Data from surveys of large samples showed the lifetime prevalence rates of bipolar disorder around 1.5%. A main question is whether the low prevalence rates of bipolar disorders are not an artefact of the over-diagnosis of depression and under-diagnosis of bipolar-II.Analysis of the clinician's logical inferential diagnostic process, confirms that the patient does not represent the sole source of useful information because many patients do not experience hypomania as distress but rather as recovery from depression or as a period during which they felt truly well.Epidemiological data are derived from interviews carried out by lay staff which only reflect the patient's point of view.The clinical monitoring study carried out alongside the ESEMED project found for the diagnosis of mood disorders, a Kappa agreement (versus clinical interview) which ranged from 0.23 in Spain to 0.49 in France.If we consider exactly what a Kappa of 0.4 implies for a disorder with an "identified" prevalence rate of 2%, we discover that the prevalence rate may have been under-diagnosed approximately 1.5-fold, so 67% of cases may not have been identified and 50% of the identified cases may be false positives.It is legitimate to surmise that the prevalence reported by recent (extremely costly) epidemiological surveys may be doubtful.Which direction should epidemiology take in dealing with the serious matter of bipolar disorders?Recently, some community surveys were carried out in the USA using the Mood Disorder Questionnaire. In the ensuing debate, one side claimed that the instrument was scarcely accurate when used in the general population, gave rise to numerous false positives and that the high prevalence reported was therefore a mere artefact. The other side defended the results reported by the research studies, on the basis that "positive" cases were homogeneous with regard to the high level of subjective distress, low social functioning and employment and with the high recourse to health care structures.It is quite probable that the problem lies at the root of the matter, in the definition of the gold standard.In the present state of our knowledge on course and response to treatment, the current diagnostic thresholds applied for mixed states and hypomanic episodes seem to be unsatisfactory.It is inconceivable that the diagnostic gold standard should be determined only on the basis of a structured interview of patients alone. But unless there is clinical consensus on the diagnostic threshold for hypomania and mixed states, there can be no consensus on the findings of epidemiological research.