We have studied, in 16 patients undergoing thoracoabdominal oesophagectomy, if two prolonged expiration manoeuvres improve prediction of arterial PCO2 (PaCO2) from end-tidal PCO2 (PE' CO2). PE' CO2, PCO2 at the end of a simple prolonged expiration (PE1 CO2), and PCO2 at the end of a prolonged expiration preceded by sustained hyperinflation of the lungs (PE2 CO2), were measured during laparotomy, in the lateral thoracotomy position during two-lung ventilation, and after transition to one-lung ventilation. (PaCO2-PE' CO2) was 1.3 (SD 0.4) kPa during laparotomy and this remained stable throughout the study. Both manoeuvres decreased the mean arterial to peak expired PCO2 difference, particularly during one-lung ventilation. However, PE1 CO2 and PE2 CO2 did not agree more closely with PaCO2 than PE' CO2 at any stage of the study. We conclude that these manoeuvres did not improve estimation of PaCO2 from PE' CO2.