We prospectively randomized 27 granulocytopenic patients who experienced a total of 30 episodes of gram-negative septicemia. The control group received an appropriate antibiotic regimen alone, whereas the "transfusion" group received infusions of granulocytes in addition to the antibiotics. Five of 14 controls survived, and 12 of 16 in the transfusion group survived, and 12 of 16 in the transfusion group survived (P less than 0.04). An important factor in the outcome of treatment was the recovery of bone-marrow function (return of peripheral granulocyte count greater than or equal to 1000 per microliter). Eighty-three per cent (five of six) of the control group and all (four of four) of the transfusion group with recovery of granulocyte levels survived the episode of sepsis. In contrast, none of the eight control patients, as compared to 67 per cent (eight of 12) of the transfusion group, survived persistent granulocytopenia (P less than 0.005). Granulocyte transfusions appear to complement appropriate antibiotic treatment of gram-negative-septicemia due to granulocytopenia.