A randomized controlled trial was designed to compare antibiotic prophylaxis with a standard combination of agents, metronidazole and gentamicin, with a single preparation, amoxycillin/clavulanate in 400 patients undergoing elective colorectal surgery. There were 41 patients who were excluded or withdrawn (wrong dose, inappropriate operation, established sepsis or concurrent disease). Abdominal wound sepsis occurred in 14% of the assessable patients in the amoxycillin/clavulanate group and in 15% of the metronidazole plus gentamicin group. Perineal sepsis occurred in 27% of the amoxycillin/clavulanate group with a perineal wound compared with 18% in the metronidazole plus gentamicin group. Intra-abdominal abscess occurred in 8% of those who received amoxycillin/clavulanate compared with 6% of those given metronidazole plus gentamicin. Only two patients in each group developed septicaemia. Postoperative diarrhoea occurred in 11 patients receiving amoxycillin/clavulanate compared with four given metronidazole plus gentamicin. Clostridium difficile was not isolated from the stool cultures in any of these cases. Thirteen of the 164 abdominal or perineal wounds were infected by 15 strains of Bacteroides spp. in the group receiving amoxycillin/clavulanate compared with only three of the 165 wounds in those given metronidazole plus gentamicin. (P less than 0.01). There was no other significant difference in the pattern of isolates between the groups.