We made MR images of 39 autopsy specimens of the temporomandibular joint at 0.3 and 1.5 T and compared the images with anatomic cryosections. Imaging time and slice thickness were the same on scans made at each field strength. The purpose was to determine which field strength provides the best scans for imaging of the joint. Additionally, we used imaging times two and four times longer on the 0.3-T scanner to assess to what extent image quality and diagnostic accuracy could be improved. The cryosections showed that 27 of the joints were normal. Twelve had disk displacements. Ten of the joints with disk displacement also had disk deformities, and seven had bony abnormalities. Investigators who analyzed the MR images had no knowledge of the findings on the cryosections. The disk position, disk configuration, and bony abnormalities were correctly diagnosed in 85%, 77%, and 100%, respectively, on 1.5-T MR images compared with 46%, 41%, and 85%, respectively, on the 0.3-T images. When the imaging time was increased by a factor of four, the accuracy rate on the 0.3-T system became comparable to that of the 1.5-T MR scanner: 73% for disk position, 67% for disk configuration. The results suggest that the diagnostic quality of MR images of the temporomandibular joint is better on scans made at 1.5 T than on those done at 0.3 T when comparable imaging times are used.