Statistics / Linguistics Seminar - Anna Laurinavichyute

Linguistics-Statistics Seminar

Event Date

Social Sciences & Humanities 2203

Speaker: Anna Laurinavichyute, University of Potsdam

Title: "A Bayesian computational model of number agreement in sentence processing"

Abstract:  An important open question in sentence processing research is how number agreement is computed. Studies on reading repeatedly show that the ungrammatical sentence "The key to the cabinets are rusty" tends to be perceived as grammatical (compare the equally ungrammatical: "The key to the cabinet are rusty"). Several theories have been proposed to explain the illusion of grammaticality in such sentences. I present several experiments that probe the predictions of these theories and show that none of the existing accounts can adequately explain the full range of patterns in both my data and the broader literature on the topic. Experimental evidence suggests that task demands may play a critical role in affecting the observed reading time patterns. To address this, I present a hierarchical Bayesian multinomial processing tree model that posits several latent processes leading to the observed patterns and discuss its implications for theories of agreement computation in reading.


Speaker's website (external link):

Seminar Date/Time: Friday March 8th, 10:00am, at SSH 2203