Magnetic resonance (MR) imaging and computed tomography (CT) were compared in 30 patients with histologically proved bladder cancer. MR imaging was accurate in depicting the presence or absence of extravesical spread in 22 patients (accuracy, 73%; sensitivity, 82%; specificity, 62%), and CT was accurate in 24 patients (accuracy, 80%; sensitivity, 94%; specificity, 62%). The MR examinations of two patients were of undiagnostic quality and therefore considered to be technical failures. Each technique resulted in five false-positive and one false-negative examination for the diagnosis of extravesical tumor spread. In 28 patients the integrity of the bladder wall was assessed with MR imaging. In 22 patients the bladder wall was disrupted, and 18 of these patients had deep muscle invasion. In six patients the bladder wall was intact, and none of these patients had evidence of deep muscle invasion at pathologic examination. In this study MR imaging was slightly inferior to CT in the delineation of invasive tumors beyond the bladder wall. However, if one excludes from analysis the two patients with undiagnostic studies, there is no significant difference in accuracy between the two techniques.