Accounting for the adjacency effects

Word order

Hungarian adpositions follow their complement in the neutral order. For dressed Ps this is the only available order. Some naked Ps can also precede their complement, and this non-neutral order is associated with emphasis on the postposition. Asbury (2008b) suggests that the P > KP order might be a result of movement of the P to a higher focus projection, while Heged˝us (2007) suggests that this order arises by movement of the P from Path to a higher p head. I capitalize on this idea and analyze the P > KP order as a result of P-movement. Such a movement thus targets a P-feature or P-features, but leaves P’s KP complement in situ. The structures before and after movement are schematized in (55) and (56) respectively. P stands for any feature in the extended PP.

(55) XP

X PP

KP

DP K

P

(56) XP

X

P X

PP KP

DP K

tP

Nanosyntax is a theory that uses post-syntactic spellout, the Lexicon is accessed only after the structure has been built. In light of this consider how (56) could be lexicalized. Dressed Ps are specified for spelling out both P and K. Theoretically, it should be possible for dressed Ps to shrink (via the Superset Principle) and spell out either P or K, leaving the other feature to be spelled out by some other lexical item. The empirical data, however, tell us that neither of these options is workable: it is not possible to separate dressed Ps from their complement. That is, dressed Ps must be unshrinkable lexical items. Since they cannot Underassociate either P or K, they can only be matched to a piece of structure in which these features are contiguous.

This is not the case in (56). Due to the movement the P-feature and K are not adjacent to each other (recall that only the highest copy of moved elements counts for spell-out purposes). This means that a representation like (56) cannot be matched to a dressed P, and so the P > KP order is ungrammatical.

(57) a the

t¨uk¨or mirror

mellett next.to

‘next to the mirror’

(58) *mellett next.to a the

t¨uk¨or mirror

‘next to the mirror’

We thus see here a clear instance of the lexicalization algorithm and the lexical entry of a morpheme having an effect on word-order: movement scatters the features that could be co-lexicalized by a dressed P, and the dressed P cannot be used to lexicalize this structure.

The separation of P from K does not pose comparable problems for naked Ps, as in this case P and K are spelt out by different morphemes. In (56) P can be matched to the lexical entry of a naked P and K can be matched to the case-marker, just like when no movement takes place.¹¹ (59) a

While in principle it is possible to analyze the P > KP order above as failure of KP to move instead of P-movement, there is a scenario that clearly involves movement of P away from KP.

The relevant data involve adpositions functioning as verbal modifiers. The immediately preverbal position in Hungarian is called the verbal modifier position. This position is open to some naked Ps, as evidenced by (63), but not to any of the dressed Ps.

(63) Mari

‘Mary went through the field’

(64) a. *Mari

‘Mary went next to the mirror.’

b. Mari

‘Mary went next to the mirror.’

Just as in (57)–(62), we see again that dressed Ps must follow their complement at all times but some naked Ps may also precede it. The analysis of (63) and (64-a) proceeds along the lines outlined above. P and K do not form a contiguous sequence after movement in either (63) or (64-a).

This is a problem only for dressed Ps, however, as these must be matched to adjacent P and K features. (63) is ruled in because in this case P and K are spelled out by different morphemes.¹² Degree modification

Degree modifiers have been argued to be harboured by a designated functional projection DegreeP in Koopman (2000); den Dikken (2010) and Svenonius (2008b, 2010). I will follow this line of thinking here. As already discussed, degree modifiers can always precede the DP. I take this to be the unmarked option which involves no movement. An example with a naked P is given below.

11Given that Hungarian adpositions in a prenominal position give rise to a contrastive reading, in (62) I tentatively assume that XP is FocP. In (62) I depict this as head-movement of P to X (i.e. Foc). However, this is only a representational convenience. Whether this extraction is best characterized in terms of head movement or phrasal movement, it does not affect the argumentation. The point is that separating P from KP does not yield the right context for lexicalization by a dressed P.

