The authors investigated the relationships between cognitive, motivational, and emotional variables across Hispanic and White students to predict mathematics performance. A theoretically based structural model fit a total sample of 666 4th- to 10th-grade students well, supporting that self-efficacy, sources of self-efficacy, and emotional feedback were all stronger predictors of mathematics performance than general mental ability. Tests of the structural model across ethnicity suggested a good fit for the White sample but not for the Hispanic sample. However, the majority of the associations in the White model were not significant. Because the model positing relationships among motivational and cognitive variables has been well established, the findings indicated that the inclusion of emotional feedback made it more complicated. (Contains 2 tables and 6 figures.)