Unlike previous attempts to implement cooperation in a prisoners’ dilemma game with an infinite horizon in the laboratory, we focus on extended prisoners’ dilemma games in which a second (pure strategy) equilibrium allows for voluntary cooperation in all but the last round. Our four main experimental treatments distinguish long versus short horizon and strict versus non-strict additional equilibrium compared to the control treatment, a standard prisoners’ dilemma. Quite surprisingly, according to our results, only a strict additional equilibrium increases cooperation rate for a given time horizon. As expected a longer time horizon promotes cooperation.