Note that adpositions in general do not give rise to a contrastive interpretation, therefore there is no Foc feature in their lexical representation. As a result, the movement in (62) could not be represented as one lexical item spanning both P and X.

12I will address the issue of why not all naked Ps allow the intervention of degree modification or can appear in the verbal modifier position in Section 5.4.4.

(65) k¨ozvetlen¨ul right

a the

h´az-on house-sup

k´ıv¨ul outside.of

‘right outside of the house’

(66) DegreeP

Degree k¨ozvetlen¨ul

PlaceP KP DP a h´az

K -on

Place k´ıv¨ul

Structures in which the degree modifier intervenes between the postposition and the complement are derived by moving KP into the specifier of DegreeP, leaving the P-features behind. This gives the representation in (66).

(67) DegreeP

KP

K DP

Degree’

Degree PlaceP tKP Place

Consider now how this structure can be lexicalized. In the case of naked Ps the P-feature and K are spelled out by different morphemes. Whether movement takes place or not, at post-syntactic spell-out it is possible to match K to the case-marker and the P-feature to the naked P (69).

(68) a the

h´az-on house-sup

k¨ozvetlen¨ul right

k´ıv¨ul outside.of

‘right outside of the house’

(69) DegreeP

KP DP a h´az

K -on

Degree’

Degree k¨ozvetlen¨ul

PlaceP tKP Place

k´ıv¨ul

Dressed Ps, however, lexicalize both the P-feature(s) and K. They can be matched to a chunk of structure in which these features form a contiguous sequence. This is not the case in (67). The movement disrupted the K – P sequence, and when it comes to post-syntactic spell-out, there is no span in this tree that is identical to the lexical entry of a dressed P. A structure like (67) thus cannot be lexicalized with a dressed P.

(70) k¨ozvetlen¨ul immediately

a the

h´az house

mellett next.to

‘immediately next to the house’

(71) *a the

t¨uk¨or mirror

k¨ozvetlen¨ul immediately

mellett next.to

‘immediately next to the mirror’

The immobility of DP

A valid objection that could be raised here is why it is not possible to move just DP to the specifier of DegP, without breaking the contiguity of K and P. If this was possible, it would produce structures like (72).

(72) DegreeP

DPi

Degree PP

P KP

K ti

This structure is perfectly lexicalizable by a dressed P: K and P are adjacent heads in the tree.

If DP-movement was a licit operation in Hungarian, then DP would be routinely separable from a dressed P. Given that this is not the case, we need to exclude the possibility of DP extraction from KP altogether. That DP cannot be extracted from under KP is a correct empirical generalization (c.f. the data in the preceding section); the question is why this is so. The reason, I suggest, is very simple: in (72) DP is not the maximal extension of the noun phrase.

Researchers often talk about DP-movement, but what is really meant by this term is the movement of the maximally extended noun phrase. Before it had been widely acknowledged that case has its own projection on the NP-spine, the maximal extension of the noun was DP indeed.

But if KP is recognized as part of the NP f-seq, ‘DP-movement’ must really be KP-movement.

Thus the term‘DP-movement’ is literally movement of DP only for those who don’t subscribe to the existence of KP, and for everybody else it really means KP-movement.

In a KP > DP decomposition nobody would think that K can be stranded by DP, and in general it is not thought to be possible to extract a phrase that is just smaller than the maximal extension of the noun in any given situation. DP is immobile in (72) because it is KP rather than DP that is the maximal extended projection of the noun.

This contrasts with Grimshaw’s view that there is no important cut-off point around DP, and the maximal extension of NP is PP. In my view there is a significant juncture at KP in the nominal functional sequence, and the projections related to the computation of location or path are found in a separate functional sequence built on KP. In Section 5.5 I will elaborate on this view in detail and lay out the fine-grained decomposition that motivates this approach.

