This study investigated the characteristics associated with smoking during pregnancy. A total of 7836 pregnant women were surveyed between 1992 and 1999 in England. Of these 27% were smoking during pregnancy. Pregnant women were more likely to smoke if they were less educated, living in rented accommodation, in unskilled manual or unemployed groups, and single or had a partner who smokes. Analysed by logistic regression, whether or not a pregnant women has a partner and, if so, his smoking status, was by far the biggest predictor of the pregnant woman's current smoking status. Thus, compared with women with partners who never smoke, the odds ratio (OR) of smoking during pregnancy for women with a partner who smokes was 2.3 (1.9-2.7) while those with no partner had an OR of 4.8 (3.8-6.0). For women exposed to passive smoke at home or at work the OR was 2.5 (2.1-3.0). Housing tenure was the most important socio-demographic predictor. Thus pregnant women living in rented council housing were nearly twice as likely (OR 1.93;1.63-2.29) as those buying their own home to be current smokers. The number of women who continued to smoke during pregnancy increased 10-fold from the least to the most deprived group. These findings highlight the importance of tobacco control strategies that address pregnant women's life circumstances and broader inequalities as well as those that focus on individual smoking behaviour.