Estimates of regional cerebral blood flow (rCBF) by non-invasive xenon methods (133-xenon inhalation, xenon-enhanced computed tomography (Xe/CT) and 133-xenon iv injection) are frequently applied in the diagnosis and evaluation of patients suffering from diseases which cause disturbances in the cerebrovascular circulation. These methods all depend on an estimate of the arterial xenon concentration curve derived non-invasively from measurements of the end-tidal xenon concentration curve and used as brain input function in the Kety equation. We have studied the influence of impaired pulmonary gas exchange on the end-tidal and arterial xenon concentration curves in nine anaesthetized pigs by simultaneously measurements of both the end-tidal xenon and arterial xenon concentration curves. Computer simulations were performed to determine the deviations in the calculated rCBF values when using the end-tidal as compared to the arterial xenon concentration curve as brain input function. The results indicated that impairment of the pulmonary gas exchange caused a significant further 'delay' in the arterial xenon concentration curve in comparison to the end-tidal xenon concentration curve. The time constants of arterial curve delay were 11.9 s in the normal pulmonary group, 21 s in the right lung atelectasis group, and 19.7 s in the left pulmonary artery occlusion group. Accordingly, computer simulations indicated a statistically significant 'underestimation' of rCBF due to: (1) pulmonary gas exchange; (2) high or low levels of rCBF; (3) partition coefficient (lambda) of gray and white matter; and (4) xenon inhalation protocols. Our results indicate that quantitative measurements of rCBF by non-invasive xenon methods are markedly affected by deviations between the end-tidal and arterial xenon concentration curve, so that estimates of flow thresholds for infarction are problematic under conditions of impaired pulmonary gas exchange.