392 patients were examined by endoscopy for acute upper gastrointestinal bleeding; 140 had ulcers containing an actively bleeding visible vessel or a non-bleeding visible vessel and were enrolled in a randomised trial of three endoscopic methods of haemostasis--adrenaline (1/10(4] alone, adrenaline plus polidocanol 1%, and adrenaline followed by yttrium-aluminium-garnet (YAG) laser photocoagulation. For patients with non-bleeding visible vessels sham treatment was significantly less effective in achieving haemostasis (8 of 20 patients) than were adrenaline plus polidocanol (18 of 20; p = 0.002) and adrenaline plus laser (16 of 20; p = 0.012). All three treatments significantly reduced total transfusion needs compared with sham treatment. For the whole group of patients, adrenaline plus polidocanol was significantly more effective than adrenaline alone in achieving permanent haemostasis; adrenaline plus laser was also more effective than adrenaline alone, but not significantly so. The efficacy of the three treatments was enhanced by repeated application on recurrence of bleeding. Since injection therapy with adrenaline and polidocanol was at least as effective as adrenaline plus laser therapy, it should be preferred over laser therapy because it is cheaper, easier to use, and perhaps also safer.