A random sample of Wisconsin dairy herds, stratified by herd size, were tested for paratuberculosis by use of an absorbed ELISA procedure. The ELISA was optimized for overall accuracy by means of receiver operating characteristic curve analysis, and had a sensitivity and specificity of 50.9 and 94.9%, respectively. Herd prevalence was analyzed for correlation with responses to a management practices questionnaire completed by the herd owners. One hundred and fifty-eight herds and 4,990 cattle were tested. Of these, 50% of herds and 7.29% of cattle had positive test results. Calculation of true prevalence from the apparent prevalence indicated that 4.79% of cattle and 34% of the Wisconsin dairy herds tested had serologic evidence of paratuberculosis. Among the 54 herds classified as positive on the basis of true prevalence estimation, the mean number of test positive cattle was 20.3%. The geographic distribution of herds with positive results was not uniform. More infected herds were found in the southern and western districts of Wisconsin than in the eastern district. The west-central district had a larger number of infected herds than did other districts. By use of chi 2 analysis, the only management factor found to be significantly associated with herd prevalence was housing of calves after weaning (P = 0.03). Specifically, in herds with higher prevalence, calves were separated after weaning into calf barns and hutches rather than into pens in the cow barn more often than in herds with lower prevalence. This factor was also considered significant by use of logistic regression analysis.(ABSTRACT TRUNCATED AT 250 WORDS)