The purpose of this study was to compare rubidium-82 PET with thallium-201 SPECT imaging in 150 patients. Both techniques followed a single dipyridamole-handgrip stress, and images were displayed using the same 3-dimensional format and quantitative colour scale. Coronary arteriography was employed to assign the correct diagnosis in situations of disparity. Results of PET and SPECT were at least partially concordant in 110 patients (73%), although 22 had more than one defect. A reversible perfusion defect was identified in 60 patients, but the scans were concordant in only 20 (33%). These disparities were chiefly due to false-negative SPECT imaging (22 patients, 55%), and probable delayed thallium redistribution (13 patients, 33%). No patients had ischaemia correctly identified by SPECT in the presence of normal PET imaging. Persistent defects were identified in 91 patients, some of whom also had reversible defects, and the results were consistent in 54 (59%). Other than the delayed thallium redistribution group, the major categories causing disparities were false-positive (6 patients, 16%), and false negative SPECT (8 patients, 22%), attributable to attenuation and scatter. PET appears able to identify smaller, less ischaemic areas subtended by milder coronary stenoses. The availability of a true resting scan with Rb-PET enhances the discrimination between ischaemia and infarction. Attenuation correction, and the high energy photons of positron annihilation, yield more accurate evaluation of inferior wall defects and greater specificity in the presence of soft tissue attenuation.