The National Institute on Aging and Reagan Institute (NIA-RI) criteria, and other neuropathologic criteria for Alzheimer's disease (AD), were compared with the clinical diagnosis of dementia in a well defined population of Catholic sisters. The 47-participant subset examined in this study were college educated and lacked complicating conditions such as brain infarcts or diffuse Lewy body disease. Sixteen participants had a clinical diagnosis of dementia. The NIA-RI criteria imply a perfect correlation between neuritic plaque (NP) density and neurofibrillary tangle distribution. However, NP density often did not coincide with tangle distribution. As a result, it was not possible to categorize many of the participants using the NIA-RI guidelines. The 'high likelihood' category of the NIA-RI criteria for AD research settings (neocortical Braak stage and frequent neocortical NP) had relatively high specificity (90% of nondemented participants did not meet this criteria). However, only half of the demented participants were in this category. Neuropathologic criteria requiring the presence of neocortical tangles (rather than neocortical Braak stage) had relatively high sensitivity, accounting for 87-94% of participants with dementia, but also included 32-35% of nondemented participants. Criteria based on neocortical NP or senile plaques had 100% sensitivity, but a majority of nondemented participants also met these criteria. The results support consideration of both tangles and NP for the neuropathologic diagnosis of AD, but indicate that refinement of the NIA-RI criteria is necessary. A possible refinement is suggested for further consideration.