We study experimentally the effect of multigame contact on cooperation, with each subject playing a pair of indefinitely repeated prisoner’s dilemmas. Multigame contact is present if a subject plays both games with a single partner, and it is absent if each of the two games is played with a different partner. In contrast to the theoretical prediction, multigame contact does not increase overall cooperation rates. Nonetheless, multigame contact systematically affects behavior and outcomes, acting like a double-edged sword, in the sense that subjects link decisions across games and, consequently, mutual cooperation and mutual defection in both games become more likely. (JEL C72, C73, C92, D91)