Presupposition Accommodation in Update Semantics

The study of presuppositions lies on the border of semantics and pragmatics. It concerns information, assumed to be part of the discourse context by the speaker, as apparent in the syntactic form of the utterance. Missing presupposed information will normally be lled in by the hearer, provided it is compatible with the context. My thesis tries to characterise this process of presupposition accommodation. part I In the rst part of this thesis I introduce some of the basic approaches to presuppositions. First, I formulate some general characteristics. Then I present the ideas behind semantic theories of presupposition: presupposition as a precondition on successful interpretation. I introduce the projection problem of presuppositions and the proposed solution by Karttunen. I discuss the innuence of context and the related pragmatic theory of presupposition : presup-positions are taken to be part of the discourse context, already. I argue for an explicit notion of presupposition accommodation: adjustment of the context when presupposed information is missing. I discuss Gazdar's approach and compare it with Heim and Van der Sandt. My approach is similar to Heim's. part II Update Semantics has provided a framework in which these basic approaches can be compared. The context, roughly the information state of the hearer, can be modelled as a set of epistemic alternatives: ways the world can be according to the information. Increasing information by updates, means eliminating alternatives. Presupposition accommodation can be modelled using updates on stacks of information states. The diierent approaches from part I are reformulated as accommodation strategies. Using only a few basic types of stack-updates, the strategies can be characterised and compared in the logical framework. part III Theoretical linguistics cannot do experiments to test its hypotheses. What we can do is formalise our theories. The next step is to try and implement the theory in a computer program. I have done this for several of the update systems deened in part II. I found a grammar that translates simple example sentences into the logical language. I have chosen to use semantic tableaux to represent information states in the program. The semantic tableau deduction method can be seen as a Constructive Update System. Branches play the role of epistemic alternatives. For propositional logic this can be implemented directly into Prolog. The greatest problem has been with the interplay between negation and discourse referents in the predicate logic version of the program. Stacks of information states …