Expecting continued play in prisoner's dilemma games: A test of several models

Several models of prisoner's dilemma interactions were tested in a series of twelve games whose termination point was determined probabilistically. A new model was introduced to discriminate among equilibrium and nonequilibrium situations on the basis of a player's expected benefits or losses for cooperating. The experiment included twelve payoff matrices, three probabilities for continuing, two opponent strategies, and the player's sex as independent variables. Results showed that both the game payoffs and the probability that the game would continue interacted to affect the rates of cooperation observed, and that the equilibrium model predicted this outcome most accurately. While the predictions of each of the models were supported, the equilibrium models appeared to be superior to the others. The discussion highlights the importance of considering the likelihood of a game terminating as a major determinant of the cooperation that can be expected in mixed motive interactions.