We investigated the effectiveness of a program of intensive postpartum support for low-income, breast-feeding women and identified potential predictors of prolonged breast-feeding in this population. Ninety-seven low-income women were randomized to receive intensive postpartum education and support for breast-feeding or to receive only the routine assistance provided by the obstetrical nurses. Both groups were telephoned 6 weeks post partum to determine the method of infant feeding then, and those still breast-feeding were contacted monthly until complete weaning had occurred. No significant difference in breast-feeding duration between the two groups was noted. There was no association between duration of nursing and race, marital status, or the need to return to work or school. Earlier age at introduction of supplement, younger maternal age, and participation in prenatal classes predicted breast-feeding duration by logistic regression.