A comparison was made of the ability of plain chest radiography, computed tomography (CT), and magnetic resonance imaging (MRI) to detect and assess the intrathoracic extent of lung cancer in 46 patients. The chest radiographs (CXR) were obtained with a high kilovoltage phototimed technique. The CT scans were obtained with a GE 9800 machine and the MRI studies with a 0.3 Tesla permanent magnet imaging system. The primary tumor was well demonstrated by all 3 imaging techniques; however, the configuration of lesions was best demonstrated by CT. MRI was superior to CXR and CT for demonstrating hilar involvement in 4 cases. CT and MRI were generally comparable for demonstrating mediastinal involvement but were superior to CXR. In 2 cases, small normal size nodes seen on CT were considered to be a single large abnormal node on MRI. Because of the paucity of signal from flowing blood, compression and displacement of vessels were easier to identify with MRI. In 1 case, a small pleural effusion was better seen with CT than with CXR or with MRI. Direct chest wall involvement in 1 case was not seen by CXR. Vertebral body abnormality in another case was seen only by MRI and not by CXR or CT. At present, MRI, with its long scanning time, motion degradation of the image, and poor spatial resolution, is inferior to CT for imaging lung cancer. For evaluation of intrathoracic extent of lung cancer, CT remains the procedure of choice after performing plain chest radiography.