A study of all the mothers in the Northern region in 1983 whose pregnancies ended at between 24 and 31 weeks' gestation was undertaken. These pregnancies accounted for 1.3% of all the births and 44% of all the fetal and neonatal deaths in pregnancies that lasted more than 23 weeks. Most of the 389 singleton deliveries without malformations between 24 and 31 weeks were caused by spontaneous premature labour (n = 119, 31%), placental abruption (n = 79, 20%), pre-eclampsia (n = 56, 14%), and premature rupture of membranes (n = 48, 12%). The percentages of babies alive at the onset of delivery who survived the neonatal period were 66, 84, 78, and 73, respectively and the percentages of neonatal survivors with severe disabilities were 19, 13, 3, and 6, respectively. A further 65 (17%) of these babies died before the onset of labour for no obvious reason. The mode of delivery did not correlate with the outcome once the period of gestation at delivery was taken into account. The results highlight the inadequacy of the perinatal mortality index currently used in the United Kingdom, which identified 10.7 deaths/1000 registered births in 1983; 20% of the babies delivered at 24-31 weeks were excluded by this index because the birth went unregistered, as were 21% of all the babies born alive who died before discharge. An index that accounted for all babies weighing 500 g or more at birth irrespective of the period of gestation, and all neonatal (0-27 days) deaths, would be more appropriate and would bring reporting more into line with recommended international practice.