The performance of disk diffusion testing of Haemophilus influenzae was evaluated in 20 laboratories. Thirteen disk-medium-breakpoint-inoculum modifications were used in Finnish clinical microbiology laboratories. The performance of various methods was evaluated by testing a susceptible control strain and one with non-β-lactamase-mediated ampicillin resistance 10 times in 16 laboratories. Gaps in millimeters were measured between these two groups of results. The strains were separated by a gap of at least 5 mm in 8/16 laboratories testing ampicillin, in 7/15 laboratories testing cefaclor, in 5/16 laboratories testing cefuroxime, and in 15/16 laboratories testing trimethoprim-sulfa. Detection of ampicillin resistance was better with 2.5μg tablets than with 10 μg disks or 33 μg tablets. For MIC-determinations, 785 isolates and their disk diffusion results were collected. None of the 12 clinical isolates with non-β-lactamase-mediated ampicillin resistance was detected as resistant in the participating laboratories. The ampicillin and cefaclor results of the isolates were no better even when a laboratory was able to separate the control strains. Cefaclor results were unreliable because of poor disk diffusion-MIC correspondence and incoherent breakpoint references. Interlaboratory variation of the zone diameters caused false intermediate results of cefuroxime-susceptible strains. When ampicillin, cefaclor and cefuroxime were tested, the discrimination of laboratories using disks and tablets was equal, whereas the laboratories using paper disks were better able to detect trimethoprim-sulfa resistance.