Community-aquired pneumonia caused by atypical bacteria or viruses was studied in a double-blind trial comparing fleroxacin 400 mg od and doxycycline 100 mg bd for 10 days. The aetiology was confirmed in 258 of 411 cases (66%), of which 133 were caused by Mycoplasma spp., Chlamydia spp. or Legionella spp.; 30 patients had viral infection, nine had pneumococcal or Haemophilus influenzae infection and 93 had mixed aetiology. In intention-to-treat analyses clinical response rates in fleroxacin-treated patients were 86% (157/182) and 75% (137/182) 2-8 days and 3-5 weeks after therapy, respectively. Corresponding results with doxycycline were 93% (177/191) and 85% (162/190), respectively. Differences between treatments seemed to be due to the lower activity of fleroxacin compared with doxycycline against mycoplasma and pneumococci. Drug-related adverse events were reported in 39% of 204 fleroxacin patients and in 34% of 207 doxycycline patients. The null hypothesis that fleroxacin was <15% inferior to doxycycline was accepted at early follow-up but rejected at later review.