The aim of this work was to determine the predictive value of 2 indicators of psychosocial isolation (living alone and feelings of loneliness) in elderly people. Twenty-two-year follow-up cohort study of 3,620 community-dwelling elderly people enrolled in the PAQUID study, a French prospective population-based study. Participants completed a battery of tests including the Centre for Epidemiological Studies Depression scale and a question regarding living conditions (living alone or living with another person). Feelings of loneliness were measured using the Item 14 ("I felt lonely") of the French version of the Centre for Epidemiological Studies Depression scale. Mortality rate was measured over a 22-year follow-up period after the baseline assessment visit. The risk of death was estimated using Cox proportional hazards models, adjusted for age, sex, and educational level. At baseline, 1,535 participants (42.4%) were living alone, and a total of 498 persons (13.8%) reported frequent feelings of loneliness (FoL). The participants reporting FoL were more frequently women (82.7%), and the mean (SD) age was 76.5 (7.1) years. At the end of the follow-up, 3,116 deaths (86.6%) occurred. Living alone and FoL were both independent predictors of death after 22 years of follow-up (hazard ratio, 1.14; 95% confidence of interval, 1.05-1.23; p = .001) and (hazard ratio, 1.20; 95% confidence of interval, 1.08-1.33; p = .001), respectively. No significant interaction was found between feelings of loneliness and living alone, suggesting that a joint exposure has only an additive effect (β = 0.08; relative risk = (0.85; 1.40); p = .48). The associations of both measures persisted after adjusting for health status. Adjusting for depression attenuated the predictive value of FoL. Living alone and FoL were independently associated with higher risk of mortality. These factors may be useful as readily available psychosocial measures to identify vulnerability in community-dwelling older adults.