There has been no major research, comparable to research on the social and emotional aspects of teaching, which focuses on teaching behavior as related to student achievement of cognitive objectives. The descriptive research of today, even if it were to correlate teaching behavior and learning, would still share a drawback with the research on noncognitive variables: studying teaching and teachers as they are restricts the quality of teaching to that of today's best teachers. What is needed, as Stolurow proposes, are models of the teaching process which would embody computer-based facilities for programed instruction and would predict learning outcomes that can be tested and corrected. Programed instruction has the benefit of being able to cope with the complexity of teaching (a factor in the difficulty of determining what in teaching behavior influences learning) and the momentary as well as permanent differences among students. Such a new approach would permit isolation of specific cognitive aspects such as oral explanations. (A 14-item reference list is included.) (LP)