An independent, inverse association between cognitive function and all-cause mortality has been reported in elderly cohorts. The purpose of this study was to determine whether the same association exists in middle-aged persons. The Atherosclerosis Risk in Communities Study is a cohort study initiated in 1987 to investigate the development of atherosclerosis in middle-aged persons. Three cognitive function measures were included in the second cohort examination conducted from 1990 to 1992 when the participants were aged 48-67 years: the Delayed Word Recall Test (DWRT), the Digit Symbol Substitution Test (DSST) (a subtest from the Wechsler Adult Intelligence Scale-Revised), and the Word Fluency Test from the Multilingual Aphasia Examination. Cox proportional hazards modeling was used to determine whether all-cause mortality ascertained through 1997 was associated with each measure after adjustment for sociodemographic, biologic, psychologic, and behavioral risk factors. Without adjustment, there was a significantly lower mortality hazard associated with higher scores on all three measures. After covariate adjustment, the hazard ratios for the DWRT and the DSST remained significant (hazard ratio1-point DWRT score increment = 0.90, 95% confidence interval: 0.84, 0.97; hazard ratio 7-point DSST score increment = 0.86, 95% confidence interval: 0.80, 0.93). Cognitive function measured in middle age appears to have prognostic importance for life expectancy similar to that reported in elderly adults.