Thirty physical science and 30 chemistry classes, which contained a total of 1332 students, were observed using the Brophy-Good Teacher-Child Dyadic Interaction System. Classroom interactions were examined for gender differences that may contribute to the underrepresentation of women in physics and engineering courses and subsequent careers. The Brophy-Good coding process allows for examination of patterns of interactions for individuals and groups of pupils. An analysis of variance of the data yielded a significant main effect for teacher praise, call outs, procedural questions, and behavioral warnings based on the sex of the student and a significant teacher-sex main effect for direct questions. Significant two-way interactions were found for the behavioral warning variable for teacher sex and subject by student sex. Female teachers warned male students significantly more than female students. Male teachers warned both genders with similar frequency. Male students also received significantly more behavioral warnings in physical science classes than female students. In chemistry classes, both male and female students received approximately the same number of behavioral warnings.