A series of 100 patients with suspected hepatic metastases was studied with magnetic resonance (MR), unenhanced computed tomography (CT), and radionuclide (RN) scintigraphy. Each set of images was read by three clinicians using a five-point scale to allow receiver operating characteristic (ROC) analysis using truth data derived from clinical review. Performance was measured by the areas under the ROC curves (0.940 +/- .018 for MR, 0.951 O +/- .013 for CT and 0.943 +/- .013 for RN) which were statistically not significantly different. We conclude that at their present level of development these three diagnostic examinations have equivalent performance and that MR is not superior in the detection of hepatic metastases.