An analysis was conducted of two large samples of 13- and 17-year-old students (N=7,928) who were assessed by the National Assessment of Educational Progress during the 1977-78 and 1981-82 mathematics assessments. A group of 18 statements about math were used to construct a math attitude index. A multivariate log-linear analysis included gender, assessment year, and age as independent variables and the math attitude index as the dependent variable. The results revealed that all three independent variables made significant contributions to the final model. A five-way analysis of variance was done using the dependent variable of math achievement score and the independent variables of math attitude index, race, gender, type of community, and racial composition of the school. The results of this analysis showed that math achievement was significantly affected by each of the independent variables. There were also three significant two-way interactions identified: (1) type of community and racial composition of the school; (2) math attitude index and racial composition of the school; and (3) math attitude index and gender. These findings suggest that while gender differences do exist in both attitudes toward math and achievement in math, these differences are less than the differences in the other variables examined. Additional multivariate studies should be done to examine this issue further. (Author/NB)