This study examines age-related differences in the association between social network characteristics and mortality for aged white women. Subjects include a community-dwelling sample of white women aged > or = 65 years (n = 806), who lived in northeast Baltimore, Maryland, in 1984. Three characteristics of social networks were measured: availability of network resources; contact with network resources; and integration into the neighborhood. The association of social network with 5-year mortality was examined with a proportional hazards model adjusting for perceived health status, impairment in physical activities of daily living, number of chronic conditions, and years of education. Analyses were stratified by age (65-74 years, > or = 75 years). Elements of social network contact and neighborhood integration were associated with reduced mortality among women aged > or = 75 years, but not among women aged 65-74 years. In the > or = 75 years group, women who had no contact with children, friends, and group organizations showed hazard ratios (95% confidence intervals (CI)) of 3.1 (1.2-7.5), 2.2 (1.0-4.9), and 2.8 (1.2-6.5), respectively. Women who had lived < or = 10 years in the neighborhood and women who had no interaction with local merchants showed hazard ratios of 2.5 (95% CI 1.3-4.8) and 2.2 (95% CI 1.2-3.9), respectively. Thus, both age and specific aspects of network structure were found to influence the association between social networks and mortality in elderly women.