Twenty-five fresh-frozen cadaveric specimens were used to evaluate the role of the syndesmotic ligaments when the ankle is loaded with external rotation torque. An apparatus was constructed that allowed pure external-rotation torque to be applied through the ankle with the foot in neutral flexion. The apparatus provided solid fixation of the tibia while allowing free movement of the fibula in all planes. The syndesmotic ligaments were incrementally sectioned, and direct measurements of anatomical diastasis were made. Mortise and lateral radiographs were made at each increment under both loaded (5.0 newton-meters) and unloaded conditions. After all structures of the syndesmosis had been divided, the syndesmosis was reduced and was repaired with one or two screws. The strength of the repair was measured with incremental increases in torque of 1.0 newton-meter. The radiographs were measured by three independent observers in a blind fashion. In order to evaluate intraobserver error, each observer was randomly given forty radiographs to reinterpret. Diastasis and rotation were found to be related to the amount of injury of the ligament (p < 0.0001). After the entire syndesmosis had been divided, application of a 5.0-newton-meter torque resulted in a mean diastasis of 7.3 millimeters. The subsequent repair of the anterior tibiofibular ligament with suture failed at a mean of 2.0 newton-meters (range, 1.0 to 6.0 newton-meters) of torque. Repair with two screws was found to be stronger than repair with one, with the first construct failing at a mean of 11.0 newton-meters (range, 5.0 to 15.0 newton-meters) and the second, at a mean of 6.2 newton-meters (range, 2.0 to 10.0 newton-meters) (p = 0.0005). Failure of the screw fixation was not associated with the maximum previous diastasis (p = 0.13). Measurements of anatomical diastasis were compared with measurements made on the mortise and lateral radiographs. Measurements on the stress mortise radiographs had a weak correlation with diastasis (r = 0.41, p < 0.0001). However, measurements on the stress lateral radiographs had a higher correlation (r = 0.81, p < 0.0001). Additionally, interobserver correlation was significantly higher for the measurements on the lateral radiographs (r = 0.87, p < 0.0001) than for those on the mortise radiographs (r = 0.56, p < 0.0001). Intraobserver correlation for the three observers was poor with regard to the measurements on the mortise radiographs (r = 0.12, 0.42, and 0.25). The respective correlations for the measurements on the lateral radiographs were r = 0.81, 0.90, and 0.89.(ABSTRACT TRUNCATED AT 400 WORDS)