The effect of class size, course level, instructor rank, campus, and subject area on 14,366 student ratings of 488 science and mathematics classes at a main university and 18 branch campuses was investigated. Class means were factor-analyzed and four resultant components of teaching performance were submitted to a 2x2x2x2x3 MANOVA. Discriminant functions of significant main effects and interactions did not sustain popular faculty notions ascribing the highest student evaluations to so-called favorable characteristics. Recommendations were made for controlling bias through norming, revision of the rating instrument, and future areas of study. The research instrument used in this investigation was a local adaption of the Student Appraisal of Teaching and is included in the appendix. (Author/RC)