A representative sample of 95 hospitalized bipolar manic-depressive patients was followed up from 1959 to 1975. The mean age of the group at the time of this study was 61 years. It was observed that female bipolar patients demonstrate depression much more frequently than mania, while male patients show a symmetric distribution of both manic and depressive syndromes. The longitudinal occurrence of syndromes remains more or less constant; for instance, individual patients do not tend to go into depression with increasing age. The study shows that even after three episodes 29% of all bipolar patients would still have been misdiagnosed as unipolar depression. An attempt is made to classify bipolar patients into three subtypes, 'preponderantly manic,' 'preponderantly depressed,' and a 'nuclear' type. Male patients belong mainly to the latter with an equal proportion of the first and third subtype. In contrast, female patients belong mainly to the depressed subtype. The findings are discussed assuming either a heterogeneity of bipolar disorders or a threshold model of affective disorders suggested by Gershon et al. (1976).