Distracter analysis was introduced as a means of investigating sex differences in mathematical performance. The data were responses to an assessment test given each year from 1976 to 1979 to all sixth-grade children enrolled in California public schools. Loglinear models were used to evaluate the consistency of response by each sex over 4 years. A significant interaction between sex and choice of distracter occurred for a large majority of the items. Hypotheses of sex differences based on Radatz's (1979) information-processing classification of errors were evaluated.