Despite widespread advocacy of breast feeding, many babies are breast fed only briefly, if at all. Mothers' decisions on how to feed are often made before the birth; so we have sought demographic, social and psychological factors that might be amenable to intervention during pregnancy. In the Avon Longitudinal Study of Pregnancy and Childhood about 12,000 women completed questionnaires in pregnancy. Univariate analyses were carried out to establish which factors were related to breast feeding intentions. All significant factors in univariate analyses were entered into logistic regression analyses. Demographic characteristics independently related to intentions to breast feed included older maternal age, more maternal education, primiparity and not smoking; in previous work all these had been associated with actual feeding behaviour. Social relationship variables had a small influence. Of the psychological variables, a notable finding was that women who were preoccupied with their body shape and those who expressed controlling, less child-centred, responses to managing an infant in the postnatal months were less likely to express intentions to breast feed. Depression did not predict breast feeding intentions once the other factors had been taken into account. Health care professionals may be able to intervene to increase breast feeding by making routine enquiries during antenatal care and targeting appropriate subgroups.