ABSTRACT
Here I explore the question of what level of representation is relevant to the determination of anaphoric relations. Of particular concern are the implications of the feature movement theory for the theory of anaphora. The overall generalization is that anaphora seems to crucially depend on S-structure configuration. This, of course, is extremely problematic for minimalism, since that framework denies the existence of S-structure as a significant level of representation. The solution to the problem is that the covert portion of the derivation of LF moves only formal features. These feature movement operations “leave behind” the semantic aspects of nominal expressions crucial to anaphora. Thus, anaphoric connection is determined at LF, but it is as if it is determined at “S-structure.” Chomsky (1995a) had argued that LF operations do affect anaphoric relations, but his only example involves control. Attempts to replicate that example with other relations invariably fails. Further, I show that even the pattern with control is not general.
