Where is pragmatics in optimality theory

This paper deals with the architectural issues of pragmatics within an overall account of natural language in optimality theory. It is argued that pragmatics can be seen as an optimisation problem described by its own constraint system which lies outside the constraint system that defines grammar (the production oriented OT models of syntax and phonology). Speaking and hearing both involve grammar and pragmatics, but in different ways. The paper argues against the popular view that grammar and interpretation should be mixed into a symmetric constraint system and connects the proposed architecture with the views that underlie the motor theory of understanding and the mirror neuron theory of understanding behaviour. 1 The Meaning of Production Optimality theory [Prince and Smolensky, 1993] can be seen as a modern version of Jakobson’s markedness theory. In the very concept of an optimisation problem, there is a concept of blocking: some regularity is broken because in the particular case there is a better solution. Gooses is ruled out by the ”better” geese and all the theorist has to do is to explain why geese is better. These explanations take the form of a system of constraints, a set of demands on outputs relative to a given input that are linearly ordered. The regularity will always exist, but stronger constraints prevent it from emerging in the particular case. Though it is not particularly hard to come up with explanations of this kind in phonology and syntax, this is not the business of this paper. For successful treatments of phonology see [Prince and Smolensky, 1993] and most of the Rutgers Optimality Archive, for syntax an interesting collection is [Dekkers et al., 2000]. The starting point of this paper is the assumption —really an assumption since many issues remain unresolved— that comprehensive treatments in optimality theory of phonology, the lexicon and syntax are possible. That successful treatment would allow mapping any meaning to its optimal pronunciation by a function F . The inverse of F would be the interpretation function F−1 and would deal with semantics and pragmatics. Linguistics would be finished! While this would be nice, there are some problems.