44 female white teachers were observed for six separate 20-min. periods for a total of 88 hr. The frequency of approving and disapproving behaviors toward boys and girls were obtained and two indices consisting of proportions of approving to total behaviors of boys and girls were derived. After observation, the Bern Sex-role Inventory was administered to the teachers, and four groups of 11 teachers each were classified by a median-split procedure as androgynous, masculine, feminine, and undifferentiated. The two proportions were then transformed to arc sine (angit), probit, and logit scales and compared to the results of the analysis of the nontransformed proportions. Differences in all analyses showed girls were more favorably perceived than boys; feminine teachers showed greatest differences and masculine teachers showed the smallest differences. All differences between approving behaviors of boys and girls were significant except for those of masculine teachers. The three transformations gave essentially similar results with approximately 4% greater non-error variance. The transformations eliminated a gross heterogeneity of variance in the proportions and the logit analysis was most sensitive to differences among types of teacher and pupil-sex groups. Implications were briefly discussed.