In a recent contribution to this journal, Sterling and Weinkam discuss the importance of the confounding of occupation and smoking when considering the contribution of smoking to mortality risk. In this report we provide estimates of the potential magnitude of this effect using data from a cohort study. It is shown that failing to consider possible confounding by socio-economic factors can lead to over-estimating of the strength of the relationship between smoking and mortality. The implications of these findings for the interpretation of results derived from observational studies are discussed.