An experiment explored the relative effectiveness of two strategies for controlling interracial aggression: observing other individuals censured for aggression and potential direct censure of the aggressor himself. Subjects were provided with an opportunity to aggress under conditions designed to maximize or reduce each type of censure. A principal components analysis of the data revealed two orthogonal variates: direct aggression and indirect aggression. It was found that both observed censure and potential direct censure were effective in modifying directly aggressive responses. However, these forms of censure interacted in their effect on indirect aggression, such that observed censure was effective in reducing indirect aggression only under conditions of potential direct censure.