The relationship between nutrient intake and pregnancy outcome (adjusted birth weight and gestational age) was investigated in randomly selected non-smokers (n97) and in heavy smokers (15 + cigarettes/d) (n72) booking for ante-natal care at a hospital in South London. Weighted dietary intakes (7d) were obtained at 28 and 36 weeks gestation. Birth weight was adjusted for gestational age, maternal height, parity and sex of infant. Compared with non-smokers, intakes of micronutrients and fibre were lower in smokers at both 28 and 36 weeks, and smokers reduced their intakes more in late pregnancy. The babies of smokers had a lower adjusted birth weight but there was no difference in length of gestation between smokers and non-smokers. After controlling for smoking, social class and alcohol consumption, nutrient intakes at 28 weeks were found to have no effect on adjusted birth weight. However, intakes of protein, zinc, riboflavin and thiamin at 36 weeks, and the change in intakes of these nutrients (plus iron) between 28 and 36 weeks, had independent positive effects on birth weight. Some of the effect of smoking on birth weight appeared to be mediated through differences in nutrient intakes. Smoking explained 14.3% of the variance in birth weight in this population and a further 2.4-7.2% was explained by change in nutrient intakes between 28 and 36 weeks. It is recommended that women in pregnancy do not reduce their dietary intakes in late pregnancy.