Cognitive impairment is associated with increased mortality rates in late life, but it is unclear whether worse cognition predicts working-age mortality. The data come from a U.S.national survey (N=3973 aged 32-84 at cognitive testing in 2004-06, mean age 56.6, 56.3% female; N=3055 retested in 2013-18 at ages 42-94, mean age 64.6, 56.6% female; mortality follow-up through 2019). We use Cox hazard models to investigate whether cognition is associated with mortality below age 65, how the magnitude of this risk compares with the risk in later life, and whether the association persists after adjusting for potential confounders. Worse cognition is associated with mortality, but the demographic-adjusted hazard ratio (HR) diminishes with age from 2.0 per standard deviation (SD; 95% confidence interval [CI], 1.7-2.4) at age 55-1.4 (95% CI, 1.3-1.6) at age 85. In the fully adjusted model, the corresponding HRs are 1.4 (95% CI, 1.2-1.7) and 1.3 (95% CI, 1.1-1.4), respectively. The absolute differences in mortality by level of cognition, however, are larger at older ages because mortality is rare at younger ages. The fully adjusted model implies a 2.7 percentage point differential in the estimated percentage dying between ages 55 and 65 for those with low cognition (1 SD below the overall mean, 5.7%) versus high cognition (1 SD above the mean, 3.0%). The corresponding differential between ages 75 and 85 is 8.4 percentage points (24.6% vs 16.2%, respectively). Cognitive function may be a valuable early warning sign of premature mortality, even at working ages, when dementia is rare.