All patients suffering from affective psychoses (ICD 296) who were admitted to the Psychiatric University Clinic of Zurich between 1959 and 1963 were studied in a follow-up investigation until 1975. Of 254 affective psychoses, 95 were bipolar patients (37.4%) and 159 were monopolar (62.6%). The sample of bipolar patients was complemented with all patients who had been admitted in the period 1959--1963 because of manic or mixed manic-depressive syndromes. This paper describes the change of diagnosis in the two diagnostic groups. In 10% (N = 20) of monopolar depression cases there was a change of diagnosis to bipolar affective illness. An analysis shows that the diagnosis of patients with three or more depressive episodes (unipolar depressives) was especially prone to change. A mathematical correction of some diagnostic errors leads to the conclusion that the ratio of unipolar depression to bipolar illness may be about 1:1. A major source of diagnostic error lies in the change of affective to schizo-affective illness. Up to now, no clinical criterion exists that would exclude this error, which was found in 6% (n=12) of the monopolar but also in 7.5% (n = 3) of the bipolar index patients. It is recommended that studies of affective disorders should be based on truly representative samples of the illness, including patients with one or two episodes, and that the term 'unipolar depression' be used synonymously with the term 'monopolar depression,' originally created by Kleist (1947) and Leonhard (1957).