Abstract

A well-developed and commonly held view of the interaction between semantics and pragmatics is the Gricean view that suggests that pragmatics is limited to operating on the output of the semantic component. This dissertation considers the impact that a certain class of "intrusion" examples has on a broadly Gricean approach. Briefly, the problem is that such examples appear to require precisely what the Gricean picture excludes: semantic operations applying to the output of pragmatic inference. ;Chapter one talks about semantics and pragmatics, and examines one general method that has been used to pull extra content into the semantics within a Gricean approach--the use of contextual slots, as commonly used in the treatment of indexicals and modals. Chapter two explores whether this solution is applicable to cases of implicature and suggests that it is unsatisfactory even for the treatment of scalar implicatures since it does not correctly explain their interaction with negation or the intuitively related implicatures that occur with many types of non-quantitative terms. Chapter three discusses whether Horn's notion of metalinguistic operation could be used to solve the intrusion problem, but suggests that it cannot be for a variety of reasons, most importantly because it would have to posit a descriptive/metalinguistic ambiguity in almost every predicate. Chapter four examines the solution to the problem of pragmatic intrusion provided by Relevance theory, but suggests that it relies on an unmotivated and unworkable notion of functional independence and fails to cover certain revealing cases of pragmatic intrusion. Chapter five proposes that modifying the Gricean framework to provide greater interleaving--so that the crucial unit for pragmatic processing is a saturated predicate rather than a whole sentence--could provide a solution that handles all of the intrusion examples satisfactorily