Interactive aspects of reference assignment in conversations

We propose a model of reference that contrasts with standard linguistic approaches in that it focuses on the role of interaction in reference, arguing that referring expressions in conversations are not designed for interchangeable audiences but rather exploit the common ground between partners. Our model also differs from psycholinguistic approaches in that it uses conversational data, since critical aspects of natural conversations are absent from laboratory tasks used so far and thus are not captured by current models. Examples from conversations are presented to demonstrate that speakers often try to manage the accessibility of a problematic referent to an addressee by presenting a context for it as well as a referring expression, even at the cost of syntactic orthodoxy. We also present examples demonstrating that partners negotiate as to what representation is good enough for present purposes and whether that has been achieved. While strategies may vary as to explicitness, we believe these negotiations underlie all formulations of referring expressions.