Empiric therapy is necessary for febrile, neutropenic patients in order to minimise morbidity and mortality. Certain agents are now available for monotherapy which offer comparable success to combinations of either an aminoglycoside with a beta-lactam or two beta-lactams. However, no regimen offers complete treatment under all circumstances in all patients. It is also apparent that febrile, neutropenic patients comprise a more heterogeneous group than just those with bacteraemia, clinically apparent infection and unexplained fever. Localized infections occur in just under a third of cases at the onset of fever and a similar number will develop during the course of fever. Mortality is higher in infections that are accompanied by bacteraemia and also those that develop subsequently, especially when related to the lung. The aetiological agent also differs with each type of infection as does the duration of fever and symptoms. Consequently modifications are required more often. The length of treatment may also differ. Therefore, during the first 3-4 days of empiric therapy, every effort should be made to identify incipient localized infections in addition to detecting bacteraemia. Changes in therapy can then be based on objective grounds rather than continued fever offering more patients individual treatment than is possible when relying only on the temperature chart.