FORMAL PRAGMATICS

In the 1950s, Chomsky and his colleagues began attempts to reduce the complexity of natural language phonology and syntax to a few general principles. It wasn’t long before philosophers, notably John Searle and H. Paul Grice, started looking for ways to do the same for rational communication (Chapman 2005). In his 1967 William James Lectures, Grice presented a loose optimization system based on his maxims of conversation. The resulting papers (especially Grice 1975) strike a fruitful balance between intuitive exploration and formal development. Though the work is not particularly formal, it marks the birth of modern formal pragmatics. Pragmatics is central to the theory of linguistic meaning because, to paraphrase Levinson (2000), the encoded content of the sentences we utter is only the barest sketch of what we actually communicate with those utterances. Utterance interpretation involves complex interactions among (i) semantic content, (ii) the context of utterance, and (iii) general pragmatic pressures (of which Grice’s maxims are one conception). The starting point for a formal pragmatics is the observation that speakers agree to a remarkable extent on the interpretations of the utterances they hear, suggesting that there are deep regularities across speakers, utterance contexts, and sentence types in how (i)–(iii) interact. An overarching challenge for pragmatic theory is that semantic content and the context of utterance influence each other. It is common, for instance, to find that the meaning of a sentence is crucially incomplete without contextual information. Indexicals and demonstratives are paradigm cases: ‘I am here now’ doesn’t have a fully specified denotation without information about who the speaker is, when he is speaking, and where he is speaking. Similarly, modal auxiliaries like must admit of a wide range of interpretations. The utterance ‘Sam must be in his office’ can be used to make a claim based on evidence (‘. . . I see the light on’), or a claim based on the laws of the land (‘. . . the boss has passed a new rule’), or any number of others. Sentenceinternal features provide some clues, but the intended interpretation cannot generally be resolved without information from the utterance context. Other examples of this form are easy to find. Context dependency is ubiquitous in language, which means that essentially all semantic theories rest on particular views of how to model contexts. Kaplan’s (1989) theory of indexicality is an early and influential approach to modelling how the context influences interpretation. For Kaplan, the context is modelled as a tuple of indices that identify the speaker, the time, the place, and so forth. These indices directly provide the meanings for indexical expressions like me, now and here. The methods are those of semantics (model-theoretic interpretation), but some interpretation happens in terms of these designated context tuples. Kaplan’s theoretical approach helps us to see how context helps determine semantic content. Its basic techniques have been used for a wide range of issues in context dependency. However, these theories have little to say about the reverse direction of influence, i.e. how the context is changed by the addition of new content. Dynamic theories of meaning attempt to model this aspect of context dependency as well. The earliest such systems in linguistics are those of Heim (1982) and Kamp (1981), who built on the insights of Karttunen (1976). The fundamental innovation of dynamic semantics is to ask, not what sentences mean, but rather how they affect, and are affected by, the flow of discourse information. For example, if I say ‘A goat entered’ with the force of an assertion then, with this utterance, I introduce a new discourse referent d into our context (this is the work of the indefinite), and I ascribe two properties to d: that of being a goat and that of entering. My language has thus changed the context. If I follow up with ‘It looks hungry’, the subject pronoun it gets it meaning from the recently introduced d. Dynamic theories provide systematic explanations for this language–context interplay. Historically, such theories have been developed to handle presupposition and discourse-mediated anaphora (Chierchia 1995; Beaver 2001), but they can be extended to many kinds of discourse information (Asher and Lascarides 2003).

Christopher Potts