Background: Depression in the elderly was found to be associated with a variety of risk-factors in cross sectional designs. Based on the vulnerability-stress model, etiologic pathways for depression have been suggested, with vulnerability modifying the effect of stress factors. The current prospective study tests an etiologic model for depression incidence, by assessing modifying effects of three types of vulnerability: genetic/familial vulnerability, organic vulnerability, and environmental vulnerability. Methods: 1940 non-depressed community-living elderly were interviewed at baseline, and at follow-up three years later. Bivariate and multivariate relationships between risk factors and incident depression (GMS- AGECAT) were studied. Results: Higher age, personal history of depression, death of spouse, health related factors and comorbid organic or anxiety syndrome showed significant bivariate associations with depression incidence. In multivariate analysis, the effect of stress factors on incident depression was not modified by a genetic/familial vulnerability, nor by an organic vulnerability. Effect modification by environmental factors was however evident; having a marital partner, and if unmarried having social support, significantly reduced the impact of functional disabilities on the incidence of depression. Limitations: The study consisted of two measurements with a three years interval, depressive episodes with a short duration may be under- represented. Conclusions: In the elderly, the effect of stress on incident depression is modified by environmental vulnerability. No evidence was found of effect modification by either genetic/familial or organic vulnerability. The results have implications for both recognition and treatment of late-life depression. (C) 2000 Elsevier Science B.V.