There is wide agreement that second twins are at increased risk of complications during labor and delivery, but large-scale observational studies going back to the 1960s disagree about the perinatal mortality risk. The present retrospective cohort study linked a national register of data on post-childbirth discharges with a register of perinatal deaths occurring in Scotland in the years 1992-1997. The study group included 4545 women who delivered twins at or after 24 weeks' gestation, 15% of whom had cesarean delivery. On univariate analysis the only independent predictors of preterm birth were parity and maternal height. Deaths of first and second twins born before 36 weeks' gestation did not differ significantly. After 36 weeks, however, there were no deaths of first twins but nine among second twins (risk ratio, 3.7; P =.004). The discordance between first and second twins was significantly different for preterm and term births. Five of the nine second-twin deaths at term were intrapartum stillbirths, and four were neonatal deaths. Intrapartum anoxia was considered responsible for all but two deaths. In six of seven term deliveries in which the second twin died of anoxia, and in all five of these deliveries where anoxia had a mechanical obstetrical origin, both twins were delivered vaginally. There were no delivery-related perinatal deaths among 454 twin pairs delivered at term by planned cesarean section. Birth-weight discrepancies between twin infants were significantly greater when the second twin died than when both twins survived. On multivariate logistic regression analysis, the relationship between birth order and gestational age was independent of maternal age, parity, smoking, height, and socioeconomic status. These findings suggest that second twins born at term are likelier than their first twin partners to die of complications relating to vaginal delivery. Planned cesarean delivery may protect against these deaths, and women bearing twins should be told this when deciding on mode of delivery.