The performance of a commercially available ELISA for detection of antibodies to Mycobacterium paratuberculosis was evaluated using sera from 1,146 cows. Samples were from uninfected cattle, infected subclinical cattle shedding low numbers of organism in feces, subclinical heavy shedders, clinical cases, and randomly selected cattle in a slaughterhouse survey for paratuberculosis. The overall sensitivity of the test, using the manufacturer's recommended cutoff was 45% +/- 4.8%, and the specificity was 99% +/- 0.9%. The ELISA result was significantly correlated with the number of colonies of M. paratuberculosis detected by fecal culturing. The sensitivity of the test was highest for clinical cases of paratuberculosis (87% +/- 8.4%), and lowest for subclinical, light-shedding cattle (15% +/- 6.6%). Changing the cutoff point did not improve performance of the test. Evaluating ELISA results with a kinetic-based method reduced plate-to-plate variation in results but did not improve performance of the test based on receiver-operating characteristic curve analysis.