Cerebral blood flow (CBF) measurement by near infrared spectroscopy (NIRS) using oxyhemoglobin (HbO2) as a tracer (CBF-HbO2) needs rapid changes in arterial oxygen saturation (SaO2) which often cannot be achieved in many sick infants. An alternative method based on the same adaptation of the Fick principle using i.v. injection of the dye indocyanine green (ICG) is described (CBF-ICG). Six mechanically ventilated infants (age 26-38 wk, birth weight 0.885-3.730 kg) requiring supplementary oxygen therapy were studied within 72 h of birth. For CBF-ICG measurements, ICG (0.1 mg x kg-1 was injected via an umbilical venous catheter, and blood ICG concentration was measured by an optical umbilical artery catheter and brain ICG concentration was measured by NIRS. For CBF-HbO2 measurements the inspired oxygen concentration was rapidly increased, blood HbO2 concentration was calculated from SaO2 measured by pulse oximetry, and brain HbO2 concentration was measured by NIRS. A series of CBF measurements were performed using each method before and after altering the arterial carbon dioxide tension (PaCO2). Mean CBF values from repeated measurements by each method at any given PaCO2 were used to compare the methods. The SD of single measurements within an individual subject by CBF-ICG was 15%, and by CBF-HbO2, 24%. The relationship between the methods was mean CBF-ICG = (1.13 x mean CBF-HbO2) - 2.76 mL x 100 g-1 x min-1 HbO2 (r = 0.93, p < 0.001). The mean difference between the methods (CBF-ICG - CBF-HbO2) was -0.25 mL x 100 g-1 x min-1 (95% confidence interval 6.30 to -6.80). The methods were in good agreement, and the use of i.v. ICG permitted rapid and repeated CBF measurements in the sickest infants at greatest risk of cerebral injury.