Although there have been numerous studies of the impact of teacher expectation on student performance, the manner in which expectations are formed has remained unexplored. The present research tested the hypothesis that teachers develop expectation of student performance by averaging information about his or her motivation and ability. In two experiements on prediction of performance in nonacademic contests (n = 60) and academic exams (n = 60), manipulations of number of pieces of similar motivation information, availability of additional pieces of average motivation information, and unavailability of information about either motivation or ability all yielded results as prescribed by the averaging model. Motivation information received not only less importance than ability information but also got discounted when it differed markedly from that of ability information. As this result differs from those previously obtained from students and managers, the second hypotheses of role differences in expectation of performance was also supported. Theoretical, methodological, and applied implications of the results were discussed. Suggestions for future research were also made.