Children with recurrent acute otitis media have been studied in a group general practice. They were seen as soon as possible after an episode of infection and the presence or absence of middle ear effusion was determined by a combination of tympanometry and pneumatic otoscopy. The sensitivity (87.5%) and specificity (90%) of this method were determined by comparing the prediction of effusions using this method with the results of myringotomy in a group of children with chronic secretory otitis media. The parents of the children with recurrent otitis media were questioned about previous aural discharge, chronic nasal symptoms, family history of atopy, numbers of siblings and breast feeding in infancy. Forty-one per cent of the 58 children studied had effusion(s) 2 months after their presenting episodes, while by 3 months this figure had fallen to 33%. Significantly more of the children with persistent effusions at 2 months had received cephalexin than of those who were effusion free at that stage (P less than 0.05). There were no other statistically significant differences between the 2 groups. Nine children had had aural discharge during the presenting episode and 5 of these developed persistent effusions.