Manetta (2010) argues that Hindi-Urdu has feature-driven overt wh-movement to Spec,vP, a position from which wh-expressions can take clausal scope. This is argued to provide a unified account of the following aspects of question formation in the language: the tendency of wh-expressions to occur in preverbal position, the possibility of overt long wh-movement, and the use of scope marking to question out of finite complements. Here, I first probe some hidden assumptions in Manetta’s account, revealing crucial gaps in the argumentation. I then discuss new facts, as well as some old ones, that challenge central components of the proposal. As a whole, the discussion establishes that the move to capture question formation in Hindi-Urdu through wh-movement to Spec,vP does not deliver the promised results.
1.1 Questioning in Hindi-Urdu
Hindi-Urdu (henceforth H-U) is an SOV language with relatively free word order. Finite complements appear on the right, often associated with an overt pronominal expression in the preverbal position (see, e.g., Davison 1984, 1988, Gurtu 1985, Mahajan 1990, Srivastav/Dayal 1991, 1996). The core facts are illustrated in (1).
anu-ne kalam/kyaa khariidaa
Anu-ERG pen/what bought
‘Anu bought a pen.’ / ‘What did Anu buy?’
anu gaaRii/kyaa calaa-naa jaantii hai
Anu car/what drive-INF knows
‘Anu knows (how) to drive a car.’ / ‘What does Anu know (how) to drive?’
siitaa (yeh) jaantii hai ki ravi-ne anu-ko/kis-ko dekhaa.
Sita this knows that Ravi-ERG Anu-ACC/who-ACC saw
‘Sita knows that Ravi saw Anu.’ / ‘Sita knows who Ravi saw.’
A point worth noting in these examples is the scope of the wh-expressions with respect to finite complements. The wh-expression in (1c) cannot be interpreted with matrix scope, leading to the view that the finite clause is an island for covert scope taking. For present purposes, the specific accounts that have been proposed for these facts are not as important as the empirical generalization they reveal.1
Given the core facts in (1), two questions arise. If word order is relatively free, what is the status of wh-movement in H-U? If the scope of wh-expressions in finite complements is confined to the local domain, how does H-U express long-distance wh-dependencies?
With regard to the first question, H-U was initially taken to be a wh-in-situ language, but this does not fit well with the observation that wh-expressions tend to occur in the preverbal position. This, however, is not a hard constraint, as (2) and (3) illustrate for questions over subject and indirect object, respectively. The (a) examples have the wh-phrase in the preverbal position; the (b) examples have it in the neutral position for subjects and indirect objects.
yeh kavitaa kis-ne likhii?
this poem who-ERG wrote
‘Who wrote this poem?’
kis-ne yeh kavitaa likhii?
tum-ne paisaa kis-ko diyaa?
you-ERG money who-DAT gave
‘Who did you give the money to?’
tum-ne kis-ko paisaa diyaa?
Both orders given above are acceptable. In fact, sometimes the base order may seem more natural than the one in which the wh-expression is in the preverbal position. This is so in (3), for example, where (3b) arguably sounds more natural than (3a). Absent a fully controlled study of the discourse conditions under which variations arise, however, it seems a reasonable working hypothesis to take wh-phrases to preferentially appear in the preverbal position. It has been claimed by Kidwai (2000), and following her Manetta (2010), that the preverbal position is a focus position to which wh-expressions move, the alternative orders being derived through scrambling.2
With regard to the second question, there are two ways of questioning into finite complements, the first of which involves overt displacement of a wh-expression.
siitaa kaun soctii hai [ki ____ aayegaa]?
Sita who thinks that will.come
‘Who does Sita think will come?’
kaun siitaa soctii hai [ki ____ aayegaa]?
Note that in (4a) the extracted wh-phrase is in the matrix preverbal position, while in (4b) it is in clause-initial position. The possibility of such displacement is contingent on there being no pronoun yeh ‘this’ in the matrix object position (cf. (1c)).
The second strategy for questioning into complements is what has been variously called “scope marking,” “partial wh-movement,” and the “wh-expletive strategy,” attested by now in a large number of languages (see Dayal 2013 for a survey). Its hallmark is that there is an invariant wh-expression corresponding to what in the matrix clause, but the question seems to be about a wh-expression in the embedded clause.
siitaa kyaa soctii hai [ki kaun aayegaa]?
