A reserve hypothesis suggests that clinical symptoms of Alzheimer's disease (AD) begin earlier in individuals with less education. Therefore, patients with less education might survive longer after diagnosis than those with more education. Two hundred forty-six subjects with probable AD were followed for 1 to 4 years. There were 78 deaths; 30 deaths occurred in the 127 patients whose education was < or = 8 years, while 48 deaths occurred in the 119 patients with > 8 years of education. Cox proportional hazards models adjusted for age, gender, and clinical dementia rating (CDR) showed that patients with more education had increased mortality (continuous variable: RR = 1.06 for each year of education; 95% confidence interval [CI], 1.01-1.11; dichotomous variable at 8 yr: RR = 1.76; CI, 1.11-2.77). This observation might at first seem counterintuitive, since groups with lower socioeconomic status are often at greater mortality risk. It implies that at any level of assessed clinical severity, the underlying pathology of AD is more advanced in patients with more education, resulting in shorter duration of diagnosed disease before death. These findings suggest either that education systematically influences global ratings of disease severity or that education provides a reserve against the clinical manifestation of AD pathology.