Brian Dillon / Four observations about dependency formation, and a proposal
Brian Dillon / Four observations about dependency formation, and a proposal
March 29, alum Brian Dillon returns to Maryland for the Linguistic Colloquium, to present his work on how the grammar of a given language influences how we encode linguistic input in working memory.
How do we form syntactic dependencies in comprehension? Despite an enormous amount of work on this topic, the answer to this question remains obscure. In this talk, I will advance four observations about the cognitive mechanisms that subserve dependency formation: One, the context of encoding matters; Two, relational constraints matter; Three, interference is not always cue-driven; Four, interference often results in illusory item-feature conjunctions. I will suggest that none of these four observations are well captured by current models of dependency formation. I will describe a new proposal that aims to capture these observations, and discuss successes and challenges for this approach.