We have measured the partial pressure of isoflurane simultaneously in inspired gas (Pl(iso)), end-expired gas (PE(iso)'), mixed-expired gas (PE(iso)BAR), arterial (Pa(iso)) and mixed venous blood (Pv(iso)BAR) in six patients (aged 57-79 yr) anaesthetized with nitrous oxide, oxygen and isoflurane before surgery and after PE(iso)' had been stable for at least 15 min. We related these changes to the various indices of pulmonary maldistribution to determine if they were sufficient to explain reported differences between PE(iso)' and Pa(iso). Alveolar deadspace dilution of end- expired gas was calculated for carbon dioxide and this dilution factor used to calculate the ''ideal'' alveolar Piso (PA(iso)) from the observed inspired and end-expired concentrations. Shunt fraction was measured for oxygen and then used to calculate the partial pressure of isoflurane in the pulmonary end-capillary blood (Pc(iso)') from the partial pressure in arterial and mixed venous blood. Mean (SE) values were: Pl(iso) 0.69 (0.05) kPa; PE(iso)' 0.52 (0.04) kPa; PA(iso) 0.50 (0.04) kPa; Pc(iso)' 0.38 (0.04) kPa; Pa(iso) 0.35 (0.03) kPa and PV(iso)BAR 0.22 (0.02) kPa; Pa(iso): PE(iso)' 0.66 (0.02) kPa. The mean ''ideal'' alveolar to pulmonary end-capillary Piso difference was 0. 12 (0.0 1) kPa and highly significant (P < 0.001). Pa(iso) was substantially less than PE(iso)' but, for isoflurane, the difference was reasonably constant (range 0. 14-0.22 kPa). The difference was attributable in part to the effects of shunt and deadspace, but also a failure of equilibration of isoflurane between the alveolar gas and pulmonary end-capillary blood. It is likely to be different for other anaesthetics. We conclude that, while PE(iso)' may adequately reflect Pa(iso) for isoflurane, it cannot be assumed that the relation between end-expiratory gas and arterial partial pressures is the same for all anaesthetics.