This study manipulates the variables of children's ethnicity, sex, and ability to ascertain the nature of the interaction relationship between teacher expectancies and student performance. The subjects were urban teachers who were asked to read case histories and then rate the child on a Likert-type family and pupil behavior rating form and a modified Osgood semantic differential scale. In addition, each subject completed a revised F scale of 30 items to assess each subjects' degree of prejudice. Analysis of the ratings given by the 386 teachers confirms the premise that the interactions between teachers' expectancies and students' performance is more complex than originally expected; the variables of ability, sex, ethnicity and prejudice of teachers were found to be of importance in understanding these interactions. It is further concluded that the global ratings of the Osgood semantic differential scale are more powerful in recording the differential ratings of the teachers than are the particular judgments of the family and pupil behavior rating form. The third conclusion is that low ability students are the particular group of students who are "victimized" by the ratings of the high prejudice teachers. These findings suggest the need for administrators to be aware of their staff's prejudices, particularly as they interact with low ability students of minority status. They also suggest the need for retraining of teachers who are prejudicial or the assignment of them to schools where they do not come into contact with low ability minority students. (Author/AM)