We have fingerprinted Clostridium difficile isolates from patients with symptomatic recurrences of infection, using random amplified polymorphic DNA (RAPD). The medical records of 55/79 patients were examined, from whom multiple C. difficile-positive faeces were received during hospitalization at least five days, but no more than two months, apart. In 20 of these cases symptoms either did not recur (i.e., absent for at least three days between episodes), or were explainable by other causes, such as laxative administration. Of the remaining 35 patients, 27 sets of C. difficile isolates (23 pairs and four triplicates) were available for RAPD fingerprinting. Differing C. difficile DNA fingerprints (at least three major bands difference) were obtained for 15/27 patients, and hence at least 56% of the clinical recurrences of infection were in fact due to re-infection as opposed to relapse. Since we found that an endemic C. difficile clone was present in 18 out of 27 patients (67%) and accounted for 53% (31/58) of all isolates, it is probable that the majority of symptomatic recurrences are in fact re-infections, with either a different or the same C. difficile strain. We conclude that more attention must be given to preventing the re-infection of C. difficile symptomatic patients. Isolation of symptomatic individuals is the preferred option for the protection of other patients, but measures must be taken to ensure that further strain acquisition by the index cases does not occur.