The Dynamics of Discourse Situations

The participants in a conversation share a lot more than information about the propositional content of statements. Among other things, they share information about each other’s intentions or about whose turn it is to speak; they also spend a lot of their time synchronizing their views of the common ground [Clark, 1996]. Utterances update this ‘pragmatic’ aspects of the common ground as well as the semantic ones more typically studied in the literature on dynamic logics. The best known formal models of the common ground tend to concentrate on one aspect of shared information only: these models focus either on the semantic update brought about by statements [Kamp and Reyle, 1993, Groenendijk and Stokhof, 1991], or on the impact of different types of speech acts on the conversants’ mental states [Cohen et al., 1990]. But there are at least two reasons for developing a unified formal theory of both the semantic and the pragmatic aspects of the common ground. One reason is that such a theory is needed by anyone trying to develop theories of the interpretationof, say, definite descriptions, since shared information of both a semantic and pragmatic nature is brought to bear in such interpretation processes (semantic information about which discourse referents are accessible; pragmatic information about the current attentional state and how the current utterance relates to the intentional structure of the conversation) [?]. A second reason is that many utterances in a conversation contribute little or nothing to the semantic content; their role is to signal that one conversant is acknowledging the other conversant’s contribution, or taking the turn, etc. Examples include utterances such as right, okay,mmm, and so forth, which constitute a very significant part of any spoken conversation. The meaning of such utterances could not be specifiedwithin a purely semantic theory of the common ground. The goal of this paper is minimal. We propose to show that certain pragmatic aspects of the common ground–specifically, the shared information about the DISCOURSE SITUATION [Barwise and Perry, 1983], i.e., which speech acts have been performed in a conversation and their hierarchical organization–can be captured using technical tools that are essentially those of Discourse Representation Theory (DRT) [Kamp and Reyle, 1993], whose ‘dynamic’ properties are also preserved. The theory of the common ground we propose gives us the tools to specify both semantic and (some) pragmatic aspects of the common ground, and can therefore be used both to capture the meaning of utterances like right and to formulate theories of interpretation. We have shown elsewhere that thisminimal account can be extended to capture additional pragmatic aspects of the common ground (Anonymous Authors, 1997). Our theory of the common ground can be sketchily characterized as follows. Our first hypothesis is that the discourse situation can be characterized by a DRS, recording (shared information about the) CONVERSATIONAL EVENTS (= speech acts) that took place. Our second hypothesis is that as speech acts are events, this information can be captured using the tools for representing events introduced in [Kamp and Reyle, 1993]. The discourse situation resulting from the dialog in (1) is thus characterized as in (2).