Forty psychiatrists in Arizona were asked to rate the dangerousness to self or others of 16 patients described in case histories and to recommend an appropriate course of action. Half the psychiatrists were given the state defining dangerousness to use in responding. Psychiatrists who used the statute summary were less consistent in their predictions of dangerousness than were those who did not use it, especially when the patient had a history of violence. The concept of cognitive dissonance is used to partially explain this paradoxical finding.