There is a growing need for a more accurate, rapid, and cost-effective alternative to conventional tests for identification of clinical isolates of Mycobacterium species. Therefore, the ability of the Sherlock Mycobacteria Identification System (SMIS; MIDI, Inc.) using computerized software and a Hewlett-Packard series 1100 high-performance liquid chromatograph to identify mycobacteria was compared to identification using phenotypic characteristics, biochemical tests, probes (Gen-Probe, Inc.), gas-liquid chromatography, and, when necessary, PCR-restriction enzyme analysis of the 65-kDa heat shock protein gene and 16S rRNA gene sequencing. Culture, harvesting, saponification, extraction, derivatization, and chromatography were performed following MIDI's instructions. Of 370 isolates and stock cultures tested, 327 (88%) were given species names by the SMIS. SMIS software correctly identified 279 of the isolates (75% of the total number of isolates and 85% of the named isolates). The overall predictive value of accuracy (correct calls divided by total calls of a species) for SMIS species identification was 85%, ranging from only 27% (3 of 11) for M. asiaticum to 100% for species or groups including M. malmoense (8 of 8), M. nonchromogenicum (11 of 11), and the M. chelonae-abscessus complex (21 of 21). By determining relative peak height ratios (RPHRs) and relative retention times (RRTs) of selected mycolic acid peaks, as well as phenotypic properties, all 48 SMIS-misidentified isolates and 39 (91%) of the 43 unidentified isolates could be correctly identified. Material and labor costs per isolate were $10.94 for SMIS, $26.58 for probes, and $42.31 for biochemical identification. The SMIS, combined with knowledge of RPHRs, RRTs, and phenotypic characteristics, offers a rapid, reasonably accurate, cost-effective alternative to more traditional methods of mycobacterial species identification.