Tying in temporal and non-spatial Ps

Most of the PP literature is preoccupied exclusively with spatial PPs. Spatial PPs provide a fertile ground for linguistic research because they are widely recognized to be compositional (Path is built from Place) and this makes it easy to test entailments and probe into the structure. Temporal Ps have received much less attention, although it is usually assumed that the structure of spatial adpositions can also account for these with no or minimal modifications. Other abstract Ps are as yet poorly researched.

One advantage of the Place > AxPart > K > DP decomposition of Svenonius (2008b, 2010), adopted here as Place > AxPart > Nplace > K > DP, is that it has been shown to be able to account for PPs with non-spatial interpretation as well. Roy and Svenonius (2009) demonstrate that (the equivalent of) Place > AxPart > K > DP can also be found in temporal and causal PPs, and that the projections of this sequence have the same type of contribution in these PPs as in spatial PPs. I will shortly summarize how the system works, in a way that already incorporates the Place > AxPart > Nplace > K > DP modification I suggested on the basis of Terzi’s and Botwinik-Rotem’s work.

It is easy to see how the Place > AxPart > NplaceP > K > DP system can be extended to temporal PPs. All that is required is that the set of points that forms the output of each projection is interpreted on a linear temporal scale instead of a three-dimensional spatial scale. This can be achieved by taking the silent noun Nplaceto have a rather general lexical-conceptual meaning, not literally meaning ‘spatial place’. Alternatively we could make use of a silent Ntime instead, c.f.

Katz and Postal, 1964 and Kayne, 2005b for silenttime. I suggest that the former track is more preferable, because it keeps the number of silent nouns smaller, but nothing crucial hinges on this choice.

Consider the dressed Pel˝ott. This adposition has a spatial interpretation, ‘in front of’, and a temporal interpretation,‘before’. We have seen above how the meaning of ‘in front of’ adds up

compositionally from the semantic contribution of the PP-projections. The‘before’ interpretation arises in a similar compositional fashion. In the phrasebefore the lecture the Groundthe lectureis first turned into a set of points occupied by the lecture on the timeline. AxPart narrows down this set to its first few points (the beginning of the lecture). Place picks out a set of points situated before the beginning of the lecture on the timeline.

Turning to causal PPs, Roy and Svenonius (2009) argue that here the Ground is the cause and the Figure is the resulting event or effect. In causal PPs the Ground is mapped onto a causal space, or in other words, onto all events that the complement may cause (motivate, force, enable, require, etc.). AxPart selects a subpart of these events, for instance the set of all consequences or all enabled events. On the basis of this Place picks out a specific (set of) event(s) that give the exact relation of the Ground and the Figure.

On this account, the internal make-up of spatial, temporal and causal PPs is identical. The difference between these PPs is thus much like the difference betweendog andcat. The functional categories involved are the same, and the difference boils down to the lexical-conceptual information associated to the projections.

The vast majority of Hungarian PPs are spatial, temporal or causal. Their structure and semantics is now accounted for. Very few remain which do not fit either category. The outliers include dressed helyett ‘instead of’, n´elk¨ul ‘without’, or naked egy¨utt ‘together with’ and k´epest

‘compared to’. Whether the Place > AxPart > Nplace > K > DP decomposition can also capture their internal structure or not remains a topic for further research. But even if the answer turns out to be no, it does not provide a direct counterargument against the Place > AxPart > Nplace >

K > DP sequence of spatial, temporal and causal PPs. Note that in this chapter I have disregarded the question of where the external argument of the PP is introduced. This must obviously be done in a projection of its own, and that projection must be integrated into the Place > AxPart >

Nplace > K > DP sequence. It is possible that the outlier Ps spell out just the external argument introducing layer.

The reason why I did not discuss the question of where the external argument or Figure of PPs is introduced is that the the existing proposals are very different, and their evaluation would take up a lot of space without contributing much to the main topic of this chapter (the relationship between case markers and adpositions on the one hand and to the characterization of dressed and naked Ps on the other).

