Case in DP

We believe the Ezafe construction suggests an answer to this question. On the picture sketched above, DP is like VP in that:

_ D selects thematic arguments.

_ DP syntax is right-descending.

_ DP modifiers are lowest complements of the head – all begin in post-head position.

Suppose now that DP is also like VP in deploying its own system of Case marking; specifically suppose that:

_ [+N] complements of D need Case – they bear a Case feature that must

be checked.

_ D/‰ can (in general) check Case on its internal argument, just as V/Ì

checks one Accusative on an internal argument of V.

Then we will have the following consequences:

_ D will in general check Case on its NP restriction.

_ DP-modifiers that do not have Case features to be checked (PPs, CPs, and

disguised CPs) will remain in situ.

_ DP-modifiers that bear Case features (APs) will be required to move to a site where they can check Case (e.g., by Concord).

So the general picture we have is as in (1). The determiner every check its one structural Case on its nominal restriction (woman), exhausting its Case-checking potential. English postnominal PPs and CPs do not bear Case features, and therefore can stay in their base position. Likewise for reduced relative clauses, which we analyze as covert CPs, following Kayne (1994).¹ However APs that do not occur inside reduced relatives cannot remain in place, and must move to a site where their Case can be checked.²

1 “Reduced relative clause” is not a uniform notion. English reduced RCs appear to be full finite clauses, as evidenced by the fact that they can contain a clausal negation licensing negative polarity items (e.g., the men [not present in any of the pictures]). Under a number of proposals, the presence of clausal negation always implies a c-commanding tense (Laka 1990; Zanuttini 1996, 1997). By contrast, in many languages, including nearly all with prenominal relatives (e.g., Turkish), reduced RCs are clearly less than full finite CPs (see Krause 2001).

2 The mechanism of case-checking for languages with prenominal APs is discussed in Larson (2006), for the eastern Indo-Iranian language Pashto, which behaves like English in relevant respects. The derivation of a Pashto nominal agha t@ge peghla ‘that thirsty girl: The crucial parametric property of Pashto (and, by extension, English) is hypothesized to be an EPP/Edge feature on little d. For details see Larson (2006).