A community sample made in the fall of 1972 in Durham County, North Carolina, and that consisted of 331 persons 65 years of age and older was assessed for adequacy of social support along three parameters: roles and available attachments, perceived social support, and frequency of social interaction. This approach to the assessment of social support complements previous approaches to the measurement of social support which have not compared with the various components of the construct. Mortality status, the outcome variable, was determined 30 months after the initial assessment. The crude relative risks of mortality were 1.96 for impaired roles and available attachments, 3.86 for impaired perceived social support, and 2.72 for impaired frequency of social interaction. Ten potential confounding variables-age, sex, race, economic status, physical health status, self-care capacity, depressive symptoms, cognitive functioning, stressful life events, and cigarette smoking-were controlled using binary linear regression analysis. The estimates of relative mortality risk were 2.04, 3.40, and 1.88, respectively, for impaired roles and available attachments, impaired perceived social support, and impaired frequency of social interaction when the control variables were included in a regression model. Therefore, these three parameters of social support significantly predicted 30-month mortality in both crude and controlled analyses in a community sample of older adults.