To investigate whether MR imaging is an accurate means of assessing mandibular invasion in patients with carcinoma.We retrospectively studied the MR scans of 22 patients with pathologic or surgical confirmation of mandibular invasion from oral and oropharyngeal cancers. The MR images were blindly analyzed using primary criteria of: (a) cortical breakdown, (b) replacement of bone marrow fat, or (c) gadopentetate dimeglumine enhancement of a mass at the bone marrow defect. Secondary criteria of: (a) contiguous soft-tissue mass, and (b) mass on both sides of the mandibular cortex were also examined. Mandibular invasion was graded as periosteal/cortical, medullary, or no invasion.Primary positive findings of cortical breakdown and abnormal bone marrow signal were highly sensitive (100%) for periosteal/cortical invasion and medullary involvement, respectively. However, a high rate of false-positive studies hampered the MR accuracy, which fell into the 73% to 77% range. A negative MR study was highly predictive, but a positive study was less valuable. Gadolinium enhancement added little to the MR study's accuracy. False-positive studies mainly occurred in the setting of prior irradiation, osteoradionecrosis, and odontogenic infections.MR imaging is a sensitive method for detecting mandibular invasion but has a low positive predictive value. A negative study virtually excludes cortical/periosteal or bone marrow invasion.