An apparent rise in incidence of clinical Johne's disease and concern within the cattle industry at the lack of accurate prevalence data prompted a survey of cull cattle slaughtered from dairy herds in South Australia from February 1990 to June 1993. A total of 2,551 mature cows culled (for a variety of reasons) from 643 herds in the principal dairy areas of South Australia were sampled for culture, histopathology and serology at slaughter. Ileum, ileocecal valve and associated lymph nodes were cultured and serological tests conducted were the absorbed ELISA and complement fixation tests. M. paratuberculosis was detected on culture in 11 (0.43%) of animals in 9 (1.4%) of herds. Infection was found in less than 1% of 2,264 culls from the three major dairy districts of the Central and Southern Hills, the Lower Murray and the South East and in none of 287 culls from the mid-North district. 188 (7.47%) of 2,516 samples were positive at 1:8 serum dilution or above in the complement fixation test, including only two of the culture-positives. This gave a specificity of 92.6%, a sensitivity of 18.2% and a predictive value (of a positive) of only 1.1% when the real prevalence is less than 0.5%. Sixteen samples (0.63% of sera tested) were positive in the absorbed ELISA, including four of the culture-positives, giving a specificity of 99.5%, sensitivity 36.4% and a predictive value of a positive of 25% at a real prevalence of less than 0.5%. The results indicate no significant increase in prevalence over a smaller, localized and less precise estimate made ten years earlier. They also confirm many previous findings that the complement fixation test is of limited value for herd screening or detection of non-clinical cattle in low-prevalence populations and that the absorbed ELISA is more useful in such populations as a herd screen and as a primary indicator of individual high-risk cattle.