A Bidirectional Optimality-based Relevance Approach To Lexical Underspecification

It is widely observed that information transferred by many words, phrases and structures is usually not equally to their literal meaning. Such a phenomenon is called lexical underspecification. Although they are underspecified semantically, addressees are often able to acquire the meanings intended by addressers in an instant. Various approaches have been proposed to explain this phenomenon, among which Relevance theorists Sperber, Wilson and Carston's lexical pragmatic project based on relevance theory is the most influential one.However, during the process of studying relevance theory, I find that relevance theory is vague and subjective in defining some key terms such as relevance, cognitive effects, and etc. To overcome the explanatory inadequacy of RT, a more precise and maneuverable mechanism of utterance interpretation is needed. Optimality Theory's strong emphasis on formal precision, specification of violable constraints and strict ordering of violability is the source of its objectivity and strong maneuverability. Yang Zi in Fudan University proposed a unidirectional optimality-based relevance theoretic model to deal with utterance interpretation. Her model degraded the status of the speaker's intention. However, intention was recognized as a decisive factor in utterance interpretation. Blunter's bidirectional OT approach used in lexical pragmatic field is very enlightening. It dealt with lexical pragmatics form both the speaker and the interpreter's perspective. I borrow the bidirectional OT approach and Yang Zi's idea of regenerating candidates of cognitive effect, and finally construct a Bidirectional Optimality-based Relevance Theoretic Framework. The new framework works in the following steps. The activation and retrieval of contextual assumptions is determined by four hierarchically ordered constraints in GEN: MAX-Acces.CA > ACCORD-ER > High-Degr.Mm.CA > High-Stren.CA. Candidates of cognitive effect are then generated and submitted to EVAL for evaluation. The constraints in EVAL will check every cognitive effect candidate and decide whether it should be eliminated or kept as the optimal reading of the utterance. If a cognitive effect candidate satisfies the absolute constraints Stren.CE!=0, !CONTRA.Str-er.CA and !CONTRA.S'I, it will survive as the optimal reading. The relative constraint CE-I'E checks the nature of the cognitive effect candidate. Finally, the interpreter can check whether his expectation of relevance is satisfied.When this new framework is constructed, its explanatory adequacy needs to be testified. A corpus of underspecified words collected from the TV series Desperate Housewives is build. A data-driven qualitative analysis approach is adopted in this study. Through the analysis of the comprehension procedure of underspecified word meanings, their superiority can be revealed. After the sample analysis by the two models, I find that when the first most accessible contextual assumption can satisfy the interpreter's expectation of relevance and meanwhile satisfy the absolute constraints in EVAL, the relevance theoretic comprehension procedure is capable of dealing with such simple underspecified words. Otherwise, it can not produce an appropriate interpretation. On the contrary, the Bidirectional Optimality-based Relevance Theoretic Framework is able to deal with all the cases of lexical underspecification. Therefore, this Bidirectional Optimality-based Relevance Theoretic Framework is indeed more objective, precise and maneuverable than the relevance theoretic comprehension procedure.