The purpose of this study was to find out if teachers' evaluations of the content of an oral response would change if the content of the response were held constant while the speech characteristics of the student giving the response were varied. A fixed set of 12 answers was established for each of two questions. Tape recordings were made of six white and six Negro ninth-grade males speaking the answers after it had been established that each student's race could be accurately identified from his voice. The content of the responses was evaluated by 62 teachers, half of whom heard one tape and half of whom heard the second tape. An analysis of ratings summed across responses to both questions showed that responses containing the same wording were rated significantly lower when given by Negroes than when given by whites. There were significant differences among ratings assigned by individual teachers and among ratings received by individual students of the same race, but no significant interaction between individual teachers and students' race was found. Bias scores for individual teachers were quite unreliable, and, probably for this reason, these scores were not associated with teachers' sex or age, years of teaching experience, most frequently taught grade level, or the percentage of Negro students most frequently taught. (Author/CK)