Neural Mechanisms of Pronoun Resolution

Abstract

To understand how pronoun resolution is implemented in the brain, one first step is to describe the algorithm that performs the task. This thesis evaluated three computational models for pronoun resolution against brain activity time-locked at every third person pronoun during naturalistic story listening. We also compared the English and Chinese populations to examine whether typological differences between English and Chinese pronouns are instantiated at the brain level. Group comparison between the activation maps for the syntax-sensitive Hobbs algorithm and the discourse-sensitive ACT-R models revealed distinct activation patterns, supporting a different weighting of information in English and Chinese pronoun resolution. Given the computational components in the Hobbs and ACT-R models, we tentatively advance a functional neuroanatomy of pronoun resolution where the left IPL is involved for maintaining multiple syntactic representations, the left MTG for morphological processing, the left Precuneus for tracking multiple referents, the left AG for integrating syntactic and semantic information and the left IFG for accessing working memory.