Saved in:
Bibliographic Details
Main Authors: Paape, Dario, Linzen, Tal, Vasishth, Shravan
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.04489
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911421793566720
author Paape, Dario
Linzen, Tal
Vasishth, Shravan
author_facet Paape, Dario
Linzen, Tal
Vasishth, Shravan
contents Using temporarily ambiguous garden-path sentences ("While the team trained the striker wondered ...") as a test case, we present a latent-process mixture model of human reading behavior across four different reading paradigms (eye tracking, uni- and bidirectional self-paced reading, Maze). The model distinguishes between garden-path probability, garden-path cost, and reanalysis cost, and yields more realistic processing cost estimates by taking into account trials with inattentive reading. We show that the model is able to reproduce empirical patterns with regard to rereading behavior, comprehension question responses, and grammaticality judgments. Cross-validation reveals that the mixture model also has better predictive fit to human reading patterns and end-of-trial task data than a mixture-free model based on GPT-2-derived surprisal values. We discuss implications for future work.
format Preprint
id arxiv_https___arxiv_org_abs_2602_04489
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Deconstructing sentence disambiguation by joint latent modeling of reading paradigms: LLM surprisal is not enough
Paape, Dario
Linzen, Tal
Vasishth, Shravan
Computation and Language
Using temporarily ambiguous garden-path sentences ("While the team trained the striker wondered ...") as a test case, we present a latent-process mixture model of human reading behavior across four different reading paradigms (eye tracking, uni- and bidirectional self-paced reading, Maze). The model distinguishes between garden-path probability, garden-path cost, and reanalysis cost, and yields more realistic processing cost estimates by taking into account trials with inattentive reading. We show that the model is able to reproduce empirical patterns with regard to rereading behavior, comprehension question responses, and grammaticality judgments. Cross-validation reveals that the mixture model also has better predictive fit to human reading patterns and end-of-trial task data than a mixture-free model based on GPT-2-derived surprisal values. We discuss implications for future work.
title Deconstructing sentence disambiguation by joint latent modeling of reading paradigms: LLM surprisal is not enough
topic Computation and Language
url https://arxiv.org/abs/2602.04489