Magnetic resonance (MR) imaging has created considerable excitement in the medical community, largely because of its great potential to diagnose and characterize many different disease processes. However, it is becoming increasingly evident that, because MR imaging is similar to computed tomography (CT) scanning in identifying structural disorders and because it is more costly and difficult to use, this highly useful technique must be judged against CT before it can become an accepted investigative tool. At present MR imaging has demonstrated diagnostic superiority over CT in a limited number of important, mostly neurologic, disorders and is complementary to CT in the diagnosis of certain other disorders. For most of the remaining organ systems its usefulness is not clear, but the lack of ionizing radiation and MR's ability to produce images in any tomographic plane may eventually prove to be advantageous. The potential of MR imaging to display in-vivo spectra, multinuclear images and blood-flow data makes it an exciting investigative technique. At present, however, MR imaging units should be installed only in medical centres equipped with the clinical and basic research facilities that are essential to evaluate the ultimate role of this technique in the care of patients.