The efficacy of nafcillin and gentamicin used alone and in combination at doses giving serum concentrations comparable to those achieved in patients was studied in rabbits with experimental Staphylococcus aureus endocarditis. The organism used was a penicillinase-producing, methicillin-susceptible, clinical isolate. The addition of gentamicin to nafcillin significantly increased the rate of killing of organisms in valvular vegetations, compared to the effect of nafcillin alone. Gentamicin alone delayed mortality but was not effective in reducing the bacterial populations of the vegetations. Bacteremia persisted in the animals treated with gentamicin alone, in contrast to the groups treated with nafcillin or the combination. Selection of a subpopulation of aminoglycoside-resistant small-colony variants occurred in animals treated with gentamicin alone. This variant was subsequently employed in the rabbit model and produced endocarditis, metastatic infection, and bacteremia comparable to those caused by the parent strain. Animals with infection produced by the variant died later than animals infected by the parent strain. Nafcillin was equally effective in reducing the population of both parent and variant strains in vitro and in therapy of the infected animals. Population studies showed the variant to be a mutant emerging at a rate of 1.9 x 10(-7). It was shown to differ from the parent strain in coagulase and hemolysin production, colonial morphology, and aminoglycoside susceptibility, but was similar by light and electron microscopy and in phage type, pigmentation of colonies, deoxyribonuclease production, mannitol fermentation, and growth rate.