Perspective Taking in Definite Reference Resolution

This thesis explores the roles of common ground and Theory of Mind in processes for definite reference resolution. It examines the proposals due to Herb Clark and colleagues that common ground is common-knowledge of discourse relevant facts and that it serves as the main constraint on definite referential processes. The thesis contrasts that view with an alternative where common ground emerges from shared interactions as a result of low-level, or automatic memory mechanisms that are constrained by what has been in shared attention. The thesis presents two experiments that verify the effect of memory-based mechanism on referential processes, and also shows that memory-based mechanisms function only where the cue to previous shared experience is itself in shared attention. Given that common ground only weakly constrains definite referential processes, it is argued that Theory of Mind inferences play a critical role in on-line referential processes where interlocutors have different perspective information. The thesis argues that research using the perspective-taking task did not distinguish between the role of common ground and inferences about the speaker’s ignorance of certain objects. Experiment 3 shows that participants have better perspective taking when the procedure highlighted the speaker’s ignorance. This effect is seen within 300ms of the on-set of the critical linguistic stimulus, demonstrating very early integration of Theory of Mind inferences. Then a follow-up experiment excludes the possibility of the effect of simple interactivity in the experiment 3, and also explores whether the co-presence of the speaker increased the degree to which the existence of private objects was shared attention. Because the existence of privately viewable objects are common ground, even though their specific identities are not. The results of experiment 4 support the proposals about common ground in this thesis by showing that participants have better performance when the speaker is not co-present