Agree at a DistanceB + Second Merge
For Chomsky (2000), XP-movement involves a two-step process, illustrated in (1). Feature checking is mediated by the operation Agree, which applies at a distance between a feature of a c-commanding head H (a probe) and a feature (e.g., wh, f) of an XP category a that is c commanded by H (a goal). a is then merged as a specifier of H to eliminate the EPP feature of H. Thus, movement to Spec (or ‘‘second Merge’’) is independent of the checking of formal features under Agree.
(1)
The second Merge step of the derivation is problematic from the point of view of the VMH. It faces the same problem as the specifier-head agreement approach in that the category is forced to occupy Spec,HP.
One could attempt to derive the effect of the VMH by simply saying that the EPP property is inactive in the case of wh-subjects, but this raises the question of why the EPP property would be allowed to be inactive only in such constructions. It could be argued that C is simply not inserted in overt syntax in questions with wh-subjects, so that all relevant movements are done in covert syntax and the clause at Spell-Out is simply TP, with no CP structure projected above it. However, this gives rise to problems with respect to the interrogative specifier analysis. We are thus left with an unresolved paradox under this approach as well.
It is important to note that in this context, Chomsky (2000) rejects the notions of feature movement and feature chains advocated in previous work (e.g., Chomsky 1995). He states:
In the absence of clear evidence to the contrary, I will assume that feature chains do not exist, hence that features cannot move or be attracted. (2000, 119)
In MP [Chomsky 1995] Agree is analyzed in terms of feature movement (Attract) and a concept of matching that is left unclear. Here we take matching to be identity and dispense with Attract, with complications it induces about extended MLIs [modified lexical items], feature chains, and other matters. (2000, 123)
In what follows, I revisit the notion of feature movement and suggest that the VMH paradox constitutes evidence in favor of the feature movement approach. The following section first outlines the two-chain theory, which incorporates feature movement, and then applies it to the observed behavior of wh-subjects.