Sita what thinks that who will.come
‘Who does Sita think will come?’
There are two possible approaches to this phenomenon, dubbed direct and indirect dependency approaches in Dayal 1994b. Very briefly, the direct dependency approach aligns scope marking with overt extraction of the kind seen in English, whereby the embedded wh-expression comes to have matrix scope. It is predicated on the view that the matrix wh-expression has no semantic role beyond marking scope. In the indirect dependency approach, on the other hand, the matrix wh-expression plays the same role that it does in a simple question like What does Sita think?, marking quantification over propositional variables. The question denoted by the CP associate of what is the restriction on this variable, telling us which set of propositions this variable should draw its value from: Which proposition in the set denoted by Who will come does Sita think? Under this view, the closest English correlate of (5) is the English sequential scope-marking construction (Dayal 1996), given in (6), rather than the translation given in (5).
(6) What does Sita think? Who will come?
It is hard to argue that who takes scope out of a syntactically separate clause. For such structures, it seems relatively uncontroversial that the indirect dependency approach must be adopted (Dayal 1996). The question is whether the same should not also apply to standard cases of scope marking, such as (5).
1.2 The Case for Feature-Driven Wh-Movement to Spec,vP
Manetta (2010), building on Rackowski and Richards 2005, takes as her starting point the view that the preverbal position in H-U, which hosts wh-expressions, is Spec,vP. Her account, in brief, posits the derivations in (7) and (8) for monoclausal questions. (7) and (8) illustrate questioning over subject and object positions, respectively. To begin with subject position:
The v head probes its domain, which includes its specifier position, and values its uninterpretable wh-feature as well as its EPP feature by interacting with the wh-expression in Spec via Move. The C head then probes its domain and values its uninterpretable wh-feature by interacting with the wh-phrase at the left edge of the vP via Agree. This also values the uninterpretable Q-feature on the wh-phrase, leading to an LF that can be interpreted as a direct question. Subsequent scrambling of the object results in the observed word order, allowing the wh-expression to surface in the preverbal position.
As illustrated in (8), questions over objects follow the same path except that it is the subject, originally merged into Spec,vP, that must be scrambled to yield the right word order.
Generalizing to the long-distance cases, Manetta gives the derivations in (9) and (10) for “extraction” and “wh-expletive” strategies, respectively. We will probe the appropriateness of these descriptive labels in section 2. For now, let us see how the observed effects are derived.
Here again, the full wh-expression ‘who-ACC’ has an uninterpretable Q-feature and an interpretable wh-feature. The probe v in the embedded clause interacts with it via Move, raising it to the embedded Spec,vP, simultaneously valuing its own uninterpretable wh-feature and satisfying its own EPP feature. Manetta gives several arguments for allowing the matrix v to probe into the embedded CP to the edge of the embedded vP, where it can value its uninterpretable wh-feature and its EPP feature, by moving the wh-phrase to the Spec of the matrix vP. The wh-phrase gets its Q-feature valued by the matrix C and ends up with matrix scope.
The expletive strategy follows along the same lines up to the embedded clause.
The main difference between (10b) and (9b) is that here the matrix v’s EPP feature is satisfied by a wh-expletive in the numeration. Since it must get accusative case, the expletive is generated in Spec,AspP, where it is assigned case by the transitive v and moves into Spec,vP to satisfy the EPP feature on v. Being an expletive, it has no wh-feature. This requires the matrix C head to have its uninterpretable wh-feature valued by a wh-phrase in some accessible position within its domain—namely, within its own phase or the edge of the immediately lower phase. It first probes the vP phase in the matrix clause, finds the expletive with no wh-feature at the edge, and continues to probe. Since the wh-feature on the matrix v is still unvalued, the matrix v gets its feature valued by the wh-phrase at the edge of the lower vP. The matrix v then values the wh-feature of the matrix C, giving the embedded wh-phrase matrix scope.
Crucial to Manetta’s account in (9) and (10) is the transparency of the embedded C in allowing the matrix v to probe into the embedded vP. Here Manetta departs from Rackowski and Richards (2005). For them, the matrix v assigns accusative case to the embedded CP, making it transparent. This then allows the matrix v to continue probing down to the lower phase, the embedded vP. Manetta’s assumptions about H-U phrase structure are a bit different. As (10b) shows, the wh-expletive is generated inside the domain of the transitive v, in Spec,AspP, where it is assigned accusative case and satisfies the EPP feature on v via Move. This, then, raises the question of the CP associate, which Manetta takes to be the true complement of V: [vP Subject [[VP CP V] v]]. Putting the two together gives the structure in (11).
