We removed Stroop trials with incorrect responses and trimmedthe reaction time data as follows. For each subject, we calculatedthe mean and standard deviation for the control trials andfor the incompatible trials (those in which the font color and themeaning of the text disagreed) separately, and then removed allresponses 3 or more standard deviations from the mean for theirrespective kind of trial. We then calculated an average reactiontime for each trial type using the remaining trials. Stroop interferencescores were calculated by subtracting the mean reactiontime for the control trials from the mean reaction time forthe incompatible trials for each subject. We searched for statisticaloutliers within each prejudice condition separately, andremoved a total of nine interference scores (three outliers ineach condition). Thus, all analyses involving Stroop interferencescores are based on an n of 237.We entered the interference scores into a 2 (subject’s race)2(evaluator’s race) 3 (prejudice condition) analysis of varianceand observed the expected interaction between subject’s raceand prejudice condition, F(2, 225)55.13, p .93, Z2 5 .03. In contrast, White subjectsexperienced significantly more interference than did Blacksubjects in the blatant-prejudice condition, F(1, 233) 5 5.89,p < .02, prep > .93, Z2 5 .03. Interference was comparable forBlack and White subjects in the no-prejudice condition (F .99, Z2 5 .08.We also observed an interaction between subject’s race andevaluator’s race, F(1, 225)57.36, p