The aim of the study was to differentiate bipolar II, bipolar I and recurrent unipolar depression by their familial load for affective disorders. Eighty bipolar, 108 unipolar, 80 control subjects and interviewed first-degree relatives were diagnosed according to Research Diagnostic Criteria using the Schedule for Affective Disorders and Schizophrenia--lifetime version. The morbid risks for bipolar I disorder were equivalent in relatives of bipolar I (3.6%) and bipolar II (3.5%) subjects and lower in relatives of unipolar subjects (1.0%). The morbid risks of relatives for bipolar II disorder distinguished bipolar II subjects (6.1%) from bipolar I subjects (1.8%), from unipolar depressives (0.3%) and from controls (0.5%). To promote further evaluation, bipolar II disorder should be included in DSM-IV as a distinct diagnostic category.