Being the maximal extended projection of NP makes KP ‘special’ in a sense. One way in which this ‘special’ nature of KP manifests is its ability to move. There is a natural and long-established parallel here with the architecture of the clause. That the structure of nominal and clausal projections show similarities has been repeatedly pointed out in the literature, for instance in Abney (1987); Lamontagne and Travis (1987); Bittner and Hale (1996); Szabolcsi (1987, 1994);

Alexiadou et al. (2007) and others, and CP has been argued to be similar either to DP or KP.

It is well known that the maximal extension of the clause, CP, can move, but the category that is just smaller than the maximal extension, i.e. TP, cannot (c.f. Abels, 2003b for examples and discussion). In a syntactic decomposition that makes use of KP, the natural nominal counterpart of CP is KP, as both are the maximal extensions of the core lexical projection in their f-seq. The immobility of DP under KP is thus the same issue as the immobility of TP under CP. Abels (2003b) provides an in-depth analysis of this phenomenon that crucially relies on phases and anti-locality.

In his theory, what I called above the‘special’ nature of KP would translate into being a phase.^13,14 In Chapters 8 and 9 I will come back to the phasehood of K and will show how it interacts with various agreement phenomena.

Another way in which KP is special is reflected in the long-standing observation that (overt) noun phrases need case. Differently put, KP is autonomous: it is the required and right size for a nominal projection to exist. In a KP > DP decomposition this means that every noun phrase

13Ables discusses the immobility of VP under vP, TP under CP, and DP under PP, but he does not discuss KP or phasehood internally to the nominal projection. Thus he makes no claim regarding the phasehood of KP.

14That noun phrases correspond to a phase (or phases) is a widely held position, c.f. for instance Radford (2000);

Svenonius (2003b); Fukui and Zushi (2008); Heck et al. (2008); Kramer (2009).

must be topped off by a KP, even if not all possible projections are projected in between N and K.

In other words, NumP or DP are not crucial for nominal projections, but KP is.

Government and Binding theory expressed the requirement for case as the Case Filter: case was seen as the tool that made nouns visible for thematic relations. Minimalism dispensed with the Case Filter, but has not yielded a deep understanding as to why case is required on nominals.

As Pesetsky and Torrego (2011, p. 72.) formulate it: "many fundamental questions, including the reason why case should exist at all, do not yet have substantive answers". This thesis is not the place to attempt to resolve the very big question of why KP is required on every nominal projection. For convenience, I will follow Gillian Ramchand’s suggestion (p.c.) that case is the piece of glue that enables nominals to bear a relation to something else in the proposition.

To sum up this section, KP represents an important juncture in nominal projections: it is autonomous and movable (and as I will argue below in connection with the (in)transitivity of Ps, it can be replaced by a null pronoun that represents an implicit ground). That DP cannot move away from and strand KP follows from the nature of KP, rather than DP. Specifically, it is KP that is the maximal extended projection of nouns, and this makes the category right under it immovable. (See Abels, 2003b for formalization of this idea.)

Wh-movement

The analysis of the data with wh-movement is identical to the account developed for other move-ments, and it should be obvious by now. The structure of a wh-question with a stranded postpo-sition is as in (73): KP is attracted to spec, FocP and the P-features stay in situ.

(73) FocP

KP

DP K

Foc’

Foc ...

PP tKP P

(73) cannot be lexicalized with a dressed P, as K and P are not adjacent, but a naked P can be matched to the P and the case-marker to K without any problems.

(74) *Mii

what

ment-´el go.past-2sg

ti alatt?

under

‘What did you go under?’

(75) Mi-ni

what-sup

ment-´el go.past-2sg

ti ´at?

through

‘What did you go through?’

(76) FocP

KP DP

mi K -n

Foc’

Foc ment´el

...

PP tKP P

´ at

If PP is pied-piped with KP, the P and K remain adjacent in the structure after movement, too.

This makes it possible for a dressed P to spell out the structure.

(77) Mi what

alatt under

ment-´el?

go.past-2sg

‘What did you go under?’

