INTRODUCTION TO DISCOURSE REPRESENTATION

Hans Kamp

The first paper of this part, A Theory of Truth and Semantic Representation (TTSR), is probably the most widely known of those in this part, and per- haps, at this point in time, of all the papers in this collection. It appears to be the document most often cited as one of the two starting points of Dynamic , as a discipline with an explicit compositional seman- tics and a well-defined logic, rather than as a mere collection of informal intuitions about how discourse works and a list of promissory notes for making those into a theory. The other document standardly mentioned in this connection is ’s The Semantics of Definite and Indefinite Noun Phrases. Often, in fact, her dissertation and TTSR have been cited in tandem; rightly so, it might be added, given that where the application domains of the two approaches overlap primarily in what they have to say about donkey , they make the same predictions about mean- ing and semantic interpretability. As with the application of Supervalua- tion Theory to the analysis of vagueness, where Kit Fine and I indepen- dently arrived at the conviction that this was an approach worth developing, Heim and I also arrived at our formulations of these mechanisms inde- pendently. But her dissertation was the far more substantial work, which, for one thing, covers definite descriptions as well as pronouns. It rightly has become the yardstick for subsequent work on the topics it addressed. It deserves emphasis in this context, however, that TTSR wasn’t meant to be just a theory of the behavior of pronouns and certain other noun- phrases. As indicated in the introduction to the first part, the paper was rather the outcome of a process that started with observations about the discourse-linking properties of tenses. I was groping for evidence at that point that this was not an isolated phenomenon. For, as Robert Stalnaker had made plain as early as the early seventies, the contributions made by a new or utterance in an ongoing discourse or text typically take advantage of information that has already been provided by earlier parts of the text or discourse. Since linguistic communication seems to rely on this to such a very large extent, it would have been remarkable if languages had not developed grammaticised mechanisms for exploit- ing this possibility of providing information cumulatively, or if the tenses 324 in languages like French and English had been the only instances of such grammaticisation. In this light the discovery that some of the main facts about donkey pro- nouns, essentially as they had been laid out by Peter Geach in Reference and Generality, could be accounted for by assuming discourse-related mecha- nisms closely reminiscent of what I thought I had observed for the tenses of French, seemed just the confirmation I was looking for. At the same time the crisp judgments about the acceptability and the truth conditions of don- key sentences—Geach’s actual target, to be distinguished from the donkey discourses that TTSR included as part of the data—made it possible to turn this small part of the discourse-oriented enterprise into a theory that could be used to make explicit predictions about robust linguistic observations. Indeed, only when the formal framework that emerged from this attempt to deal with these particular data was in place, was the road free to deal in simi- larly formally precise ways with the sentence-transcending semantic effects of tenses and other parts of temporal vocabulary. This is also the reason why difficulties with some of the predictions made by TTSR and The Semantics ofDefiniteandIndefiniteNounPhrases did not seem as troubling to the gen- eral DRT enterprise, as they may have seemed to Heim. In particular, there are the recalcitrant facts about donkey discourses like A woman came into the lobby. She went up to the reception desk and got herself a double room with a view of the river. or A fat man was pushing his bicycle. He seemed at the end of his tether. There is a strong intuition that such discourses are inappropri- ate, unless the speaker is talking about some particular woman or fat man she could identify upon demand. Of course, these data, and others that have to do with donkey sentences rather than with donkey discourses, have to be taken seriously by anybody who wants to use DRT as a framework. But once one is convinced that dynamics is an important aspect of linguistic mean- ing and interpretation and that it would have been a miracle if things had been otherwise, the natural response to these counterexamples is not to go and look for an alternative framework. Instead one will be inclined to see such data as that this part of the theory isn’t quite right, so it needs rethink- ing and adjusting. If there are other parts of the general approach that do not seem directly affected by the difficulties observed, then concluding that the approach was wrong after all in toto, that it should scuttled and that we ought to go back to a well-tried, well-established form of sentence semantics such as , seems overly alarmist and overly defeatist. It is now more than thirty years after TTSR appeared and in the course of that time the scientific landscape has changed considerably. , in a broad sense of the term, according to which it includes DRT