The early diagnosis of invasive fungal infection in granulocytopenic patients remains unreliable. Granulocytopenic patients who are persistently or recurrently febrile despite therapy with appropriate antibacterial agents are at high risk for the development of such infection. Two randomized clinical trials demonstrated that the empiric administration of amphotericin B to persistently or recurrently febrile granulocytopenic patients decreased the frequency, morbidity, and mortality of invasive fungal infection; these effects were especially marked in profoundly granulocytopenic patients who were not receiving antifungal prophylaxis. Current studies continue to indicate that prompt empiric administration of amphotericin B to persistently or recurrently febrile granulocytopenic patients ensures earlier treatment of deep mycoses. The roles of newer antifungal triazole compounds and of liposomal and lipid complexes of amphotericin B in empiric antifungal therapy must be investigated further in thoughtfully designed, randomized clinical trials.