Pronoun resolution and discourse models

Abstract

Psychological investigations of pronoun resolution have implicitly assumed that the processes involved automatically provide a unique referent for every pronoun. We challenge this assumption and propose a new framework for studying pronoun resolution. Drawing on advances in discourse representation and global memory modeling, this framework suggests that automatic processes may not always identify a unique referent for a pronoun. In 9 experiments, we demonstrate that, unlike noun anaphors, pronouns sometimes do not produce relative facilitation of their referents in comparison with nonreferents. We argue that research on pronoun resolution must consider the discourse contexts in which pronouns are likely to occur.