Using Constraints and Reference in Task-oriented Dialogue.

This paper presents an analysis of linguistic data that stems from a task-oriented dialogue. We demonstrate that certain referring expressions used in this setting are potentially ambiguous and indeterminate but do not lead to referential errors such as under– or over–population of the discourse representation. Three studies are reported which show that it is the skilled use of non-linguistic constraints, present in the task which facilitates successful reference. When these non-linguistic constraints are removed, skilled speakers are able to make a compensatory adjustment in the precision of their referring expressions. The ability of people to make this compensatory adjustment is an aspect of referential skill not revealed by more traditional task–oriented assessments.