Do priming effects in dialogue reflect partner- or task-based expectations?

It is now well established that people in conversations repeat each other's words and structures. Does doing so reflect dialogue participants' expectations that their own choices of words or structures will be repeated back to them? In two experiments, subjects and confederates (purportedly) took turns describing pictures to each other. On critical trials, we measured response latencies to choose pictures when labels (e.g., stroller) or syntactic structures (a prepositional dative) that subjects had just produced were repeated back to them, versus when they heard reasonable alternatives (baby carriage or a double-object structure). Experiment 1 showed that repeated words and syntactic structures both elicit faster responses. Experiment 2 showed that the effect happens even when subjects hear descriptions from computers, instead of from their addressees, and that the repeated-word effect was not due to preferences for labels. These observations suggest that dialogue participants expect their own word and structure choices to be repeated back to them, and this is general to the task situation rather than specific to their communicative partners.