Mortality data from the OPCS Longitudinal Study were used to determine whether the conventional classification of married women by their husband's occupation under-estimates the extent of social differences in lung cancer among this group. Differences existed for social class measures but alternatives based on housing tenure and car access defined socio-economic differences wider than any other previously recorded for England and Wales: married women living in rented housing and without a car were two and a half times as likely to die from lung cancer than those in owner occupied housing with access to a car. In 1957 and 1974 mothers of children included in the 1958 cohort study showed parallel socio-economic differences in smoking patterns as well as in uptake and cessation rates. Data from the General Household Survey for 1982 similarly suggest that cigarette smoking is more sharply differentiated using household rather than occupational measures of class. This suggests that wide differences in mortality are likely to persist through the eighties and beyond.