The neuropathological staging model of Alzheimer's disease proposed by Braak and Braak [Acta Neuropathol (1991) 82:259] requires that the evolution of neurofibrillary pathology follows a predictable pattern that can be ordered in a regular regional hierarchy. We have operationalized the neuropathological staging system to permit testing of its validity. Forty-two cases were derived from an epidemiological study of cognitive function in an elderly population for which post-mortem brain tissue was collected. Cases with neuropathological diagnoses other than Alzheimer's disease and normal aging were excluded. Neurofibrillary tangle counts were determined in all cortical laminae and regions used for staging. There was a significant correlation between the overall extent of neurofibrillary pathology and the number of regions affected. There were frequent order violations in the proposed hierarchy: 19 instances (45%) involving entorhinal and transentorhinal cortices, and 16 instances (38%) involving CA1 of hippocampus and entorhinal cortex. Only 6 out of 42 cases conformed in all regions to the expected hierarchy. Nevertheless, 90% of the cases had 2 order violations or less, supporting the approximate validity of the hierarchy.