Cheating, operationally defined as the falsification of scores on a word construction task, was found, as predicted, to be influenced by achievement anxiety and knowledge of the performance of a peer reference group in 111 high school subjects (Ss). However, achievement anxiety was positively correlated with cheating only when knowledge of reference group performance was provided. Likewise, providing Ss with knowledge of the reference group's superior or inferior performance elicited cheating only at high anxiety levels. The results are interpreted in terms of the general hypothesis that cheating is a response instrumental to the avoidance of aversive social consequences. (AUTHOR)