In 45 patients with cancerous pleural effusion, thoracocentesis was performed according to the following technique : after local anaesthesia (5 ml of 2% lidocaine) in the posterior part of the 6th intercostal space, thoracocentesis was carried out with a blunt trocar connected to a water manometer. The pleural pressure was measured before removal of the fluid (Po). Then three different techniques of drainage (active suction at -- 80 or -- 40 mmHg or underwater sealed drainage) were randomized. Pleural pressure was measured after removal of every 100 ml and at the end of the procedure (PT). The mean Po +/- SD was respectively -- 4.08 +/- 2.95, -- 3.81 +/- 3.12 and -- 2.53 +/- 2.42 cmH2O for the three different groups of drainage. After drainage, PT was decreased in the three groups, especially after -- 80 mmHg aspiration. In 14 patients, side effects occurred during thoracocentesis and the procedure was stopped. Symptoms, chiefly cough, occurred after a mean removal of 1.1 +/- 0.7 1. PT in these patients was statistically lower (-- 18 +/- 5.59 cmH2O) than in the 31 asymptomatic patients (--11.40 +/- 5.75 cmH2O) (p less than 0.001). In conclusion, the assessment of the pleural pressure during thoracocentesis seems to be convenient in order to prevent any complication of pleural evacuation : pleural evacuation should be stopped if the pleural pressure decreased below -- 18 cmH2O. A depression at -- 80 mmHg is too dangerous for pleural aspiration; -- 40 mmHg or under water sealed drainage are both convenient for a safe and complete evacuation.