The relationship between social ties, stage of disease, and survival was analyzed in a population-based sample of 525 black and 486 white women with newly diagnosed breast cancer. There were significant differences between the two race groups in reported social ties. Using logistic regression to adjust for the effects of age, race, study area, education, and the presence of symptoms, there was little or no evidence for an association between individual network measures of social ties and stage of disease. However, a summary measure of social networks was found to be associated modestly with late stage disease, attributable in part to significantly more advanced disease among black, but not white, women reporting few friends and relatives [relative risk (RR) = 1.8; 95% confidence interval (CI) = 1.1-3.0]. With adjustments for differences in stage of disease and other covariates, and with the use of Cox proportional hazards modeling to estimate hazard ratios, the absence of close ties and perceived sources of emotional support were associated significantly with an increased breast cancer death rate. White women in the lowest quartile of reported close friends and relatives had twice the breast cancer death rate of white women in the highest quartile (RR = 2.1; 95% CI = 1.1-4.4). Notably, both black and white women reporting few sources of emotional support had a higher death rate from their disease during the 5-year period of follow-up (RR = 1.8; 95% CI = 1.3-2.5).(ABSTRACT TRUNCATED AT 250 WORDS)