Although the presence of amyloid deposits is required to establish the neuropathologic diagnosis of Alzheimer's disease, from a clinical point of view, a direct contribution of these cerebral lesions to cognitive deficits is still controversial. The development and standardization of quantitative and accurate biochemical and neuropathologic methods may be critical to improve the postmortem diagnosis and clinicopathologic correlations. Here, we used a point counting method, based on the Cavalieri principle, to estimate the volume occupied by amyloid deposits in a discrete region of the prefrontal cortex and in the hippocampal formation, in brains from patients with cognitive status ranging from normal to severely demented. We demonstrate that the assessment of the total volume occupied by the amyloid deposits in the entorhinal cortex and subiculum can be considered an effective predictor of dementia severity. We also reveal the existence of a high degree of regional and interindividual heterogeneity in amyloid distribution and relative volume. Our data suggest that even though a correlation was observed between the stereologic point counting method and a non-stereologic random field thresholding approach, in most cases non-stereologic methods may not provide adequate samples of the tissue and may lead to unreliable estimates of amyloid burden due to the inhomogeneous distribution of amyloid in the cerebral cortex and the large variability among brains.