha-be ruTi1] Sune-chilam The reason for the ungrammaticality in (ii) seems to be that the extraposed DP destroys the adjacency relation between the clause final C-position and the matrix verb. For further details see Bayer (1995).
Our analysis of je resembles our analysis of German daß. An open question is why daß-CPs do appear in topic position. Clearly, the daß-CP cannot be moved from its position to the right of the head. As we have seen in chapter 6, it has a limited distribution in the canonical object position. In my view, the answer should be sought in the feature structure that distinguishes daß from je. As we have seen earlier, daß-clauses must be [+N], while V2-clauses must be [-N]. The latter have a distribution somewhat like je-clauses i.e., they are never permitted in situ, and only in rare cases (involving the subjunctive) they can move to topic position. A fuller account of this lies outside the scope of this chapter. See Davison’s (1993: 41) discussion of differences between Hindi ki-clauses and German daß-clauses.
In German, there are NPIs like mehr (“any more”) which must be c-commanded by a negator that appears in the same clause, while others like eine Träne nachweinen (“to shed tears for”) can be licensed from the matrix clause: (i) a. Kein Kind mußte mehr hungern no child must any more starve No child had to starve any more b. *Kein Mensch glaubt, daß die Kinder mehr hungern müssen (ii) a. Keiner wird uns eine Träne nachweinen no-one will us a tear after-weep No one will shed tears for us b. Keiner glaubt, daß Gustav uns eine Träne nachweint No one believes that Gustav sheds tears for us One would want to be sure that the Hindi cases discussed by Dwivedi do not belong to the first class.
17 Reinhart (1991) argues on the basis of sentences with elliptic conjunctions (comparative ellipsis, except-phrases etc.) that subjacency must hold at LF in full generality, and that in principle any phrase (not just “quantified” phrases) may undergo abstract movement.
Wh-expressions in Bengali have /k/ in their onset e.g., ke (“who”), ki (“what”, or yes/no question marker), kEmon (“how”), kEno (“why”) etc. These forms derive from the Proto Indo-European interrogative form *k”i, *k’“o. Dasgupta (1980) calls them K-words. For simplicity, I will stick to the familiar Wh terminology. Of course, “Wh” will then address the semantic and not the phonological side.
In Bengali, non-interrogative operators of this kind can be derived from k-words by affixing the emphatic element -o which we have already seen in example (31). Consider the following examples where -o is subject to allophony: (i) ke (“who”) + o — keu (“someone”, nom.) (ii) ka +o+ ke kauke (“someone”, obj.)
The Q-morpheme ki has a distribution somewhat like internal je and other clitic elements. Thus besides tumi ki ca kine-cho? (“Did you buy tea”) we also find tumi ca ki kine-cho?,and more marginally tumi ca kine-cho ki?. There is no indication that ki is in a sentence peripheral C-position.
For accounts of scrambling in Bengali, see van der Wurff (1989) and Sengupta (1990).
E.g. her (26b): (i) ek laRkii jis -ko ki sab pasand karte hai one girl who ACC that all like AUX A girl whom everybody likes
The Bengali data corresponding to those in (40) are: (i) *ami jani na [je bill kal ka-ke dekhe-che] (ii) ami jani na [bill kal ka-ke dekhe-che] (iii) *ami jani na [ka-ke; je bill kal e; dekhe-che]
No claims about “LF word order” should be derived from the fact that ke is represented in a lefthand SpecCP position.
The complementizer je must have rather different features. In note 23, we saw that in Bengali the cornplementizer je prevents the appearance of a Wh-operator in its specifier position.
See Lasnik and Saito (1984). If this were the case, the sentence in (i) should be out, while the one in (ii) should be in: (i) Who; do you believe kPe; [e; won the race]]? (ii) *Who; do you wonder [ape; [e; won the race]]?
According to Probal Dasgupta and Ayesha Kidwai (p.c.), there are contexts in which bhab-may select a + Wh complement, and then show a meaning similar to ponder or wonder. The same is true for English think, as can be seen in examples like I was just thinking how on earth he could have escaped from this dungeon. In spite of this complication, (46b) is ungrammatical for all the speakers I could consult.
44b) turns out to be grammatical if the Q-morpheme ki is inserted in the matrix clause: (i) tumi ki bhab -cho [ke baRi kor -be] ou Q think-2 who house build-FUT3 ho do you think will build a house?
