Three social network measures were obtained from a randomly sampled cohort of 2603 HMO members in 1970-71. Mortality and first incidence of ischemic heart disease, cancer, stroke, and hypertension were assessed over the next 15 years. Outcome data were adjusted for age, sex, smoking, SES, and baseline health status. Persons with histories of a given morbidity at or prior to baseline were excluded from the analyses of that morbidity. Social network measures, particularly network scope (a measure of the number of different domains in which a person has social contacts), were powerful predictors of 15-year mortality hazard, but weak predictors of incident disease. Only network scope predicted IHD incidence, and none of the other morbidities was predicted by the social network measures. However, social network measures were strong predictors of both cause-specific and all-cause mortality among persons who had incident cases of IHD, cancer, and stroke. These data suggest that social networks may be more effective in supporting recovery after illness has occurred than in preventing the incidence of new disease.