Concord within DP argues that movement is driven by uninterpretable features of either the target or the moved item, contra Chomsky 1995. The uninterpretable f-features of which concord consists must be eliminated by LF, to satisfy Full Interpretation. But raising of inflected APs and KPs into checking relations with N0 cannot be motivated, in Chomsky's system, since N0 has no uninterpretable features that these items can check. Assuming Kayne's (1994, 1998) proposal for APs, the problem can be partially overcome, but inflected “of” constructions still lack an account. Chomsky's (1998) probe-goal approach applied to concord also encounters difficulties, avoided under revision of the (1995) system: if the f-features of APs and KPs drive them to raise for checking, correct results are obtained.

No abstract

I argue against Chomsky's (1999Chomsky's ( , 2000 proposal that Case deletion correlates with the u-completeness of probes, based on (i) the omission of gender in subject agreement in, for example, Romance languages; and (ii) the inclusion of full u-features in subject agreement in Bantu, repeated on all verbal heads within a clause. I propose instead a return to the traditional view that certain categories are Case ''assigners,'' such that Agree deletes the goal's Case only if the probe has an intrinsic structural Case value. Finally, I show that Agree so modified accounts for concord in noun phrases, including concord on 'of' in African languages, reflecting u-features of head nouns. Crucial to this account is a structural analysis in which 'of' is merged with a nominal constituent that includes the head noun but excludes the surface 'of' object, be it possessor, agent, or theme. The Agree RelationIt is fairly common in syntactic theory to suppose that a relationship exists between Case and agreement. A recent instantiation of this idea is Chomsky's (1999Chomsky's ( , 2000 proposal that uninterpretable u-features delete Case through the relation called Agree. The u-features of an agreeing element a, which are called a probe, seek a goal element b with interpretable u-features and an unchecked structural Case feature, under Chomsky's theory. When the search succeeds, the Agree relation results, deleting a's u-features and b's Case. This is necessary because both Case and the uninterpretable u-features of agreement are illicit at LF, owing to their lack of interpretations.Once its Case is eliminated, b ceases to be a candidate for further Agree relations and accordingly for A-movement, which is tied to instances of Agree.As Chomsky notes, not all agreement relations can be taken to delete Case. To see why this is so, consider (1a). Here the subject is a deep object and enters into two Agree relations before raising to its surface position. The first is motivated by past participle agreement (PPA) on morte, the features of which are, by assumption, deleted in an Agree relation with elle. The Case of elle appears not to be deleted in this relation, however, as a second Agree relation is then established between elle and subject agreement (SA) on T 0 , here instantiated as the auxiliary est. This deletes both the nominative Case of elle and the u-features of est. Elle then raises to [Spec,TP] to satisfy the EPP feature of est.

Agree(X, Subj) accounts for all agreement in West Germanic: complementizer agreement (CA) results from an Agree relation between uninterpretable φ-features of Fin0 (Rizzi 1997) and φ-features of the subject; subject-verb agreement (SA) spells out uninterpretable φ-features of T0 on V0 raised to T0, even in OV clauses (Haegeman 2000). Although DPs need Case to participate in Agree relations (Chomsky 2000), deletion-marked Case remains syntactically accessible until the next strong phase (Pesetsky and Torrego 2001), allowing CA and SA to cooccur. In Frisian, ‘that’ cannot agree in embedded VO clauses because it is in Force; the verb is in Fin0, bearing CA (contra Zwart 1997).

There are at least four opposing views on the directionality and configuration of Agree relations.