Three members of family A, who had diarrhoea on 20 October, lived on a small arable farm which had 10 cattle. Manure from the animals was used to fertilize the ground for growing potatoes which were then offered for retail sale, unwashed, directly from the farm. The mother from family B bought potatoes, which were covered with manure, from family A in early November and over the subsequent 10 days she became ill with diarrhoea and her daughter and son both became ill with bloody diarrhoea. The mother from family C visited family B while the daughter from the latter family was symptomatic; the mother developed diarrhoea several days later. The mother and two sons from family D visited family B while the son from the latter family was symptomatic; the first son developed bloody diarrhoea 6 days later which progressed to development of haemolytic-uraemic syndrome. Direct culture of faecal samples onto cefixime rhamnose sorbitol MacConkey agar failed to isolate E. coli O157 from any of the symptomatic patients, and direct culture onto cefixime tellurite sorbitol MacConkey agar isolated the organism from only one patient. In contrast, a combination of isolation of E. coli O157 by immunomagnetic separation and detection of E. coli O157-specific secretory IgA, suggested E. coli O157 infection in all eight symptomatic patients, but not in any of the family members who were not ill. Two children who excreted the organism for 60 and 89 days respectively were the only two patients who did not develop a secretory IgA response. E. coli O157 was not isolated from potatoes from the farm and faecal samples from the farm animals were not available for examination. The study illustrates the need to use the most sensitive methods available during the investigation and follow up of cases of E. coli O157 infection.