We assessed the accuracy of the family history (FH-RDC) and family study (SADS-L) methods for obtaining information about the presence of psychopathology in 274 first-degree relatives of patients with psychotic disorders. The family history method had only modest sensitivity, 40.8% for affective disorders and 58.6% for psychotic disorders, but high specificity, 94.1% for affective disorders and 98.7% for psychotic disorders. For both disorders, sensitivity was higher for relatives who had had previous psychiatric admissions. However, with the family study method, we found that relatives with affective disorder were more likely to be interviewed than those relatives with other disorders. Hence, the family study method may be prone to selection bias that distorts morbid risk estimates. We conclude that the best way of collecting information regarding family psychopathology is to interview directly as many relatives as possible and to collect supplementary family history information on unavailable relatives.