Recent experimental studies have accumulated evidence about self-punishment. In accordance with the evolutionary perspective that shame has a reputation-maintenance function, we speculated that shame would promote self-punishment. Accordingly, we tested whether proneness to shame would predict self-punishment. In the first phase of the experiment, 98 undergraduates completed the Test of Self-Conscious Affect (TOSCA), a standard measure of proneness to shame and guilt. About 2 months later, 50 of the original participants took part in a self-punishment experiment, in which they all unintentionally made an unfair resource allocation, and then had the opportunity to inflict self-punishment by abandoning some of the money they had allocated to themselves. The amount of money the participants relinquished was significantly correlated with their shame-proneness. The intensity of posttransgression shame mediated the effect of shame-proneness on self-punishment. These results provide support for the evolutionary theorization of shame as a reputation-maintenance emotion.