A stratified random sample of 84 student teachers in secondary school academic subject areas was distributed into four treatment groups according to a trichotomization of their scores on the Minnesota Teacher Attitude Inventory (MTAI). Each student teacher taught two 20-minute lessons with the same content to different classes of five pupils. On group supervised itself using videotape and Flanders' system of interaction analysis; another group was supervised in a more traditional manner (supervisor present). The two remaining groups experienced both self-supervision and teacher supervision. Upon program completion the MTAI was retaken (in one group retesting was at delayed intervals). Analysis of variance indicated the following: (1) No significant relationship exists between attitudes and teaching behavior before supervisory treatment. (2) Supervisory treatment tends to promote a significant relationship between attitudes and teaching behavior. (3) Self-supervision tends to promote indirect teaching and higher MTAI scores. (4) Estimated by student teachers of the precentage of indirect teaching they exhibit in their lessons are very inaccurate under both traditional and self-supervision. (5) No significant relationship exists between time and attitude change in student teachers supervised in a traditional manner. (Author/LP)