The relationship between birthweight, gestation and obstetric and neonatal course was examined for 444 live births at 25-31 weeks gestation in 5 centres from 1982 to 1984 and comparative data were obtained from 121 infants below 32 weeks born in 1968-1972. The rate of 'interventive' deliveries by elective caesarean section, at this gestation, performed largely for conditions associated with poor intrauterine growth, has risen over 10-fold in the past decade, now accounting for 42% of deliveries at 30 weeks. This may explain in part a secular trend towards reduced birthweight for gestation in preterm infants. Electively delivered infants were significantly lighter than those delivered spontaneously and were responsible for substantial skewing and lowering of overall birthweight centiles. In contrast, the infant's sex, whether or not the infant survived following live birth, and whether gestational age was assessed by maternal dates or by clinical estimation (including ultrasound), all made no significant difference to birthweight centiles. A new birthweight centile chart is presented for infants less than 32 weeks, based only on spontaneous deliveries. This chart differs considerably from those commonly used in Britain and, we suggest, provides a realistic standard for identifying abnormal intrauterine growth.