In a prospective, randomized trial of 195 febrile episodes in granulocytopenic patients short course aminoglycoside treatment (initial tobramycin and cefoperazone followed by tobramycin discontinuation at day four of therapy) was compared with two regimens (tobramycin plus cefoperazone and tobramycin plus mezlocillin) in which both drugs were continued for up to 26 days. All regimens were successful as empirical therapy with comparable response rates of just over seventy per cent. Fifty-three per cent of the initial episodes of fever were related to documented infections which responded less well (P = 0.007) than unexplained fever. Patients with bacteraemia, pneumonia or Gram-positive aerobic or Pseudomonas aeruginosa infections responded poorly to all regimens. The recovery from granulocytopenia was the most important determinant of successful response. Aminoglycoside discontinuation followed by cefoperazone monotherapy after day four was statistically as effective as the combination regimens. Short course tobramycin therapy eliminated the nephrotoxicity seen in the combination limbs. The use of cefoperazone was not associated with an increased incidence of hypoprothrombinemia; however, the only three bleeding episodes occurred in patients given cefoperazone but not vitamin K. Short course aminoglycoside therapy will reduce cost and nephrotoxicity when compared with prolonged combination therapy and should be further explored in this setting, with use of different agents and comparison with monotherapy.