The article examines the association of disability and sociodemographic background with the type of social network in which elders are embedded and considers the implications of such associations for network continuity. METHODS. Secondary analysis of data from a national probability sample of Jewish adults aged 60 and older in Israel (N = 2,079) was used. Multinomial logistic regression analysis was employed to predict membership in different network types. RESULTS. Higher education predicted membership in friends networks. Female gender, Asian/African origin, low income, and disability were associated with neighbors networks. Family networks were predicted by Asian/African origin, less education, younger age, and disability. Restricted networks were associated with lower income and disability, and particularly by the interaction of age and disability. DISCUSSION. The findings suggest that network stability is most jeopardized when the oldest old become disabled. The social care needs of this latter group will continue to demand public attention.