Detection of the source of Clostridium difficile strains is of importance for the control of the nosocomial spread of this microorganism. For this purpose, vaginal and rectal swabs from 183 mothers, duplicate fecal samples (taken on days 1 and 4 after birth) from 183 neonates, and 94 environmental samples were cultured for C. difficile. The microorganism was never detected in the meconium obtained on day 1 after birth. On the other hand, an incidence of 17% C. difficile positivity was noted in the fecal samples obtained on day 4 after birth. Forty-two percent of the 31 colonized neonates had been delivered with complications. The bacteria were never encountered in the rectal swabs of the mothers, and C. difficile was identified in only one vaginal swab. In contrast, 13% of the environmental samples were positive for C. difficile. No major difference was encountered between patient and environmental isolates with respect to toxigenicity (58 to 65% toxigenic isolates). All strains were subsequently typed by PCR amplification of the 16S-23S ribosomal intergenic spacer regions and by arbitrarily primed PCR (AP-PCR) with different primers and combinations thereof. All environmental isolates and 11 of 31 neonatal strains were of a single type. The vaginal strain was unique, and among the maternity ward- and neonate-related isolates, only two additional AP-PCR types were identified. When a collection of C. difficile strains from patients hospitalized in other institutions and suffering from antibiotic-associated diarrhea or pseudomembranous colitis was analyzed in a similar manner, it appeared that the strain from the maternity ward was unique. The other strain commonly encountered among the neonates was also identified frequently among the isolates from patients with antibiotic-associated diarrhea or pseudomembranous colitis, indicating its general occurrence. On the basis of both epidemiological studies and PCR-mediated genotyping, it was shown that the environment and not the birth canal is the major source of C. difficile acquisition by neonates in this maternity hospital setting. Furthermore, AP-PCR appears to be a fast and useful method for epidemiologically relevant typing of C. difficile isolates.