I present a model of renegotiation in infinitely repeated games that incorporates the assumption that if a player cheats early in a game, then other players are likely to distrust his promises not to cheat in the future. I show how this can allow much more cooperation in some games than allowable in the approach taken by Bernheim and Ray (1989) and Farrell and Maskin (1989). In other games, my solution concept rules out Pareto-efficient subgame-perfect equilibria, even as the discount factor approaches 1. This contrasts (generically) with existing models of renegotiation.