Presented are the results of multivariate regression analysis of the variations in birth weight of term babies associated with the socioeconomic, physical, and medical factors recorded in obstetric case notes at a London teaching hospital. The results show that the sex of the baby, parity, maternal smoking during the pregnancy, maternal height, weight, marital status, and race, and gestation (after allowing for the foregoing characteristics) were all important and significant factors. Socioeconomic status, per se, was not, nor was age. In particular, the results negate the suggestion that the loss of birth weight associated with smoking during the pregnancy might be due to the smoker, rather than to smoking itself. The results indicate that moderate smoking (one to 15 cigarettes per day) causes a loss of 107 gm, and that heavy smoking (16+) causes a loss of 158 gm, after all other factors have been taken into account. An investigation of how these figures would be affected by failing to control for one or more of the other factors led to the conclusion that the estimates of the loss of birth weight as a result of smoking in completely uncontrolled studies may be reasonably accurate, whereas the estimates in partially controlled studies are probably biased.