The purpose of this study was to develop and assess a causal model of loneliness among a national probability sample of noninstitutionalized elderly adults. Path analysis was used to test the causal model. The findings indicated that marital status, self-perceived health status, income, social activity levels, and a sense of social fulfillment directly predict feelings of loneliness. The social fulfillment variable, however, was the most important predictor. An interesting finding was that age had no notable direct or indirect effect on feelings of loneliness.