Overall evidence from 19 case-control studies of Alzheimer's disease (AD) and smoking shows a highly significant (p < 0.001) negative association [ever/never smokers, relative risk (RR) 0.64, 95% confidence interval (CI) 0.54-0.76]. Some studies have apparent design faults but the association is clearly evident in those which do not (RR 0.60, 95% CI 0.46-0.78). A report of a positive relationship from a prospective study can be dismissed due to unreliability of AD diagnosis on death certificates, and other study weaknesses. Although more data are needed on dose response and to rule out possible confounding, the negative association is consistent with other data suggesting nicotine protects against AD.