OBJECTIVE--To measure the time to spontaneous resolution of severe chronic otitis media with effusion (glue ear) in children and study the effects of adenoidectomy, adenotonsillectomy, and ventilation tubes (grommets). DESIGN--Randomised controlled study over 12 years. SETTING--Paediatric otorhinolaryngology clinics and in-patient unit. SUBJECTS--228 children aged 2-9 years with pronounced hearing loss from glue ear and persistent bilateral middle ear effusions confirmed on three occasions over three months. INTERVENTIONS--Children were randomly allocated to adenotonsillectomy, adenoidectomy, or neither procedure. In all groups a Shepard type ventilation tube was inserted in one randomly chosen ear. Follow up was annually for five years and then less often for up to seven years four months. For analysis the two operated groups were combined. MAIN OUTCOME MEASURES--Otoscopic clearance of fluid, change in tympanogram, and improvement in mean audiometric hearing threshold. RESULTS--Survival analysis showed appreciable otoscopic and tympanometric resolution of fluid with ventilation tubes alone and adenoidectomy alone compared with no surgery. Further improvement was seen after combination of both treatments. Mean audiometric hearing thresholds improved with fluid resolution. Resolution was delayed in younger children and in those whose parents smoked, irrespective of treatment. Whereas a single insertion of a Shepard tube resolved the glue for a mean (SD) period of 9.5 (5.2) months, the effect of adenoidectomy was sustained throughout follow up. CONCLUSIONS--Treatment of glue ear considerably shortened the time to fluid resolution, combined adenoidectomy and tube insertion being better than either procedure alone. Resolution was longer in younger children and those whose parent(s) smoked, irrespective of treatment.