An enzyme-linked immunosorbent assay (ELISA) was developed and optimised using a Mycobacterium avium subspecies paratuberculosis (MAP) antigen prepared from a C strain (316v) passed through a French press. The optimised assay was evaluated with a panel of sera from MAP infected (n = 66) and uninfected (n = 1,092) sheep. Animals in the MAP infected category were positive on either tissue culture or histopathology but were of unknown serum antibody status. The diagnostic performance and cost of the assay were compared with those of a commercial ELISA (IDEXX). At 99.8% diagnostic specificity the assay showed a diagnostic sensitivity of 23% (95% CI: 15.1-35.8) compared with 36.4% (95% CI: 25.8-48.4) for the commercial ELISA (McNemar's test: chi-square 5.82, p < 0.05). The sensitivities were 5.9% (95% CI: 1-26.9), 27.9% (95% CI: 14.7-45.7) and 35% (95% CI: 18.1-56.7), for low grade, paucibacillary and multibacillary lesion grades, respectively. The cost of the commercial assay kit was 2.7 to 5.2 times greater than that of the 316v ELISA for an equivalent number of tests, the multiple depending on the number of plates processed per run. For flock-level surveillance, to account for the lower sensitivity of the 316v ELISA compared with the commercial ELISA, sample sizes would be increased but the test cost would still be lower. The 316v assay will be useful for diagnosis of Johne's disease in sheep flocks, particularly in developing countries where labour costs are low relative to the cost of consumables.