The objective of this study was to evaluate the effect of lateral body position on gas exchange in patients with unilateral pleural effusion, with special reference to the influence of effusion volume. Thirty consecutive patients with unilateral pleural effusion, without evidence of parenchymal pulmonary involvement, were entered into the study. Arterial blood gas tensions ( PaO 2, PaCO 2) were randomly measured in both right and left lateral decubitus body positions, while breathing room air. To assess the influence of the effusion volume, roentgenographic and functional parameters were used. Among the latter, FVC, FEV 1, TLC and RV were determined. The influence of the presence or absence of pleuritic pain on gas exchange was also assessed. There was no significant difference in PaCO 2 between right and left lateral decubitus body positions (31·1 ± 4·2 vs. 31·0 ± 4·5 mmHg). The differences in PaO 2 between the two body positions ranged from 0·5–2·5 mmHg (mean 9·3 ± 6·6 mmHg). Mean PaO 2 with the normal-side (control) down ( PaO 2-N) (81·4 ± 8·5 mmHg) was higher, but without significant statistical difference, than mean PaO 2 with the effusion-side down ( PaO 2-E) (78·0 ± 12·5 mmHg). PaO 2-N was higher than PaO 2-E in 22 of 30 patients (conventional), and lower in eight patients (paradoxical). No consistent relationship was found for alterations in PaO 2 in different positions with the volume of effusion, either when estimated by a roentgenographic method or when using spirometric or plethysmographic values. The mean difference between PaO 2 values in the two positions ( ΔPaO 2 N-E) (6·3 ± 9·8 mmHg) in patients with chest pain was significantly greater ( P<0·05) than the ΔPaO 2 N-E in patients without pain (−3·3 ± 10·7 mmHg). Moreover, only pleuritic chest pain showed a significant statistical influence on ΔPaO 2 N-E, when all factors were analysed simultaneously. Although gas exchange is improved in most patients in the lateral decubitus position with normal-side down, some patients behave paradoxically. This difference does not appear to be related to the effusion volume, but seems to be influenced by the presence of pleural pain.