Although smoking has been linked to various causes of death, there is no systematic account of the underlying and multiple cause-of-death distributions associated with various smoking statuses. We analyze such patterns by age and gender for the USA in 1986. Our study is based on a one-percent random sample of decedents 25 and over in the USA for whom survey data from informants were linked to death certificate data. Smoking is related to several underlying causes of death, the most common being circulatory diseases. Lung cancer is less prevalent than circulatory diseases or other cancers among ever smokers. Multiple medical conditions are common for both smokers and nonsmokers, but particular combinations vary among persons with different smoking statuses. Former smokers who quit soon before death and were under frequent medical care are most likely to have had lung cancer. Amount of smoking is tied to variations in cause-of-death patterns. Differences by age and gender are not substantial, although other cancers appear frequently for both smokers and non-smokers among women. The distribution of medical causes of death for ever smokers is not radically different from that of never smokers. However, differences in cause patterns are seen when smoking statuses are detailed by amount of smoking and timing of quitting. These similarities and differences in cause patterns must be related to the fundamental fact that the average smoker will die earlier than the average nonsmoker. Such findings should especially influence programs for diseases whose links to smoking have been underestimated.