The accuracy of high-resolution ultrasonography (HR-US) in detecting disk displacement and condylar erosion of the temporomandibular joint (TMJ) was evaluated, using corresponding cryosections as a "gold standard". HR-US of the TMJ was performed with a high frequency 12 MHz transducer on 30 preserved autopsy specimens. Succeeding sonography, the autopsy specimens were deep-frozen and cut in paracoronal planes corresponding to the sonographic images. HR-US diagnoses were compared with cryosectional findings in a blinded fashion. HR-US detected 19 (95%) of 20 instances of condylar erosion and 16 (73%) of 22 instances of disk displacement. There were one false-positive finding for condylar erosion and two false-positive findings for disk displacement. The accuracy of HR-US evaluating condylar erosion and disk displacement rated 93% and 73%, respectively. In conclusion, condylar erosion was reliably assessed by HR-US, but the evaluation of disk position was less accurate.