Background: Rates of preterm delivery, a major proximate cause of perinatal morbidity and mortality, have been increasing. Prenatal alcohol exposure has been implicated in preterm delivery, although results have been inconsistent due to inadequate control for confounding factors, insufficient power, unreliable and inaccurate assessment of both exposure and gestational age, and lack of stratification of prematurity into severity levels. The purpose of this study was to examine the relation between maternal alcohol, cocaine and cigarette use during pregnancy, and extreme and mild preterm birth. Methods: Three thousand one hundred thirty consecutive gravidas were followed prospectively for antenatal substance use and had ultrasound confirmed pregnancy dating. Results: Alcohol and cocaine, but not cigarette use, were associated with increased risk of extreme preterm delivery after control for potential confounders. For every unit increase in alcohol exposure, risk of extreme preterm delivery increased significantly [odds ratio (OR) 34.8]. Furthermore, in women aged 30+, alcohol exposure was associated with mild prematurity. Abstention from alcohol while continuing to use cocaine and tobacco was related to a decrease in extreme prematurity of 41%. Conclusions: The risk of extreme preterm delivery associated with alcohol use is substantial and similar in magnitude to other well-recognized risks. Increased accuracy in identifying exposure and the use of ultrasound to confirm gestational age dating likely contributed to the findings of the current study. These findings suggest that eliminating pregnancy alcohol use might substantially reduce the risk of preterm delivery.