The role of implicit arguments in sentence processing

The goal of this dissertation is to develop a framework for characterizing routine inferencing during immediate sentence comprehension. Whether our understanding of sentences includes unexpressed information has been the subject of considerable debate and at least two radically different views have been proposed. The "constructivist" view suggests that the representations we form in understanding a sentence are routinely enriched with specific information from stored conceptual knowledge. Support for this position has come largely from off-line paradigms that interrogate subjects' memory for passages that they have encountered. Based on the results of studies employing more on-line methodologies, the "minimalist" view posits that our immediately-formed representations of sentences contain little information from stored knowledge and are instead very close to the input we receive. The framework developed here captures the constructivist intuition that comprehension is typically quite rich and yet in compatible with the kinds of representations that minimalist experiments have suggested that we form. The lead idea is that, in addition to consulting the explicitly expressed content of a sentence and stored conceptual knowledge, comprehenders consult an additional source of information in forming their representations of sentences. This additional source of information is derived from the linguistic forms of sentences, more specifically from the schematic semantic information that characterizes what the linguistically necessary participants of an event denoted by the verb are. This schematic information corresponds to the thematic roles that linguists have argued are associated with the explicitly expressed arguments of a verb. The hypothesis pursued is that this information need not be associated with an explicitly expressed argument to be interpreted. Since this information is intrinsic to the meanings of verbs, it is natural that it should be included in our representations of sentences regardless of whether it is specified by lexical content. The possibility that our representations of sentences may routinely include implicit arguments is compatible with both the constructivist intuition that our understanding of sentences typically includes more information than that which is explicitly expressed and the minimalist findings that suggest that we do not infer specific information from conceptual memory during immediate comprehension.