What little we know about relationships between specific classroom behavior of teachers and relevant pupil outcomes has been obtained almost entirely from coreelational studies. Yet if we are ever to apply knowledge in this area to teacher education, we must carry out experimental studies in which teachers are trained to emit specific behaviors that are found to be related to pupil outcomes. In order to evaluate how specific changes in teacher behavior can bring about changes in related pupil outcomes, Utah State University created three sets of protocol modules employed as experimental treatments. Through these studies, it was determined that relationships between specific teacher behaviors and pupil outcomes tend to be higher in correlational studies than in experimental studies. Four variables seem to be the cause of this difference. First, general teaching competence operates more strongly in correlational studies and probably leads to spuriously high correlations between specific teacher behaviors and pupil outcomes. Second, the length of pupil exposure to teaching behaviors studied may be longer in correlational studies than in experimental studies. Third, when teachers adopt new behaviors, there is a lag in the development by pupils of appropriate responses to these behaviors. Fourth, when teachers are trained in new behaviors, time is not often allowed to incorporate the skill into teacher performance. (Author/JS)