Routine determination of antimicrobial susceptibility of anaerobic clinical isolates is difficult. The E-test is a practical alternative technique that we evaluated while testing clinical isolates in a multicenter study. The susceptibility to 9 antibiotics (penicillin, amoxycillin/clavulanate, ticarcillin/clavulanate, piperacillin/tazobactam, imipenem, cefoxitin, metronidazole, clindamycin, chloramphenicol) of 351 strains belonging to 63 different species was determined by the NCCLS reference agar dilution procedure using Wilkins-Chalgren agar medium with 5% sheep blood and was compared to the E-test performed on the same medium and using manufacturer's recommendations. The MIC values obtained with the E-test were generally one dilution lower than those obtained with the reference technique, 87.1% of the results being within two dilutions. In terms of susceptibility categories, 95.1% agreement was observed with 3.8% minor errors and only 0.5% major and 0.6% very major errors. With some Fusobacterium spp. and Clostridium spp. strains, the E-test was difficult to read or not interpretable because of the presence of growth within the inhibition zone of all beta-lactam antibiotics, representing a trailing phenomenon. We conclude that, if some interpretation difficulties are taken into account, the E-test is a convenient and reliable technique that can be applied in all clinical laboratories. It could be used for the individual testing of important anaerobes in certain clinical situations but cannot yet be considered as a reference technique. Its utility is emphasised by the increased resistance rate against clindamycin and the appearance of a few strains in the B. fragilis group with a reduced susceptibility against metronidazole.