The febrile neutropenic patient with cancer requires the prompt administration of empirically chosen antibiotic therapy. For many years the choice has usually been an aminoglycoside plus either a cephalosporin or an antipseudomonal penicillin; the exact combination was dependent on local patterns of infection and especially on local patterns of microbial resistance. In the absence of resistance problems, one normally would choose the least expensive aminoglycoside and, probably, the least expensive antipseudomonal penicillin. A new approach to be considered is the use of single agents with very broad spectra such as ceftazidime or imipenem. However, such an approach is fraught with danger--that of leaving a resistant pathogen "uncovered" and of eliminating the possible benefits of synergistic activity against bacteremia due to gram-negative organisms in the profoundly neutropenic patient. Combinations of two beta-lactam agents have the advantage of not being nephro- or ototoxic, although with a few such combinations antagonism has been demonstrated. A final approach is the use of regimens that include aztreonam, ceftazidime, or imipenem, which have little or no suppressive effect on stool anaerobes. Presumably, with these agents colonization resistance will be preserved and the opportunities for secondary infection with a resistant organism will be reduced.