Planar pulmonary scintigraphy is currently the standard investigation for the diagnosis of pulmonary embolism. There are a number of problems with the technique, particularly in patients with an intermediate scan report under the PIOPED criteria. The technique is also under threat from the increasing use of spiral CT angiography. A putative improvement may be gained by use of tomography. The incremental value of tomography over planar studies was therefore evaluated in a virtual model of pulmonary scintigraphy. A model of the segmental anatomy of the lungs was developed from computed tomography, cadaveric human lungs and available anatomical texts. Counts were generated within the phantom by Monte Carlo simulation of photon emission. Eighteen single segmental lesions were interspersed with 47 subsegmental defects and displayed on an Icon reporting station. These were presented in the transaxial, sagittal and coronal planes to four experienced reporters to obtain assessment of defect size. Planar studies of the same defects were displayed to the same observers in the standard eight views with a normal study for comparison. With planar studies, the accuracy of estimation of defect size was 51% compared with 97% using tomographic studies. Defects in the medial basal segment of the right lower lobe were not identified in planar studies but were easily seen by all observers in the tomographic study. It is concluded that there is marked improvement in the accuracy of determination of defect size for tomographic studies over the planar equivalents. This is especially important in the lung bases, the most common reported site of pulmonary emboli. Tomography permits visualisation of defects in the medial basal segment of the right lung, which are not seen in planar studies.