Process-product research in which the investigator observes in teachers' classrooms and tries to relate process measures of teaching behavior to product measures of student outcome has face validity appeal and common sense logic. This research approach appears to be the simplest and most direct way to identify teaching behaviors which discriminate successful from unsuccessful teachers. Once identified, these process behaviors can be manipulated experimentally in order to establish their causal relationship to product outcomes. All of this is straightforward in theory, but it has not worked well in practice. Until recently, process-product research in education has continually failed to establish dependable relationships between teaching process behaviors and student outcome measures (Rosenshine & Furst 1973; Dunkin & Biddle, 1974). The dismal results of process-product classroom research have led many to conclude that teaching is a complex and mystical art, not an applied science which can be objectively analyzed and measured. The authors reject this view, however, believing that the problem lies not in the inherent elusiveness of the phenomenon but in a number of identifiable and correctable weaknesses in the research approaches which have been applied to date. This paper reports the authors' research in the Texas Teacher Effectiveness Project that has been designed explicitly to test out these assumptions and to seek correlates of effective teaching using research design and data collection methods which represent improvements in several ways over previous work. [This article was originally published in "Journal of Classroom Interaction," v9 n2 1974.]