Subjects playing the role of psychiatrists (actors) engaged in a simulated medical decision-making task in which they attempted to bring the value of a patient indicator variable into a desired range. For each treatment recommended by the actor, both the actor and an observer subject playing the role of a nurse assessed the probability that the treatment would be effective. Both actors and observers were overconfident. Actors were more confident in their treatment recommendations than were observers, but this difference was eliminated when observers were given the opportunity to offer their own alternative recommendation. Under the latter circumstances, actors and observers were equally confident in the actors' decisions but observers were more confident than actors in the observers' decisions. These findings suggest that while control over the outcome of the decision has little influence on actor-observer differences in confidence, feedback regarding this outcome plays a crucial role.