On theta role assignment by feature checking
Given the dominant role of the generalized syntactic operation of feature checking in contemporary Minimalist theory, the question arises as to whether feature checking is the primary mechanism whereby argument DPs get assigned their rightful theta roles in any syntactic derivation. This paper revisits the proposal, adopted either directly or indirectly in recent literature (Manzini and Roussou 2000, Fanselow 2001), that argument DPs in a sentence get theta roles assigned to them through a checking mechanism for theta features in keeping with the spirit of Minimalist theory (Chomsky 1995, 1999, 2000). It seeks to point out that, while the arguments presented by Fanselow (2001) against the movement analysis of scrambling of argument DPs in German are mostly valid, they do not conclusively establish that such argument DPs are necessarily assigned theta roles through a checking mechanism for theta features. In particular, it is demonstrated in this paper that there are problems with Fanselow’s arguments against scrambling as movement based on scrambled DPs as islands for movement and on parasitic gaps, and that, moreover, an account of scrambling as discourse-function-driven movement along the lines of, e.g., Kidwai (2000) has not been taken into account by these arguments against scrambling as movement. Furthermore, the operation Attract proposed by Manzini and Roussou (2000) as being driven by argument DPs and directed at predicates, it is argued, does not need to be based on a feature-checking account of theta-role assignment. Finally, it is pointed out that serious problems with motivating the operation Merge between predicate expressions and argument DPs arise under a feature-checking account of theta role assignment, and that <i>wh</i>-movement, in particular, can be accounted for only in costlier terms under such an account. The paper presents additional arguments, in terms of the nature of lexical entries in the lexicon component of the faculty of language (notably with reference to lexical decomposition) and the role of argument structure at LF (the conceptual-intentional interface level), that theta-features – if they exist at all – are features of a very different kind from the usual kind of features that are involved in the operations Move and Agree.