We have developed a method to estimate the critical closing pressure (CrCP) of the cerebral circulation based on the intrinsic variability of arterial blood pressure (BP) around stable values of mean arterial pressure (MAP). A consecutive cohort of 33 premature newborns was studied at 6, 12, 24, 48 and 72 hours of age. Cerebral blood flow velocity (CBFV) was measured with Doppler ultrasound in the middle cerebral artery and BP was recorded in the abdominal aorta or in a peripheral artery. Continuous measurements lasting five minutes were recorded on digital magnetic tape and signals were digitized at a rate of 200 samples/seconds for processing on a digital computer. Mean values of BP (mBP) and CBFV (mBV) were computed for each cardiac cycle and CrCP was determined as the pressure axis intercept of the regression line of mBV as a function of mBP using 100 sequential heart beats. The resistance-area product (RAP) was obtained from the slope of the regression line. For 57 records (30 patients) the mean +/- SD values of CrCP and RAP were 23.9 +/- 11.6 mmHg and 4.07 +/- 1.83 x 10(4) kg.m-2.s-1, respectively. CrCP has a highly significant correlation with mean arterial pressure (p < 0.001) but RAP has not. Neither CrCP nor RAP are significantly correlated (p > 0.05) with PO2, PCO2, pH, haematocrit, gestational age, birthweight, postnatal age, heart rate on Pourcelot's resistance index. Our results suggest that cerebral perfusion pressure should be defined as MAP-CrCP for normal values of intracranial pressure.