This study was designed to investigate whether spatial skill, math anxiety, and math self-confidence functioned as mediators of a significant gender difference in the Mathematics Scholastic Aptitude Test (SAT-M) among the top third of a college-bound sample. Using path analytic techniques, the decomposition of the significant gender-SAT-M correlation into direct and indirect effects indicated that there were no direct effects of gender on SAT-M. Mental rotation and math self-confidence showed indirect effects, mediating the gender-SAT-M relationship; math anxiety did not. Of these indirect effects, 36% was mediated by math self-confidence; 64% by mental rotation. For both these variables, most of the mediational effects of the gender-SAT-M relationship did not occur by way of the causal pathway leading through geometry grades. Thus, the mediational effects cannot simply be attributed to the presence of geometry items on the SAT-M or to math self-confidence acquired during prior geometry coursework.