Two experiments tested whether stereotype threat can undermine the acquisition of academic knowledge and thus harm performance even in nonthreatening settings. In Experiment 1, Black and White students studied rare words in either nonthreatening or threatening conditions. One to two weeks later, participants recalled word definitions, half in a nonthreatening "warm-up" and half in a threatening "test." Replicating past research, Black students performed worse on the test than on the warm-up. But importantly, Black students who had studied in the threatening rather than nonthreatening environment performed worse even on the warm-up. White students were unaffected. In Experiment 2, a value affirmation eliminated the learning-threat effect and provided evidence of psychological process. The results suggest that stereotype threat causes a form of "double jeopardy" whereby threat can undermine both learning and performance. The discussion addresses implications for the interpretation of group differences and for understanding how brief threat-reducing interventions can produce long-lasting benefits.