Though the CP is the complement of the matrix V, Manetta argues that its postverbal position is due to a postsyntactic linearization rule. She posits that the phase boundary of the embedded C is transparent to v, which can then continue to probe down to the next phase edge.3 Note that this also holds for (9), which does not have an expletive. The matrix v can probe through C to the edge of the embedded vP to satisfy its wh-feature and EPP feature.
To sum up, Manetta claims that three very disparate properties of H-U question formation have at their core a single source. Wh-expressions are obligatorily moved to Spec,vP, and in this position their features are transparent for valuation from C.
1.3 Probing Assumptions about Complementation
The status of the embedded CP is a much-discussed topic in the syntax of H-U. Here, we will look at some assumptions, explicit or implicit, that Manetta makes in this regard. As (11) illustrates, Manetta considers the finite clause to be a true complement of V but one that is not assigned case by the transitive v. When the wh-expletive is present in the structure, v assigns accusative case to it and also values its EPP feature in the process. I take it that a sentence like (1c), repeated here as (12a), would have a derivation along the lines of (12b). I am assuming that the non-wh-expletive has no features and merely satisfies the EPP features of the matrix transitive v. In this case, the embedded CP is interpreted as an ordinary root question and an indirect question interpretation is obtained.
An issue that remains unaddressed is the reason for the impossibility of a direct question interpretation in such cases. Suppose there is no overt wh-expletive (i.e., no kyaa ‘what’) in the sentence. Either yeh ‘this’ satisfies the EPP feature of the matrix v or the subject does. Now, suppose the matrix C has Q- and wh-features, interpretable and uninterpretable, respectively. We would get a structure like (13b) for (13a), modeling it after Manetta’s account of the corresponding scope-marking construction (10b), repeated in (13c) for ease of comparison.4
In this derivation, the matrix v needs to value its uninterpretable wh-feature, which it can do by probing into the embedded clause. The matrix C can now use the features of v to value its own uninterpretable wh-feature. Given the logical possibilities available within Manetta’s system, then, the prediction is that H-U should allow wh-expressions inside finite complements to have matrix scope, a prediction that the facts do not support.
As we can see, Manetta’s account leaves a crucial gap: it does not nail down the status of the finite complement as a scope island. Since all accounts of long-distance wh-dependencies in H-U build on this fact, the omission is not a trivial one.
2 The Three Facets of Question Formation in Hindi-Urdu
Let us turn now to the strategies H-U employs in forming questions. We will start with local wh-movement, which is known to favor the preverbal position. We will see that Manetta’s account does not capture this fact. We will then consider two strategies used to override the ban on questioning out of finite clauses, long-distance scrambling and scope marking. Manetta argues that the first involves extraction and the second an expletive wh-strategy. We will see that these positions are problematic when the full range of facts, new as well as old, is taken into consideration.
2.1 Local Wh-Movement
We have seen the role that Spec,vP and the theory of feature checking play in the analysis of H-U question formation proposed by Manetta. Let us focus on local wh-movement to see if it captures the adjacency of the wh-expression and the verb, the primary goal of Manetta’s article. As Manetta herself notes (see, e.g., pp. 14–15), positing a vP phase with the EPP feature that attracts a wh-expression does not guarantee this. A theory of obligatory scrambling is needed to derive the relevant adjacency. Spelling out such a theory is not a trivial enterprise, as a look at Kidwai 2000 makes clear. Without a similarly explicit specification of the details by which scrambling would be ensured, Manetta’s explanation for the preverbal position of H-U wh-expressions in terms of wh-movement to Spec,vP cannot deliver on its promise.
