In a previous analysis from the Alameda County Study, it was observed that although men had higher heart disease mortality rates than women, there was no male excess in the prevalence of self-reported heart disease morbidity at baseline or in new reports of morbidity 9 years past baseline. This apparent contradiction might occur because women report less severe heart disease than men. In the present study, this hypothesis was evaluated by examining whether self-reported heart trouble was more strongly associated with subsequent heart disease mortality for men than for women in a representative sample of the population of Alameda County, California, selected in 1965 and followed for mortality for 19 years (n = 3,742). In a time-dependent Cox model, self-reported heart trouble was a stronger predictor of heart disease mortality for men, but only during the early years of follow-up (p = 0.00). This effect was due to a shorter time to death for men who reported heart trouble. The relative hazard for men reporting heart trouble was 6.6 (95% confidence interval (CI) 3.7-11.6) at baseline, declining to 3.2 (95% CI 2.2-4.5) by 5 years past baseline and 1.5 (95% CI 0.9-2.5) by 10 years past baseline. Self-reported heart trouble was a consistent predictor of subsequent heart disease mortality for women over the 19-year follow-up period (relative hazard = 2.0, 95% CI 1.4-2.8). Sex differences in the prognosis of self-reported heart trouble were masked in non-time-dependent analyses. These results illustrate that consideration of time dependence may be required for meaningful analysis of long-term cohort studies. Possible explanations of the shorter time to death for men who reported heart trouble are discussed.