This study investigated stereotype threat in a simulated selection exercise so that stereotype threat and face validity could both be manipulated. Black and White undergraduates completed a test to "understand how to select people for retail managerial positions", and were informed that the top 15% of performers would receive $20. Before beginning the test, students were either told that it was diagnostic of intelligence (stereotype threat for Blacks), a good indicator of "skills as a retail manager" (non-diagnostic), or that the test was difficult (control). The test itself was either high or low in face validity (i.e., problems were either framed in terms that would be appropriate for a retail manager or were framed in a way where the link to retail management was tenuous). Overall, performance was higher for Whites than for Blacks, and higher in the control than in the diagnostic condition. Race did not interact with diagnosticity, as predicted by stereotype threat theory. However, an analysis focusing on only students highly identified with their race did produce an effect consistent with stereotype threat. Specifically, Blacks scored best in the control condition when the test was low in face validity. When the test was high in face validity, however, Blacks scored best in the non-diagnostic condition. When the test was low in face validity and students were that the test was difficult, performance of Black students was as good as White students. When the test was high in face validity, however, Blacks performed best if intelligence was not mentioned. These findings suggest that Blacks might assume that face valid tests are diagnostic unless they are assured to the contrary. Correlational analyses indicated that differences in perceptions of threat influence face validity perceptions, motivation, and anxiety, and that these factors partially mediated performance under stereotype threat.