The relationship between mortality and marital status has long been recognized, but only a small number of investigations consider also the association with cohabitation status. Moreover, age and gender differences have not been sufficiently clarified. In addition, little is known on this matter about the Italian elderly population. The aim of this study is to examine differentials in survival with respect to marital status and cohabitation status in order to evaluate their possible predictive value on mortality of an Italian elderly cohort. This paper employs data from the Italian Longitudinal Study on Aging (ILSA), an extensive epidemiologic project on subjects aged 65-84 years. Of the 5376 individuals followed-up from 1992 to 2002, 1977 died, and 1492 were lost during follow-up period. The baseline interview was administered to 84% of the 5376 individuals and 65% of them underwent biological and instrumental examination. Relative risks of mortality for marital (married vs. non-married) and cohabitation (not living alone vs. living alone) categories are estimated through hazard ratios (HR), obtained by means of the Cox proportional hazards regression model, adjusting for age and several other potentially confounding variables. Non-married men (HR=1.25; 95% CI: 1.03-1.52) and those living alone (HR=1.42; 95% CI: 1.05-1.92) show a statistically significant increased mortality risk compared to their married or cohabiting counterparts. After age-adjustment, women's survival is influenced neither by marital status nor by cohabitation status. None of the other covariates significantly alters the observed differences in mortality, in either gender. Neither marital nor cohabitation status are independent predictors of mortality among Italian women 65+, while among men living alone is a predictor of mortality even stronger than not being married. These results suggest that Italian men benefit more than women from the protective effect of living with someone.