The selection of appropriate antimicrobial therapy for patients with sepsis is discussed. Antimicrobial selection is based on the most likely source of infection, the most common pathogens at that site, knowledge of antimicrobial susceptibility patterns in the local community and institution, and host factors. Prompt initiation of appropriate antimicrobial therapy is essential for achieving the best possible outcomes. Defining optimal antimicrobial therapy for sepsis is difficult because few clinical studies have specifically addressed this. Sepsis has traditionally been associated with gram-negative infections, but studies indicate that gram-positive pathogens commonly cause sepsis and are associated with high mortality. Enterobacteriaceae, species, Pseudomonas aeruginosa, and Staphylococcus aureus are the most common pathogens in sepsis patients and should be considered in the selection of empirical therapy. Antimicrobial therapy should be initiated as soon as samples have been obtained for culture; empirical therapy should cover a very broad spectrum of organisms, with consideration given to the most likely pathogens for the infection site. An antipseudomonal beta-lactam with or without an aminoglycoside is recommended for initial therapy in most patients. The role of monotherapy versus combination therapy is controversial. Routine use of vancomycin and antifungal therapy as part of initial regimens should be discouraged but may be justified in specific circumstances.