The reproducibility and predictive power of MRI diagnosis was evaluated in a group of 70 patients with limited mouth opening. The predictive reliability of the MRI findings was assessed by comparing the images with a final clinical diagnosis that was based on the initial clinical diagnosis, adjusted and refined, when possible, by findings made during the treatment period (conservative treatment in 51 patients, conservative and surgical treatment in 19 patients). The overall reproducibility of the MRI interpretation by the radiologist appeared very good. The predictive reliability was excellent for normal disc position and fair for anterior disc displacement with reduction. For disc displacement without reduction, however, the predictive power was questioned because of false-positive MRI findings.