Lung cancer area mortality rates for the period 1980-1983 in England and Wales followed the pattern observed for previous years, with high rates concentrated in urban districts and low rates in remote rural districts. Using data for 401 local government districts, estimates of smoking prevalence were made from socio-economic distributions and regional smoking variations. Confounding effects of migration and errors in mortality records were investigated by comparing the pattern of recorded lung cancer mortality with that of all other cancers. Holding these variables constant by statistical means reduced the correlation between lung cancer SMRs and population density from 0.65 to 0.53 for males and from 0.54 to 0.46 for females, so a large proportion of the urban-rural gradient remained unexplained. It is likely that a combination of several effects rather than a single urban risk factor is responsible for the urban concentration of lung cancer mortality.