Lexicalizing the positions above D

5.3.1 Recapitulation: disruption effects and Underassociation

Although this has already been discussed in the previous chapters, I would like to briefly recapit-ulate how movement and base-generated interveners change the spell-out possibilities of syntactic structures. Suppose that a lexical itemLI can span the features A, B and C. If these features are contiguous in the syntactic representation, thenLI can spell them out at once, as in (45).

(44) LI ⇔{A, B, C}

(45) A

B C

LI1

If, however, movement or a base-generated intervener breaks the contiguity, then A, B and C will have to be spelled out by two different lexical items. (46) shows a movement scenario where C is separated from A and B and cannot be spelled out together with them.

(46)

CP . . . C. . .

X A

B C

How A, B and C are spelled out in a structure like (46) depends on what kind of vocabulary items are available in the language. Suppose thatLI can Underassociate (i.e. not function as the exponent of) A and B but not C. Then it can shrink down to spell out C alone. Alternatively, if there is a lexical item that can spell out only C, then it is employed as the exponent of C. What happens to A and B? Again, this depends on the lexical items available in the language: they may be spelled out separately or together.

I have mentioned in the previous chapters that the empirical evidence tells us that lexical items spanning multiple terminals sometimes simply cannot underassociate some of their features (Svenonius, 2009; Starke, 2011). Consider now what happens in a structure like (46) ifLIcannot Underassociate its highest and lowest features (in our example, A and C). This means that this lexical item is entirely unshrinkable and consequently the spellout of a structure like (46) cannot containLI (it cannot spell out A, B and C at once but it cannot shrink down to C or A and B either).

In what follows I will argue that this is exactly what happens with dressed Ps. To anticipate the analysis, I will propose that dressed Ps spell out both case and P features, but they cannot shrink at all. Any kind of movement would separate their two features from each other, and thus yield a structure that cannot be lexicalized with a dressed P. As a result, these adpositions are always contiguous to their complement and cannot be separated from it.

5.3.2 Dressed Ps span case, naked Ps don’t

Recall that the consensual view of (spatial) case markers in the Hungarian literature is that they are suffixal P elements; and that along with many other authors, I take naked Ps to be true adpositions that occupy some projection in the extended PP. Taken together, this means that the PP must be decomposed into two layers, at the very least: a higher layer that is spelled out by naked Ps and a lower layer that is spelled out by the spatial case makers. I will label the layer spelled out by spatial case markers as KP, and the higher layer spelled out by naked Ps as PP.

Their fine-grained decomposition and semantics will be thoroughly discussed in Section 5.5.

(47) [[[DP a the

hat´ar]-on border-sup

KP] t´ul beyond

P P]

‘beyond the border’

I suggest that the crucial difference between dressed and naked Ps is whether P and K are lexicalized by the same lexical item or not. Let us begin with naked Ps. These adpositions spell out some material above KP, inside PP, but they do not spell out KP. KP is spelled out by an independent morpheme, the case-marker. As there are Place, Path and non-spatial naked Ps as well, there must be variation among the individual naked postpositions as to how many and exactly which features they spell out in the P-domain. In the trees below, X and Y stand for any projection in the extended PP. These structures do not intend to suggest that naked Ps spell out exactly two features, this is merely a representational convenience.

(48) Lexical entry of a naked P XP YP

Y X

naked P

(49) PP with a naked P YP XP KP

DP K

case-marker X

Y

naked P

A specific example:

(50) a the

fal-on wall-sup

t´ul beyond

‘beyond the wall’

(51) PlaceP

KP DP a f al

K -on

Place t´ul

Let us turn to dressed postpositions. The complement of these postpositions does not bear morphologically visible case. This fact has been interpreted in the literature in two ways. Mar´acz (1986, 1989) suggest that these complements bear the morphologically null Nominative case, while E. Kiss (2002) and Asbury (2008b) argue that they are caseless. I propose that the feature shared´ by all dressed Ps, spatial and non-spatial, is that in addition to P, they spell out K as well.

(52) Lexical entry of a dressed P YP XP KP

K X

Y

dressed P

(53) PP with a dressed P YP XP KP

DP K

X Y

dressed P A specific example:

(54) a the

fal wall

mellett next.to

‘next to the wall’

PlaceP KP DP a f al

K

Place

mellett

This means that there is a K in the structure of dressed PPs, but that K is swallowed by the P, leaving the complement to be the spell-out of merely DP. This captures the intuitions of both Mar´acz on the one hand and ´E. Kiss and Asbury on the other. ´E. Kiss and Asbury are right in that there is no Nominative case layer in between the dressed P and its complement. However, Mar´acz is right in that there is a case-related syntactic projection in the representation of dressed PPs.

This proposal amounts to saying that the very same chunk of structure can be lexicalized in two ways: by a dressed P alone, or by a case marker plus naked P combination. Therefore here it is highly relevant that the effect of the Maximize Span economy principle can be obviated if a

less economic lexicalization gives rise to a different meaning. Dressed Ps on the one hand and case marker plus naked P combinations on the other give rise to different meanings indeed. There are Place and Path denoting Ps in both classes, but each adposition adds a different lexical-conceptual meaning to the syntactic-grammatical features Place or Path, thus they all yield a different kind of Place or Path. For instance, dressedk¨oz¨ott ‘between’ and naked szemben + Comitative case

‘opposite to’ both project up to PlaceP and yield a place denotation, but the different lexical-conceptual meaning associated to these Ps guarantees that they do not compete with each other.

It would not be possible to use a dressed P and get a PlaceP with the meaning ‘opposite to’, therefore dressed Ps don’t blockszemben + Comitative case, or any other case marker plus naked P string.