There is still controversy concerning the beneficial aspects of 'dynamic analgesia' (i.e. pain while coughing or moving) on the reduction of postoperative atelectasis. In this study, we tested the hypothesis that thoracic epidural analgesia (TEA) prevents these abnormalities as opposed to multimodal analgesia with i.v. patient controlled analgesia (i.v. PCA) after thoracotomy. Fifty-four patients undergoing thoracotomy (lung cancer) were randomly assigned to one of the two groups. Clinical respiratory characteristics, arterial blood gas, and pulmonary function tests (forced vital capacity and forced expiratory volume in 1 s) were obtained before surgery and on the next 3 postoperative days. Atelectasis was compared between the two groups by performing computed tomography (CT) scan of the chest at day 3. Postoperative respiratory function and arterial blood gas values were reduced compared with preoperative values (mean (SD) FEV1 day 0: 1.1 (0.3) litre; 1.3 (0.4) litre) but there was no significant difference between groups at any time. PCA and TEA provided a good level of analgesia at rest (VAS day 0: 21 (15/100); 8 (9/100)), but TEA was more effective for analgesia during mobilization (VAS day 0: 52 (3/100); 25 (17/100)). CT scans revealed comparable amounts of atelectasis (expressed as a percentage of total lung volume) in the TEA (7.1 (2.8)%) and in the i.v. PCA group (6.71 (3.2)%). There was no statistical difference in the number of patients presenting with at least one atelectasis of various types (lamellar, plate, segmental, lobar).