BACKGROUND: Studies in mixed-aged populations show differences between the predictors of a relapse and those of a long-term course of depression, supporting the hypothesis about similar differences among the aged. AIM: The aim was to identify the factors predicting or related to a relapse of depression among the Finnish elderly having recovered from depression during treatment. MATERIAL AND METHODS: The population consisted of 70 depressed (DSM-III criteria) elderly (60 yr-) Finns having recovered from depression during treatment as determined 15 months after baseline. By the 4-year follow-up after the recovery, 20 patients had relapsed and 50 persons were non-depressed. RESULTS: The logistic regression model showed major depression and psychomotor retardation to be independent predictors. Relapses were not related to stressors in life or physical illnesses occurring during the follow-up. CONCLUSIONS: Major depressive elderly patients have a high risk for relapses without the occurrence of the stressors or physical illnesses. In clinical practice, major depressive elderly patients should be followed up in order to detect and treat potential relapses as early as possible. Cooperation between psychiatrists and general practitioners is needed in the follow-up. Theoretically, the results suggest the assumption of a biochemical aetiology of major depression.