| Creating a model of language comprehension requires understanding how representations from a variety of subsystems are accessed, constructed, and coordinated in real time. One type of interpretation that is pervasive during comprehension is that of definite reference, which requires a restricted domain of interpretation. While the language-as-action tradition considers the domain of interpretation to be defined by the mutual knowledge or common ground between interlocutors, little work has addressed the time course with which common ground affects reference resolution. Fine-grained time course analyses abound in the language-as-product tradition, but this literature has largely ignored complex contextual factors such as common ground.; The work presented here combines the action and product traditions in four head-mounted eyetracking, referential communication experiments in order to: (1) determine what constitutes the appropriate context or domain of interpretation for definite reference; and (2) determine the time course with which this domain has its restrictive effects. The experiments explored three types of constraints on definite reference resolution: the form of an anaphoric referring expression, the common ground between conversational participants, and the perspective of a conversational partner.; Contrary to accounts that postulate a delay in the restrictive effects of common ground, Experiments 1 and 2 found that while common ground was less effective than a pronoun at restricting an addressee's domain of interpretation, it did have immediate probabilistic effects. Experiments 3 and 4 made use of situations where the perspectives of the addressee and speaker did not match, and found that addressees were able to quickly take into account a speaker's differing perspective during reference resolution, showing little processing difficulty with expressions that were ambiguous from the addressee's perspective but unambiguous from the speaker's perspective.; These results taken together provide support for accounts in which the common ground between conversational participants, including the ability to take into account the perspective of the speaker, is a source of information that is integrated with other constraints during the earliest moments of processing. |
