The purpose of this research was to develop stable categories of college teacher performance. The data for developing these categories were collected by means of the critical incident technique. In three separate conditions, college students were asked to give critical incidents that they had observed of professors. In the first condition, critical incidents were requested by use of a very general question. Conditions two and three were designed to provide checks on possible sources of bias or distortion in condition one. Upon qualitative sort of the incidents, nine categories (three of which were course-specific) were determined. Through use of a rank-order correlation to compare the relative frequency of reporting incidents falling into each category, it was found that all classes, freshman through senior, maintained the same rank order. Females, however, tended to report incidents relating to personal relationships more frequently than males. The effects of three personality variables were also checked, and found not to affect the category of incident reported. The critical incident taxonomies developed here offer an advance toward a behaviorally based rating scale for teachers. (Author/JLL)