In the current controversy over accountability in education, educators concerned with the affective aspects of education seem to be in danger of losing the battle of "behavioral objectives" on two fronts: (1) demonstrating that the affective behavior of the teacher does make a difference in educational outcomes, and (2) demonstrating how affective educational outcomes can be satisfactorily evaluated. The National Consortium for Humanizing Education addressed itself to the first of these questions; this document reports on three large N studies of teacher behavior. The first two studies indicate that in describing relationships with affective teacher behavior, it is necessary to move into polynomial models. The third study demonstrates that in predicting student outcomes, factors other than the linear term of the mean of the teacher's behavior add considerably to the predictive power of the model. Additionally, the fact that the teacher's stability of affective behavior is a significant predictor of student outcomes offers some implications which help explain why polynomial models of affective teacher behavior seemed to be more adequate than linear models. Extensive data graphs, tables, and charts are included. (Author/PC)