A model of surgical wound infection that uses low inocula of bacteria and closely simulates clinical infection involved inoculating suspensions of Staphylococcus aureus and dextran microbeads into intermuscular sites on the dorsum of guinea pigs, harvesting lesions at 72–96 h, identifying as a positive end point lesions yielding staphylococci on subculture, and using logistic regression for data analysis. Prophylaxis was placebo, ampicillin, or cefazolin, and three representative strains of S. aureus were used. A highly significant correlation (P < .001) was observed between inoculum sizes and infection rates. Without antimicrobial prophylaxis, ID50 for each strain was 50 was significantly higher. Differences in the virulence of strains and in the efficacy of the antimicrobial regimens also were observed. The model should prove useful for understanding mechanisms of virulence among pathogenic bacteria and for elucidating subtle but important differences in efficacy among antibiotics used in prophylaxis.