EVALUATING THE DIRECT ASSOCIATION HYPOTHESIS - A REPLY

Recent investigations of filler-gap dependencies in sentence processin g have generally assumed that the parser must compute an antecedent-tr ace relation in which the trace site is identical to the canonical pos ition of the moved phrase. In a recent article, Pickering and Barry (1 991) challenge this view and present arguments against a processing ac count involving traces (or empty categories) and propose a ''direct as sociation hypothesis'' (DAH) in which a preposed phrase is directly as sociated with its subcategoriser (either verb or preposition). But the apparent validity of the arguments against the empty category analysi s stems from the fact that only a narrow range of examples was discuss ed. An examination of a wider range of structures involving filler-gap dependencies reveals that the arguments presented by Pickering and Ba rry fail to support the DAH.