I had a chance to discuss this during my lectures at the International Summer School in Syntax (ISIS) held at the Central Institute of English and Foreign Languages, Hyderabad, 1994. The following Tamil example is taken from Savio (1991): (i) enna jaan kaTaiyil neeRRu saappiTTaan enRu meeri soonaL? what John shop yesterday ate COMP Mary said What did Mary say that John ate in the shop yesterday. See also Mohanan (1984) on Malayalam.
See, for example, Subbarao (1984), Davison (1988; 1993), Srivastav (1989; 1991a,b) and Mahajan (1990).
Mahajan (1994) presents a series of arguments against this proposal which I cannot review here.
Notice, however, that Davison (1993) presents some serious problems for an extraposition analysis of Hindi ki-clauses which I will not review here.
The problem cannot be connected to a that-trace effect. Such an effect seems to be entirely absent in Bengali. The same result would be obtained with Wh-objects such as ka-ke (“whom”) or any other Wh-phrase.
The same applies in Hindi, as shown by Alice Davison’s example in (i), here quoted from Srivastav (1989). (i) *tum e; cahte ho [PRO kyaa kamaa]; you want AUX what do The reason why sentences like (i) and (55b) do not even allow narrow scope of the Wh-item may be sought in the absence of a landing site for the Wh-operator in the infinitive. Notice that this is the case in German too: (ii) *Ich weiß nicht [wohin PRO zu gehen] I know not where to go I don’t know where to go
We will return to these cases in more detail in section 7.5.
Davison (1993) argues against the extraposition of ki-clauses in Hindi, but assumes their base generation in a right-adjoined position. She claims that “[t]he external finite clause is syntactically external to normal argument position, though it has an argument interpretation [emphasis J.B.]” (p.42). Since it is not quite clear what “argument interpretation” should mean, it seems to me more desirable to explore the more restrictive solution according to which L-marking is achieved by head government.
Various Indo-Aryan languages like Bengali provide a view of the historical development of clausal subordination that is less obvious when we look at genetically related modern European languages where clausal subordination in complementation and in relativization seems to be fully grammaticalized. The je-construction in Bengali and ki-clauses in Hindi show a picture in which subordination (hypotaxis) exists side by side with adjunction (parataxis). It seems to be agreed upon that subordinate clauses arose from parataxis. See Delbrück (1919: 238ff.), Hettrich (1988) and Kiparsky (1990), and for German among others Lenerz (1984) and Abraham (1993). The modern Indo-Aryan languages show effects of this transition in their synchronic grammars.
This had very clearly been noticed by Hermann Paul (1880) who writes: “Ein wichtiger Schritt zur Erzeugung komplizierter [Satz-,J. B.] Gebilde war, dass das Objektsverhältnis auf einen Satz übertragen wurde. “(p.145) [An important step towards the creation of complex sentences was that the object relation was extended to the sentence].
I will not repeat the technicalities of this proposal. They have been discussed sufficiently in chapters 5 and 6.
To my knowledge, Wali (1988) is the only work on Indo-Aryan that explicitly suggests explaining these facts in terms of Kayne’s (1983) theory of connectedness.
The slight awkwardness of the b.-examples may have to do with a problem for tense linking. I will ignore this factor.
Bengali uses reduplication for pluralization and also for the expression of continuity of action. Roughly, the reduplication in the examples in (63) signals that the two activities described proceeded simultaneously.
For Bengali, see Sengupta (1990), for Hindi see Mahajan (1990) among others.
See Bayer (1990c: ch.6) and Bayer and Lahiri (1990).
I have concrete evidence from Hindi, Tamil and Turkish. Consider the following Tamil examples from Savio (1991: 56): (i) avan saappiTala saatam he did-not-eat rice (ii) *avan saappiTTaan enna? he ate what
I do not want to suggest that “rightward scrambling” is in any sense comparable to regular (leftward) scrambling, or that it forms chains. Thus, there may be no problem for proper binding. Alternatively, there could be a predicate raising analysis to the specifier of some empty head H. Consider, for example, a structure for (70b) as in (i) [HP [radha-ke bhalobaSe]; [H. H [vp ke e,]]] in which case ke (“who”) would be included in HP and thus be prevented from raising to an operator position.
We ignore the fact that these DPs retain their morphological shape under these different conditions.
I am indebted to Jogamaya Bayer for discussions of the pragmatics of postposed Wh-elements and the role of the clitic -Ta.
See Bierwisch (1980) for relevant discussion; also Bayer (1991) for considerations of the syntax-pragmatics interface.
80b) shows that besides ki also kEno (“why”) can be postposed, but there are other K-words that cannot be postposed. A more detailed investigation than is possible here would be required to explore these restrictions.
