A total of 536 infants and children with acute otitis media were randomly assigned to one of six consistent year-long regimens involving the treatment of nonsevere episodes with either amoxicillin or placebo, and severe episodes with either amoxicillin, amoxicillin and myringotomy, or, in children aged 2 years or older, placebo and myringotomy. Nonsevere episodes had more favorable outcomes in subjects assigned to treatment with amoxicillin than with placebo, as measured by the proportions that resulted in initial treatment failure (3.9% vs 7.7%, P = .009) and the proportions in which middle-ear effusion was present at 2 and 6 weeks after onset (46.9% vs 62.5%, P less than .001; and 45.9% vs 51.5%, P = .09, respectively). In subjects whose entry episode was non-severe, those assigned to amoxicillin treatment had less average time with effusion during the succeeding year than those assigned to placebo treatment (36.0% vs 44.4%, P = .004), but recurrence rates of acute otitis media in the two groups were similar. In the 2-year-and-older age group, severe episodes resulted in more initial treatment failures in subjects assigned to receive myringotomy alone than in subjects assigned to receive amoxicillin with, or without, myringotomy (23.5% vs 3.1% vs 4.1%, P = .006). In the study population as a whole, severe episodes in subjects assigned to receive amoxicillin alone, and amoxicillin with myringotomy, had comparable outcomes. It is concluded that children with acute otitis media should routinely be treated with amoxicillin (or an equivalent antimicrobial drug). The data provide no support for the routine use of myringotomy either alone or adjunctively.