2.2 Long-Distance Scrambling
Let us turn now to cases involving finite complements. Gurtu (1985), and following her Mahajan (1990), discusses examples of overt wh-movement in H-U of the kind discussed in section 1 under the term extraction. As I noted in Srivastav/Dayal 1991, 1996, such questions were not initially accepted by many speakers. As I also pointed out there, however, the sentences become fully acceptable under certain intonational contours. (In her footnote 11, Manetta herself notes a similar problem in Kashmiri, but she does not mention this fact for H-U.) At any rate, the conclusion I offered for the resistance to these data was that the cases under discussion do not instantiate the neutral/primary strategy for long-distance dependencies in the language, as we might expect them to if they represented extraction of the kind seen in English. Instead, I characterized them as cases of long-distance scrambling.
Briefly, there are two pieces of evidence that support the analysis of long-distance wh-displacement as scrambling. First, such questions are not neutral requests for information but show the kind of discourse sensitivity typically associated with scrambling. Stressing the matrix subject in (14) makes the question acceptable but also infuses it with a contrastive reading. The information sought is about the addressee’s thoughts as opposed to someone else’s.
kaun TUM socte ho ki ____ aayegaa?
who you think that will.come
‘Who do YOU think will come?’
Second, as also discussed in Dayal 1996, while a direct question interpretation is possible for questions like (14), it is not required. In the context of people wondering what Anu is doing, as in (15), an indirect question interpretation is natural. This is a well-attested fact about scrambling (Saito 1992).
A: anu kyaa kar rahii hai?
Anu what is.doing
‘What is Anu doing?’
B: kaun anu soc rahii hai ki ____ aayegaa.
who Anu is.thinking that will.come
‘Anu is wondering who will come.’
Let me emphasize that (15) is not, in and of itself, an argument against Manetta’s account of wh-movement in H-U. It merely establishes the existence of long-distance scrambling in the language, that is, of instances of discourse-sensitive long wh-movement that do not necessarily lead to matrix scope interpretations. In doing so, it sets the stage for evidence that does argue against Manetta’s claim that the landing site of long-distance wh-movement is the matrix Spec,vP position. Put simply, long wh-movement does not show the preference for preverbal position that the intended analysis is designed to capture. In this, it differs from local wh-movement.
(16a) shows that it is possible, if somewhat marginal, to move a wh-phrase out of the complement of a ditransitive verb like bataanaa ‘tell’. Notably, this is possible if the wh-expression moves to clause-initial position, not if it immediately precedes the matrix verb as in (16b) (or even the indirect object and the matrix verb as in (16c)). Note that a pause after the wh-phrase helps the acceptability of (16a), but no such modulation redeems (16b) or (16c).5
?kaun, anu-ne umaa-ko bataayaa ki ____ yehaaN rahtaa hai?
who Anu-ERG Uma-DAT told that here lives
‘Who is such that Anu told Uma that he lives here?’
*anu-ne umaa-ko kaun bataayaa ki ____ yehaaN rahtaa hai?
*anu-ne kaun umaa-ko bataayaa ki ____ yehaaN rahtaa hai?
Another, perhaps clearer contrast emerges in (17), where the displaced expression is the nominative-marked object of an embedded dative subject construction.
?kaun, siitaa-ne kahaa ki umaa-ko ____ pasand hai?
who Sita-ERG said that Uma-DAT is.appealing
‘Who did Sita say Uma likes?’
*siitaa-ne kaun kahaa ki umaa-ko ____ pasand hai?
Similarly telling examples are given in (18)–(19). The embedded agentive subject, with the same case as the matrix subject, can be scrambled to the clause-initial position, as shown in the (a) cases, but not to the matrix preverbal position, as shown in the (b) cases.
kis-ne siitaa-ne socaa ki ____ ravi-ko dekhaa?
who-ERG Sita-ERG thought that Ravi-ACC saw
‘Who did Sita think saw Ravi?’
*siitaa-ne kis-ne socaa ki ____ ravi-ko dekhaa?
kaun siitaa soctii hai ki ____ yehaaN rahtaa hai?
who Sita thinks that here lives
‘Who does Sita think lives here?’
*siitaa kaun soctii hai ki ____ yehaaN rahtaa hai?
Finally, the preference for scrambling to clause-initial position is corroborated by the pairs of sentences in (20) and (21), due to Rajesh Bhatt (pers. comm.). The matrix subject and the scrambled embedded object in (20) are both marked with -ko, though the matrix subject involves dative case and the wh-phrase the homophonous accusative case. By varying the form of the dative subject from tum-ko to tum-heN in (21), we eliminate the possibility that the contrast in (20) is a low-level surface Obligatory Contour Principle–type effect.
kis-ko anu-ko lagtaa hai ki ravi ____ pasand kartaa hai?
who-ACC Anu-DAT seems that Ravi likes
‘Who does it seem to Anu that Ravi likes?’