Svenonius (2003a) and Svenonius (2010), for instance, argue that the Figure is introduced rather high up in the structure by aphead, which is merged between Deg (of Place) and Path. Botwinik-Rotem (2008b), on the other hand, argues that the source of the external semantic role is the silent NP place. The proposals of Svenonius (2010) and Botwinik-Rotem (2008b) thus represent two extremes: the former suggests that the external argument is introduced in the highest projection within PlaceP, and the latter suggests that it is introduced in the lowermost projection of PlaceP.

Given that the position of the Figure does not directly bear on the main issues of this chapter, I will not pursue this issue here.

5.6 Summary

Internally to the Hungarian PP, case markers instantiate the lexicalization problem. On their own, they contribute a spatial Place or Path meaning, but combined with naked Ps they lose this meaning and merely serve as a grammatical glue between the noun and the adposition. At the same time, they occupy related positions in the extended PP in both cases. As in the previous chapters, I used the Superset Principle to tackle this problem. I suggested that in their free-standing use, case markers spell out the range of projections from K to Place/Path. When embedded under naked Ps, on the other hand, they spell out just a subset of these projections, corresponding to K and the lower P-projections. The loss of Place and Path meaning in this case is due to the Underassociation of the higher category features. The proposed structures for dressed and naked Ps as well as free-standing case markers is summarized in (173) through (175).

(173) Place denoting dressed P

PlaceP

AxPartP NplaceP

KP

DP K

Nplace

AxPart

Place

dressed P (174) Place denoting case marker

PlaceP

AxPartP NplaceP

KP

DP K

Nplace

AxPart

Place

case

(175) Place denoting naked P

PlaceP

AxPartP NplaceP

KP

DP K

Nplace

AxPart

Place naked P

case marker

The crucial driving factor behind this proposal is the syntax-semantics mapping I assume. DPs with place denoting dressed Ps, place denoting free-standing case markers and place denoting case + naked P combinations are all analyzed as structures with the same size: PlaceP. I rejected the idea of merging naked Ps higher than this as well as other alternatives on the basis of compositional semantics.

At the beginning of this chapter I asked two fundamental questions. First: what is the relation-ship of adpositions and case markers? The answer to this question is that the functional sequence of adpositional DPs and case marked DPs consist of the same projections, Path > Place > AxPart

> Nplace > K. Second: how should we characterize the dressed vs. naked divide in Hungarian ad-positions? The answer to this question is that both dressed and naked Ps are situated on the spine of the extended PP projection. Specifically, both of them are contained in PlacePs and PathPs.

Dressed Ps and naked Ps, however, lexicalize the same zone of the functional sequence differently.

Specifically, dressed Ps lexicalize K in addition to adpositional projections. The difference in their size was shown to have important repercussions for what kind of structures they can be used to lexicalize.

As far as the broader theoretical issues of general PP structure are concerned, I argued that Place does not directly combine with DP. Instead, their relationship is mediated by intermediate projections. The lowest of these is a garden variety KP, which is the maximal extension of the noun phrase corresponding to the Ground. The next higher projection encodes a possessive relationship, whereby the Ground is the possessor of a phonologically null noun Nplace. The semantic output of this possessive relationship is the eigenspace of the Ground, which is further modified and narrowed down by AxPart before Place and Path are merged into the structure. Table 5.6 summarizes the semantic contribution of each projection. Table 5.7 shows how the Place semantics of the phrase above the box is built up gradually.

Place region picked on the basis of the subset of points denoted by KP AxPart subset of points denoted by KP

Nplace set of all points denoted by KP, the place of KP KP

Table 5.6: The semantic contribution of P-projections

Place above the box

AxPart set of points on top of the box

Nplace set of all points denoted by the box, the place of the box

KP the box

Table 5.7: The derivation ofabove the box