This paper rejects the hardware-based "levels of interaction" made popular in interactive video literature to describe human-machine interaction in favor of a new taxonomy of learner-media interaction based on the type of cognitive engagement experienced by learners. Interaction can be described on three levels, based on the quality of the interaction. A reactive interaction is a response to presented stimuli, such as an answer to a specific question. Proactive interaction emphasizes learner construction and generative activity. The learner goes beyond selecting or responding to existing structures and begins to generate unique constructions and elaborations beyond designer-imposed rules. Mutual interaction is characterized by artificial intelligence or virtual reality designs in which the learner and system are mutually adaptive, each capable of changing based on encounters with the other. Reactive, proactive, and mutual interactivity can be described at five functional levels: confirmation, pacing, navigation, inquiry, and elaboration. The transactions (mechanics of how interaction is accomplished) can also be described in terms of their functions and levels of interactivity. Although several transactions can be employed at all levels of interaction, as interaction reaches for higher levels of engagement with learners, generative transactions are required. One of the major implications this taxonomy carries for instructional design relates to learner control. As levels of interaction are ascended by the instructional developer and reflected in the design of interaction, the amount of control abdicated to the learner changes. At the reactive level, the instructional developer retains almost complete control over the content, its presentation, sequence, and level of practice. While research in the area of learner control is relatively new, some tentative advice is available from the literature. Inherent in this emerging literature is the concept of learner control, an issue which will occupy a central position in multimedia research during this decase. (Contains 27 references.) (KRN)