The Cavalieri and Vertical Sections methods of design based stereology were applied in combination with 3 tesla (i.e. 3T) Magnetic Resonance Imaging (MRI) to estimate cortical and subcortical volume, area of the pial surface, area of the grey-white matter boundary, and thickness of the cerebral cortex. The material comprises eight human cadaveric cerebri which had been separated into sixteen cerebral hemisphere specimens prior to embedding in agar gel. The results from MRI were compared with corresponding ‘gold standard’ values subsequently obtained by application of the same methodology using physical sectioning of the specimens. 95% agreement intervals revealed poor agreement between MR imaging and physical sectioning, specially for pial surface and thickness, as well as cerebral cortex and subcortex volumes. On average, pial surface area was estimated to be almost half the extent using MRI compared to physical sectioning (i.e. 45%, p<0.05) and the average thickness of the cerebral cortex was calculated to be much greater (by 60.9%) on the MR images compared to the physical sections (3.7mm versus 2.3mm, p<0.001). The main cause of the discrepancies is that the resolution of the MR images is not sufficient to always allow reliable depiction of the cerebral sulci on 2D image sections. Accurate application of manual stereological methods for measuring the cortical surface area thus requires higher resolution MR imaging than is typically applied at 3T.