When subjects interact in continuous time, their ability to cooperate may dramatically increase. In an experiment, we study the impact of different time horizons on cooperation in (quasi) continuous time prisoner’s dilemmas. We find that cooperation levels are similar or higher when the horizon is deterministic rather than stochastic. Moreover, a deterministic duration generates different aggregate patterns and individual strategies than a stochastic one. For instance, under a deterministic horizon subjects show high initial cooperation and a strong end-of-period reversal to defection. Moreover, they do not learn to apply backward induction but to postpone defection closer to the end.