Postoperative infection accounts for significant morbidity and mortality following penetrating abdominal trauma. During a 2 1/2-year period, December 1980 through June 1983, 257 patients sustaining penetrating abdominal injury were initially treated at Parkland Memorial Hospital in Dallas. Following the patient's written consent, they were prospectively randomized to receive, prior to surgery, intravenous clindamycin 600 mg every 6 hours and tobramycin 1.2 mg/kg every 6 hours (CT), or cefamandole 1 gm every 4 hours (M), or cefoxitin 1 gm every 4 hours (C). The antibiotics were continued for 48 hours. Major organ injuries in the three groups were comparable. The overall infection rate was significantly less in the cefoxitin group (13%), compared to cefamandole at 29%, and was comparable to the combination of clindamycin/tobramycin at 20%. The most significant difference followed colon injury. There were 96 patients who sustained colon injuries and the infection rate was CT 33%, M 62%, and C 19% (p = 0.002). If nonoperative wound infections were excluded from the colon group and only severe infections were evaluated, the infection rate was CT 18%, M 38%, and C 13% (p = 0.021). The infection rate was higher in the shock patients and tended to increase as age increased. Enterococcus, Escherichia coli, and Klebsiella pneumoniae were the most frequent aerobes isolated along with anaerobes. Five of six Bacteroides isolates from major infections occurred in the cefamandole group; two of which were in bacteremic patients. The hospital stay corresponded with infection rates, being 11.4 days (CT), 13.1 days (M), and 9.4 days (C). The results of this study indicate that cefoxitin is comparable to the combination of clindamycin/tobramycin and superior to cefamandole when used before surgery in patients sustaining penetrating abdominal trauma. The study suggests that antibiotic coverage should be against aerobes and anaerobes. Routine administration of an aminoglycoside is unnecessary.