This experiment involved highly math-identified women who completed a difficult math test in a room either with two females or two male accomplices of the experimenter who posed as fellow test-takers. Half the women were led to believe that their scores would be shared at the end of the testing session, and half were told that their scores would be anonymous both to the fellow students and the experimenter. Analyses of the math scores showed that women performed more poorly in the presence of the males, and this was true regardless of whether the scores were private or were to be shared. This study shows that stereotype threat can occur even when the behavior under consideration cannot be observed or even known by others.