Processing for relevance : a pragmatically based account of how we process natural language
This thesis presents an account of some of the mental mechanisms and processes that take the addressee from a linguistic input to the interpretation of that input. Because on-line interpretation involves our knowledge of language, the relation between input processing and grammar is evaluated. The full interpretation of a linguistic input also involves pragmatic, i.e. central cognitive processes, but these processes are the least well understood within psycholinguistics. Relevance theory (Sperber & Wilson, 1986) gives us a way of making our understanding of these processes more explicit. However, Relevance theory claims turn out to be incompatible with psycholinguistic models which postulate an autonomous syntactic parser, such as the 'Garden-path' model. A review of the experimental literature reveals that the findings claimed to support the 'Garden-path' model do not in fact support it. Likewise, the principle of Lexical Preference, proposed to account for how verb subcategorization frames are accessed, turns out not to be supported by the experimental evidence. Full interpretation involves computing a conceptual representation, and an account is given of what constitutes conceptual structure. This leads to the proposal that verbs are represented as structured concepts. This view of verb representation together with Relevance theory can account for when arguments of verbs can be left implicit. Finally, an account is given of how the addressee computes the propositional form communicated by an utterance, by building hypotheses about the conceptual structure of the proposition on-line. These hypotheses are based on structural information stored under the concepts referred to by the utterance. This proposal can account for psycholinguistic research findings, with pragmatics playing an integral role in the explanations: it is no longer grafted onto the model as a psycholinguistic afterthought.