O

The area of pragmatics is characterized by defaults and preferences. This is exactly the reason to be interested in the application of optimality theory (OT) to pragmatics. Interpretation is clearly an optimization problem: what is the best interpretation of an utterance in a context? If pragmatics were formalizable using optimality theory, it would mean that a system of ranked constraints could deal with the optimization problem, capturing exactly the preferences and defaults that human language users use when interpreting utterances. It would also lead to an integrated pragmatics, in which the waste-basket structure of the field is overcome and in which a single account given by a small number of principles predicts conventional and conversational implicature, explicature, presupposition, disambiguation, pronoun resolution, rhetorical structure, politeness marking and other aspects of interpretation. At the same time, there is a sense in which

there is no place for semantics and pragmatics in a complete optimality theoretic account of language. Production OT is the original OT, developed for phonology as an account of what is the appropriate surface phonological structure for a given abstract phonological representation. And the focus of the optimization problem on what is the appropriate pronunciation for a given abstractly specified linguistic structure is no accident – that is where natural generalizations are achievable, not in the other direction. The same model has been followed in optimality theoretic syntax where the aim has

been to assign a surface structure to some characterization of the semantics (at various depths of abstraction), because this is where natural generalizations are possible. If production OT is the right account, the theory of interpretation can be read off from the relevant components of production: themeaning is given by the inverse of the relation between input and output characterized by the phonological and syntactical constraint systems. The first point to be made here is that this is

not a notion of meaning that can be ignored. It is necessary for any pragmatics to be able to account for unusual modes of expression such as ‘How late is it?’ for ‘What time is it?’ or ‘cause to die’ for ‘kill’, given that these can be blocked by other forms or receive an extra meaning in the context. OT production syntax excels at blocking of forms and at neutralization: giving the same pronunciation to different words or at assigning the same form to two different meanings. Yet, it is not difficult to show that it is not

enough. To see this, consider a central principle of OT pragmatics known as DOAP (‘Do not overlook anaphoric possibilities’) (Hendriks and de Hoop 2001; Williams 1997), or as *ACCOMMODATE (Blutner 2000) or as *NEW (Zeevat 2001). This principle militates against interpretationsthat do not identify enough material.