An analysis of all perinatal deaths occurring in twin pregnancies in Dundee women from 1956 to 1983 was performed. The uncorrected perinatal mortality rate fell from 116/1000 births in 1956-60 to 16/1000 births in 1981-83, this fall almost entirely taking place after 1975. Causes of death were identified using the Aberdeen Classification and a reduction in deaths in all cause groupings occurred. About half of the deaths were in the Premature, Cause Unknown group and a marked decrease in deaths in this group made the largest contribution to the improved perinatal mortality rate. This fall was partly due to a reduction in the incidence of extreme prematurity and low birth weight. Changes in obstetric management which may have influenced outcome included the introduction of routine early pregnancy ultrasound scanning, the use of tocolytic drugs, intrapartum fetal monitoring, epidural analgesia and an increase in Caesarean section rate from 2% in 1956-60 to 39% in 1981-83.