Control data from 1529 infants studied in a multicentre case-control study of sudden infant death in New Zealand were analysed to identify factors that might hinder the establishment and duration of breast feeding. Although 1300 infants (85%) were exclusively breast-fed at discharge from the obstetric hospital, this fell to 940 (61%) by 4 weeks. Logistic regression was used to identify factors that might adversely influence breast feeding 'at discharge', 'at 4 weeks' and the overall 'duration' of breast feeding. When adjusted for confounding factors, not exclusive breast feeding 'at discharge' was significantly associated with: twin pregnancy, being a Pacific Islander, mother not bedsharing, subsequent dummy use, birthweight less than 2500 g, heavy maternal smoking, not attending antenatal classes and mother less than 20 years old at first pregnancy. Mothers smoking more than 20 cigarettes a day were nearly twice as likely to not exclusively breast feed on discharge compared to those who did not smoke. A 'dose response' was apparent with the heaviest smokers having the least likelihood of establishing exclusive breast feeding. Being exclusively breast-fed at discharge but not 'at 4 weeks' was associated with: twin pregnancy, admission to a neonatal intensive care unit, subsequent dummy use and not being married. A shorter overall 'duration' of breast feeding was associated with maternal smoking, subsequent dummy use, mother not bedsharing, twin pregnancy, mother less than 20 years old at first pregnancy, low occupational status and not attending antenatal classes. These effects persisted when social and demographic factors, including birthweight, were taken into account.(ABSTRACT TRUNCATED AT 250 WORDS)