The epidemiology of twin pregnancies was studied with the use of data on 7,001 live-born twins from the Georgia neonatal surveillance network for the period 1974-1978. A cesarean section did not appreciably reduce the risk of neonatal death for a twin with a vertex presentation. The relative risk of such a neonatal death was 1.4 after adjustment was made for birth weight. A cesarean section did improve the outcome for breech and other presentations. Twins had a sixfold higher neonatal mortality rate than had singleton infants (p less than 0.001), which can be explained on the basis of distribution of birth weights. Twins had a weight-specific mortality rate equivalent to or significantly less than that for singletons after adjustment was made for birth weight. The relative risk of neonatal death for Twin 2 compared with Twin 1 was not significant. Breech presentation was more common in twins than in singletons, and for Twin 2 more than for Twin 1. As birth weight increased, the number of breech presentations decreased for Twin 2 but not for Twin 1. To reduce the high neonatal mortality rate for twins, the objective should be to reduce the incidence of low-birth-weight twins, rather than to increase the cesarean section rate for them.