Abstract

The study of conversation is notoriously difficult, because naturally occurring conversations take place against the rich background of virtually every aspect of human existence. A theory of conversations will therefore inevitably have its shortcomings. For instance, in this book the focus is entirely on information exchange in conversations. No attention is paid to the atmosphere in which a conversation is conducted (e.g., hostile or friendly), ethical considerations (honesty, truthfulness), etc.

There is still an abundant background that needs to be taken into account, even if we narrow down our attention to information exchange in conversations. Utterances depend on the context for their interpretation and also change the context. In this respect, conversations are very similar to common board games such as chess. Utterances bear a likeness to the moves in such a game; a move in chess changes the context (i.e., the distribution of the chess pieces on the board) and is at the same time dependent on the context for its interpretation. A move of a particular piece from, say, square A1 to A3, will in some contexts be interpreted as beating another piece, in other contexts as checkmate, etc.

We have exploited the aforementioned resemblance by modelling conversation as a conversational game. In order to specify a conversational game, we needed to say what its contexts, i.e., conversational states, are and what the rules are that govern the interpretation of moves and the update of the conversational states.

We assume that an important principle governing conversational games is the quest for a consistent conversational state. This hypothesis legitimizes the use of a formal system of logic to model conversational states. Consistency is one of the basic notions of logic. The particular formal system that we chose to employ is Pure Type Systems (PTS).

PTS was chosen for several reasons. Most important are the following three. Firstly, it can be seen as a higher-order generalization of Discourse Representation Theory (DRT). DRT has proven to be very succesful for modelling anaphora. Anaphora play a key role in information exchange by means of natural language. Secondly, a PTS allows a formalization of conversational states that can be processed by a computer. Thirdly, in PTS proofs for propositions are explicitly represented. These proofs can employed for consistency maintenance of contexts. In particular, they can be used to locate the information that gave rise to an inconsistency.

PTS is introduced in part I. We exploit the similarities between DRT and PTS to explain the basic notions of PTS. Subsequently, we extend PTS with a framework for reasoning about normal objects. In particular, the extension allows for derivation of a conclusion about an object on the basis of the assumption that the object is a normal object of its kind. We demonstrate that the extension covers several forms of default reasoning. The underlying motivation for our attention to this subject is the fact that in daily life jumping to conclusions on the basis of the assumption that everything is normal is the rule rather than the exception.

In part II, the formal tools that have been introduced in part I are applied to three different areas in the study of conversation: presupposition, answerhood and accent. We demonstrate that these apparently different phenomena can be modelled along similar lines. The notion of an informational gap plays a key role. 

The chapter on presuppositions provides a deductive framework for modelling the influence of world knowledge on presuppositions. We address the projection of presuppositions in conditionals and disjunctions, and supply an account of the so-called bridging examples. We report on a small observational study which supports our hypothesis that the quest for consistency might sometimes provide a good explanation for phenomena in the realm of presupposition resolution that have been linked to such intractable notions as `relevance'.

Our account of answerhood provides a proof-theoretic characterization of a variety of answers (indirect answers, negative answers, preventive answers, etc.). We show that our approach brings together the theories that start from the idea that answerhood should be explicated in terms of the possibility to unify a question and its answer and those theories that deal with the dynamic and contextual side of answerhood.

In the chapter on accents, we provide an account of accents which covers both contrastive accent (e.g., `The children were taken to the circus. The small children enjoyed it' implicates that it is not the case that the `non-small' children enjoyed it and newness accent. 

In part III, we examine longer stretches of conversation. The aim is to explain the patterns that such stretches exhibit in terms of an underlying model of changes to the conversational state. We propose some simple rules for updating the conversational state in the course of a conversation. We show that our system can be used to model a range of conversational structures that have been reported by conversation and discourse analyst. In particular, our account provides an incremental account of subdialogues, as opposed to the mainly non-incremental accounts from the literature and deals with the satisfaction of presuppositions in conversations.

Finally, the problem of conveyed meanings is addressed. The conveyed meaning of an utterance is the information that is communicated by means of the utterance without being literally said.  We defend the distinction between literal and conveyed meaning and account for the relation between conveyed meaning, in particular the conveyed meaning which is associated with so-called indirect speech acts, and literal meaning in terms of a system of defeasible rules of thumb. We show that a simple system of such rules can account for a wide range of conveyed meanings.

In short, in this dissertation a number of problems concerning the meaning and use of utterances in conversations are addressed in detail. We show that the proposals that we put forward improve on the current state of the art in the literature. At the same time, we have tried to reveal to some extent the unity which underlies the different topics that are addressed.