The ELISA test, although lacking sensitivity for subclinical infections, is a very convenient diagnositic method because it is quick and inexpensive. In order to determine the relative performance of each of four commercially available ELISA protocols, we carried out an assay where four of these tests (A, B, C, D) plus a locally modified PPA3 based ELISA (NEIKER) were compared. We used two sets of sera. One consisted of 20 suspect fecal culture confirmed cases, and 174 animals from a herd with no history of clinical paratuberculosis, nor serological reactors in the last 5 years. The other set was composed of 201 adult cattle killed in a slaughterhouse on which tissue culture and PCR and histopathology of the ileocecal valve and lymph node were used for classification into infected or non-infected. All the 395 sera were submitted to the NEIKER test, but only the first set was used for the five tests comparison. Overall, the NEIKER ELISA showed a sensitivity of 54.4% and an specificity of 93.2%, with an 87.6% agreement (kappa: 0.49, p<0.05) with the reference scoring. On the first set, our ELISA showed 100% sensitivity and specificity, but only ELISA A was close to the performance of NEIKER test showing a sensitivity of 95.0% and an specificity of 100% with a 95.5% (kappa: 0.97, p<0.05) agreement. The other tests showed a sensitivity ranging from 32.3% to 88.9% and an specificity ranging from 30.0% to 98.9%, with an agreement ranging from 47.5% to 96.9%. These results indicate that there might be great variability in ELISA performance according to the antigen and procedure used.