Lung densities (atelectasis) and pulmonary gas exchange were studied in 13 supine patients with no apparent lung disease, the former by transverse computerized tomography (CT) and the latter by a multiple inert gas elimination technique for assessment of the distribution of ventilation/perfusion ratios. In the awake state no patient had clear signs of atelectasis on the CT scan. Lung ventilation and perfusion were well matched in most of the patients. Three patients had shunts corresponding to 2-5% of cardiac output, and in one patient there was low perfusion of poorly ventilated regions. CT scans after 15 min of halothane anesthesia and mechanical ventilation showed densities in dependent lung regions in 11 patients. A shunt was present in all patients, ranging from 1% in two patients (unchanged from the awake state) to 17%. Ventilation of poorly perfused regions was noted in nine patients, ranging from 1-19% of total ventilation. The magnitude of the shunt significantly correlated to the size of dependent densities (r = 0.84, P less than 0.001). Five patients studied during spontaneous breathing under anesthesia displayed both densities in dependent regions and a shunt, although of fairly small magnitude (1.8% and 3.7%, respectively). Both the density area and the shunt increased after muscle paralysis. PEEP reduced the density area in all patients but did not consistently alter the shunt. It is concluded that the development of atelectasis in dependent lung regions is a major cause of gas exchange impairment during halothane anesthesia, during both spontaneous breathing and mechanical ventilation, and that PEEP diminishes the atelectasis, but not necessarily the shunt.