During a 10-month period in 1999, 994 serum and tissue samples were collected from dairy cows at slaughter in eastern Canada. The sources of these cattle were from all four Atlantic Canadian provinces along with some cows from the state of Maine. The sera were used to assess the agreement of three commercially available ELISAs for Mycobacterium avium subsp. paratuberculosis. Two ELISAs were indirect absorbed ELISAs licensed for use in North America, the third was an indirect non-absorbed ELISA licensed for use in Europe. Overall, there was poor agreement between the three ELISAs. The highest and lowest kappa values were 0.33 and 0.18, which is fair and poor agreement, respectively. However, when only tissue culture-positive cattle were compared, the ELISAs had better agreement (kappa=0.37-0.51). The proportions of positive tests, however, were significantly different among the three ELISAs. The poor agreement among the three ELISAs is as concerning as the fact that these tests have low sensitivity. The implications are greatest when the tests are used at the cow level to make individual animal decisions, which is not the recommended method on the product labels. At the cow level, if the result obtained from one ELISA is positive, using a different ELISA in a pre-clinical animal has a high likelihood of giving a different result due to low predictive values of positive test results.