Topicalization of Hindi ki-clauses is also completely impossible. Jaklin Komfilt (p.c.) informs me that the same is true for Turkish CPs which normally appear to the right of V. These are also introduced by ki (a borrowing from Persian). The present theory seems to offer a natural explanation of this fact. See also note 15.
See Huang (1982) and Cheng (1991) among others.
ki is a clitic and in the sense that it resembles the floating internal je. For details on this class of elements see Dasgupta (1984; 1987), who calls ki, je and a number of other elements anchors,referring to their property of clinging to arbitrary maximal constituents in the clause. See also section 7.3.3.
This should not be taken to mean that all speakers are very comfortable with the scope marking construction. What is relevant for my point is that every speaker finds a clear difference between (83) and (84).
For German see van Riemsdijk (1983) and McDaniel (1989), for Hindi see Davison (1988), Mahajan (1990), Srivastav (Dayal) (1989; 1991a,b; 1994) and others, for Hungarian see Marâcz (1987; 1988; 1989) and de Meij and Marâcz (1986), for Iraqi Arabic see Wahba (1991), For Kashmiri and Marathi see Wali (1988), for Romani see McDaniel (1989).
Proposals of this kind have already been made in the literature e.g., in Aoun and Li (1993).
Chomsky has shown that such insertion must be assumed for independent reasons. See my discussion in chapter 6.
See Cheng (1991).
See Davison (1988), Mahajan (1990) and Srivastav (1989; 1991a) as well as literature quoted therein, some of which was not accessible to me. Davison gives arguments against syntactic Wh-movement in Hindi showing that the process is not as pervasive as it is in languages which entirely follow the movement schema. Srivastav points out that many speakers of Hindi do not accept cases of overt movement. My own investigations of Bengali revealed that (a) there are quite a number of speakers who accept cases of overt movement, and (b) that there are hardly any restrictions against certain categories (adjuncts, for instance) which Davison found could not be moved in Hindi. One element which does not like to appear in trans-CP chains is kEno (“why”). The reason for this is quite obviously that the comprehension system almost mandatorily creates the “upstairs” construal. The same is true for English and German. But this could be a processing factor that should be kept aside from issues of grammar proper. Notice also that Mahajan (1994) presents two examples of adjunct extraction in Hindi as fully grammatical.
The speakers I consulted prefer the presence of the complementizerje here, although there are no clear intuitions about ill-formedness of the examples when it is absent.
The appearance of the displaced Wh-element in sentence initial position is accidental. It could equally well occur between the subject turn or turne and the verb.
As has repeatedly been stated, cases of Wh-scrambling must not be misinterpreted as movement to a Wh-checking position.
Since Hindi has only C-initial finite complements, we cannot see topicalization, but Bengali bole-clauses undergo topicalization almost invariably. The same seems to be true for C-final complements in other Indo-Aryan and in the Dravidian languages. Davison (1993), who defends an adjunction analysis of kiclauses on the basis of a special version of Case Resistance,has noticed the problem. She offers a “prohibition against non-branching projections” according to which a ki-clause must not be attached on the left side of the projection which is the recursive side in Hindi i.e., (i) is permitted, but not (ii): (i) [xr WP [xr YP [x. ZP X]]](ii) *[xp WP [xr 0 6, 0 X]]]
See also chapter 6.
I leave open the possibility that there is no LF-reconstruction of the restrictive part of the Wh-phrase. Reinhart (1993) shows that in the sentence Who will be insulted if we invite which philosopher? the restrictor “x is a philosopher” must be outside the if-clause if we want to avoid the consequence that Lucy, if we invite Donald Duck could be a licit answer (assuming that Donald Duck is not a philosopher).
This cannot hold for Bengali because (90) is grammatical for many speakers although kEnton kore (literally “how having done”) is an adjunct and cannot be considered to bind a pronoun of the same category.
Notice that in my (though not in Sinha’s or Davison’s) analysis the ki/je-clause would be under V’. Since CP does not need to undergo Case checking, there is no question of movement to SpecAGRoP, even independently of my own proposal. Assuming that AGRo may nevertheless be projected, there will be a position in which the clause-internal topic can be Case-licensed.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1996 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Bayer, J. (1996). Complementation and the Scope of Wh in Bengali. In: Directionality and Logical Form. Studies in Natural Language and Linguistic Theory, vol 34. Springer, Dordrecht. https://doi.org/10.1007/978-94-017-1272-9_8
Download citation
DOI: https://doi.org/10.1007/978-94-017-1272-9_8
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-4623-9
Online ISBN: 978-94-017-1272-9
eBook Packages: Springer Book Archive