For 11 AD cases and four normal elderly controls, post mortem volumes of the hippocampal subdivisions were calculated by using magnetic resonance imaging and histological sections. After at least six weeks of fixation in formalin, brains were examined on a 1.5-T Philips Gyroscan imager producing T₁-weighted coronal images with a 3-mm slice thickness. Brains were then processed and embedded in paraffin. Serial coronal sections, 3mm apart and stained with Cresyl Violet, were used for the planimetry and unbiased estimation of the total numbers of neurons in the hippocampal subdivisions. For all 15 cases, magnetic resonance imaging- and histology-based measurements were performed along the whole rostrocaudal extent of the hippocampal formation and included three subvolumes: (i) the hippocampus (CA1–CA4 and the dentate gyrus); (ii) hippocampus/subiculum; and (iii) hippocampus/parahippocampal gyrus. After controlling for shrinkage, strong correlations were found between magnetic resonance imaging and histological measurements for the hippocampus (r=0.97, P<0.001), hippocampus/subiculum (r=0.95, P<0.001) and hippocampus/parahippocampal gyrus (r=0.89, P<0.001). We also calculated the total number of neurons in the hippocampus and hippocampus/subiculum subvolumes. Strong correlations between the magnetic resonance imaging subvolumes and neuronal counts were found for the hippocampus (r=0.90, P<0.001) and the hippocampus/subiculum subvolume (r=0.84, P<0.001). We conclude that very accurate volumetric measurements of the whole hippocampal formation can be obtained by using a magnetic resonance imaging protocol. Moreover, the strong correlations between magnetic resonance imaging-based hippocampal volumes and neuronal numbers suggest the anatomical validity of magnetic resonance imaging volume measurements.