In a reproducible infected rat burn model, subeschar infusion of antibiotics failed to protect the animals from death via burn wound invasion excepting those animals receiving carbenicillin. Subcutaneous injection of maximal doses of carbenicillin at a distance from the burn wound protected these animals equally well. Some advantage was defined for the subeschar route of administration with suboptimal doses of carbenicillin. More important is the fact that prospective selection of an effective antibiotic could not be made on the basis of in vitro antibiotic sensitivity tests.