Our goal was to measure the pregnancy-related reduction in the prevalence of reported binge drinking (>/=5 alcoholic drinks per occasion) and to characterize binge drinkers among pregnant and nonpregnant women aged 18-44 years, in the United States, 1991-1995. We used the Behavioral Risk Factor Surveillance System data from 46 states. We used the prevalence rate ratio between pregnant and nonpregnant women to determine the magnitude of the reduction in reported binge drinking and multiple logistic regression models to identify characteristics associated with binge drinking. Between 1991 and 1995, the prevalence of binge drinking among pregnant women increased significantly from 0.7% (95% confidence interval 0.2-0.9) to 2.9% (95% confidence interval 2.2-3.6), whereas among nonpregnant women the prevalence changed little (11.3% vs 11.2%). Over the study period pregnant women were one fifth (prevalence rate ratio 0.2, 95% confidence interval 0.1-0.2) as likely as nonpregnant women to binge drink. Among various population subgroups of women, pregnancy-related reduction in binge drinking was smallest among black women and largest among women aged </=30 years and among those who had quit smoking. Among pregnant women binge drinking was independently associated with being unmarried, being employed, and current smoking. Among nonpregnant women binge drinking was independently associated with age </=30 years, nonblack race, college level education, being unmarried, being employed or a student, and current smoking. Clinicians serving women of childbearing age need to be aware of the recent rise in reported binge drinking during pregnancy, as well as the known risk factors for binge drinking.