An experiment assessed the prisoner's dilemma game (PDG) choices of a set of three persons who interacted with another supposed set of three persons. There were four conditions: (1) group-on-group (both three-person sets constrained by a majority vote), (2) group-on-one (only own three-person set constrained by a majority vote), (3) one-on-group (only other three-person set constrained by a majority vote), (4) one-on-one (neither three-person set constrained by a majority vote). The four conditions were compared with three orthogonal contrasts. Consistent with the assumption that interindividual interactions are less competitive than interactions involving groups, the first contrast indicated that there were fewer competitive choices in the one-on-one condition than in the other three conditions pooled. Consistent with the assumption that competitiveness can flow either from acting as a group or interacting with a group, the second contrast found no significant difference between the group-on-one and one-on-group conditions. Finally, consistent with the assumption that the interindividual-intergroup discontinuity effect is a joint function of acting as a group and interacting with a group, the third contrast revealed that there were more competitive choices in the group-on-group condition than in the one-on-group and group-on-one conditions pooled. Copyright © 2006 John Wiley & Sons, Ltd.