We use a novel experimental design to disentangle strategically- and non-strategically-motivated cooperation. By using contingent responses in a repeated sequential prisoners' dilemma with a known probabilistic end, we differentiate end-game behavior from continuation behavior within individuals while controlling for expectations. This design allows us to determine the extent to which strategically-cooperating individuals are responsible for the so-called end-game effect. Experiments with two different subject pools indicate that the most common motive for cooperation in repeated games is strategic and that the extent to which end-game effects are driven by strategically-cooperating individuals depends on the profitability of cooperation.