The effects of infant suckling patterns on the post-partum resumption of ovulation and on birth-spacing are investigated among the Gainj of highland New Guinea. Based on hormonal evidence, the median duration of lactational anovulation is 20·4 months, accounting for about 75% of the median interval between live birth and next successful conception (i.e. resulting in live birth). Throughout lactation, suckling episodes are short and frequent, the interval changing slowly over time, from 24 minutes in newborns to 80 minutes in 3-year olds. Maternal serum prolactin concentrations decline in parallel with the changes in suckling patterns, approaching the level observed in non-nursing women by about 24 months post-partum. A path analysis indicates that the interval between suckling episodes is the principal determinant of maternal prolactin concentration, with time since parturition affecting prolactin secretion only in so far as it affects suckling frequency. The extremely prolonged contraceptive effect of breast-feeding in this population thus appears to be due to (i) a slow decline in suckling frequency with time since parturition and (ii) absence of a decline over time in hypothalamic–pituitary responsiveness to the suckling stimulus.