Magnetic resonance imaging (MRI) was compared to computed tomography (CT) of the mediastinum and/or hila in 37 patients with bronchogenic carcinoma (35 unresectable for cure) and 11 patients with other masses. Spin-echo pulse sequences using a short pulse repetition rate (TR) and short echo delay (TE) were most helpful for detection of abnormal soft-tissue mediastinal and hilar masses. The accuracy of MRI and CT in staging bronchogenic carcinoma for curative resectability/nonresectability was comparable. CT staged 35 of 37 cases appropriately, while MRI correctly staged 36 of 37 cases. Several pitfalls in MRI evaluation of the mediastinum were identified. By MRI the esophagus may be misinterpreted as an enlarged retrotracheal lymph node unless serial scans are studied. Scattered calcifications in enlarged mediastinal and hilar lymph nodes due to old granulomatous disease are not detectable by MRI. Small adjacent lymph nodes shown individually by CT may appear as a single enlarged lymph node by MRI due to partial-volume averaging. Small lung nodules may be undetected by MRI due to respiratory motion and partial-volume averaging. Certain patients are unsuitable for MR scanning. Because of the requirement for patient selection and the identified pitfalls of MRI, CT remains the radiologic procedure of choice in the staging of patients with bronchogenic carcinoma and the evaluation of other mediastinal and hilar masses at present. However, because of the ability to show blood vessels without an intravascular contrast agent, MRI is useful in evaluating patients with potential contrast allergy and solving diagnostic problems not solved by CT.