Caffeine is the most widely consumed behaviourally active substance in the western world. Neuroprotective effects of caffeine in low doses, chronically administered, have been shown in different experimental models. If caffeine intake could protect against neurodegeneration in Alzheimer's disease (AD), then higher levels of caffeine consumption in normal subjects as compared with AD patients should be detectable in the presumably long period before diagnosis when insidious pathogenic changes are taking place. A case–control study was used: cases were 54 patients with probable AD fulfilling the National Institute of Neurologic and Communicative Disorders and Stroke and the AD and Related Disorders Association criteria, in a Dementia Clinics setting. Controls were 54 accompanying persons, cognitively normal, matched for age (±3 years) and sex. Patients with AD had an average daily caffeine intake of 73.9 ± 97.9 mg during the 20 years that preceded diagnosis of AD, whereas the controls had an average daily caffeine intake of 198.7 ± 135.7 mg during the corresponding 20 years of their lifetimes (P < 0.001, Wilcoxon signed ranks test). Using a logistic regression model, caffeine exposure during this period was found to be significantly inversely associated with AD (odds ratio=0.40, 95% confidence interval=0.25–0.67), whereas hypertension, diabetes, stroke, head trauma, smoking habits, alcohol consumption, non-steroid anti-inflammatory drugs, vitamin E, gastric disorders, heart disease, education and family history of dementia were not statistically significantly associated with AD. Caffeine intake was associated with a significantly lower risk for AD, independently of other possible confounding variables. These results, if confirmed with future prospective studies, may have a major impact on the prevention of AD.