The purpose of this prospective study in patients undergoing tonsillectomy was to determine whether perioperative bleeding could be predicted by use of a standardized questionnaire concerning bleeding risk combined with measurement of the activated partial thromboplastin time and prothrombin time. Of the 96 patients enrolled in the study, none had a history of a severe bleeding disorder, but 6 (6%) had histories suggestive of a mild bleeding disorder. Further evaluation showed possible von Willebrand's disease in one of these patients. Of the 90 patients with negative questionnaires, 16% had prolongation of the activated partial thromboplastin time. One of these patients had possible von Willebrand's disease. However, none of the patients with positive questionnaires or a prolonged activated partial thromboplastin time bled after surgery. Bleeding that resulted in additional hospital use occurred in 2% of patients, and blood-tinged sputum was described by 4% after discharge; all of these patients had negative questionnaires and normal screening studies. The data provide further evidence that routine measurement of the activated partial thromboplastin time and prothrombin time in asymptomatic patients is not useful for predicting postoperative bleeding. In addition, histories suggestive of a mild bleeding disorder are also not accurate predictors of postoperative bleeding. Excessive bleeding associated with tonsillectomy is usually not a result of an identifiable coagulation disorder.