The authors studied the extent to which preterm birth and perinatal mortality are dependent on the gestational ages of previous births within sibships. The study was based on data collected by the Medical Birth Registry of Norway from 1967 to 1995. Newborns were linked to their mothers through Norway's unique personal identification number, yielding 429,554 pairs of mothers and first and second singleton newborns with gestational ages of 22-46 weeks, based on menstrual dates. Siblings' gestational ages were significantly correlated (r = 0.26). The risk of having a preterm second birth was nearly 10 times higher among mothers whose firstborn child had been delivered before 32 weeks' gestation than among mothers whose first child had been born at 40 weeks. However, perinatal mortality in preterm second births was significantly higher among mothers whose first infant had been born at term, compared with mothers whose firstborn child was delivered at 32-37 weeks. Since perinatal mortality among preterm infants is dependent on the gestational age in the mother's previous birth, a common threshold of 37 weeks' gestation for defining preterm birth as a risk factor for perinatal death may not be appropriate for all births to all mothers.