Providing effective feedback and supervision for student teachers is an important area of concern. The two primary sources of feedback, the cooperating teacher and the university supervisor, are somewhat limited in scope. At the same time, various approaches used successfully with experienced classroom teachers have been rarely tried with student teachers. This paper describes a 1986 pilot study with 20 secondary student teachers assigned to three groups to examine the efficacy of three training approaches. (1) The control group received ordinary supervisor feedback. (2) The "experimental feedback" group received feedback from a classroom analysis scheme--the Stallings Observation System (SOS). (3) The "experimental seminar" group also received feedback from the SOS, but with guidance and seminar discussion on how to use it for self-improvement. This group also learned to use the Low Inference Self-Assessment Measure. Research results indicate that student teachers in the experimental (third) group improved their teaching in desired directions, exhibiting more interactive teaching behaviors and fewer organizing statements and off-task behaviors, whereas the first and second groups generally did not improve their teaching behaviors. Also, the first and second groups had similar profiles that differed markedly from those of the experimental group. The results indicate that the combination of systematic feedback from a classroom analysis system, self-analysis, and collegial feedback and support is effective for improving student teachers' classroom instruction. A more difficult problem is providing student teachers with sufficient opportunity to practice and modify their behaviors. Included are 1 table and 14 references. (MLH)