The aim of this paper is to assess the influence of selected psychosocial factors as predictors of stroke incidence in a probability sample of noninstitutionalized elderly. The main psychosocial factor of interest was depression. Marital status, social support, social networks, and religiousness were also assessed as potential antecedent or mediating factors. The data were obtained from a prospective longitudinal study based on 2,812 individuals aged 65 years and over living in New Haven, Connecticut. The incidence of stroke was monitored from the baseline interview in 1982 until December 1988. Depression, measured by the Center for Epidemiologic Studies Depression Scale (CES-D), was measured at baseline as were other predictor variables. Univariate Cox regression analyses revealed that higher CES-D scores were predictive of greater stroke incidence (p < 0.05). More frequent attendance at religious services was associated with lower incidence (p < 0.001). CES-D scores were also correlated with many measures of sociodemographic, health, and physical function factors in our multivariate analysis (p < 0.05). When combined with other significant predictor variables such as age, sex, hypertension, diabetes, physical function, and smoking, neither depression nor religious attendance retained its significance.