To determine the relationship between 16 background variables and students' evaluations of instruction, a questionnaire was completed in 511 undergraduate courses at the University of Southern California, Los Angeles. Student variables, including grade point average, class size, expected grade, and prior subject interest, rarely explained 10% of the variance in any student ratings and generally explained less than 5%. Different statistical techniques, however, suggested that 12% to 14% of the variance in the student ratings could be predicted by the set of background variables. The variables most important in predicting evaluations were prior subject interest, expected grade, workload/difficulty, and perhaps, percent taking course for interest only. Of these, prior subject interest was the most important and was better interpreted as a variable affecting quality of education. Background variables did have a small relationship to the evaluations, but results argue against bias interpretation. Workload/difficulty was correlated in the opposite direction as would be expected from a bias effect. No single variable was related to a majority of the evaluation scores. The scores most likely to be biased (overall rating and instructor enthusiasm) were not the scores most related to the background variables. A sample summary of the instructor's rating, and the questionnaire are appended. (Author/JAG)