Focus movement out of TP?

Before concluding this section, let us examine the possibility of focus movement outside of the TP projection where the focus element originates. First, consider (35b) as a response to the question in (35a).

(35) a. Úgy tudom, János csak EGY PÁR FILMET nézett meg

‘I thought John watched only a couple of movies.’

b. Nem. János [TP MINDEN FILMET [TP meg nézett ... ]]

no J.-nom every film-acc PRT watched-3sg

‘No, John watched EVERY FILM.’

Here the NS is found on a fronted i-QP adjoined to TP, and it functions as the focus of the answer. In Section 5.1 above it was argued that overt focus movement targets the left edge of the IntP that corresponds to the TP category that excludes any adjuncts to TP. In (35b) above, however, an element functioning both semantically and prosodically as the focus of the clause is overtly raised outside of TP proper to an adjunct position.

It can be shown that this incongruity is merely apparent. The i-QP in (35b) is not interpreted as an id-focus. That it cannot be interpreted as an identificational predicate may be inferred from its inability to appear in the immediately pre-verbal Spec,TP position followed by an inverted V > VM word order. One could entertain the possibility that perhaps it cannot function as an identificational predicate only when occupying the Spec,TP position. That this is not so is evidenced by the fact that i-QP in a pre-TP position cannot function as a contrastive id-focus in constructions where it is preceded by an adjunct negation (id-foci may generally be used as contrastive, cf. É. Kiss 1998a), see (36a). The movement to the pre-verbal field of an i-QP bearing the NS is optionally overt or covert, similarly to non-NS-bearing i-QPs (see Section 2.2): (35a) can also be felicitously answered by (36b), on the same interpretation as that assigned to (35b).

(36) a. *Nem MINDEN FILMET meg nézett,

not every film-acc PRT watched-3sg hanem csak EGY PÁR FILMET nézett meg but only a couple film-acc watched-3sg PRT

intended: ‘He didn’t watch EVERY FILM, he only watched a couple of films.’

b. Nem. János [TP meg nézett MINDEN FILMET ... ] no J.-nom PRT watched-3sg every film-acc

We can conclude that the i-QP bearing the NS in (35b) and in (36b) is ordinary focus, rather than id-focus, and its syntactic movement is due to what is responsible for these properties of i-QPs generally: QR. As both (35b) and (36b) involve main stress shift (given that the NS is not located on the VM element in Spec,TP, but on i-QP), neither is more costly in that regard (cf. (21)). This determines neither the overt nor the covert spell out pattern of QR to be less economical.

However, this cannot be the whole story. Scope Transparency (31) would favor overt QR. Covert movement, at the same time, is less costly than overt movement, hence is preferable. One could entertain an account on which the optionality of the overt versus covert status of the QR movement is precisely due to these two antagonistic economy principles: if one is satisfied, the other will be inevitably violated. Unfortunately, this line of argument would falsely predict that the movement of id-focus to the pre-verbal position should also be optionally overt or covert, as the same economy principles would have the same violation profiles in each of the two cases. It can be conjectured that the movement of id-focus and that of i-QPs must differ in some regard that we have not considered yet.

I suggest that the relevant difference lies in a prosodic requirement that characterizes id-focus, but not i-QPs:

(37) An id-focus and its background (domain) must belong to the same intonational phrase.

This requirement may very well be related to the phenomenon of post-focal stress eradication, i.e., the reduction of stresses following the id-focus until the end of the intonational phrase.

(37) allows for scenarios where id-focus and its background together form an IntP with no IntP boundary between them, and it only permits a prosodic structure where there is an IntP boundary between the id-focus and (the whole or part of) the background, if that IntP boundary falls within the IntP that both the id-focus and its background are part of (i.e., a

recursively embedded IntP structure).³⁸ I suggest that (37) gives rise to the difference between id-focus movement and i-QP raising to the pre-verbal field in an indirect way.

I base my account on a sharpened version of the Focus–Stress Correspondence condition in (22) above, and two further assumptions pertaining to the syntax–prosody mapping (part of the syntax–PHON interface). First, this mapping has a preference for simpler prosodic structures over more complex ones. As a particular principle, I adopt (38).

