Over a 12-month period 501 children (age range 11 months to 15 years) underwent surgery for a possible middle ear effusion. All had tympanometry performed within 2 h of surgery. The results of tympanometry were correlated with the surgical findings in 955 ears. A type-B tympanogram has a high sensitivity (0.91) in predicting middle ear effusion with good specificity (0.79). A type-A tympanogram has a very high specificity (0.99) in predicting a dry middle ear but low sensitivity (0.34). Both the positive (0.91) and negative (0.84) predictive values of a type-A tympanogram are high. The addition of a type-C tympanogram increases the sensitivity of predicting a dry middle ear to 0.79. The positive predictive value of a peaked (type-A or -C) tympanogram is 0.71 and should be considered strong evidence that the middle ear is dry. Tympanometry is the best clinical test for the presence or absence of a middle ear effusion, and on the basis of preoperative tympanometry alone the need for surgery should be carefully reassessed.