Seventeen-year mortality data from the Alameda County Study are used to examine the relative importance of social ties as predictors of survival at different ages, ranging from 38-94 years at baseline. Previous analyses of Alameda County data by Berkman and Syme (Am J Epidemiol 1979;109:186-204) have shown that such ties are significant predictors of lower nine-year mortality risk for persons aged less than 70 years at baseline. Proportional hazard analyses indicate that social ties are also significant predictors of lower 17-year mortality risks for those aged 70 and older after adjusting for age, sex, race, baseline health status, perceived health, depression, and health practices (relative hazard = 1.49 for Berkman-Syme Social Network Index; 95% confidence interval (CI) = 1.09-2.05). Comparisons of the relative importance of four types of social ties reveal an interesting shift across the age groups. Marital status assumes primary importance for those aged less than 60 years at baseline (relative hazard = 1.6 and 1.4 for those aged 38-49 and 50-59, respectively; 95% CI = 1.12-2.29 and 1.02-1.91, respectively). However, ties with close friends and/or relatives assume greater importance for those aged 60 and older (relative hazard = 1.17 comparing those reporting five or more contacts per month to the more socially isolated who report less than five such contacts per month; 95% CI = 0.98-1.89).