The majority of clinical microbiology laboratories in the UK use comparative disc diffusion methods based on the Stokes' method to determine antibiotic susceptibility. The technical validity of the results obtained from the modified Stokes' method of disc testing and how they relate to MIC data are not known. We studied susceptibility testing using a modified Stokes' disc diffusion method for a wide range of clinical isolates against which MICs had been determined by collaborators not involved with the disc testing evaluation. Results indicated that for 1301 organism-antibiotic combinations the number of major errors (where resistant strains were reported as sensitive) was 21/468 (4.4%) and the number of minor errors (where sensitive strains were reported as resistant) was 14/713 (1.9%) using ciprofloxacin breakpoints of 0.5 and 2 mg/L. There was good correlation between the disc susceptibility test and the MIC for 119 isolates of Enterobacteriaceae tested with the exception of Serratia spp. Excluding Serratia spp. the number of major errors for Enterobacteriaceae was 1/200 (0.5%). Data revealed 2/25 (8%) major errors for Pseudomonas aeruginosa and 1/45 (2.2%) for Acinetobacter spp. Haemophilus influenzae showed a number of unexpected categorization errors. The modified Stokes' method performed accurately for Staphylococcus aureus and coagulase-negative staphylococci when tested for susceptibility to gentamicin, erythromycin, teicoplanin and vancomycin. No major errors were reported for Streptococcus pneumoniae and beta-haemolytic streptococci. Problems occurred with the detection of antibiotic resistance in Enterococcus spp. Major errors were seen for ampicillin (2/12 strains), teicoplanin (5/6 strains) and vancomycin (5/13 strains) using a 30 microg disc but only 1/13 strains using a 5 microg disc. Overall, from our data, the modified Stokes' disc diffusion antibiotic susceptibility test showed an unacceptable number of major errors but an acceptable number of minor errors.