Antimicrobial susceptibilities of artificially created bacterial mixtures were investigated using an agar dilution technique with a standardized inoculum. Three types of combinations, ie, mixtures containing a susceptible and a resistant organism, and those containing two susceptible or two resistant organisms were tested. Organisms used in the study were previously isolated from a variety of clinical specimens. Both gram-positive and gram-negative organisms were included in the evaluation. Of the 42 combinations containing a susceptible and a resistant organism, 30 (73.8 percent) appeared as resistant, and 12 (26.2 percent) as susceptible. Of the 50 susceptible-susceptible combinations, 8 (16 percent) appeared as resistant and 42 (84 percent) as susceptible. However, in the case of 17 resistant-resistant combinations, 16 (94.2 percent) appeared as resistant and only 1 (5.8 percent) as susceptible. "Direct" or "primary" antimicrobial susceptibility testing of the clinical specimens, which are likely to contain a variety of microorganisms, can be unreliable and should not be used as a guide to antimicrobial therapy.