This study was designed to investigate the effect of posture on oxygen saturation during fibre-optic bronchoscopy (FOB). Thirty-eight consecutive patients requiring diagnostic FOB were randomized into two groups according to the initial posture in which the FOB was performed. In group 1 (20 patients), FOB was commenced supine, and in group 2 (18 patients) in a semi-recumbent position (45° from horizontal). Sedation with midazolam was titrated according to clinical response. All patients received atropine 0·6 mg intravenously and topical lignocaine. Observations of peak, trough and plateau oxygen saturation and pulse rate were recorded during six study periods, each lasting 3 min. Periods 1 and 2 were pre- and post-sedation without supplemental oxygen, respectively. The bronchoscope was then inserted into the distal end of the trachea and observations taken during periods 3 and 4 (no supplemental oxygen) and periods 5 and 6 (2 l oxygen by nasal cannulae). In group 1, posture was changed from supine to semi-recumbent from periods 3–4 and reversed in periods 5 and 6. In group 2, posture changes were in reverse sequence. Patients with initial oxygen saturation of less than 90% or showing a fall below 85% during FOB were excluded. Five patients from each group were withdrawn because of hypoxia. In both groups, oxygen saturation fell significantly (P<0·001) following sedation. There was no significant change in saturation (peak, trough or plateau) with change in posture from supine to semi-recumbency (group 1) or the reverse (group 2). These correspond to periods 3–4 and 5–6 in both groups. Supplemental oxygen was associated with a significant rise in oxygen saturation in both postures, attaining levels close to presedation levels.