Experimental pulmonary emboli, consisting of tantalum-labeled autologous blood clots and barium-labeled 3-mm plastic spheres that did not produce an MR signal, were introduced through the femoral vein into each of five dogs. The sensitivity of MR to detect autologous clots of known size was assessed, and the size of the clot on MR was compared with its actual size. Emboli were localized by using chest radiographs made in multiple projections. Cardiac-synchronized MR imaging was performed on a 0.35-T superconducting magnet and was followed by a 99mTc macroaggregated lung scan. All 12 blood clots larger than 3 mm, and three of 20 clots less than 3 mm in transverse diameter were correctly visualized with MR. Five of the 15 clots seen on MR were larger than 150% of the actual size. There were nine false-positive emboli on MR. Two of six plastic spheres resulted in an abnormal signal on MR. MR signal from pulmonary emboli results from the thrombus itself and probably also from slow blood flow proximal to the obstruction. MR may be of value in detecting pulmonary emboli; clinical trials to evaluate its usefulness should be carried out.