A review of epidemiological studies has been undertaken to examine the relationship between limited education and the risk of developing Alzheimer's disease in later life. On the basis of a large number of international studies of prevalence and incidence, as well as clinic- and community-based case-control studies, it seems that the relationship is strongest when the diagnostic criteria applied emphasize 'cognitive impairment'. It is least evident in longer-term prospective studies of incidence based upon clinically verified caseness. Education is considered as a socializing process promoting certain lifelong learning strategies, encouraging children to develop forms of decontextualized thinking. These learned habits of 'thinking on demand' enable educated people to adapt to and perform more competently on the neuropsychological and mental state tests that make up screening batteries and case ascertainment methodologies in epidemiological research on Alzheimer's disease. As a result, surveys of elderly poorly educated populations in both developed and developing countries may lead to significant over-diagnosis of clinical dementia-and by implication a possible under-diagnosis of dementia in relatively well-educated populations.