In finitely repeated prisoner dilemma games, two-person teams start with significantly less cooperation than individuals, consistent with results from the psychology literature. This quickly gives way to teams cooperating more than individuals. Team dialogues show increased payoffs from cooperation, along with anticipating opponents’ recognition of same, provides the basis for cooperation, even while fully anticipating defection near the end game. A strong status quo bias in defecting across super-games limits unraveling. Defecting typically occurs’ one round earlier across super-games, consistent with low marginal, or even negative, benefits of more than one-step-ahead defection. We briefly discuss the effect of cheap talk between opponents on outcomes, and the driving forces for the results observed with cheap talk.