Speaker interaction is a central feature of human dialogue, one with a powerful influence on its discourse structure and performance efficiency. The present study examined two speech modalities that represent opposites on the spectrum of speaker interaction — the telephone dialogue and audiotape monologue. Experts provided instructions by either telephone or audiotape as their novice partner completed an assembly task. Within this task framework, a comprehensive analysis is provided of the basic differences in discourse organization, referential characteristics, and performance efficiency for these two spoken modalities. The outlined distinctions are interpreted with special reference to the role of confirmation feedback in promoting dialogue efficiency. Implications are discussed for the development of prospective speech systems designed to be habitable, high quality, and relatively enduring. Finally, a theoretical model of collaborative dialogue is proposed from which several features of interactive and noninteractive speech can be derived.