In this article, we examine the hypothesis that acoustic variation (e.g., reduced vs. prominent forms) results from audience design. Bard et al. (Journal of Memory and Language 42:1-22, 2000) have argued that acoustic prominence is unaffected by the speaker's estimate of addressee knowledge, using paradigms that contrast speaker and addressee knowledge. This question was tested in a novel paradigm, focusing on the effects of addressees' feedback about their understanding of the speaker's intended message. Speakers gave instructions to addressees about where to place objects (e.g., the teapot goes on red). The addressee either anticipated the object, by picking it up before the instruction, or waited for the instruction. For anticipating addressees, speakers began speaking more quickly and pronounced the word the with shorter duration, demonstrating effects of audience design. However, no effects appeared on the head noun (e.g., teapot), as measured by duration, amplitude, and perceived intelligibility. These results are consistent with a mechanism in which evidence about addressee understanding facilitates production processes, as opposed to triggering particular acoustic forms.