Interpretation of brain positron emission tomography (PET) in terms of function vs. structure is ambiguous owing to the partial volume effect (PVE). Therefore, observed differences in tracer distribution could reflect differences in either activity or volume, a problem that applies principally to gray matter (GM) since white matter (WM) virtually always has uniform activity. To assess the contribution of GM volume vs. activity, we implemented a method to directly compare PET images with underlying structure, and applied it to resting-state (18)Fluoro-deoxy-glucose-PET (FDG) of healthy subjects. Methods. Average GM and WM PVE-corrected mean FDG uptake values were applied onto co-registered segmented magnetic resonance imaging data sets to generate a "virtual PET" in which activity is proportional to GM volume and resolution set to that of PET. The raw PET and virtual PET values were then compared across the sample of subjects, first voxel-wise to detect clusters with significant activity-volume mismatch, and second within regions-of-interest (ROI) to quantify mismatches between unsmoothed voxel values. Results. Relative to volume, there was significant hyperactivity of most GM structures of the dorsal brain-except the thalamus-and significant hypoactivity of the temporal lobe, hippocampal region, and cerebellum, consistent across the voxel- and ROI-based analyses. Conclusion. As applied to normals, our method documented the expected contribution of functional activity independently of local differences in GM volume in the normal pattern of FDG uptake, and disclosed marked heterogeneities in functional activity per unit GM volume among structures. This generic method should find applications in pathological states as well as for other PET and SPECT radiotracers.