The system of Gustilo and Anderson for the classification of open fractures is commonly used as a basis for treatment decisions and for comparison of the published results of treatment. The reliability of this classification system was tested on the basis of the responses of 245 orthopaedic surgeons to a survey. The respondents were asked to provide data about their age, type of practice, and type of training; the number of open fractures of the tibia that they treated each year; and their use of the Gustilo-Anderson classification system. They were also asked to classify twelve open fractures of the tibia on the basis of a series of videotaped case presentations. Each case presentation on the color videotape included demographic data on the patient, a history of the injury, the results of the physical examination, the appearance and dimensions of the open wound before the operation, preoperative radiographs, and selected portions of the operative debridement with narration. The level of agreement for the classification of each fracture was determined according to the largest percentage of observers who chose a single classification type. The average agreement among the observers for all twelve fractures was 60 percent. The over-all agreement for each fracture ranged from 42 to 94 percent. The average agreement in the subgroup of surgeons who were considered to have the least experience (residents and fellows) was 59 per cent (range, 33 to 94 percent).(ABSTRACT TRUNCATED AT 250 WORDS)