Near-infrared spectroscopy combined with partial jugular venous occlusion (JVO) offers promise for determining cerebral venous saturation (CSvO(2)) in sick preterm infants, but has not been validated in the newborn brain or under conditions of hypoxaemia. We assessed the accuracy of the CSvO(2) estimate using cerebral venous oxygen saturation in superior sagittal sinus blood (SSSO(2)) as the 'gold standard'. Comparisons were made in seven newborn lambs over a wide range of arterial oxygen saturations (SaO(2)) of 20% to 100%. Overall, median (range) CSvO(2) was 49.8% (10.6% to 88.5%), whereas SSSO(2) was 45.5% (4.3% to 76.6%); Bland-Altman analysis revealed a mean difference (CSvO(2)-SSSO(2)) of 5.1% and limits of agreement of +/-27.4%. The change in cerebral blood volume (DeltaCBV) induced by JVO increased with SaO(2) (P<0.05). In addition, the strength of the correlation of CSvO(2) with SSSO(2) progressively improved with increasing change in total haemoglobin concentration (DeltaHbT) induced by JVO. With Bland-Altman analysis repeated for data with DeltaHbT >30 micromol cm, the mean difference (CSvO(2)-SSSO(2)) decreased to 2.4% with limits of agreement of +/-18.8%. We conclude that the accuracy of estimating CSvO(2) varies with the DeltaCBV induced by JVO. Potential differences of optical properties between the head of the lamb and the human infant suggest that caution be exercised in directly applying these data to the human newborn. Nevertheless, this critical aspect of the JVO technique needs to be taken into consideration in developing an accurate measurement for sick preterm human infants.