The effect of maternal alcohol consumption in the first trimester on birth weight was studied prospectively in 34,660 pregnancies from the Walnut Creek Birth Defects Study (NICHD). Complete medical and demographic information was available in 28,698 live births. Drinking showed a dose-response effect on birth weight; mean birth weight decreased as daily alcohol consumption increased. When adjustment was made for confounding factors (smoking, parity, age, etc.), multiple linear regression revealed an estimated decrease in birth weight of 16 grams for women consuming <1 drink, 86 grams for 1-2 drinks, 169 grams for 3-5 drinks and 127 grams for 6 or more drinks daily compared to non-drinkers (R2=.26, p<.0001). When birth weight below the 10th percentile for gestational age, sex, and race was examined, rates were 6.0%, 6.8%, 11.5%, 16.8%, and 17.7% in non-drinkers, and those drinking <1, 1-2, 3-5, and 6 or more drinks per day, respectively. The trend was significant (p<.0001). This relationship remained significant (p<.005) after multiple logistic regression was used to adjust for confounding factors. Drinkers did not deliver significantly earlier than non-drinkers. Because of the large number of subjects, the study was able to demonstrate a statistically significant effect of alcohol on birth weight independent of other confounding factors. Women should be advised that alcohol is associated with intrauterine growth retardation, but that the effect in light drinkers is clinically small.