We study experimentally whether and when a shared history of signals affects players’ incentives to cooperate in infinitely repeated prisoner’s dilemmas. We compare games where the player can privately observe imperfect signals about her partner’s actions to games where signals are perfect and games where signals are imperfect but public. Based on evidence from multiple specifications of stage payoffs, we find that when the return to cooperation is low, subjects are more likely to cooperate in games with public signals than in games with private signals. However, when the return to cooperation is high, the difference is not significant. Furthermore, we show evidence that strategies are more lenient with public signals than with private signals when the return to cooperation is low. The results suggest that a shared history of signals could increase players’ incentives to cooperate when the return to cooperation is low.