The in vivo relevance of the paradoxical bactericidal effect (the Eagle effect) is not evident. We found in vitro a paradoxical bactericidal effect of amoxicillin on 2 strains of nontoxigenic Corynebacterium diphtheriae. Then, using an experimental rabbit model of endocarditis, we evaluated the in vivo relevance of this phenomenon. Rabbits were assigned to the following groups: no treatment (control group), continuous amoxicillin infusion simulating a dosage of 200 mg/kg/day in humans, and continuous amoxicillin infusion simulating a dosage of 20 mg/kg/day in humans. The low dosage (20 mg/kg/day) was significantly more effective than the high dosage (200 mg/kg/day) against both strains (P<.025), confirming the paradoxical bactericidal effect observed in vitro.