We run an experiment in which two subjects play a two-round minimum effort game inthe presence of a third player (principal) who is the only one informed about past effortchoices and benefits from a higher minimum effort of the others. Sanctions introduced in thesecond round by the experimenter lead to more optimistic beliefs and higher efforts. This isnot true when sanctions have been imposed by the principal. The possibility that the choiceof a sanction is a signal of low effort levels causes players who chose high effort in the firstround to be less optimistic.