AIMS--To evaluate adenosine triphosphate (ATP) bioluminescence as a rapid technique for antimicrobial susceptibility testing of Mycobacterium spp by comparing it with conventional and radiometric methods, and to assess its potential for use in clinical microbiology laboratories. METHODS--115 clinical isolates from a wide range of mycobacterial species and four control organisms of known susceptibility were tested against six antimicrobial agents. Minimum inhibitory concentrations (MICs) were determined after 4-6 weeks' incubation on Middlebrook 7H10 agar. Susceptibility was also determined radiometrically using a Bactec 460, and by bioluminescent assay of ATP using a 1250 luminometer (LKB-Wallac). RESULTS--Susceptibility results after 7 days showed excellent correlation with conventionally determined MICs. 714 susceptibility tests were performed by both techniques, with seven major discrepancies between the two systems. For pyrazinamide, agreement was 100%, but five strains of M tuberculosis, including one control, and 11 mycobacteria other than M tuberculosis (MOTT) failed to grow on Middlebrook agar at pH 5.5. 606 tests were performed by radiometry, with four major discrepancies between this technique and ATP bioluminescence. No particular species of Mycobacterium gave aberrant results. Contamination was a problem; 12 of the 119 strains tested were contaminated at day 1 and had to be repeated before results were obtained. Contamination of individual tests increased significantly after 7 days of incubation. CONCLUSIONS--ATP bioluminescence can be used to monitor mycobacterial growth in fluid culture media; the technique has considerable potential for rapid susceptibility testing. Advantages include lower initial cost of analytical equipment, lower reagent cost per test, and the use of non-radioactive substrates.