A relatively high prevalence and incidence of dementia have been found in population strata with low levels of education in comparison to population strata with high levels of education. However, doubt remains whether this may be an artifact of education bias in the screening tests used. To investigate this matter, we analyzed results of two Dutch population surveys in which unbiased measures of memory decline were used. In the Longitudinal Aging Study Amsterdam (n = 1774) the percentage of words retained in a verbal learning test was found to be disproportionately low in the oldest age cohort (80-85 years) with less than 11 years of education. The Amsterdam Study of the Elderly (n = 4051) found a "dose-response" relationship between education and dementia prevalence. Cross-sectional and longitudinal results showed that, in less educated people, memory decline is faster and sets in at an earlier age. These findings indicate that the relationship between dementia and education is not just an artifact of case detection methods.