Using structured psychiatric interviews, we searched for differences between depressed patients with bipolar affective disorder, and with other types of depression. The patients were taken from 2 large series, collected in London and Chicago. Some differences were found, but few more than would be expected by chance, in view of the large number of variables examined. The possible reasons for this failure are discussed. They include contamination of the groups, and the insensitivity of structured mental state inventories.