???anu-ko kis-ko lagtaa hai ki ravi ____ pasand kartaa hai?
kis-ko tum-heN lagtaa hai ki ravi ____ pasand kartaa hai?
who-ACC you-DAT seems that Ravi likes
‘Who does it seem to you that Ravi likes?’
(?)tumheN kis-ko lagtaa hai ki ravi pasand kartaa hai?
While (21b) is much improved, it remains somewhat marginal. The important point, though, is that in every case the order in Which the displaced phrase is clause-initial is perfect, modulo intonation. Instead, the one in which it is in preverbal matrix position is either completely unacceptable or degraded to some extent. It appears that the default landing position for long wh-movement is the matrix clause-initial, not preverbal, position.6
Given what we have seen here, long-distance wh-movement in H-U is clearly not analogous to extraction in English, as it allows for indirect question readings. And its landing site is clearly not the Spec of the matrix vP, as it shows a preference for clause-initial over preverbal matrix position.7 Any account used to explain local wh-movement cannot be extended to long wh-movement, as they are clearly distinct phenomena.
2.3 Scope Marking
As we saw in section 1.2, Manetta’s account of scope marking is a direct dependency account. However, it bears mentioning that there is no inherent conflict between indirect dependency and the clausal architecture that Manetta assumes. In fact, Rackowski and Richards (2005), whose work Manetta’s analysis builds on, endorse the indirect dependency account for scope marking. Manetta raises several objections to the indirect dependency approach to H-U scope marking. Most stem from misconceptions about the connection between the semantics of indirect dependency and its syntactic manifestations, but to go into them here would take us too far afield.8 I will therefore restrict my discussion to scope freezing, generally thought to follow from the indirect dependency approach. Manetta, however, offers a solution within the direct dependency approach she proposes that merits consideration.
To appreciate what is at stake, consider the extraction structure in (22a). Kroch (1998) noted that such questions are ambiguous between a referential reading (‘What is the size of the set of books that Bill said John read?’) and an amount reading (‘What is the number such that Bill said John read that many books?’). Lahiri (2002) made the important observation that H-U scope-marking structures only have the amount reading—in other words, the reading that monoclausal amount questions have. Here, I demonstrate the contrast using the English extraction and sequential scope-marking structures in (22a) and (22b), respectively.
How many books did Bill say John read?
What did Bill say? How many books did John read?
As Lahiri (2002) argues, the scope-freezing effect follows from an indirect dependency approach since the question in (22b) effectively asks, Which proposition in the set denoted by How many books did John read? is such that Bill said that proposition?
Manetta addresses the challenge that scope freezing poses for the direct dependency account she proposes in (10).9 She adopts Reinhart’s (1998) choice-functional analysis of long-distance dependencies, where the wh-expression stays inside the embedded clause. She suggests that overt extraction allows for ambiguity while covert scope taking does not. While this seems plausible enough on the surface, it does not hold up under scrutiny. Consider a language like Japanese, which mirrors precisely the type of LF that Manetta posits for H-U scope marking, one where the wh-expression remains in the embedded clause and is bound at a distance by an (unselective) Q-binder in the matrix Spec,CP.10
[John-ga [Bill-ga nan-satsu-no hon-o yonda to] omotteiru-ka]
John-NOM Bill-NOM how-many-GEN book-ACC read C think Q
‘How many books does John think Bill read?’
Ken-wa [Sato-sensei-ga kinoo nannin-no kanja-o shinsatsu-shita]-to
Ken-TOP Sato-Dr.-NOM yesterday how.many-GEN patient-ACC see-did C
‘How many patients does Ken think that Dr. Sato saw yesterday?’
The crucial point about these examples is that they do not display the relevant scope-freezing effect. (23a) and (23b) are ambiguous in exactly the way that the corresponding English extraction question in (22a) is. We can conclude, then, that the scope freezing in H-U scope marking is evidence of indirect dependency at work. Manetta’s account in terms of a long-distance dependency is not empirically supported.
