Structural covariance models that would explain the correlations observed among mathematics achievement and participation measures and related cognitive and affective variables were developed. A sample of college calculus students (N=268; 124 females and 144 males) was administered a battery of cognitive tests (including measures of spatial-visual ability and field dependence/independence) and a battery of affective scales (including measures of attitude toward mathematics, confidence, perceived usefulness of mathematics, and effectance motivation). Measures of previous academic achievement and participation in mathematics and science, and achievement in the calculus course itself were recorded. The sample was then split into two subsamples. Using one subsample, a test of the hypothesis that males and females have the same covariance structure for the measures was performed. A significant difference was found (p less than .0001). Separate structural models were developed for males and females based on the sample correlations from this subsample. The second subsample was used to evaluate the adequacy of the developed models by application of factorial modeling, a confirmatory tool for such models. There was evidence of a reasonable fit for both models. Directions for further research with the aim of improving the model is discussed. Extensive statistical tables are appended. (Author/JN)