The purpose of this dissertation is to extend our knowledge of how speakers plan what they are saying to meet the needs of their audience. Three experiments were conducted to help connect our understanding of audience design with general cognitive processes of attention, memory, and learning. The first experiment tested how quickly speakers adapt to new common ground, and whether this adaptation is related to their eye gaze. It found that speakers often used speech repairs to integrate new common ground information indicating that they had difficulty adjusting their perspective. The second experiment focused on the question of whether audience design can be seen as a form of expert performance, and under what circumstances speakers are more or less likely to rely on this expertise. Speakers were observed relying heavily on memory routines established during a training phase, resulting in frequent misspecification of referents. In addition, self and other prompted speaker adaptation were considered in light of the potential for feedback by the addressee, and evidence was found that utterance planning was influenced by the interactional affordances of the situation. The final experiment examined how speakers learn information about their partners' perspective over time, and how this learning is impacted by the availability of feedback and role constancy. Successful speakers integrated their partner's privileged knowledge, mostly through direct feedback but to a lesser degree also from the experience of taking turns as the director and addressee. Taken together, these experiments demonstrate that successful audience design is best understood as a process that weaves together many strategies and mechanisms. These range from a strong reliance on established linguistic routines to strategies such as selectively attending to privileged information and the opportunistic use of feedback.