The paper explores the cooperation rate in one-shot Prisoner’s Dilemma games in three cases – when the game is played among friends, among foes, and among a mixture of friends and foes. The paper checks empirically the prediction that Simpson’s paradox like effects are to be expected in this situation (Chater, Vlaev, & Grinberg, 2008). The existence of a bias for cooperation when playing with friends and for defection when playing with foes, is expected to lead to a reinforcement of cooperation based on higher average payoffs in the mixed condition, i.e. when playing with friends and foes together. At the same time, the average payoff for cooperation remains lower for games with friends and foes taken separately (Simpson’s paradox). There results of the experiment support the existence of such effects and suggest that further exploration is worthwhile.