Pooja Paul (Harvard Linguistics)

Date: 

Thursday, February 23, 2017, 12:00pm to 1:30pm

Location: 

WJH 765

It's not just what you said, but why (I think) you said it: Psychological Constraints on Linguistic Meaning

The compositionality of language (the assumed tight correspondence between the meaning of a complex expression and the internal structure and meanings of its constituents) has been a guiding principle in contemporary linguistic theory, and is indeed a shared feature of generative models elsewhere in the cognitive sciences. While observations of context-sensitivity in linguistic meaning have at times been argued to be antithetical to compositionality, this tension has been largely reconciled within the generative framework by relegating context-sensitive effects to post-compositional, pragmatic processes (e.g. treatments of scalar implicature). However, the temporal delineation between semantic (=compositional) and contextual processes in the construction of meaning implied by the classical model is increasingly at odds with evidence for predictive coding/top-down effects in human cognition, for example in vision and memory.

In this talk, I present a line of research showing that adults and children are indeed able to recruit both high-level semantic representations and extra-linguistic information online during discourse comprehension to predict upcoming linguistic material (as measured by anticipatory eye-movements). Our findings in support of predictive processing in language lead us to revisit prevailing assumptions about the nature of the interface between the linguistic system and closely linked conceptual levels. I also present some ongoing work exploring the role of implicit scale-based representations in guiding people’s naive intuitions in a range of (ostensibly linguistic) reasoning tasks. I end with preliminary findings from a study with pre-linguistic infants that seems to support such a link.