Mechanisms to explain the "inoculum effect" have not been elucidated in gram-positive infections. A mouse model of group A streptococcal myositis was used to compare the efficacies of two beta-lactams, penicillin and ceftriaxone, and a protein synthesis inhibitor, clindamycin, at three different inoculum sizes. beta-lactams were more susceptible to inoculum effects than was clindamycin both in vivo and in vitro (P < .05). The large inocula were hypothesized to reach stationary phase of growth sooner than smaller inocula both in vitro and in vivo. The penicillin-binding protein (PBP) patterns from membrane proteins isolated from mid-log-phase and stationary-phase cultures of Streptococcus pyogenes were compared. Binding of radiolabeled penicillin by all PBPs was decreased in stationary cells; however, PBPs 1 and 4 were undetectable at 36 h. Thus, the loss of certain PBPs during stationary-phase growth in vitro may be responsible for the inoculum effect observed in vivo and may account for the failure of penicillin in both experimental and human cases of severe streptococcal infection.