In a double blind study 175 patients with acute otitis media were randomized into 2 treatment groups: 10 days of therapy with cefaclor or 5 days of therapy followed by 5 days of placebo. The dosage of cefaclor was 40 mg/kg/day administered orally in equally divided doses at 12-hour intervals. Tympanocentesis before treatment yielded specimens that contained Streptococcus pneumoniae or Haemophilus influenzae or both in 55% of specimens. Branhamella catarrhalis was isolated from 21% of specimens. Culture of material from the ear canal of patients with spontaneous perforation of the tympanic membrane of less than 24 hours duration yielded pneumococci or H. influenzae or both in 38% of specimens and staphylococci in 31%. Patients were scheduled for follow-up examinations at 5 or 6, 10, 30, 60 and 90 days. Of the 175 children 151 were evaluable at 10 days. There were 123 patients with both tympanic membranes intact at the time of diagnosis. There were 6 (10%) treatment failures of therapy in the 59 patients assigned to 5 days of therapy and 4 (6%) failures and 1 (2%) early relapse in the 64 assigned to 10 days of therapy (difference not significant). There were 28 evaluable patients with spontaneous perforation. There were 8 (53%) failures in the 15 children assigned to 5 days of therapy and only 1 (8%) failure in the 13 children assigned to receive 10 days of therapy (P = 0.016, Fisher exact test). Rates of reinfection and persistent middle ear effusion at 10, 30, 60 and 90 days follow-up were not significant different in patients assigned to 5 to 10 days of therapy. In patients with acute otitis media with intact tympanic membranes we have not been able to show any advantage of the standard duration of 10 days of therapy over a shortened course of 5 days. A 5-day course of antibiotic therapy does not appear to be sufficient for children with acute otitis media and spontaneous purulent drainage.