AbstractSixty Japanese female students were asked to exchange shocks in electric roulette games with female opponents. The subjects were assigned to either the 80%, 50%, or 20% win conditions. Half of them were then led to judge that the power unbalance was legitimate by being informed that the assignment was based on a prior performance contest, with the good performer being assigned to the advantageous position. The other subjects were led to perceive the power imbalance as illegitimate by being informed that the assignment was randomly decided. The opponents always severely attacked them. The retaliation by the subjects was analyzed by a two‐way ANOVA with Power Imbalance and Legitimacy. It was found that the subjects both in the 20% and 50% win conditions set more intense shocks to their opponents than those in the 80% win condition. This is not consistent with the fear of retaliation hypothesis which had predicted that the subjects would refrain from intensely aggressing against the opponent who had a greater aggressive capacity. It was also found that in the 80% win condition, the subjects set more intense shocks when the unbalance had been determined by their performance than when determined by luck, whereas in the 20% win condition, they set more intense shocks when the imbalance had been determined by luck than when determined by their performance. These suggest that retaliation depended upon perceived justifiability of aggression which was predicated on the legitimacy of the power imbalance. Finding in the 50% win condition that the shock settings were higher when the power imbalance had been determined by their performance than when determined by luck was interpreted in terms of their heightened competitivenes