Large discrepancies are usually found when different ELISAs for the diagnosis of pig salmonellosis are compared. Thus, our main goal was to estimate the diagnostic accuracy through Bayesian approaches of two commercial assays (Svanovir "test A" and HerdCheck "test B") for the detection of antibodies to Salmonella spp. in slaughter pigs. Previously, we estimated the agreement between both tests and their relative sensitivity (Se) and specificity (Sp) with respect to bacteriology on caecal content and ileocaecal lymph nodes. Test A, at a cut-off OD%>or=20%, indicated higher prevalence than test B (OD%>or=10%) (14.6% vs. 8.6%). Relative Se with respect to overall bacteriology was low (approximately 30%) and similar for both tests, but the relative Sp was significantly lower for test A compared to B (88% vs. 95%). Both tests failed to detect some pigs infected with Salmonella serogroups B and C1, which they were supposed to identify. In general, tests showed only fair-to-moderate agreement when they were compared (kappa: 0.41). In the Bayesian models, Se of test A varied between 63% and 77%, while Se of test B was 73%. Sp of A was always lower than that of test B (89% vs. 95%). The implications derived from the use of these imperfect serological tests will have to be accounted for in large-scale Salmonella-control programs.