To examine the association and interaction between cognitive impairment and social support and mortality. Prospective cohort study. Fifty-three family practices in the United Kingdom. Community-dwelling people (aged > or =75) participating in the Medical Research Council Trial of the Assessment and Management of Older People in the Community (10,720 individual subjects analyzed). Cognition was measured using the Mini-Mental State Examination. Social engagement was assessed using categorical data on marital status, living situation, availability of assistance, availability of a confidant, and frequency of social contact. The primary outcome was all-cause mortality. The prevalence of cognitive impairment was 13.0% (mild) and 2.0% (moderate to severe). In Cox survival models (fully adjusted for physical health, lifestyle, daily function, and depression), there was a consistent association between greater cognitive impairment and mortality risk (mild cognitive impairment, hazard ratio (HR)=1.31, 95% confidence interval (CI)=1.21-1.40; moderate to severe cognitive impairment, HR=1.64, 95% CI=1.41-1.93. Mortality risk was greater in the medium (HR=1.09, 95% CI=1.02-1.16) and low social engagement groups (HR=1.17, 95% CI=1.05-1.29) than in those with the highest level of social engagement. Lower social engagement did not increase mortality risk in those who were more cognitively impaired. Cognitive impairment and social support are independent risk factors for mortality. Interventions that promote early identification and management of cognitive impairment and enhance social support for older people may decrease mortality and produce public health benefits.