Regular 6 monthly testing by absorbed ELISA and g-interferon was undertaken in an M. paratuberculosis-infected stud beef cattle herd in NSW over the 9 year period 1987 to 1995. Necropsy examination of 835 cattle was undertaken on test reactors, relatives of infected cattle and routine culls, representing 75% of all animals culled from the herd. 133 infected cattle were identified at slaughter, with 96 of these receiving full necropsy examination. Of these 96, 79% were confirmed histologically and all culturally. On the basis that part examination (histology without culture) had a sensitivity of 79%, and with the exclusion of 19 cattle with equivocal histology, it was assumed that 656 cattle were uninfected and 140 infected. Based on these assumptions, the apparent sensitivity, specificity and positive predictive value (PPV) of a range of criteria for two ELISAs and a GIF assay based on johnin PPD were determined at different time points during the program. ELISA specificity appeared to remain high, but its sensitivity declined to under 40% by year 910. GIF specificity at a fixed criterion (500 johnin-nil units) tended to remain at 80% until year 9 (90%), and the increased specificity was accompanied by a decreased sensitivity equivalent to ELISA. The PPV decreased for both tests throughout the program, being less than 20% for GIF and 30% for ELISA after year 7. For one cut-off parameter for each test, the tandem use of ELISA and GIF together had a combined sensitivity of 74%, a specificity of 79% and a PPV of 37%.