Current theories of communication yield predictions about the expression choice of overhearers as well as primary discourse participants. We discuss three such theories and evaluate them with reference to new data on object naming elicited through a confederate priming paradigm. Our results show that participants adopt primed referring expressions if they are highly involved in the task, but mere exposure to the object labels yields very limited priming effects. Also, common ground is a relatively marginal factor in expression choice here. We interpret these results as supportive of the importance of grounding and challenging for interactive alignment-based accounts of expression choice.