We report from a well-established cerebral palsy (CP) register the changes in CP rates by gestational age for singleton births over a 25-year period in north-east England. The gestational ages of numerators and denominators are of high accuracy back to 1970 because academic units in paediatrics and obstetrics were studying the assessment of gestational age in individual infants, and the distribution of gestational age across all births in the north-east from the 1960s. The rate of CP rose between 1970-75 and 1990-94 from 1.6 to 2.3 per 1000 singleton neonatal survivors, a rise of 0.7/1000 [95% CI 0.2, 1.3]. There was little change in the rate of CP in term infants whereas in preterm infants (<37 weeks) it rose from 5.5 to 16.8, a rise of 11.3/1000 [95% CI 5.9, 16.8]. Rises occurred in the three preterm gestational age bands <28, 28-31, 32-36 weeks with the most marked rise in those <28 weeks from 0 to 112.7. The proportion of all cases of CP arising in the preterm group rose from 19% to 45%; and the proportion of the severest cases arising in the preterm group rose from 8% to 55%. In those born after 32 weeks, there is a preponderance of small-for-gestation infants, with 10% more than two standard deviations below the mean. All types of CP are more common in infants below average weight for gestation and this is most marked for the non-spastic types that are almost only seen in term, small-for-gestation infants. Gestational age is the crucial determinant of rate of CP and the increase in prevalence seen over the past 25 years is due to increased rates in preterm infants, not term infants. Both conclusions, suspected from birthweight analyses, are now demonstrated conclusively, with the contribution coming from those 32-36 weeks gestation as well as very preterm infants.