Event Detail

Abstract: Normally, uttering a sentence changes the state of a conversation. Suppose we thought a natural language as determining a state transition system, where each state corresponds to a state of a possible conversation, and each sentence corresponds to a label inducing a state transition. Then we can ask: what properties are characteristic of the state transition systems appropriate to modeling conversation in natural language (or various fragments thereof)? A leading idea in the philosophy of language, and in linguistic semantics and pragmatics, is that the central effect of uttering a declarative sentence in conversation is to add a proposition to the common background assumptions of the interlocutors. We give a representation theorem indicating what formal properties are characteristic of this effect from a state transition system perspective. We then go on to discuss some fragments of language that appear to lack these properties. I connect the results to the debate about how ‘dynamic’ natural language is, and to debates about what a compositional theory of meaning should look like. (This work is joint with Daniel Rothschild, All Souls College, Oxford.)