When do speakers take into account common ground?

What role does common ground play in the production of utterances? We outline and test two models. One model assumes that common ground is involved in initial utterance planning, while the other model assumes that it only plays a role in monitoring. To compare these models, we focus on common ground as evidenced in physical co-presence. We had speakers describe objects for listeners in a modified version of the referential communication task. While descriptions under no time constraints appeared to incorporate common ground with the listener, common ground was not used when the speakers were under time pressure. These results suggest that speakers do not engage in audience design in the initial planning of utterances; instead, they monitor those plans for violations of common ground.