Five diagnostic tests based on ELISA technology for bovine paratuberculosis were evaluated using individual serum or milk samples from 359 dairy cattle in seven paratuberculosis-free herds and 2,094 dairy cattle in seven M. paratuberculosis-infected dairy herds. Three independent laboratories using three different culture procedures completed fecal cultures for M. paratuberculosis on these cattle and found 417 cows to be shedding M. paratuberculosis in their feces. An animal that was fecal culture-positive for M. paratuberculosis by any of the three laboratories was considered a confirmed case of infection. The specificity of three ELISAs (two on serum and one on milk) was 99.8%. The specificity of the remaining two ELISAs, both done on serum, was 94.9% and 84.7%. Four of the five ELISAs evaluated produced similar sensitivity in detecting fecal culture-positive cattle (27.8% - 28.9%). Serum ELISA "D" had the lowest specificity (84.7%) and the highest sensitivity (44.5%). If the case definition for M. paratuberculosis infection was based on the culture results of a single laboratory instead of the combined results of three laboratories, ELISA sensitivity estimates were 45.7% to 50.0%. With the exception of ELISA D, assay agreement was high (kappa 0.66 to 0.85) for categorical assay interpretations (positive or negative) but linear regression of quantitative results showed low correlation coefficients (r² = 0.40 to 0.68) due to the fact that ELISA results for some cows were high in one assay but low in another assay. Likelihood ratio analysis showed a direct relationship between the magnitude of ELISA result and the odds of a cow shedding M. paratuberculosis in its feces. If used judiciously and interpreted quantitatively, these ELISAs are useful tools in support of paratuberculosis control programs in dairy herds.