The sensitivity and specificity of an ELISA for the detection of bovine IgG anti-Mycobacterium bovis antibodies were 73.6% and 94.1%, respectively, as determined in 53 bacteriologically confirmed tuberculous cattle and 101 healthy cattle from a tuberculosis-free area. In addition, the results of ELISA and tuberculin tests in 149 cattle were compared with those of subsequent necropsy studies. Both tests failed to detect 2 animals with tuberculous lesions and positive culture; 3/12 cattle with M. bovis isolation and no lesions, and 2/7 with atypical mycobacterial infection reacted to tuberculin, but none had antibodies; in 128 cattle with neither lesions nor mycobacterial isolation, 6 were tuberculin reactors and 7 others had antibodies. Negative results were obtained by ELISA in 21/22 paratuberculous cattle. Antibodies were not detected in 88.9% to 96.4% of 697 cattle from two tuberculin negative herds of an endemic area. In a herd with proved M. bovis infection, distribution of seropositive animals in tuberculin and non-tuberculin reactors was similar. Antibody responses to cutaneous tuberculin stimuli were observed in 4 experimentally infected cattle, but only in 2/10 healthy controls after repeated PPD stimuli. Nine controls which had either received a single tuberculin dose or none showed no increase in antibody levels. The low sensitivity of this ELISA limits its usefulness as a diagnostic tool for bovine tuberculosis eradication campaigns. However, it could be helpful in epidemiological surveillance if its efficiency to identify infected herds is demonstrated.