Our previous cerebral ultrasound study of antecedents of periventricular haemorrhage in infants weighing 1250 g or less at birth suggested that neonatal events that caused increased or fluctuating cerebral blood flow lead to periventricular haemorrhage. As the risk period for this type of haemorrhage was the first four days of life strict guidelines were introduced to avoid the previously identified neonatal risk factors. No attempt was made to modify obstetric practice. Over the next two years, although the obstetric risk profile, the frequency and severity of hyaline membrane disease, and the gestation, birth weight, and sex distributions of a similar cohort of infants did not change, the incidence of periventricular haemorrhage decreased significantly from 60% to 36%. Significant antecedents of haemorrhage similar to those found in the previous study included severe bruising, low arterial:fractional inspiratory oxygen ratio and low packed cell volume on admission, hyaline membrane disease, hypercarbia, and hypoxaemia. Assisted ventilation, pneumothorax, treatment with tubocurarine, and hypotension were no longer significant risk factors for periventricular haemorrhage. A multivariate discriminant analysis correctly predicted haemorrhage in 86% of the study group when bruising, hypercarbia, hypoxaemia, hyaline membrane disease, and low gestation were considered. These results suggest that changes in neonatal practices can reduce the incidence of periventricular haemorrhage and that drug studies indicating similar reduction in haemorrhage need to be evaluated carefully to ensure that placebo and treated groups are in fact comparable.