The observation that, in rats, metronidazole exhibited antimicrobial activity against resistant E. coli (when this was accompanied by susceptible B. fragilis) prompted an attempt to reproduce this phenomenon in another species and under other experimental conditions. In experiment 1, mice injected i.p. with an E. coli/B. fragilis mixture were treated with metronidazole, 250 mg/kg given by mouth at 0 and 10 h, or left untreated. At 24 h, viable counts of bacteria in blood and peritoneal washings were determined. In experiment 2, mice with 5 day old subcutaneous abscesses containing E. coli and B. fragilis were also either given metronidazole as above or left untreated. At 24 h, viable counts of bacteria in pus were determined. Metronidazole affected neither the frequency with which E. coli persisted at the 3 sites, nor the viability of E. coli at these sites. This was so despite the fact that, in each of the sites, a significant B. fragilis kill was registered. Thus, pending elucidation of the mechanism by which the phenomenon operates, its non-appearance in these experiments cautions against the extrapolation of the original observations beyond the circumstances under which they were first made.