The dynamics of perinatal mortality rates (PNMR) and causes of death in twin pregnancies over 13 years in the Northern Region of the National Health Service in England is described. All twin perinatal deaths occurring between 1982-1994 were identified from the Northern Region Perinatal Mortality Survey. The twinning rate increased from 9.9 per 1000 maternities in 1982 to 12.0 in 1994. There was a total of 10,734 twin pregnancies and of these 421 resulted in 530 perinatal deaths. The perinatal mortality rate in twins significantly decreased over time (1982-87, 55.4 per 1000; 1988-94, 44.4 per 1000; P = 0.01). The PNMR was significantly higher for twins from like-sexed than from unlike-sexed pairs (53.5 and 34.4 per 1000 respectively, P < 0.001). Despite no improvement in birthweight distribution in the twin population, birthweight-specific perinatal mortality rates for both like and unlike-sexed twins decreased for each birthweight category in 1988-94 compared with 1982-87. Twins with very low birthweight (< 1500 g) comprised 69%, and preterm twins (< 37 completed weeks of gestation) 74.9% of all twin perinatal deaths. The major immediate cause of early neonatal death was pulmonary immaturity (63%); antepartum anoxia caused 76.9% of antenatal deaths. Unexplained preterm labour and intrauterine death were the leading obstetric factors underlying death in twins. Despite a decrease over the 13 years, the perinatal mortality rate in twins in the Northern Region remains high. Continued monitoring of trends in twinning and mortality rates is needed to inform health care planning.