Identical measures of social ties obtained from three community-based cohorts aged 65 and over from East Boston, MA; New Haven, CT; and two rural counties in Iowa permit the first direct cross-community comparison of the hypothesis that social isolation increases 5-year mortality risks (1982 to 1987) for older men and women. In sex-specific proportional hazards analyses, social ties were significantly and inversely related to mortality independently of age in all three cohorts (e.g., relative hazard (RH) = 1.97 to 3.06 for men and women, comparing those with no ties to those with four types of ties). After controlling for age, pack-years of smoking, body mass, chronic conditions, angina, and physical and cognitive disability, social ties remain significant predictors of mortality risk for the men and women in New Haven (RH = 2.4 and 1.8) and for women in Iowa (RH = 1.9). For the men in Iowa (RH = 1.4) and the men and women in East Boston (RH = 1.0 and 1.3), the associations are weaker and nonsignificant.