The number of band differences in DNA macrorestriction profiles required to distinguish unrelated strains from an index strain varies in an outbreak with the species and restriction enzyme used. In order to define this difference for epidemiological studies of Serratia marcescens, we produced DNA fingerprints from 57 isolates of the organism using the restriction enzyme XbaI and pulsed-field gel electrophoresis (PFGE). The isolates were selected on the basis of their epidemiology, serotype and phage-typing patterns to include 28 unrelated strains and 29 representatives from 2 distinct outbreaks. One of the outbreaks was prolonged, lasting for several years. Electrophoretic profiles consisting of 20 or more clearly resolved bands were obtained for all isolates. Twenty-six of the unrelated strains had unique profiles with over 10 band differences from all other strains, while 27 of the outbreak representatives could be assigned to the appropriate outbreak with confidence. The majority of the outbreak isolates had none or 2 band differences from the index profile, although 3 isolates differed by 5–7 bands. The 2 exceptions among the unrelated strains differed by 4 bands, and 3 phage typing reactions, and were isolated from London and Berlin 3 years apart, while the 2 exceptions among the outbreak collection had clearly unique profiles with over 20 band differences from each other and the outbreak profiles. Cluster analysis using Dice coefficient and UPGMA gave cut-off values of 75–78% similarity overall for related isolates, while the closest similarity for unrelated strains was 70%. The results of this study together with those of the 6 previous reports of PFGE for S. marcescens (which used either enzymes XbaI or SpeI) confirm that this technique is of value for this species and that with XbaI at least, most epidemiologically related strains will only differ by 3–4 bands. However, on occasion up to 7 band differences can be found within an apparent outbreak, which may be suggestive of genetic drift.