Adapting referring expressions to the task environment

When people refer to objects linguistically, they must choose properties of the object that make it possible for the listener to identify the intended referent. We show that this selection of properties not only depends on the task environment but also changes over the course of time. We find that the salient feature color is used less often over time because of its limited utility in our task, while other features with high utility are used more often. We also find that the speaker does not change his/her behavior because of feedback from the interlocutor but because of experience gained when the roles in the task are reversed.