Comparison of the Millon Clinical Multiaxial Inventory (MCMI) and clinician-generated DSM-III diagnoses for a sample of psychiatric outpatients (N = 72) indicated only chance agreement. The MCMI was of limited use as a screening device because of overdiagnosis and failure to identify individuals who prematurely terminated treatment. Other potential problems with the instrument included substantial intercorrelation among scales and possible pathologizing of stereotypic feminine traits. The characteristic structure of the MCMI appeared stable across patient samples; there was evidence for identification of a mean profile for psychiatric patients who voluntarily seek treatment for subjective distress.