The accuracy of three diagnostic tests for paratuberculosis was evaluated using maximum-likelihood estimation of sensitivity and specificity. We also explored the variety of estimates that can be obtained if the tests are to be used in populations of different composition with regard to infection and disease states. Two enzyme-linked immunosorbent assays (ELISAs) were evaluated separately with the faecal culture (FC). The study was carried out as a cross-sectional field study to cover all likely states of infection with Mycobacterium avium subsp. paratuberculosis. The three basic assumptions for the maximum-likelihood technique were evaluated to validate the results. Our accuracy estimates for the ELISAs were not very different from those previously published, but those for faecal culture differed if a different cut-off value was chosen for the ELISA. If faecal culture was used for screening in a Danish dairy region where the median ELISA reading was a measure of the general disease situation, the sensitivity of the faecal culture was 20-25%. If faecal culture was used as a confirmatory test on cows with a high ELISA reading (and thus high level of antibodies), the sensitivity of the faecal culture would be in the range 60-70%. These results emphasise the importance of the composition of a target population before selecting a specific diagnostic test for a given purpose. We concluded that faecal culture is useful for confirmation but not for screening purposes.