OBJECTIVE This analysis aimed to quantify the long-term stability of distinctions between nonbipolar, bipolar II, and bipolar I affective disorders and to determine the predictors of shifts in patients' diagnoses among these categories. METHODS Probands entered the study as they sought treatment for manic, major depressive, or schizoaffective disorder diagnosed according to the Research Diagnostic Criteria. After thorough baseline evaluations, 605 patients with nonbipolar major depressive disorder or schizoaffective disorder, depressed type; 96 with bipolar II disorder; and 231 with bipolar I disorder or schizoaffective disorder, manic type, began the follow-up study. Direct interviews took place at 6-month intervals for the first 5 years and annually thereafter. RESULTS Only 20 (5.2%) of the 381 initially nonbipolar probands who completed 10 years of follow-up developed mania during that time, and only 19 (5.0%) developed hypomania. A slightly higher proportion of the 67 who began with bipolar II disorder developed mania during the 10 years. Although 101 (66.4%) of the 152 bipolar I or schizoaffective manic probands developed subsequent manic episodes, only 11 (7.2%) developed hypomanic episodes and no mania. Young age at intake and at onset and chronicity of the index episode predicted shifts from nonbipolar to bipolar II disorder. Psychosis and a family history of mania predicted shifts from nonbipolar to bipolar I disorder. CONCLUSIONS The high stability of baseline distinctions between nonbipolar, bipolar II, and bipolar I disorders, in combination with previously described family study data, strongly supports the separation of these disorders for both clinical and research purposes.