Samples of rectal faeces were collected immediately after slaughter from 400 cattle each month for a 1-year period and from 1000 each of sheep, pigs and poultry over the same period. Samples were examined for Escherichia coli O157 by enrichment culture in buffered peptone water with vancomycin, cefixime and cefsulodin followed by immunomagnetic separation and culture of magnetic particles onto cefixime tellurite sorbitol MacConkey agar. E. coli O157 was isolated from 752 (15·7%) of 4800 cattle, 22 (2·2%) of 1000 sheep and from 4 (0·4%) of 1000 pigs, but not from any of 1000 chickens. Of the cattle sampled, 1840 (38·4%) were prime beef animals, 1661 (34·6%) were dairy animals being culled and the status could not be determined for the other 1299 (27%) animals. E. coli O157 was found in 246 (13·4%) of the 1840 beef cattle and 268 (16·1%) of the 1661 dairy cattle. The monthly prevalence of E. coli O157 in cattle was 4·8–36·8% and was at its highest in spring and late summer. Seventeen of the 22 isolates from sheep were also made over the summer period. All E. coli O157 isolates from sheep and 749 (99·6%) of the 752 E. coli O157 isolates from cattle were verocytotoxigenic as determined by Vero cell assay and DNA hybridization, eaeA gene positive, contained a 92 kb plasmid and were thus typical of strains causing infections in man. In contrast isolates from pigs were non-toxigenic, eaeA gene negative and did not contain a 92 kb plasmid and would, therefore, be unlikely to be a source of infection for man.