Seven commercial systems for antimicrobial susceptibility testing of Streptococcus pneumoniae were evaluated by using a challenge set of 55 pneumococcal isolates with a variety of resistance phenotypes and genotypes. Overall, the results produced by the Pasco and Etest methods were found to be acceptable for all drugs tested except for trimethoprim-sulfamethoxazole testing by the Etest. The Just One system for penicillin MIC testing was also judged to be acceptable (minor error rate, 5.5%). Although the Sensititre and MicroTech methods both produced 12.7% minor errors with penicillin, the Sensititre method classified penicillin-intermediate strains as resistant or vice versa, while four of MicroTech's errors were among intermediate strains that were classified as susceptible. The MicroMedia (minor error rate, 16.4%) and MicroScan Rapid (minor error rate, 63.6%) methods produced unacceptably high levels of errors when testing penicillin. Minor error rates for cefotaxime and ceftriaxone ranged from a low of 12.7% (Etest and Sensititre) to a high of 28% (MicroMedia). Error rates were low for erythromycin, tetracycline, and chloramphenicol by most methods with the exception of the MicroScan method, which had a high very major error rate for erythromycin (34.6%). For testing of beta-lactam drugs, the Pasco, Etest, and Just One tests for penicillin are the most accurate methods; the Sensititre method also provided acceptable results.