In a two-part experiment, Black and White students studied the definitions of 24 obscure English words, and were later tested, in threatening or non-threatening environments. In the threatening study environment, students were told that the task would assess their "learning abilities and limitations" and "how well people from different backgrounds learn”. In the non-threatening environment, students were told that the study focused on identifying "different learning styles". When tested one to two weeks later, students were first given a low-stress warm-up exercise with half of the word definitions. Then, in order to evoke concerns about stereotypes, a test was given which was described as evaluating "your ability to learn verbal information and your performance on problems requiring verbal reasoning ability".
The effect of these different environments on the Black students was dramatic. On the non-threatening warm-up test, Black students who had studied in the threatening learning environment performed about 50% worse than Black students who had studied in the non-threatening environment. But on the ‘real’ test, for which stereotypes had been evoked, all the Blacks — including those who had done fine on the warm-up — did poorly.
In the second experiment, only Black students were involved, and they all studied in the threatening environment. This time, however, half of the students were asked to begin with a "value affirmation" exercise, during which they chose values that mattered most to them and explained why. The other students were asked to write about a value that mattered little to them. A week later, students did the warm-up and the test. Black students who had written about a meaningful value scored nearly 70% better on the warm-up than black students who had written about other values.
(2011). Stereotype Threat Undermines Academic Learning.
Personality and Social Psychology Bulletin. 37(8), 1055 - 1067.