Cranial ultrasounds were performed on 218 (96%) of 227 liveborn infants of birthweight 500-1500 g delivered in the Royal Women's Hospital, Melbourne, Australia, in an 18-month period concluding in March 1982. Seventy-two (31.7%) of the children died; 28 children (38.9%) had cerebroventricular haemorrhage, 35 (48.6%) showed no bleeding and there were nine (12.5%) with no data. Paired necropsy and ultrasound data were congruent in 22 (88%) of 25 children. One hundred and forty-eight (95.5%) of 155 survivors were seen at 2 years of age. Forty-one (28%) had cerebroventricular haemorrhage; nine children (6%) had both ventricular dilatation and haemorrhage and two had ventricular dilatation alone. Apart from a marginal advance in gestation and higher number of immigrant and less educated mothers in children without cerebroventricular haemorrhage, all other perinatal, biographical and social variables between those with haemorrhage and those without were similar. The major handicap rate overall was 14.2% (21 patients). The children with cerebroventricular haemorrhage had a trend for greater prevalence of handicap and lower mean Bayley psychological scores. This was even more evident with ventricular dilatation being present. Of children with major handicap 57.1% (12/21) had normal serial ultrasound findings during their primary hospitalization. Major handicap occurred in 15% (3/20) of children with grade 1 haemorrhage, 23.5% (4/17) with grade 2 or 3 bleeds and 25% (1/4) of those with grade 4 haemorrhage. Laterality of cerebral palsy did not correlate with ultrasound findings. Ultrasound findings did not improve statistical prediction of deaths or major handicap.