A locally revised measure of self-concept and racial attitudes was used to determine if factor analysis could detect possible item bias. Participants were 2,800 Hispanic Americans, Whites, and Blacks in fifth or eighth grade. A classical factor analysis (using commonality estimates) with Varimax rotation defined the factor structure of all non-demographic items, for each ethnic group separately and for all groups combined. Only the fifth grade Hispanic group and the eighth grade White group failed to match the combined factor structure. The remaining four groups differed only by order of factors extracted during rotation. Although item bias was not present, the importance of factor analysis cannot be overemphasized because grouping by sex, socioeconomic status, or ethnicity may effect test validity. (CP)