A combined clinical and molecular epidemiologic analysis of 46 strains of Clostridium difficile, including 16 nosocomial isolates from one ward (outbreak ward) plus 17 other nosocomial isolates and 13 community-acquired isolates, was performed. HindIII digests of total cellular DNA were analyzed by restriction enzyme analysis (REA) and ribotyping; SmaI digests were analyzed by pulsed-field gel electrophoresis (PFGE). Isolates were assigned to typing groups on the basis of the profiles detected; isolates with closely related profiles were assigned to subgroups. The 16 isolates from the outbreak ward were resolved by both REA and PFGE into five distinct groups; 13 isolates represented two REA groups and three PFGE groups and two isolates were resolved as distinct groups by both techniques. DNA obtained from one isolate was persistently partially degraded, precluding analysis by PFGE. Seventeen sporadic nosocomial isolates were resolved by REA and PFGE into comparable numbers of groups (i.e., nine groups) and subgroups (i.e., 15 and 14 subgroups, respectively), with two isolates not evaluable by PFGE. The 13 epidemiologically unrelated community-acquired isolates were assigned to 11 groups by REA and to 12 groups by PFGE. Overall, ribotyping identified only nine groups among the 46 isolates. We conclude that REA and PFGE have comparable discriminatory powers for epidemiologic typing of C. difficile isolates and that ribotyping is appreciably less discriminatory. For a few isolates, partial DNA degradation prevented analysis by PFGE but not by REA or ribotyping; the cause of the degradation is unknown.