Objectives were to estimate percentages of seropositive herds with cows shedding Mycobacterium paratuberculosis in feces and milk, and to estimate sensitivity, specificity, and predictive value of an ELISA relative to fecal culture. Dairy cows (n = 712) were randomly selected from 61 herds previously identified by ELISA as positive for Johne's disease. Fecal and bulk tank milk samples (n = 52 of 61 herds) were obtained from 10 states in the United States. Fecal samples were processed by a double centrifugation, double decontamination culture procedure. Milk samples were processed for both culture and DNA analysis by using polymerase chain reaction (PCR). Of 24 herds with at least three cows that had tested ELISA-positive, 79% were also culture-positive, compared with 18 of 37 herds with one or two ELISA-positive cows. Both fecal-culture and ELISA results were available on 651 cows; only 25% of cows that were fecal-culture positive also tested positive by ELISA and over 6% of cows that were fecal-culture negative tested ELISA-positive. Milk samples all cultured negative, but analysis of milk samples by PCR resulted in 68% of herds positive for M. paratuberculosis DNA including 24 of 31 herds with positive fecal cultures and 11 of 21 herds with negative fecal cultures. Sensitivity and specificity of the ELISA compared with fecal culture is lower than previously reported and perhaps best used in screening herds because of limited efficacy to predict infection in individual cows. In addition, contamination of bulk tank milk samples with M. paratuberculosis does occur in seropositive herds, even in some with negative fecal cultures.