Three sandwich-ELISAs, two of which are commercially available (Tecra and Locate), and one developed at the Veterinary Sciences Division, Stormont and a 3-step culture protocol, were compared for the detection of salmonella in 1000 animal specimens. Eight hundred and fifty of these were new submissions and the remainder were frozen portions from specimens previously shown to contain salmonellas by culture. The incidence of ELISA false-negative and false-positive results was highest for the Stormont and Locate kits respectively although the differences in sensitivity and specificity between the three ELISAs was not statistically significant. On 16 occasions all ELISA methods indicated the presence of salmonellas when none were isolated by initial culture, eight of these specimens contained salmonellas when reinvestigated by culture.