In this article, Clark's theory of language use is applied to the study of conversations of augmented speakers and their addressees. Discussion focuses on how shared meaning--called common ground--is achieved, the process of grounding utterances under real-time constraints, and how the media characteristics of devices affect the grounding process. A joint action analysis of grounding will be applied to examples of word-board and Voice Output Communication Aid (VOCA)-mediated conversations to show how participants organize their talk around specific media constraints to conduct their conversations and solve communication problems. The authors argue that this model of device-mediated communication performance, or some variety of it, has the potential to reconcile many of the individual research findings in this field within a single explanatory framework. If successful, this approach could be used to evaluate future research with an empirically-based model of communication performance.