An outbreak of diarrhoea occurred in an acute geriatric ward of a hospital (A). It affected six patients initially and was found to be due to a single strain of Clostridium difficile. There was little evidence for asymptomatic carriage of this strain or others in the hospital patients. The following three months saw an increase in the number of symptomatic cases of C. difficile disease in two peripheral hospitals, B and C. Patients had been moved from the outbreak ward to these hospitals. Of 18 cases in hospital B all 15 isolates saved for typing were of the 'outbreak' strain. Of three cases occurring in hospital C, only one was the 'outbreak' strain (a relapsed patient who was part of the original episode). There were seven further cases in geriatric and medical wards of hospital A. All six typed isolates were also the outbreak strain. By chance, four isolates from hospital A and four from hospital B pre-dating the outbreak were also available for typing and seven of these were found to be identical to the outbreak strain. This suggests that one strain of C. difficile was endemic in geriatric and medical facilities on two sites and was responsible for nosocomial diarrhoea over at least one year. The problems of optimal management and infection control in this situation are discussed.