The present study investigated the effect of a commons simulation and fines on a generalization test that incorporated several features important in the real world. Two hundred and seventy-five volunteer college students in groups of seven participated in this study. Approximately one-half received exposure to the commons simulation and one-half did not. One-half of the groups in each treatment level received two posttests with a fine option available and the other one-half received two posttests without this option. The two posttests differed in that one was played with a large unknown referent group and the other was played with the immediate group of seven. The results showed that some generalization from the commons simulation to the large posttest does occur. However, subjects in the large group do not cooperate (act in the common group interest) more or defect (act in the individual interest) less, but become more cautious as a result of the simulation exposure and withdraw from the commons when playing with a large and unknown referent group.