A 23 kDa protein of the M. paratuberculosis A36 complex is immunodominant in Johne's disease. The DNA fragment corresponding to the carboxylic end of the protein coded for a polypeptide containing B-cell epitopes present in all tested M. paratuberculosis strains¹. This a362 polypeptide, which represents the first specific recombinant antigen of M. paratuberculosis, was used to develop an ELISA test for paratuberculosis serology. The a362-ELISA was applied to the sera from 300 randomly chosen bovine representative of a large Belgian region. The distribution of the ELISA values frequency classes was then analyzed according to a mixture population model². A bimodal distribution was found to be the most appropriate to accommodate our results: it included 87.7% of low titer sera and 12.3% of high level sera. This 12% prevalence of a362-seropositivity in the Belgian population was close to those estimated by others for paratuberculosis in different countries. The optimum cut-off value separating these two populations was estimated by the method of Anderson³, assuming that the prevalence of antibodies directed to peptide a362 is equal to the proportion of the high lg titer population (p=0.12) and that the cost of a false negative diagnosis is twice as great as that of a false positive one. Accordingly, five sera with ELISA-absorbance values corresponding to the determined cut-off were chosen as reference points in each ELISA test. The serological test was then applied to some 30 reference sera from paratuberculous-certified cattle belonging to the U.S. National Repository (MT Collins, Univ. Wisconsin) and to five Belgian herds: two herds of healthy animals (H1=110, H2=65), two herds in which paratuberculous cattle were found (P1-144, P2=64) and one herd with tuberculosis (T=38). Sensitivity of the a362-assay was 70% according to the reference sera. For the healthy and tuberculous herds, the distribution of anti-a362 lg followed a monomodal model, whereas a bimodal distribution was found for the herds with paratuberculous animals, indicating that some 40% were found positive to the assay. This figure is probably an underestimation of the a362-ELISA sensitivity because not all animals were infected by M. paratuberculosis. The sensitivity of the assay is estimated to be in the range of 40.8 to 70%. From the percentage of false positive results in herds H and T, we can estimate that the specificity of the a362-ELISA with respect to healthy and tuberculous animals is 95.0%. In conclusion, the sensitivity of the a362-ELISA, based on a recombinant product, is practically similar to that of other ELISA tests, based on natural antigens or on mixtures of mycobacterial components. ¹ J Clin Microbiol, 1993, 31:947-954. ² J Virol Meth, 1990, 27:135-14; ³ An introduction to multivariate analysis. 1958. Wiley NY, 126-131.