We used a model, originally applied in the context of occupational cohort studies to evaluate confounding from smoking, to estimate lung cancer death rates in the Italian population not attributable to active smoking (background rates). Relevant parameters were the percentages of current and former smokers in the population, as derived from national interview surveys, and rate ratios for lung cancer and smoking. We studied differences over time (1956-1958 to 1987-1989) and place (urban/rural) in the background lung cancer death rates while allowing for changes in the average daily amount of tobacco consumption and in duration of smoking. We applied two functions proposed by Whittemore to obtain rate ratios for smokers: one assumes a linear relation between cumulative number of cigarettes smoked and lung cancer death rates; the other is closer to the multistage carcinogenesis theory. All our calculations indicate that the estimated background rates increased in Italy from the period 1956-1958 to the period 1987-1989; the increase was stronger in males than in females. In 1980-1982, higher background rates were observed in heavily urbanized areas than in rural areas. The results did not change when we considered cohort effects in smoking prevalence and when the key parameters were perturbed in a sensitivity analysis. It appears that factors other than smoking play an important role in causing lung cancer in Italy.