In this study, the authors evaluated whether the association between low educational level and increased risk of Alzheimer's disease (AD) and dementia may be explained by occupation-based socioeconomic status (SES). A cohort of 931 nondemented subjects aged > or = 75 years from the Kungsholmen Project, Stockholm, Sweden, was followed for 3 years between 1987 and 1993. A total of 101 incident cases of dementia, 76 involving AD, were detected. Less-educated subjects had an adjusted relative risk of developing AD of 3.4 (95% confidence interval: 2.0, 6.0), and subjects with lower SES had an adjusted relative risk of 1.6 (95% confidence interval: 1.0, 2.5). When both education and SES were introduced into the same model, only education remained significantly associated with AD. Combinations of low education with low or high SES were associated with similar increased risks of AD, but well-educated subjects with low SES were not at high risk. Low SES at 20 years of age, even when SES was high at age 40 or 60 years, was associated with increased risk; however, this increase disappeared when education was entered into the model. In conclusion, the association between low education and increased AD risk was not mediated by adult SES or socioeconomic mobility. This suggests that early life factors may be relevant.