Using confirmatory factor analysis, this study examines the validity of the Usher and Pajares Sources of Self-Efficacy in Mathematics (SSEM) for prospective or eventual first-generation college students while still in middle school, a population not studied with the SSEM heretofore. Studying this population is especially noteworthy as educators seek to increase college going rates for this group, especially in science, technology, engineering, and math. After minor modifications, we achieved an acceptable model fit for both first-generation and non-first-generation students, although the model was not invariant across groups. The most notable difference was the significantly higher level of stress and negative physiological response reported by first-generation students. Knowing this important difference can be important for middle grades educators in helping these students successfully access foundational math curricula. Additionally, we successfully utilized the SSEM as a predictor of math course grades for first-generation and non-first-generation students showing a significant positive relationship between math self-efficacy and math course grades. The positive relationship indicated an increase in self-efficacy related to an increase in math performance and was not significantly different between groups.