Using a step-level public good game, we analyze the effects on contributions of having played under a sanctioning regime. We find that ”educational” effects, in terms of learning a particular way to coordinate towards ”good” equilibria, are more relevant than motivational ”crowding out” effects, whereby cooperating to avoid sanctions spoils intrinsic incentives. If groups vote, they decide to remove the costly sanctioning regime; then they cooperate as much as in automatic removal only when this decision entails a clear ”trust” message.Peer reviewed