Previous studies indicated that unmarried persons are subject to higher mortality than the married, and that the differentials are more marked for male than for females. There are two major approaches to explaining the marital status differentials in mortality ; selection function and protection function of marriage. Following protection fucntion, this study develops the new "instrumental / expressive sex-role" hypothesis in order to explain why marriage protects males more against death. The hypothesis expects that male"s instrumental role and female"s expressive role have direct effect as well as indirect effect through social integration on sex differential mortality by marital status. for the hypothesis testing, Korea and US vital statistics and census data are used to compute age-specific , age-adjusted mortality rates and their ratios for persons in different marital status. Major findings are as follows. 1)For both Korea and US being married is more advantageous to males than females, ad being widowed, divorced, and separated is more disadvantageous to males, while being never-married is more disadvantageous to females, 2) For Korea, the never married men and women have the highest mortality rates, 3) For US the never married women have the highest mortality rate, while the divorced, separated, and widowed men have the highest mortality rate. Fro both Korea and US data, selection function is rejected, but instrumental/expressive sex-role hypothesis succeeds in accounting for the sex and marital status differential in mortality.