This study investigated the relationship between the stressfulness of each state's social environment, smoking, and mortality rates for respiratory cancer. It was based on a health behavior model which assumed that under conditions of high stress some people fail to exercise normal prudence in either protecting their health or engage in practices inimical to health. Two types of social stress were conceptualized and measured at the state level: one based on life changes requiring adaptation and measured by an index in which negative personal life events in 15 categories were aggregated for each state using macro measures, the second based on the idea of chronic stressful conditions and measured through the Measure of Status Integration. Smoking was measured by tobacco sales and by a survey of percent smoking by state. Both stressful events and stressful conditions were correlated with all indicators of smoking at the state level, 12 of 14 correlations being in the theoretically expected direction. Correlations were enhanced when age, urbanicity, the percentage of blacks, education, and low income were controlled. The data were compatible with a causal model suggesting that socially generated stress in states increases the level of smoking behavior which in turn soon leads to higher death rates for cancer of the respiratory system. Although stress and smoking have been previously linked at the individual level and the smoking/cancer link has long been established, this study appears to be the first to link living in stressful geographic locales with the increased health risk of respiratory cancer. (Author/AA)