In a prospective study 1256 pregnant women at their first antenatal visit were asked if they drank alcohol every day, occasionally, or not at all, both at that time and before the diagnosis of pregnancy. There were no significant differences, among the three alcohol groups, in the proportions having miscarriages, perinatal deaths, congenital anomalies, or premature births. Women who reported drinking every day before pregnancy was diagnosed had heavier babies than those who did not drink at all, but there were no differences in mean birth weight according to alcohol use at the first antenatal visit. There were statistically significant differences among the three groups in important background characteristics, such as social class, age, smoking, and marital status. When adjustment was made for these factors, both occasional and daily consumption of alcohol before pregnancy appeared to have significantly positive effects on birth weight. These results illustrate the difficulty of evaluating the effects of alcohol in observational studies, and imply a need for caution in attributing either negative or positive effects of alcohol on the outcome of pregnancy.