Studies on the relationship between depression and mortality in elderly community populations have yielded contradictory findings, although an association frequently is found in studies of elderly psychiatric patients. These different results may be due to differences in the measures of depression, the populations under study, the covariates in the analysis, or to sample attrition. In this study of elderly residents of an urban neighborhood, depressive symptoms are measured at two time points. People are classified as consistently nonsymptomatic (N-N), with emergent symptoms (N-D), in remission (D-N), or persistently symptomatic (D-D). Symptoms of depression, sociodemographic characteristics, and measures of changes in health, functional status, number of chronic medical conditions, and social support are examined in relation to mortality in multivariate cox regression models. Although symptoms of depression are not found to be related to time-to-death, older people, those with declines in health and functional status, and men have greater relative risks of mortality over a three-year follow-up