IRCS Conference Room
Rutgers Center for Cognitive Science (RUCCS)
Department of Computer Science
Grammar, Strategy and Implicature
Because interlocutors use language purposefully, utterance interpretaion draws both on our knowledge of language and our understanding of the speaker's choices. The broad sweep of interpretive reasoning is epitomized by Paul Grice's famous theory of conversational implicature, according to which interlocutors draw on pragmatic principles of cooperation to exploit and enrich linguistic meanings. Grice's view is not inevitable, however: another philosophical tradition, associated for example with David Lewis, sees linguistic meaning itself as an outgrowth and abstraction of collaborative language use. This talk will lay out the dialectic and sketch some empirical and philosophical arguments against the pragmatic enrichment of linguistic meaning.
We start by presenting an empirical model of lexical choice in corpora. The model assumes speakers make heuristic decisions to provide true information and otherwise match what others tend to say. The model offers a close fit to our data, but, as in Lewis's approach to signaling, its meanings originate in equilibria from which no implicatures can follow. What then explains the intuitive appeal of Gricean explanations? We use the linguistics of information structure to undercut some key intuitions by showcasing the role for linguistic rules in deriving apparently indirect interpretations. In particular, we argue that in many cases it is grammar itself that characterizes the relationship between the speaker's semantic contribution and the issue it addresses.
The talk draws on the arguments in our book Imagination and Convention (Oxford, 2015) and on results from McMahan and Stone (TACL, 2015).