In dialogue, a conversational partner's behavior is not solely the product of one individual mind, but instead reflects a process of mutual coordination between both conversational partners. This project investigates how speakers' and addressees' behavior is shaped by addressees' informational needs and by speakers' prior expectations of addressees' informational needs. In two experimental settings pairs of speakers and addressees were observed while giving directions (Experiment 1), or narrating short stories (Experiments 2a & 2b). Addressees had either high informational needs (the information they were receiving was new to them) or low informational needs (they were already familiar with the information). Speakers' expectations of addressees' informational needs were informed through prior experiences with the addressee (Experiment 1), or explicit information about addressees' knowledge made available prior to the interaction (Experiments 2a & 2b). Results show that speakers' behavior was shaped by addressees' behavior: Corresponding with how much feedback they received, speakers shifted how clearly and deliberately they articulated target referring expressions, how many details they provided, whether they introduced salient information as new or old, and where they positioned this information syntactically. This suggests that addressee feedback is one important cue through which speakers monitor addressees' needs, and supports the assumption that information about the conversational partner can influence different levels of linguistic processing. As for addressees, their behavior was shaped not only by their informational needs, but also by speakers' expectations of addressees' needs. This suggests that speakers' expectations modulate addressees' expression of their actual needs, perhaps by feeding back into the opportunities addressees have to give feedback, or by contributing to an implicit agreement on how to accomplish the task at hand. These findings have theoretical implications for understanding dialogue as a collaborative process in which conversational partners mutually shape each other, integrating bottom-up information available online in the conversational situation with top-down expectations that are brought into the conversation. There are also methodological implications for how to go about studying dialogue, in particular with respect to the practice of replacing na ve addressees with experimental confederates who typically have very limited informational needs, thereby changing the nature of the interaction and the behavior under study.