Children with and without spastic hemiplegia were identified among a representative sample of 1048 low-birthweight survivors by clinical assessment after three years of age. The hypothesis that hemiplegia was predetermined at birth was tested by estimating the probability of hemiplegia for each infant by logistic regression analysis, using data from hospital records on conditions known at the time of birth. 16 of 42 children with cerebral palsy had spastic hemiplegia. Allowing for the lower birthweights of hemiplegic children, increased prevalence was associated with previous reproductive loss, breech vaginal delivery, later birth-order, prolonged second stage of delivery, emergency caesarean section, and low Apgar scores. These variables identified correctly most children as having a higher or lower estimated probability of hemiplegia. Hemiplegia was also associated with prolonged respiratory disease and intraventricular haemorrhage. In this population it is likely that intrapartum events were closely related to the pathogenesis of hemiplegia; their effects may have been mediated by postnatal events.