Of 4583 patients in the Large Bowel Cancer Project, 713 (16 per cent) were obstructed. The site of greatest risk was the splenic flexure (49 per cent). Advanced stage was neither the full reason why some patients obstructed nor for their subsequent poor prospects (age-adjusted 5-year survival: not obstructed, 45 per cent; obstructed, 25 per cent). Also, there was no greater risk of vascular invasion, no heavier lymph node burden and no worse tumour differentiation in patients with obstruction. In-hospital mortality was high (23 per cent), was not reduced by either a policy of primary or staged resection and was not influenced by the site of obstruction. There was no survival advantage for either policy, but hospital stay after primary resection was half that of staged. Immediate anastomosis in the obstructed left colon had a high clinical leak rate (18 per cent versus 6 per cent elective; P less than 0.001). Both registrars and consultants had similar mortality rates for elective primary resection and for the management of obstruction itself (as evidenced by results after the first stage of a staged resection). Selection probably accounts for the very much better results achieved by consultants for primary resection in the presence of obstruction (in-hospital mortality: consultants, 13 per cent; registrars, 24 per cent).