The author examines the current state of research on teaching and argues that the technique of "grounded theory," a method for discovering theory from data, would be an effective means for improving theoretical bases of teaching. The meanings of "theory" and "theory building" are discussed, including what theory is, the forms it can take, and alternative methods for its construction. The paucity of theoretical effort in current research on teaching is attributed to misconceptions of its scope (i.e., that theory must be all inclusive) and role (a product rather than a process). Glaser and Strauss are cited as the foremost proponents of grounded theory, and the basic components of their research/theory process are described. Two examples of the use of grounded theory in recent educational research are given. A concluding argument is presented: that grounded theory, in addition to being an effective descriptive tool, may also be an effective method for the generation of prescriptive theory. Use of the process in this latter role would aid in the establishment of a dialectic between educational research and educational practice. (MJB)