Second, following the Transfer-based approach to the syntax–prosody mapping in Kratzer and Selkirk (2007), I assume that an adjunct phrase in a clause is mapped by default to its own IntP. Finally, I adopt a sharpened, and I believe, more accurate formulation of the Focus–

Stress Correspondence condition suggested by Reinhart (1995) and others (see Reinhart 1995:

62), which relativizes the prosodic prominence requirement to IntPs; see (40).

(38) Minimize the number of IntPs.

(39) Adjunct phrases are mapped by default to an independent IntP.

(40) Focus–Stress Correspondence

A focus constituent contains the prosodically most prominent syllable of the intonational phrase it is contained in.

In this work I have adopted the standard view of QR that it creates an adjunction structure, i.e., QR-ed i-QPs are adjuncts, in the case at hand, to TP. As a result, when QR is overt, the i-QP forms an IntP of its own. When QR to an adjunct position above the pre-verbal VM is covert, i-QP will spelled out in situ, where it will fall inside the IntP corresponding to TP. These two options are given schematically below:

(41) a. (_IntP QP ) (_IntP VM V . . . ) b. (IntP VM V . . . QP . . .)

If QP functions as a focus, then, as is by now familiar, stress strengthening (probably even stress shift, if Focus–Stress Correspondence is to be properly adhered to) needs to apply in (41b). Stress strengthening does not need to take place in (41a), however, unlike we assumed immediately above. This is because the QP, which, by assumption is a focus in the cases at issue, forms an IntP of its own, therefore it will necessarily contain the most prominent stress in its own IntP (though not necessarily the NS of the utterance, which by default falls in Hungarian on the prosodic head of the rightmost IntP of the Utterance Phrase). At the same

time the interface economy principle in (38) favors (41b), as it involves only one IntP, given that in (41b) QP does not form an IntP of its own. Therefore we have the prosodically dispreferred stress strengthening (stress shift) and the syntactically preferred covert movement, along with a violation of the Scope Transparency principle in (41b), against an extra IntP, disfavored by (38), and overt movement, a costly syntactic operation, and adherence to Scope Transparency in (41a). We can then understand the optionality between overt and covert realizations of QR in terms of these two very different violation profiles, as a lack of an overall preference of the grammar of either form over the other.

The relevance of (37) above should now be clear. The requirement in (37) does not allow FOC in (42a) and FOC1 in (42b) to form an IntP of their own. (42a–b) would be mapped from a syntactic structure where FOC and FOC1 are adjoined to TP – a structure that (37) excludes.

(42) a. *(IntP FOC ) (IntP VM V . . . ) b. *(_IntP FOC1 ) (_IntP FOC2 V . . . )

Ultimately, then, i-QPs are fronted optionally overtly or covertly because they form an IntP of their own by default (qua adjuncts), while id-focus must move to the pre-verbal position in overt syntax because of the prosodic condition that it belong to the same IntP as its background (or domain), where post-focal stress eradication applies.

As expected based on this account, an i-QP functioning as focus may overtly front to the left of a fronted id-focus too (43). In this case we have multiple foci. The prosodic structure corresponding to (43) is analogous to that in (41a) with id-focus replacing VM. If i-QP is a focus, then given the structure of (43), it has id-focus in its domain. Even though i-i-QP has its own main stress within its own IntP, this stress is perceived as stronger than that of id-focus. This may be the result of post-focal stress eradication following i-QP functioning as a focus, or to the more general requirement that stress eradication facilitates, namely that a focus needs to be the prosodically most prominent element in its domain (see Truckenbrodt 1995, 1999).

(43) [TP MINDEN FILMET [TP JÁNOS nézett [AspP meg ...]]

every film-acc J.-nom watched-3sg PRT

‘JOHN watched EVERY FILM.’

To conclude the discussion, it appears that the model can be successfully extended to cover cases like (35b), (36b) and (43). Note that according to the analysis presented here, apparent focus-movement from inside TP to outside TP proper of i-QPs does not involve bona fide focus movement; rather, it is an overt form of QR. It is its overtness that is derived from the (ordinary) focus property of the i-QPs entering these constructions. We briefly turn to an analogous scenario involving clausal negation, before we turn to non-focus i-QPs.

