To what reference point do people calibrate cost-free, third-party punishment?

To what reference point(s) do third parties calibrate punishments to be inflicted upon unknown wrongdoers? We introduce a novel method that allows direct comparison of preferred punishments (and compensations) to victim loss and perpetrator gain. In two experiments, minimalist scenarios indicated various monetary gains for a thief and costs for a victim. Participants indicated a fine for the thief (the victim was uncompensated). We found that victim loss and perpetrator gain had about equal influence on punishment preferences. However, analysis of individual differences indicated a substantial number of participants (about 25% to 40%) preferred relatively large punishments (i.e., greater than the outcome differential-the sum of perpetrator gain and victim loss), and in both experiments the mean of preferred punishments was greater than the outcome differential. A third experiment used identical scenarios but instead had participants indicate a compensation for the victim. In contrast to punishment preferences, only about 2% of participants preferred victim compensation greater than the outcome differential and the mean compensation was less than the outcome differential.