A unified analysis of seemingly diverse properties of a language is a worthy goal, but I have shown here that the attempt in Manetta 2010 to provide such an analysis of H-U question formation falls short. There are assumptions on which the account rests, such as the status of finite clauses as scope islands in H-U, that are not clearly articulated. Focusing on the specific issue of wh-movement to Spec,vP, I have shown that an explicit account of the correlation between obligatory scrambling and preverbal focus position in H-U is needed. A proper evaluation of the role of movement to the edge of the vP in H-U must therefore await these elaborations. I have also addressed specific claims about long-distance wh-dependencies in H-U—long wh-movement as well as scope marking. I have shown that long wh-movement is not an instance of standard extraction. And I have shown that the embedded wh-phrase in scope-marking constructions does not take matrix scope. My conclusions were based on both old and new facts about these structures.
To end this discussion, we might ask, Does H-U have feature-driven wh-movement to Spec,vP? The answer is that it well may as far as local wh-movement is concerned, but we have no evidence one way or the other at this point. As far as wh-dependencies across clauses are concerned, we have seen evidence both against long wh-movement landing in Spec,vP and against a direct long-distance dependency between the matrix C and the embedded wh-phrase.
I am indebted to Mark Baker, Josef Bayer, Rajesh Bhatt, Miriam Butt, Lisa Cheng, Ayesha Kidwai, the participants of Syntax 3 at Rutgers University in Fall 2013, and three anonymous reviewers for much useful feedback. All remaining errors and omissions are my own.
1 Let me state at the outset that these facts are uncontroversial and hold across dialects of H-U. The same holds for other facts discussed in this article.
2Manetta (2010) also notes the possibility of focused non-wh-phrases appearing in other positions in the sentence. The data here highlight that the same holds for wh-phrases.
3Manetta (2010:20) states, “Following Rackowski and Richards’s (2005) approach, the matrix v head must have some feature that requires it to Agree with the embedded C, just as it might interact and agree with a direct object. In Tagalog, the presence of this feature has overt morphophonological consequences; in Hindi-Urdu, it does not. A consequence of this relation is that the phase boundary of the embedded C becomes transparent to v, and v can continue probing down to the next phase edge.”
4 It is not entirely clear to me whether the matrix v needs to have an uninterpretable Q-feature. Manetta takes an interpretable Q-feature to correspond to an unselective existential binder over choice function variables. Given standard assumptions, such binding can occur across islands. At any rate, regardless of whether the matrix v should have an uninterpretable Q-feature, the unavailable reading is predicted within the terms of Manetta’s account. Thanks to an anonymous reviewer and Rajesh Bhatt for helpful discussion.
5 Let me reiterate the point made in footnote 1, this time with respect to (16)–(21). These data have been verified by several speakers of H-U from different parts of India and Pakistan. The contrasts hold across speakers and dialects. An anonymous reviewer notes, for example, that the difference between (16a) and (16b) is more of a contrast than a ? versus * judgment but agrees that (17) is clearer and furthermore that the conclusion drawn on the basis of these data is valid.
6 Another anonymous reviewer agrees that (16)–(21), as well as the other observations, are robust but suggests that they may be compatible with long wh-movement to the Spec of the matrix vP. The suggestion is that the propensity for wh-phrases to appear in clause-initial position may be due to the triggering of further movement by C-related features. This would prevent the embedded wh-phrase from stopping in Spec,vP. However, this suggestion distinguishes long wh-movement in H-U from extraction, a suggestion that goes counter to Manetta’s proposal where movement to clause-initial position is not required for scope. Also, it does not explain the possibility of the indirect question interpretation found in (15).
The reviewer also suggests that movement triggered by C-related features may explain the well-known incompatibility of matrix yeh ‘this’ and long wh-movement.
(i) a. siitaa (*yeh) kaun soctii hai [ki ____ ayegaa]?
Sita this who thinks that will.come
‘Who does Sita think will come?’
b. kaun siitaa (*yeh) soctii hai [ki ____ ayegaa]?
I believe this conclusion is premature. Consider the following perfectly acceptable sentences, which contain both a yeh ‘this’ associated with the finite complement and a wh-expression that is an argument of the matrix verb:
(ii) a. kis-ne yeh siitaa-ko bataayaa ki ravi nahiiN aayegaa?
who-ERG this Sita-DAT told that Ravi not will.come
‘Who told Sita that Ravi won’t come?’
b. kis-ne siitaa-ko yeh bataayaa ki ravi nahiiN aayegaa?
