This study investigated the ability of middle school teachers to use descriptive feedback from their students in changing their teaching behavior. One homeroom group of twenty-five students was observed in interaction with nine teachers of math, English, social studies, and science over a one-year period to elicit both quantifiable and qualitative data. Classroom observation techniques included sequential records of student-teacher interaction, running accounts of student behavior, and a questionnaire describing student attitudes toward school and self. Pupils were surveyed in September, mid-January, and late May. In mid-January, teachers were given descriptive information based upon the observation data, including information about pupil attention rates, the quantity of teacher-initiated interaction, student-initiated interaction, disciplinary comments, praise, criticism, and extended interaction. General teacher reaction to this type of nonevaluative, descriptive feedback was extremely receptive, though each teacher's reaction and behavior change during the second semester was different, as illustrated in separate vignettes of each. Teachers expressed a desire for more frequent feedback and seemed eager to establish a dialogue with the observer-researcher. The observation-feedback method used was viewed as generally useful for improving the ability of teachers to improve classroom life. (MB)