A group of 150 psychiatric patients were administered the Diagnostic Interview Schedule (DIS) on two occasions, once by a trained layperson and once using a computerized interview format in which the patient interacted directly with the computer. Agreement between the two methods on 15 diagnoses was relatively modest, as indicated by a mean kappa score of .51, but was similar to agreement rates reported in other studies of the DIS. The discrepancies may have been due to the high number of acutely ill inpatients studied, patients' reporting more symptoms in one of the interviews, and difficulty translating some of the DIS questions to the computer. Patients had positive feelings about both methods, but a significant majority liked the computer interview better and found it less embarrassing. The authors conclude that computerized administration of the DIS is as reliable as other methods but that exclusive reliance on the DIS for clinical diagnosis is inappropriate.