(78) FocP PP

KP DP mi

K P

Foc’

Foc ment´el

...

tP P

alatt

We have now accounted for all the data points that show that naked Ps are separable from their complement but dressed Ps are not. All sentences with extraction from a dressed PP were ruled out on account of K and the rest of P being separated. Given that dressed Ps cannot shrink, movement of either K or P yields a structure which cannot be matched to the lexical entry of a dressed P. We can see that in the lexicalization algorithm of Nanosyntax, lexical representations can constrain word-order possibilities. Specifically, certain movements are ruled out not because they violate syntactic principles (e.g. locality) and lead to a crash in narrow syntax, but because they yield structures which cannot be properly matched to the lexical items we want to use, and so cannot be spelled out with these lexical items.

Transitivity

Some naked Ps can be used intransitively, but dressed Ps must have a complement. How does this fact follow from the proposed analysis? Let us consider the structure of intransitive Ps. In D´ek´any (2009) I suggested that intransitively used Ps lack a syntactic complement. An intransitive naked P thus has the structure in (79), and an intransitive dressed P looks like (80).

(79) PP

P

(80) PP

KP K

P

I suggested that the structure in (80) is uninterpretable because it contains a K ‘hanging in the air’. Case allows DPs to surface in the clause, all DPs and only DPs need case. Conversely, KP can only be erected on top of a DP. With a DP radically missing from the structure, K has no function at all, and this causes (80) to be uninterpretable. By way of contrast, naked Ps do not spell out K, and when they appear without a complement, not only the DP but the whole KP is absent from the structure. Such a PP is entirely interpretable.

Here I would like to propose a different analysis. Consider the meaning conveyed by an intran-sitively used P. An example is given (81).

(81) a. Az The

iskola school

k¨ozel close

van be.3sg

a the

hat´ar-hoz.

boarder-allat

‘The school is close to the boarder.’

b. Az The

iskola school

a close

k¨ozel be.3sg

van.

‘The school is close (by).’

(81-b) has no overt Ground, nevertheless the sentence implies the presence of an implicit Ground. In particular, k¨ozel ‘close to’ is interpreted with respect to the utterance’s deictic center, which functions as the Ground of the seemingly intransitive P.

The reference of the deictic center must be inferred from the context: it could be located with respect to the speaker’s present or past position or with respect to a third party who is the topic of the conversation. In other words, the location that close refers to in the school is close (by) entirely depends on the context in which the sentence is uttered. This is reminiscent of how the reference of third person pronouns is computed: the reference ofhe, she and it must always be fixed with the help of the context. I suggest that the deictic center is represented in the syntax by a phonologically covert pronoun, prodeic, and this is in turn the complement of the P in (81-b).

The question now is why this pronoun cannot be the complement of dressed Ps. I suggest that prodeic is a pro-KP, in other words, it replaces KP in syntax. This follows naturally from the form of the PP in (81-b): not only the complement is missing, but also the case-marker that the P subcategorizes for (Allative in the example at hand). Note that case markers do not require a phonologically overt host. When the complement of a case marker is a pronoun, the pronoun can stay covert. In this case the case marker appears to function as the stem and it hosts person-number agreement with the covert pronoun.

(82) hozz-´am, allat-1sg

r´ol-ad, delat-2sg

t˝ol-e, ablat-3sg

n´al-unk, adess-1pl

hozz-´atok, allat-2sg

benn-¨uk iness-3pl

‘to me, from/off o you, from him, to us, at you, in them’

(82) clearly shows that pronouns don’t need an overt host. If prodeicreplaced DP, we would expect the case marker required by the naked P to surface overtly, cliticize onto the covert pronoun and show agreement with it, as in (83).¹⁵ However, (83) does not have a reading in which the Ground is the deictic center inferred from the context. In this PP the Ground can only be understood as a third person personal pronoun.

(83) Az

In document A profile of the Hungarian DP (Pldal 135-141)