Concerns about intercenter variation in methods and interpretation prompted CERAD investigators to examine standardization of the neuropathological assessment of Alzheimer's disease (AD). Contiguous frontal lobe sections derived from autopsy brains of eight patients clinically diagnosed as having probable AD and two cognitively normal individuals were distributed to 24 neuropathologists from 18 medical centers in the United States and Canada. Using their routine staining method(s), neuropathologists determined the rank order of severity of AD neuropathology in these cases, as well as semiquantitative and quantitative senile plaque and neurofibrillary tangle frequencies. Ranking of the ten cases revealed 75% inter-rater reliability among the 24 raters. Semiquantitative analyses showed reasonable inter-rater agreement, whereas quantitative measures yielded significant differences between raters for plaque and tangle counts (p < 0.0001). These differences reflected variation in stain sensitivity, staining technique (even when the same stain was used), and interpretation of the histological findings. Ratings on the cases with the highest proportions of diffuse plaques showed the greatest dependence upon stain sensitivity and variability in interpretation. This study indicates that greater attention to quality improvement is needed for the neuropathological evaluation of AD, particularly when pooling data in multicenter studies such as CERAD.