Two hundred fifty-eight patients with suspected sepsis were treated with tobramycin or gentamicin in a prospective, randomized, double-blind trial. One hundred forty-six patients received nine or more doses, had serial determinations of serum creatinine, and were evaluated for nephrotoxicity; 91 were able to cooperate with audiometry and were evaluated for auditory toxicity. Auditory toxicity developed in five of 47 (10 per cent) given gentamicin and five of 44 (11 per cent) given tobramycin. Nephrotoxicity developed in 19 of 72 (26 per cent) given gentamicin and nine of 74 (12 per cent) given tobramycin (P less than 0.025). The severity of the nephrotoxicity was not different; the mean increase in creatinine was 1.3 mg per 100 ml (114.9 mumol per liter) in both groups. Both the tobramycin and gentamicin groups had a similar mean age, initial serum creatinine level, total dose, serum aminoglycoside level, and duration of therapy. We conclude that tobramycin causes nephrotoxicity less frequently than does gentamicin.