The association between dementia and education was studied in 143 twin pairs discordant for dementia, using a matched-pair design, and in 221 dementia cases and 442 unrelated controls from the same twin registry, using a case-control design. Low education was defined as 6 years or less of schooling. Case-control analyses with prevalent cases showed low education to be a risk for Alzheimer's disease but not dementia in general. Low education did not significantly predict incident cases. In the matched-pairs analysis, which controls for genetic and other familial influences, differences in education between demented twins and twin partners were not statistically significant. However, for Alzheimer's disease, odds ratios resulting from matched pairs and case-control analyses were similar. Twins' comparative reports about intellectual involvement earlier in their lives suggest a long-standing difference on this dimension, with less involvement by the twin who became demented.