In the light of the preceding discussion, the fact that a fronted id-focus can be preceded by a clausal negation bearing a major stress falls into place. Clausal negation located in the same adjoined position that i-QP occupies in (43) can receive its own prominence in the same way as i-QP does (compare (25) in Section 4.3 above), i.e., by forming its own IntP. As clausal negation gets to a TP-adjoined position independently of what goes on inside TP proper, the same analysis and predictions apply to it as to (43), correctly, it appears.³⁹

(44) [TP NEM [TP JÁNOS nézett [AspP meg egy pár filmet ...]]]

not J.-nom watched-3sg PRT a couple film-acc

‘It’s not John who watched a couple of films.’

Finally, a note on the impossibility of raising id-focus phrases outside of TP. As should be clear, any elements in the clause that are to the left of the default NS position at the left edge of TP proper may bear NS only by stress strengthening, or by a recursive IntP structure where This also explains why an id-focus cannot be extracted from TP proper to some position in the clause preceding the leftmost PhonP of TP proper, say, to an adjunct position of TP, as in (45):

(45) *[TP JÁNOS [TP minden moziban [TP meg nézett [AspP ... egy filmet...]]]]

J.-nom every cinema-in PRT watched-3sg one film-acc intended: ‘It’s John who watched a film in every cinema.’

This hypothetical derivation involves stress shift ‘across’ an IntP boundary to the focus phrase that appears outside the IntP corresponding to TP proper. Granting for the sake of the argument that such stress shift is permitted, an alternative derivation where the movement of the focus targets Spec,TP, raising i-QP to some scope position below Spec,TP (see (46) below), does not incur stress shift, and therefore blocks (45).⁴⁰

(46) [_TP JÁNOS nézett [_AspP minden moziban [_AspP meg [ ... egy filmet...]]]]

J.-nom watched-3sg every cinema-in PRT a film-acc

In short, the model predicts, correctly it seems, that bona fide focus movement to a position within the clause that is outside TP proper (=IntP) is excluded.

There is some evidence that fronted i-QPs may be realized prosodically as topics. This is not wholly unexpected, given that those i-QPs that are sufficiently rich descriptively or are D-linked, are acceptable even in a syntactic topic position (e.g., Surányi 2003):⁴¹, ⁴²

(47) Minden diák, aki most itt ül tegnap egy buliban volt every student-nom who-nom now here sits yesterday a party-in was

‘Every student who’s now sitting here was at a party yesterday.’

Furthermore, in a small-scale prosodic experiment carried out together with Shinichiro Ishihara (see Ishihara and Surányi 2009), we found that non-focus fronted i-QPs tend to bear either H*, H*L or L*H pitch accents. Topic phrases were found to bear all these accents (and also the rise-fall L%H*L).⁴³ It appears then that – even though simple i-QPs are not in a syntactic topic position – fronted i-QPs can bear topic pitch accents, and are likely to be interpreted as topics. The movement of topics is restricted to overt syntax in Hungarian. I will suggest in Chapter 5 below that this is due to the following prosodic requirement of topic interpretation (essentially, the inverse of the relevant requirement applying to id-foci):

(48) Topics must not belong to the same IntP as the comment.

Consider now i-QPs that are to be interpreted as topics. If such an i-QP undergoes QR but is spelled out in situ, then it can only obey (48) if IntP boundaries are inserted around it (or at least at its left edge, if it is in a final position). If, however, it is spelled out in the higher position of its QR chain, then, as an adjunct, it will be mapped to its own IntP by default, as we saw above. It follows then that if an i-QP is to function as a topic, its QR must be overt.

In our experiment we did not find any occurrences of a fronted i-QP that could not be categorized either as having a topic interpretation or having a focus interpretation. I therefore suggest that the overtness of the QR of i-QPs is due to indirectly to the prosodic requirements of either one of these two interpretations. In all other cases, i.e., when non interpreted either

as a focus or as a topic, QR in Hungarian remains covert, just as it is in familiar languages of the English type.