Married adults are generally healthier than unmarried adults. It has been hypothesized that marriage is associated with good health because marriage has beneficial effects on health (marriage protection effects) and/or because healthier individuals are more likely to marry and to stay married (marriage selection effects). To investigate these hypotheses, this study analyzes prospective panel data for a large national sample of women in the U.S. (the National Longitudinal Surveys of Young Women). The women were aged 24-34 yr at the beginning of two successive five-year follow-up intervals. Analyses of the prospective data indicate that there were significant marriage protection effects, but only among women who were not employed. Specifically, for women who were not employed, married women had better health trends than unmarried women in each follow-up interval. It appears that marriage had beneficial effects on health for women who did not have a job which could provide an alternative source of financial resources and social support. In addition, analyses of the prospective data provide limited evidence for marriage selection effects. Specifically, women who had better health initially were more likely to marry and less likely to experience marital dissolution, but only for women who were not employed full-time and only during the first follow-up interval. Thus, the prospective evidence suggests that, for women who were not employed, both marriage protection and marriage selection effects contributed to the marital status differential in health observed in cross-sectional data. In contrast, neither marriage protection nor marriage selection effects were observed for women who were employed full-time. As would be expected, the cross-sectional data show that marital status differentials in health were large and highly significant for women who were not employed, whereas marital status differentials in health were much smaller and often not significant for employed women. Women who were neither married nor employed had particularly poor health. Additional evidence indicates that the women who were neither married nor employed suffered from multiple interacting disadvantages, including poor health, low incomes, and sociodemographic characteristics which contributed to difficulty in obtaining employment.