We studied the efficacy of tonsillectomy, or tonsillectomy with adenoidectomy, in 187 children severely affected with recurrent throat infection. Ninety-one of the children were assigned randomly to either surgical or nonsurgical treatment groups, and 96 were assigned according to parental preference. In both the randomized and nonrandomized trials, the effects of tonsillectomy and of tonsillectomy with adenoidectomy were similar. By various measures, the incidence of throat infection during the first two years of follow-up was significantly lower (P less than or equal to 0.05) in the surgical groups than in the corresponding nonsurgical groups. Third-year differences, although in most cases not significant, also consistently favored the surgical groups. On the other hand, in each follow-up year many subjects in the nonsurgical groups had fewer than three episodes of infection, and most episodes among subjects in the nonsurgical groups were mild. Of the 95 subjects treated with surgery, 13 (14 per cent) had surgery-related complications, all of which were readily managed or self-limited. These results warrant the election of tonsillectomy for children meeting the trials' stringent eligibility criteria, but also provide support for nonsurgical management. Treatment for such children must therefore be individualized.