This study presents evidence that differences in mathematical career outcomes of males and females are the result of early socialization in the family. One reason fewer females than males elect to take high school mathematics courses is that during middle school years females perceive low parental evaluations of their mathematical aptitude and develop low self-concepts of mathematical ability. Traditional independent variables, intelligence quotient (IQ) and socioeconomic status (SES), do not explain the behavioral difference between the sexes in higher mathematics. Data are taken from a longitudinal study of 588 students in grades 8-12. Data analysis is based on path analysis models and tables of percentages which describe differences between males and females. Tables present information on completed mathematics courses, perceived parental evaluation of mathematical ability, and self-evaluation. Three intervening variables are examined for path analysis: sex, IQ, and SES. Path analysis indicates that the intervening variables, perceived parental evaluations, and self-concept of mathematical ability are strongly correlated with subsequent behavior. That is, females are programmed by both parents and themselves for lower performance in mathematics than males, and they subsequently behave according to the self and parental concepts. Implications for social change are considered. (Author/ND)