Purpose: The purpose of this study was to assess whether there is an enduring gender difference in the ability of self-rated health to predict mortality and investigate whether self-reported physical health problems account for this difference. Design and Methods: Cox models for 4-year survival were fitted to data from successive cohorts aged 75-81 years registered with a primary care practice in the U.K. Midlands surveyed in 1981, 1988, and 1993-1995. Results: Self-rated health was consistently a stronger predictor in men (hazard ratio [HR] = 2.7; 95% confidence interval [CI] = 2.1-3.5) than it was in women ([HR] = 1.9; 95% CI = 1.5-2.4). Women surveyed in 1993-1995 were more likely than men to report problems that were disabling but not life-threatening, whereas men were more likely to report potentially life-threatening problems. However, these differences did not explain the association of self-rated health with mortality. More than half of those who reported a potentially life-threatening problem said that their health was good. Implications: Self-rated health is more strongly associated with mortality in men, but this is unlikely to be explained by differences in the nature of their physical health problems.