Analyses of two large data sets collected in real-world settings were used to test predictions derived from theorizing on stereotype threat. Specifically, the notions that stereotype threat should emerge only among those highly identified with a domain and only with tests that push the limits of individuals abilities was used to develop hypothesized models of the relation between testing and performance on a criterion. Both assumptions suggest that the test-performance relationship should be nonlinear for members of minority groups and that prediction of performance for minorities should be poorest at the highest levels of test performance. These assumptions were tested using the SAT to predict academic performance (focusing on women and Blacks in math) and with data from the Armed Services Vocational Aptitude Battery (ASVAB) to predict military job performance (focusing on Blacks in technical proficiency). Results using SAT data showed no evidence of nonlinear relationships, and equivalent effects for women, men, and Blacks. Analyses of the ASVAB revealed a linear test-criterion relationship throughout range of scores for one test and a stronger test-performance relationship for soldiers in the top half of the test score distribution on a second test, regardless of race. The authors speculate that the failure to obtain effects inferred from stereotype threat theory might reflect either the need for explicit highlighting of minority status or gender prior to testing to produce stereotype threat effects, or the fact that stereotype threat can be overcome by increased effort and motivation.