To determine whether mechanical bowel preparation influences the incidence of anastomotic dehiscence following colorectal surgery, 186 patients undergoing elective left colonic or rectal resection were randomized before surgery to bowel preparation (n = 89) or no bowel preparation (n = 97). Surgical technique was standardized and no patient had a defunctioning colostomy. Seventeen patients were excluded (seven with preparation, ten without). Indications for surgery in the remaining 169 patients were carcinoma (133 patients), diverticular disease (26), inflammatory bowel disease (six) and miscellaneous conditions (four). Operations performed were left colonic resection or reversal of Hartmann's procedure (26 with preparation, 28 without) and anterior resection (56 versus 59). The overall morbidity rate (18 per cent) was similar in the two groups. All seven clinical anastomotic leaks occurred after low anterior resection, in three of the 39 patients who had undergone bowel preparation and four of the 36 who had not (P > 0.9). Two deaths occurred, both of patients who had received bowel preparation, one being secondary to anastomotic leakage. Bowel preparation does not influence outcome after elective colorectal surgery.