The basis for true registrations, and therefore a basis for finding the difference between the measured and correct transformations, may be obtained from real scans if external fiducials are present. This has previously been suggested and implemented for the validation of the surface-fitting technique elsewhere in the literature [ABKC90][HZJ+90][TJP+93][TJG+91]. Point simulations alone, even with the more realistically bound spherically symmetric configurations discussed above, are still insufficient because it is difficult to obtain realistically variable homology errors. Studies using real scans avoid the need to mathematically simulate anisotropic homology error distributions and clinically realistic point configurations. A problem arises though when it is considered that most existing real scans do not conveniently have external fiducials. This is the reason why many validation studies of retrospective registration techniques use simulations, where the true transformations are implicit, to indirectly obtain measurements of registration error. Clearly, only real scans of the brain can properly yield clinically insightful results with anisotropic homology errors and realistic point configurations. Suitable real scans with external fiducials may be obtained by using an appropriate brain phantom without the unethical imposition of the additional morbidity associated with requesting external marker systems on real patients. This is done in the real scan studies with the brain phantom imaged in MR and SPECT. It shall be shown that the results from the studies with point simulations and real scans based on the Hoffman brain phantom are complementary and therefore doubly strengthen both the viability of the separate validation methods and the quantitative results of the validation of SPECT/MR registration using the ILM technique.