We have studied prospectively the effect of 10 cm H2O of PEEP on the incidence of venous air embolism and on the cardiovascular response to change from the supine to the seated position in a large neurosurgical population. Patients were allocated randomly to receive either PEEP (10 cm H2O, n = 45) or conventional (control, n = 44) ventilation. Cardiovascular and respiratory variables were measured in the supine and sitting positions, and monitoring included precordial Doppler probe, pulmonary artery pressure and expiratory carbon dioxide concentration. Venous air embolism was assumed if changes in precordial Doppler sounds occurred, end-tidal carbon dioxide concentration decreased or air could be retrieved from a central venous multi-orifice catheter. The incidence of venous air embolism (26%) did not differ between patients undergoing conventional ventilation and those undergoing ventilation with 10 cm H2O of PEEP. Venous air embolism was always detected first by alterations in Doppler sounds. Cardiac output was significantly higher in patients undergoing conventional ventilation than in those undergoing ventilation with PEEP in the supine but not in the sitting position. Furthermore, pulmonary vascular resistance increased significantly only in the upright position in those undergoing ventilation with PEEP. The pulmonary artery wedge pressure to central venous pressure gradient did not attain negative values with PEEP or with upright positioning. We conclude that the use of PEEP during neurosurgical procedures performed in the sitting position should be abandoned as it does not decrease the incidence of venous air embolism but is associated with significant adverse cardiovascular effects.