Ling 848A, Fall 2025: Predictive Processing and Sentence Structure

Course Description

This readings and discussion seminar will focus on two closely related facts about language understanding. First, processing sentences fundamentally involves relating sequential input with underlying hierarchical structure; and second, there is a large body of work suggesting that sentence processing involves an interplay between top-down predictions and bottom-up evidence. The papers we cover will target interests in human sentence processing from multiple perspectives, including psycholinguistics, neurolinguistics, and computational modeling. This means that some time will be spent establishing some common ground on essential background, including for example computational perspectives on grammars, parsing and the relationship between them, as well as psycholinguistics and neurolinguistics methods and hypotheses related to sentence structure. The "Bayesian brain" hypothesis and its realization as predictive coding will also figure prominently in parts of the course.

The seminar is intended to be of interest both to people who are interested in computational models of human parsing, and people who are interested in computationally oriented discussions of prediction in language processing. It can be viewed as a descendant of my Fall 2016 seminar on computational models of human parsing, but now shifting to incorporate key ideas from my Spring 2022 seminar, Topics in Computational Linguistics and the Cognitive Neuroscience of Language, and the Fall 2024 seminar that Ellen Lau and I taught on Computing and the Brain. From the perspective of interests in human parsing, this evolution is a recognition of how much our understanding of sentence structure processing involves prediction; conversely, from the perspective of prediction, the seminar topic highlights the extent to which sentence structure, broadly construed, introduces additional challenges beyond traditional mainstays of predictive processing research like visual perception or speech perception.

This is a graduate seminar, but by design it's also going to involve people from a diversity of intellectual background. No programming will be expected. Students from all departments are welcome, but students should come in with some knowledge about computation, psycholinguistics/neurolinguistics, or both. If you're not sure, talk with me.

Logistical info

Course Times: Wednesdays 2:00pm-4:30pm

Classroom: Jimenez (JMZ) 3122 [info] (Yay, lots of windows!)

Syllabus. TBD, but you can expect it to be substantially similar to seminars I've taught before.

Schedule of Topics. TBD.

Class discussions and announcements will take place on Piazza, link TBD.