Pragmatic reasoning allows people to adapt their language to better fit their communicative goals. Consider scalar implicatures, e.g. the inference that "Some of the students passed the test" means that not all of them passed. Without this pragmatic strengthening, the only way that a speaker could communicate this meaning is by using the longer and clumsier phrase, "Some but not all". The speaker in this example can be confident that the listener will draw the correct inference, because they share a simple maxim of conversation: be informative. If the speaker had known that all of the students had passed, then saying "All" would have been more informative than saying "Some" the listener can therefore conclude that not all of the students passed. This type of Gricean reasoning has recently been formalized in models of recursive social reasoning (Franke, 2009; Frank and Goodman, 2012; Jager, 2012), and used to predict quantitative judgments in pragmatic reasoning tasks.

I will discuss recent work on pragmatic inferences which require more than just the assumption of speaker informativeness. This includes a diverse set of phenomena: exaggeration, embedded implicatures, and the interpretive effects of prosodic stress. Drawing on experimental evidence and computational modeling, I will argue that each of these phenomena corresponds to a natural way of augmenting pragmatic reasoning with additional knowledge about the world or the structure of social intentions. These phenomena illustrate both the sophistication of people’s pragmatic reasoning, and how people leverage this reasoning to improve the efficiency of their language use.

If you would like to meet with the speaker, please contact Vera Demberg.