The reliability of the two most widely used course evaluation instruments at Indiana University was examined. The 38-item Form A, a modification of a course-evaluation instrument reported by Hildebrand and Wilson, at the University of California at Davis, asked students to rate instructors on a seven-point scale, from excellent to poor. The 40-item Form B, developed by the School of Business, asked for ratings of both course and instructor on a seven-point agree/disagree scale. Separate studies investigated four types of reliability: (1) stability of scores when two items were randomly selected on each Form and repeated near the end (lie scale date); (2) internal consistency or homogeneity of the items within and between subscales; (3) consistency of student evaluations of each instructor evaluated; and (4) ability of the Forms to discriminate between good and poor teachers. Thirty-eight instructors were evaluated by 914 undergraduates, 16 instructors with both Forms. Students were not consistent in their responses and the instruments did not adequately separate one teacher from another. Despite the lack of reliability of the Forms, the need for improving the evaluation procedure is critical. Suggestions include isolating students who give reliable responses; using computer scorable instruments only to isolate the very good or very poor teachers; using open-ended hand-written responses; and, finally, providing adequate administration time and stressing the importance of the evaluation. Forms A and B are appended. (Author/CP)