Background. The long-term course of seasonal affective disorder has not been well studied. Method. Using the Structured Clinical Interview for DSM-III-R, we interviewed 75% of a sample of 124 subjects diagnosed from five to eight years previously as fulfilling DSM-III-R criteria for recurrent major affective disorder, seasonal pattern. Results. In the follow-up period, 38% of the sample continued to fulfil DSM-III-R criteria for seasonal illness; 28% had recurrent major depressive disorder, but no longer displayed a seasonal pattern; 18% were completely well with no further depression: 6% had subsyndromal symptoms; and 5%, although not meeting DSM-III-R criteria for seasonal illness, were still displaying constant periodicity. A short duration of index episode and a high frequency of illness predicted a continuing seasonal course of illness. Conclusion. Diagnostic criteria for seasonal affective disorder need to be further refined, possibly restrictively, if they are to be used to predict the future course of seasonal illness.