Each of four laboratories conducted three runs of the IDEXX Johne's Antibody assay. Each run consisted of two plates: one was from a kit lot used at all laboratories, and the other from a kit lot unique to each laboratory. A panel of 10 sera was tested in duplicate in all plates at each laboratory. Also, 108 sera were chosen and run by each lab with 1/3 of the sera tested in duplicate in each of the 2 plates per run. Thus, 432 sera were used to evaluate between-duplicate variability, and ten serum standards were used to assess variability between runs of the assay as well as between kit lots within the same run of the assay. For the kit lot used among all labs, the between-duplicate median coefficient of variation (CV) for the 432 samples was 3.4% (range between labs was 2.6% to 4.9%). The overall between-run CV, based on optical density (OD) readings of the 10 standards, averaged 17.3% (range between labs was 4.1% to 27.6%). Such variation in raw ODs between runs is not unusual in ELISA. Accordingly, ODs are usually adjusted (normalised or indexed) to a serum control(s), reducing the variation between runs. In the IDEXX assay, a ratio [(sample OD - neg control OD)/(pos control OD - neg control OD)], is calculated for this purpose. When this was done, the variability unexpectedly increased to a mean CV among laboratories of 22.0% (range = 11.1% to 31.7%; outliers not included to reduce bias) for the 10 serum standards. The negative control OD was eliminated from the calculation of ratios. The CVs for the resultant S/P ratios (18.6%) indicated that factors other than the negative control were responsible for the variation. Since the positive control had low activity (mean OD of 0.320), one of the 10 standard sera (mean OD of 1.288) was substituted for the positive control in the S/P calculation to reduce the impact of a potentially large systematic and/or random error factor. Unfortunately, this did not change the CVs significantly. The IDEXX Johne's antibody assay is variable, resulting in possible misclassification of animals having ratios near the assay's cutoff. The extent of this potential problem currently is being determined by evaluating over 10,000 results from routine runs of the assay. These data will be discussed.