Amikacin plus penicillin (A+P) was compared to amikacin plus carbenicillin (A+C) in a double-blind study. Therapy with one of these combinations was given, as soon as servere infection was suspected, to 117 patients with proved gram negative infection, none of whom was granulocytopenic. Gram negative bacteremia was documented retrospectively in 52 patients; 25 had received A+P and 27 had been treated with A+C. All the isolated gram negative pathogens were sensitive to amikacin (MIC less than 12 microng/ml). In the A+P group, 55 per cent of the patients responded favorably while in the A+C group 63 per cent did respond; the difference was more striking for bacteremic patients: 52 per cent responded in the A+P group and 70 per cent in the A+C group. This difference, however, was not statistically significant. The outcome of patients whose infection was treated by synergistic combinations against the offending pathogen was better (66 per cent) than that observed in patients who received nonsynergistic combinations (48 per cent) (p less than 0.05). Once again the results were more striking in the bacteremic patients (p less than 0.01). A favorable outcome was associated also with a high (larger than or equal to 1/8) bactericidal activity of the diluted serum of the treated patient against the offending pathogen (p less than 0.05). This study suggests that the optimal therapy in gram negative septicemia might be the administration of synergistic combinations of antibiotics.