We examined the efficacy of ciprofloxacin as an empirical treatment for fever in 97 neutropenic patients in a randomized study of ciprofloxacin and benzylpenicillin versus netilmicin and piperacillin. Benzylpenicillin was included because of evidence of in-vitro resistance to ciprofloxacin in some streptococci. Clinical response rate was similar in the two groups (46% resolution for ciprofloxacin/benzylpenicillin and 52% for netilmicin/piperacillin). Microbiological assessment revealed more pathogens eradicated by ciprofloxacin and benzylpenicillin (66%) and fewer persisting (3%) than in patients receiving netilmicin and piperacillin (52% and 13% respectively). Staphylococcus epidermidis was the commonest pathogen, accounting for 38% of all isolates and 30% of all treatment failures. There were no treatment failures or superinfections due to streptococci. More therapy-related adverse reactions were seen in patients on netilmicin and piperacillin (28%) compared with those on ciprofloxacin and benzylpenicillin (10%). The combination of ciprofloxacin and benzylpenicillin is as effective as a standard regimen of netilmicin and piperacillin, with fewer adverse effects, and is highly attractive as empirical therapy for the febrile, neutropenic host. The inclusion of benzylpenicillin prevents streptococcal-associated treatment failure.