Background: This paper presents drinking patterns in a prospective study of a population-based cohort of 1570 pregnant women using a combination of dose and timing to give best estimates of prenatal alcohol exposure (PAE). Novel assessments include women's special occasion drinking and alcohol use prior to pregnancy recognition. Methods: Information on up to nine types of alcoholic drink, with separate frequencies and volumes, including drinking on special occasions outside a 'usual' pattern, was collected for the periconceptional period and at four pregnancy time points. Weekly total and maximum alcohol consumption on any one occasion was calculated and categorised. Drinking patterns are described in the context of predictive maternal characteristics. Results: 41.3 % of women did not drink during pregnancy, 27 % drank in first trimester only; most of whom stopped once they realised they were pregnant (87 %). When compared to women who abstained from alcohol when pregnant, those who drank in the first trimester only were more likely to have an unplanned pregnancy and not feel the effects of alcohol quickly. Almost a third of women continued to drink alcohol at some level throughout pregnancy (27 %), around half of whom never drank more than at low or moderate levels. When compared with abstainers and to women who only drank in trimester one, those who drank throughout pregnancy tended to be in their early to mid-thirties, smoke, have a higher income and educational attainment. Overall, almost one in five women (18.5 %) binge drank prior to pregnancy recognition, a third of whom were identified with a question about 'special occasion' drinking. Women whose age at first intoxication was less than 18 years (the legal drinking age in Australia), were significantly more likely to drink in pregnancy and at binge levels prior to pregnancy recognition. Conclusions: We have identified characteristics of pregnant women who either abstain, drink until pregnancy awareness or drink throughout pregnancy. These may assist in targeting strategies to enhance adherence to an abstinence policy and ultimately allow for appropriate follow-up and interpretation of adverse child outcomes. Our methodology also produced important information to reduce misclassification of occasional binge drinking episodes and ensure clearly defined comparison groups.