Role-specific self-concept was assessed in 357 geometry students in Utah and North Carolina and analyzed in a causal-comparative study of the influences of success (geometry letter grade from previous semester), teacher, and sex of student. Research literature suggests that variations in self-concept might be expected as a result of, or from interaction with, these independent variables. Self-concept was assessed as a rank order correlation coefficient of two Q-sorts (ideal mathematics student, actual self) of 24 student descriptor items. Success and teacher were found to be statistically significant influences; sex was not. No interactions were found to be significant. The three independent variables accounted for 19.5% of the observed variance in self-concept scores. The results confirm previous findings about the role of success in learning, have implications for teacher recruitment and performance, and raise questions about sex differences in mathematics education. (Author)