Under Manetta’s account, the wh-expression kis-ne and the so-called expletive yeh would both have a claim to the preverbal position. Whatever mechanisms would be harnessed to navigate these cases would apply equally to the unacceptable (ia–b), and whatever mechanisms are used to rule out (ia–b) would apply equally to the acceptable (iia–b).
7 The landing site of long movement is an old problem in the literature on scrambling. See Mahajan 1990, Dayal 1994a, and Kidwai 2000 for discussion specific to H-U. Here, I restrict myself to the obvious conclusion that the landing site is not the final scope position but a position from which reconstruction to the embedded Spec,CP is possible (see (15)).
8 For example, Manetta classifies Lahiri’s (2002) account of variable binding in scope marking using Skolem functions as a semantic workaround (p. 28). Lahiri’s solution relies on independently motivated and well-established principles in the semantics of natural language and can hardly be classified as a workaround. Manetta also claims that her account answers several questions that are mysterious in earlier accounts, such as the role of the wh-expletive (pp. 25–26). The indirect dependency approach makes explicit and precise claims on this point (Dayal 1994b), leaving little room for mystery. Furthermore, there is greater flexibility in the syntax of scope marking within indirect dependency than Manetta recognizes (see, in particular, Dayal 2000 on this).
9 Manetta (pp. 30–31) presents the following paradigm to argue against direct dependency, citing Bhatt 2003 for (ib).
(i) a. mujhe [yah khabar [ki ve log nahiiN aa paaeNge]] kal milii.
I.DAT this news that those people not come will.be.able yesterday got
‘I got this news that those people will not be able to come yesterday.’
b. mujhe [yah khabar kal milii [ki ve log nahiiN aa paaeNge]].
(ii) *siitaa-ne [kyaa [ki ravi-ne kis-ko dekhaa]] socaa?
Sita-ERG what that Ravi-ERG who-ACC saw thought
‘Who did Sita think that Ravi saw?’
(ia–b) are the non-wh-counterparts of scope marking. The CP complement can occur inside the noun phrase in preverbal position (ia) or at the right edge of the clause (ib). The scope-marking construction (ii) is ungrammatical with the CP complement inside the noun phrase. Since the base form is missing, Manetta argues, there is no source under indirect dependency for the attested form where the CP occurs at the right edge.
However, Manetta’s paradigm is incomplete. The non-wh-counterpart of (ii) is (iii), without a head noun. It is unacceptable, just like (ii). The CP complement is bad in preverbal position, good at the right edge.
(iii) maiN [yah [*ki ve log aayeNge]] jaantii thii [ki ve log aayeNge].
I this that those people will.come knew that those people will.come
‘I knew (it) that those people will come.’
The true generalization about scope marking is that a full noun-complement structure is not possible with the wh-counterpart, regardless of position.
(iv) a. *siitaa-ne [kyaa khabar [ki ravi-ne kis-ko dekhaa]] socaa?
Sita-ERG what news that Ravi-ERG who-ACC saw thought
b. *siitaa-ne [kyaa khabar] socaa [ki ravi-ne kis-ko dekhaa]?
Why yeh ‘this’ and kyaa ‘what’ differ in this respect is an interesting independent issue, which needs to be settled before implications for direct versus indirect dependency can be evaluated. It may well pose a problem also for Manetta’s conception of H-U complementation (see (11)).
As a final aside, note that (v) with a genitive possessive and an NP head is possible.
(v) tumhaaraa kyaa khyaal hai ki kaun ayegaa?
your what thought is that who will.come
‘What are your thoughts about who will come?’
10 Thanks to Satoshi Tomioka and Kunio Kinjo for help with these data. Note that for the relevant ambiguity to arise, the embedded wh-phrase must not be of the form nan-satsu hon-o (lit. ‘how-many book-ACC’), which has a floating numeral quantifier, because such phrases tend to be nonspecific and favor narrow scope. Satoshi Tomioka points out, however, that this is just a tendency, not a hard fact. The data in (23) do not pose such problems. (23b) can mean ‘What is the size of the group of patients that Ken thinks Dr. Sato saw?’ or ‘What is the number such that Ken thinks Dr. Sato saw that many patients?’.