This paper discusses the definition and subclassification of cerebral palsy, its ascertainment, birth prevalence, and aetiology. Cerebral palsy is a description defined by clinical observation rather than a diagnosis informative about aetiology, pathology, or prognosis. As a disease entity it is aetiologically and pathologically diverse. The cerebral palsy grouping is useful for management and service planning but the combining of several individually rare congenital and early childhood conditions creates an epidemiological challenge. The birth prevalence of cerebral palsy in the developed world has not declined since the mid-1950s, in contrast to perinatal mortality. The proportion of children with cerebral palsy that are born very preterm has increased with the advent of neonatal intensive care, and refinement of the techniques of neonatal intensive care may be necessary before its benefits can be fully realised. Since cerebral palsy is, by definition, incurable, epidemiology strives to elucidate causation as a necessary step towards prevention. Definitive allocation of cause before the post-neonatal period is usually impossible and a large proportion of all cases of cerebral palsy are not adequately explained. Recognised risk factors may be single or multiple and can occur at any stage from before conception to infancy. In developed countries, the majority of aetiological paths are likely to commence before labour and delivery. Approximately 10% of cases are believed to now be causally associated with adverse intrapartum events, less than either the proportion traditionally assumed or the proportion caused post-neonatally. The proportions of children with cerebral palsy resulting from an intrapartum cause and from a post-neonatal cause are both believed to be higher in developing countries. Since post-neonatally acquired brain damage has the most easily recognised and understood causes, postneonatally acquired cerebral palsy should be the most preventable.