In 52 patients we compared the accuracy of standard anteroposterior (AP) radiography, mortise radiography and MRI with arthroscopy of the ankle for the diagnosis of a tear of the tibiofibular syndesmosis. In comparison with arthroscopy, the sensitivity, specificity and accuracy were 44.1%, 100% and 63.5% for standard AP radiography and 58.3%, 100% and 71.2% for mortise radiography. For MRI they were 100%, 93.1% and 96.2% for a tear of the anterior inferior tibiofibular ligament and 100%, 100% and 100% for a tear of the posterior inferior tibiofibular ligament. Standard AP and mortise radiography did not always provide a correct diagnosis. MRI was useful although there were two-false positive cases. We suggest that arthroscopy of the ankle is indispensable for the accurate diagnosis of a tear of the tibiofibular syndesmosis.