The paper shows that efficient cooperation processes exhibit gradualism when each player does not know the likelihood that the other player exercises a stochastically available outside option. Two players, asymmetrically informed on this likelihood, play an infinitely repeated prisoner's dilemma...