The possible effect of caffeine as an enhancer of cognitive performance, particularly that on abstract reasoning, has never been studied in an epidemiological setting, especially in relation to -163C>A polymorphism of CYP1A2 gene, largely controlling caffeine metabolism. Aim of this study was to ascertain whether in general population free chronic caffeine intake modifies abstract reasoning, and if this effect is influenced by the above mentioned genotype, by age, schooling, ethanol intake and smoking habits.We studied 1374 unselected men and women aged 51 ± 15 years (range 18-89) from a general population. Daily caffeine intake deriving from coffee, tea, chocolate or cola was calculated from an anamnestic questionnaire and from a 7-day dietary diary. Abstract reasoning was measured in the frame of a neuropsychological assessment as the ability to find a concept linking two words indicating objects or actions and explaining how they were connected.In age-schooling-adjusted linear regression, the higher the caffeine intake, the better the abstraction score. Abstract reasoning depended on caffeine in the -163C>A CC homozygous only (so-called slow metabolizers), where it was higher in the 3rd tertile of caffeine intake. Age and ethanol reduced while smoking and schooling enhanced this association. The interaction term between caffeine and the -163C>A polymorphism was accepted in linear regressions. Caffeine consumption resulted innocuous for the A-carriers (so-called fast metabolizers).In general population, a positive association between caffeine intake and abstract reasoning exists in the CC homozygous of the -163C>A polymorphism of CYP1A2 gene.