Knowledge of the course and outcome of major affective illness has clinical as well as theoretical implications. In understanding the pathophysiology of the major affective disorders, an essential question in the interplay between biological, psychological and social factors is whether the individual is changed biologically by experiencing an affective episode or not. A biological change may be reflected in a changed risk of experiencing new episodes and changed chances of recovery from these episodes for the individual, and may possibly also be reflected in persisting altered cognitive function as an expression of brain function affected during a longer period. Previous studies of the course of affective episodes are flawed by a number of drawbacks such as various definitions of recovery and recurrence, various kinds of bias and confounders, low statistical power, and statistical analyses conducted without survival models and without paying attention to diagnostic instability or the individual heterogeneity of the course of episodes. Totally, these drawbacks and pitfalls affect the results of previous studies in unpredictable ways and make it hazardous to draw conclusions about the effect of prior affective episodes on the subsequent course of unipolar and bipolar disorder. The present thesis avoided most of these pitfalls or adjusted for them in analyses of hospital data from the Danish Psychiatric Central Register, collected nationwide from 1971 to 1993. Hospitalisation was used as an expression of an affective episode. On average, a progressive course with increasing risk of recurrence with every new episode was found for unipolar and bipolar affective disorders. Initially, the two types of disorders followed markedly different courses, but later in the course of the illness the risk of recurrence was the same for the two disorders. However, analyses with frailty models revealed that for unipolar men, this progressive course was due to a subgroup of patients who had a great liability to recurrence from the beginning of the illness, whereas the remainder of the unipolar men did not have a progressive course of episodes. Initially in the course of affective disorders, socio-demographic variables such as gender, age at onset, and marital status and co-morbidity with alcoholism acted as risk factors for further recurrence. Later, however, particularly variables related to the previous course of illness played a role. The chances of recovery from an episode were found not to change during the course of unipolar or bipolar disorder. In contrast, a review of studies from the era before active treatment revealed that the duration of untreated episodes seemed to increase during the course of illness. Further case register studies and a clinical follow-up study by the author showed in accordance with previous studies that unipolar and bipolar affective disorders seem to be associated with increased risk of developing cognitive impairment and dementia and that the risk seems to increase with the number and the frequency of affective episodes. Overall, the thesis illustrates: 1) that case register studies can supplement prospective clinical studies in longitudinal research of major affective disorders: 2) the importance of taking the number of episodes into account in the analyses of data: 3) that affective episodes persistently increase the liability to illness for the individual, and that a cerebral component seems to bee present in some patients with major affective disorder: 4) that the treatment methods used so far seem to prevent the progression in the duration of affective episodes but not the progression in the frequency of episodes. It seems that among the existing pathophysiological models the sensitisation and kindling paradigm is the one that fits these epidemiological data best. Studies for future epidemiological research on the course of affective disorder are suggested.