The control of paratuberculosis is becoming a major issue for French breeders. The Laboratoire Vétérinaire Départemental of Haute-Vienne has been involved for years in routine controls of the status of high genetic value herds by serological testing and coproculture. The aim of this study is to compare the performances of three commercially available ELISA tests that may be used in large scale control programs. The specificity was evaluated on sera (n=250) coming from closed herds declared as negative following controls by coproculture every six months that did not gave any positive result for at least 6 years. Tests A, B and C gave specificities of 100%, 97.7% and 67% respectively (doubtful considered as negative). As there is no perfect method allowing paratuberculosis diagnosis on living animals, the sensitivity was evaluated on sera coming from all the animals (n=90) of a herd considered as infected on the basis of frequent coproculture positive results within the herd. Test A detected 23 positive sera, test B detected 32, test C detected 56 (doubtful considered as negative). Test A and B did not detect always the same sera in the infected herd suggesting some differences on antigens used for the preparation of the plates. Distribution of optical densities suggested that a better sensitivity (36 sera detected instead of 23) could be obtained for test A without altering specificity just by lowering the cut-off. The apparently good sensitivity of test C is difficult to interpret because of its very poor specificity: a lot of the extra positive sera may be false positive. The detectability was evaluated by serial dilutions of a pool of sera coming from coproculture positive animals. Test A detected of a two fold higher dilution of the pool of positive sera than test B and C which were equivalent. This work suggests that strong efforts must be done to standardise serological reagents for paratuberculosis and create panel of reference sera for the evaluation of performances of the different techniques.