Departmental Colloquium: Top-down prediction in language comprehension: a handful of facts and a mystery - Jesse Snedeker (Harvard)

Over the last thirty years, psycholinguists have converged on a model of language comprehension in which perceptual input (e.g., sounds) incrementally activates higher level representations (e.g., possible messages) resulting in predictions of upcoming information at the lower levels (e.g., words or even sounds). I’ll report on some studies that probe the development, universality and limits of this system by looking at prediction in naturalistic tasks with diverse populations (children, signers, and bilinguals).  I’ll introduce a new method of collecting EEG data. Along the way I’ll try to integrate the developmental studies with my earlier work using the visual world paradigm.  But I’ll end with as many questions as answers.