Ticarcillin was used in combination with either cephalothin or gentamicin as initial empiric antibiotic therapy for 127 patient trials of suspected infection in granulocytopenic cancer patients. Bacteremia was present in 20%, nonbacteremic microbiologically documented infections in 21%, clinically documented infections in 23%, and possible infections in 5%; infection was doubtful in 31%. Although Staphylococcus aureus was the most common single organism isolated (23%), gram-negative bacilli accounted for 54% of all pathogens. Both antibiotic regimens were highly efficacious, with complete resolution in 46% of bacteremias, 88% of nonbacteremic microbiologically documented infections, and 95% of clinically documented infections. Among bacteremias, 8 of 9 caused by S. aureus but only 4 of 15 (27%) caused by gram-negative bacilli were completely resolved with these antibiotic combinations. Reasons for nonresponse in bacteremias were persistent granulocytopenia, mixed infection and, in two patients, antibiotic-resistant organisms. Toxicities other than hypokalemia were minimal. Although the rate of further infections was high overall (18/127), only one occurred among the 39 patients with <4 days of antibiotic therapy. Ticarcillin in combination with either cephalothin or gentamicin was effective as initial empiric therapy of suspected infection in granulocytopenic cancer patients.