A five year semi-annual follow-up of patients with non-bipolar (N = 442), bipolar II (N = 64) and bipolar I (N = 53) major depression tracked the courses of prospectively observed major depressive, hypomanic and manic syndromes. In all three groups, depression was much more likely in any given week than was hypomania or mania. However, during the majority of weeks, no full syndrome was present and none of the groups exhibited evidence of continuing psychosocial deterioration. Though all three groups exhibited similar times to recovery from index and subsequent major depressive episodes, both bipolar groups had substantially higher relapse rates and developed more episodes of major depression, hypomania and mania. The two bipolar groups, in turn, differed by the severity of manic-like syndromes and thus remained diagnostically stable; the bipolar II patients were much less likely to develop full manic syndromes or to be hospitalized during follow-up. In conjunction with family study data showing that bipolar II disorder breeds true, these data support the separation of bipolar I and bipolar II affective disorder.