Magnetic resonance angiography (MRA) is increasingly used as a noninvasive means to assess internal carotid artery (ICA) stenosis. When used alone, however, MRA may not be sufficiently accurate in certain settings to determine whether ICA disease meets surgical criteria. Although MRA has been recognized to overestimate the degree of stenosis, the authors present two cases in which it severely underestimated arterial stenosis. Two male patients, 70 and 40 years old, respectively, were admitted with crescendo transient ischemic attacks. Their MRA studies suggested nonsurgical lesions of the ICA. After the patients continued to demonstrate clinical evidence of embolic disease, digital subtraction angiography (DSA) was performed on one patient, and the other received a gadolinium contrast-enhanced MRA. These tests revealed critical stenosis in each patient. Each was taken to the operating room for awake carotid endarterectomy with heparin anticoagulation and electroencephalographic monitoring. At surgery, both patients were found to have severely stenosed ICAs with complex plaques. MRA to determine whether ICA stenosis meets surgical criteria may not be sufficiently accurate in certain clinical settings. Additional imaging studies, such as confirmatory digital ultrasonography, MRA with gadolinium contrast, or DSA, may be required to determine the extent of carotid artery stenosis accurately.