As part of an interventional study to determine glycopeptide-resistant enterococci (GRE) acquisition on a three-ward haematology unit, rectal swabs were taken weekly from 293 patients recruited to the study between June 1995 and December 1996. The GRE isolates obtained from the first positive rectal swab from 120 colonized patients, the isolates from 7 patients with clinical infection and 43 isolates obtained from the ward environment were compared by pulsed-field gel electrophoresis (PFGE). Sixty-three of 120 patients were colonized by one of strains A-H, while 49 were colonized by unique strains. The first 18 weeks were associated with the highest prevalence of GRE by rectal swab, with a single strain A responsible for 52% of acquisitions on ward 2, 22% on ward 3 and 36% on ward 4. Other smaller ward associated clusters were evident. Environmental sampling of ward 2 during this time showed that all but 2 of 30 isolates were indistinguishable from strain A. As the GRE prevalence fell, rectal swab and environmental isolates became more heterogeneous, and strain A disappeared after week 55. GRE prevalence rose again in the final 15 weeks of the study, and a new predominant strain B emerged on ward 2 responsible for 50% of new acquisitions. In the seven patients with clinical infection with GRE, the clinical isolates were compared with the contemporaneous rectal swab isolate, and were found to be the same in only two cases. An analysis of five long-term carriers colonized for a median of 19 weeks (range 11–34) showed colonization with at least two and in one case six distinct strains, raising the question of how many strains may be colonizing a patient at any one time, and suggesting that multiple colonies should be analysed. These data suggest that cross-infection was an important factor in the spread of GRE when the colonization rate was high.