A longitudinal study based on consecutive admissions to all residential homes for the elderly was conducted in the industrial city of Mannheim (Germany) and in the London borough of Camden (England). Inclusion criteria were that the elderly persons (65 years old and older) came directly from their own home or, if transferred from a hospital, had been there for less than 3 months. At each site, 60 home residents were interviewed at admission and 3 months and 8 months later. Depression and dementia were assessed with the aid of the Brief Assessment Scale. The prevalence of depression (Mannheim: 34.6%; Camden: 47.9%) was already high at admission and did not change significantly over time. Residents in Camden were more demented and more impaired in their activities of daily living at the time of admission, and the percentage of those who died or were transferred to a hospital or nursing home within 8 months thereafter was higher in Camden (30%) than in Mannheim (5%). Multiple regression analysis revealed that, in both study areas, depression at baseline was the best predictor for depression 3 months and 8 months later. This relationship was particularly strong in Camden, where a high percentage of the depressed at admission showed a chronic course of illness. Sex, age, home visits, social isolation, activities of daily living, cognitive impairment, and somatic symptoms at the time of admission were not significantly associated with depression 3 months later. Eight months after admission, a similar pattern was found in Mannheim. In Camden, however, in addition to depression, a lack of home visits by relatives and friends, and somatic symptoms at baseline, were significant predictors of depression 8 months after admission.