An extensive body of research indicates that men on the average achieve higher scores in mathematics than women. Despite this extensive research, conclusions about sex differences in mathematics achievement have suffered from various inadequacies in the use of measures of association among coursework experiences, sex, and other correlates of mathematics achievement. This paper addresses these issues by estimating a latent-construct causal model of mathematics achievement with a mixed matrix of tetrachoric, polyserial, and product-moment correlations. This model of mathematics achievement is a block-recursive model. Mathematics achievement of high school seniors is considered to be a function of sophomore mathematics and verbal abilities, attitudes toward mathematics, and exposure to mathematics. The model was tested using data from the first follow-up of the nationwide, longitudinal study "High School and Beyond." Based on a covariance-structures model of the process of mathematics achievement and appropriate measures of the associations among the variables in the model, it was found that this process differs for men and women. (Author/BW)