A randomized trial of ticarcillin plus gentamicin (group 1), ticarcillin plus amikacin (group 2) and ticarcillin plus netilmicin (group 3) as empiric antibiotic therapy in patients with granulocytopenia and cancer was carried out at the Baltimore Cancer Research Center. The response rate for all infections was 97 per cent in group 1, 91 per cent in group 2 and 95 per cent in group 3. Patients with bacteremias showed improvement in 93 per cent (group 1), 78 per cent (group 2) and 82 per cent (group 3) of cases. All failures were among patients with gram-negative bacteremias. Both antibiotic susceptibility of the bacteremic organism and granulocyte recovery correlated with patient improvement. Nephrotoxicity and ototoxicity were rare and were not significantly different in three groups of patients. Therefore, ticarcillin plus gentamicin, ticarcillin plus amikacin and ticarcillin plus netilmicin appear to be equally efficacious and minimally toxic in this patient population. Excellent over-all results can be expected with these combinations provided the etiologic agent is susceptible.