An item classification scheme developed by A. M. Gallagher (1990) was refined, resulting in a more accurate prediction of sex differences in the mathematical sections of the Scholastic Aptitude Test (SAT). Differential Item Functioning (DIF) procedures for examinees scoring over 650 indicated that the majority of items favoring males required the use of mathematical insight, while all the items flagged as favoring females required standard algorithmic solutions. Structured interviews were conducted with 25 male and 22 female high school juniors and seniors from the examinee group to study differences in strategy use. There was a substantial overlap in strategies used by males and females, but analysis across all items indicated that females were more likely than males to use algorithmic strategies, and males were more likely than females to use insightful strategies. Questionnaire data obtained from the students interviewed indicated a positive relationship for both males and females between SAT mathematical performance and positive attitudes toward mathematics. The use of algorithmic strategies was correlated with negative attitudes toward mathematics. Implications of these and other findings are discussed. Eleven tables present study findings. Four appendixes contain the following: (1) prototypical items for classification categories; (2) problems used for think-aloud protocols; (3) attitude and background questionnaire; and (4) background variables and correlations between SAT mathematics or strategy and questionnaire items. (Author/SLD)