Automated microdilution MIC results, obtained with the Autoreader (Sensititre, Inc., Salem, N.H.) following 5 h of incubation, were compared with manually read, concurrent control MICs following 18 h of incubation in a three-laboratory comparative study. A total of 704 members of the family Enterobacteriaceae or similar gram-negative organisms were tested against 17 antimicrobial agents. Autoreader MICs were within 1 doubling dilution of control values in 92.9% of instances. Discrepancies of +/- 2 doubling dilutions and +/- 3 or greater doubling dilutions were noted in 4.5 and 2.6%, respectively, of the 7,687 drug-organism combinations analyzed. The majority of errors occurred when beta-lactam antimicrobial agents were tested with a variety of different species. MICs at 5 h, when Pseudomonas aeruginosa was used, were possible in only half the isolates tested and yielded data on only a limited number of drugs in the remaining instances. Excluding results obtained with penicillin and ampicillin, which were uniformly poor, Staphylococcus aureus Autoreader values were within +/- 1 doubling dilution of control values in 93.6% of instances, 5.4% varied by +/- 2 dilutions, and only 1% of test values by +/- 3 or more dilutions from control values for 82 isolates tested against nine antimicrobial agents. Of eight additional S. aureus isolates tested that were resistant to methicillin, only one was read correctly by the Autoreader, with results on the remaining seven appearing as either insufficient growth or as total resistance to all drugs tested. Interlaboratory reproducibility was excellent for selected isolates of S. aureus and gram-negative bacilli. The accuracy of the Sensititre Autoreader MIC results was comparable to that of other same-day quantitative systems for members of the family Enterobacteriaceae and S. aureus, while the economic and procedural advantages of the broth microdilution method was maintained.