The epidemiology of Clostridium difficile was studied prospectively in 451 newborn infants by daily screening of fecal samples. Colonization rates in three postnatal wards ranged from 2% to 52%. Many colonizations were sporadic, but on two wards there was evidence of clustering. On one of these occasions prospective environmental sampling yielded C. difficile organisms from a potential common source. Mothers were shown not to be the sources of their infants' organisms. Both toxin-producing and non-toxigenic strains were common; differentiation according to toxin type was epidemiologically useful. Cross contamination is the most likely explanation of the spread of C. difficile among hospitalized infants; the organism could spread among adults who are at risk of developing antibiotic-associated colitis in a similar manner.