The authors' primary objective is to outline the phenomenology, importance, and available data on issues concerning the boundaries between bipolar disorder and diagnoses such as schizophrenia, unipolar depression, and personality disorders. In addition, by illuminating the many difficulties with the boundaries of one of psychiatry's more robust diagnoses, they hope to awaken in the reader a healthy skepticism about current psychiatric nosology. For a topic of this scope, a literature review must be selective. For each boundary area, a mixture of classic and recent papers covering a range of validating criteria is included whenever possible. Good summary data are cited when available, as are a selection of relevant theoretical papers. The review indicates that current diagnostic criteria for bipolar disorder are generally reasonable, but there are many problem areas, most of which cannot be solved by changes in criteria. Notable among these are 1) the possibility of future manic episodes in unipolar disorder, 2) schizoaffective disorder, bipolar type, and 3) borderline personality disorder with prominent mood swings. The disputes concerning the boundaries of bipolar disorder illustrate the limitations of categorical diagnosis which result from the implementation of diagnostic criteria, the criteria themselves, the fundamental nosologic process, and the phenomena themselves. If these limitations are to be extended, it may be necessary to explore alternative ways of defining psychiatric diagnoses for different settings in research and clinical practice.