Clostridium difficile strains were detected in 14 of 210 (6.7 %) faecal samples from children in Rio de Janeiro, Brazil, by cultivating faeces on cycloserine/cefoxitin/fructose agar after alcohol-shock. Two main groups of children were studied: inpatients (n = 96) and outpatients (n = 114). The inpatient group consisted of children on antibiotics or immunosuppressors who presented with diarrhoea and other children who did not present with diarrhoea and were not under an antibiotic or chemotherapeutic regimen. Among the outpatients, two groups were examined: namely, a group that comprised children who presented with diarrhoea and were occasionally under an antibiotic regimen and another group that comprised patients who were not taking antibiotics. After cytotoxic assay, toxigenic C. difficile (Cd tox+) strains were detected in 4.2 % of inpatients and 3.5 % of outpatients. Exclusion of other infectious causes of diarrhoea indicated a typical case of C. difficile-associated paediatric diarrhoea in the community. Among Cd tox+ isolates, no variations were detected by PCR for toxin A that employed primers NK9 and NKVO11. No resistance was found to metronidazole or vancomycin among strains that were isolated from children who presented with diarrhoea, but the MIC(50) and MIC(90) values for clindamycin were 6-8 and 16 microg ml(-1), respectively. Resistance to clindamycin seems to be more disseminated in strains from outpatients than in those from inpatients (P < 0.05). In conclusion, these data suggest that investigation for C. difficile infection should be taken into account in paediatric diarrhoea in both inpatients and outpatients in developing countries.