No abstract

Factor analysis of student ratings of teacher effectiveness may reveal more about the semantic similarity of items than it does about actual covariation in instructor behavior. To test this hypothesis, 28 graduate students were asked to rate several hypothetical instructor profiles, constructed by systematically manipulating information about the instructors' classroom behaviors. Factor analyses of the student ratings were performed before and after all behavioral information in the instructor profiles had been statistically removed. The results revealed no substantive change in the underlying factor structure. In both analyses the factors represented clusters of semantically equivalent items. Thus, students appeared to be imposing an implicit semantic organization on their ratings, apart from any covariance among the instructors' actual classroom behavior.