Environmental samples collected from each of 3 locations on 23 large California dairies were cultured to evaluate the utility of this approach for identifying herds infected with Mycobacterium avium ssp. paratuberculosis. Results were compared with concurrent ELISA testing of > or = 60 animals in each herd, and with previously performed individual and pooled fecal cultures of 60 animals. The estimated proportions of infected herds did not differ significantly among the testing methods (environmental sampling, 74%; previous fecal culture, 70%; and concurrent ELISA testing, 65%). Measures of agreement between environmental sampling and the results of previous fecal cultures were 70% (observed agreement), 85% (positive agreement), 62% (negative agreement), and 0.47 (kappa), whereas agreement between environmental sampling and concurrent ELISA testing was 65, 75, and 43%, and 0.19, for the same measures, respectively. The proportion of positive environmental samples on each farm was significantly correlated with the proportion of seropositive animals (r = 0.53), suggesting that environmental sampling may also provide a qualitative estimate of within-herd prevalence. Of the sampling locations that were evaluated, samples of lagoon water (15/23; 65%) were significantly more likely to yield a positive result than were composite manure samples (8/22; 36%) collected from the sick/fresh cow pen or from the alleyway (9/23; 39%) where cows exited from the milking parlor. Environmental sampling was an effective and inexpensive method of identifying herds infected with Mycobacterium avium ssp. paratuberculosis.