Reported Relative Risks associated with smoking differ between studies; these differences may reflect true biological differences between populations or may be research artifacts introduced by differences in factors such as amount smoked or smoking duration. The authors reviewed the literature published before June 1992 on relative risks associated with smoking for heart disease, stroke, lung cancer, and chronic obstructive lung disease. They quantified the effect of variables such as age, amount smoked, and smoking duration on reported relative risks. The main reasons for the variation in reported relative risks were: misclassification of former smokers as never smokers, the use of mortality rate ratios rather than incidence rate ratios, a possible period effect suggesting increasing relative risks over time, and differences in the amounts smoked. It is far more likely that these factors are responsible for the observed variation between studies than that the variations reflect true biological differences between populations. Using relative risks from other studies is therefore justified in calculating a population attributable risk if the studies are carefully selected and address factors such as amount smoked and period effects.