The social basis of referential communication: Speakers construct reference based on listeners' expected visual search.

A foundational assumption of human communication is that speakers should say as much as necessary, but no more. In referential communication, the pressure to be efficient is typically formalized as an egocentric bias where speakers aim to minimize production costs. While intuitive, this view has failed to explain why people routinely produce redundant adjectives, particularly color words, or why this phenomenon varies cross-linguistically. Here we propose an alternative view of referential efficiency, whereby speakers create referential expressions designed to facilitate the listener's visual search for the referent as they process words in real time. We present a computational model of our account, the Incremental Communicative Efficiency (ICE) model, which generates referential expressions by considering listeners' expected visual search during online language processing. Our model captures a number of known effects in the literature, including cross-linguistic differences in speakers' propensity to over-specify. Moreover, our model predicts graded acceptability judgments with quantitative accuracy, systematically outperforming an alternative, brevity-based model. Our findings suggest that reference production is best understood as driven by a cooperative goal to help the listener identify the intended referent, rather than by an egocentric effort to minimize utterance length.