This study examined the prevalence and predictors of smoking by pregnant women attending a public antenatal clinic. The prevalence of smoking in this population (n=2577) was found to be 38.0% (95% CI 36.1-39.9%), A review of previous research investigating variables associated with smoking in pregnancy indicated that only three of 42 studies had used multivariate analysis, Using step-wise logistic regression analysis, five variables were found to be independent predictors of smoking in pregnancy: education (having 4 years or less high school), marital status (being unmarried), gravidity (being multigravida), age (being under 25 years) and language spoken at home (speaking English), The model correctly predicted 63.7% of cases, The knowledge and attitudes of pregnant smokers were also investigated using data from a sub-sample of consenting subjects. Three-quarters of the women claimed that they had reduced their smoking since discovering they were pregnant. However, their mean intake of 13.7 cigarettes daily remained at a hazardous level, Approximately half(51%) these smokers claimed to have tried to quit smoking in the current pregnancy, Most (61%) women said they believed smoking was definitely harmful to the unborn child, However, awareness and acceptance of specific risks were inadequate, Of the women in a current relationship, 72% said their partner was a regular smoker, Less than half(45%) the continuing smokers who had seen a doctor about their current pregnancy could recall being advised to stop smoking, There is a need for health care providers to adopt a more systematic and tailored approach to smoking cessation counselling, Efforts to convert quit attempts in pregnancy into sustained cessation represent a priority area of programme development and evaluation.