A new miniaturized 4-h method for the identification of anaerobic bacteria, ATB 32 A (API System SA, Montalieu Vercieu, France), was evaluated against conventional methods of identification. The evaluation was done by using 260 recent clinical isolates and 21 reference strains of anaerobic bacteria. All reference strains were correctly identified and did not figure in the detailed analysis. Of the 140 gram-negative bacilli, 90.6% of Bacteroides spp. and 95.5% of Fusobacterium spp. were correctly identified to the species level, with an additional 8.4% of the Bacteroides spp. being identified to the genus level. Clostridia were correctly identified in 85.9% of cases, with an additional 9.9% being identified to the genus level. Peptostreptococci were correctly identified in 91.6% of cases. The 4 strains of Actinomyces spp. were all identified correctly, as were 10 of the 11 strains of Propionibacterium spp. A total of 3.1% of strains were not identified by ATB 32 A, while for 1.9% of strains, completely false identifications were obtained. Estimation of the individual preformed enzyme results may pose problems, although these decrease with familiarity with the system. With certain enzyme profiles, additional testing was necessary to arrive at an identification; however, there was no requirement for gas-liquid chromatography. If certain additions are made to the data base and the difficulties of determination of organisms to the species level among the non-Bacteroides fragilis (sensu stricto) members of the B. fragilis group can be reduced, this system holds promise as a reliable standardized alternative for the identification of anaerobic bacteria in clinical laboratories.