Despite an abundance of older and more recent retrospective and considerably fewer prospective-longitudinal studies in bipolar disorders I and II, there are still remarkable deficits with regard to our knowledge about the natural course and burden. The considerable general and diagnosis-specific challenges posed by the nature of bipolar disorders are specified, highlighting in particular problems in diagnostic and symptom assessment, shifts in diagnostic conventions and the broadening of the diagnostic concept by including bipolar spectrum disorders. As a consequence it still remains difficult to agree on several core features of bipolar disorders, such as when they begin, how many remit spontaneously and how many take a chronic course. On the basis of clinical and epidemiological findings this paper summarizes (i) a significant need to extend the study of the natural course of bipolar disorder in clinical samples beyond the snapshot of acute episodes to the study of the mid-term and long-term symptom course, associated comorbidities and the associated burden of the disease. (ii) In terms of epidemiological studies, that are also of key importance for resolving the critical issues of threshold definitions in the context of the bipolar spectrum concept, there is a clear need for identifying the most relevant risk factors for the first onset and those for the further illness progression in early stages. Since there are some indications that these critical processes might start as early as adolescence, such studies might concentrate on young cohorts and clearly before these prospective patients come to clinical attention. (iii) The value of both types of studies might be enhanced, if beyond the use of standardized diagnostic interview, special attempts are made to use prospective life- and episode-charting methods for bipolar illnesses.