Although genetic studies clearly implicate beta-amyloid peptide (Abeta) as a pathogenic agent in Alzheimer disease (AD), it is puzzling that the total amount of Abeta immunoreactivity does not correlate closely with neuronal loss or degree of dementia. We hypothesized that Abeta deposits could vary in the extent to which they disrupt the neuropil, and that the degree to which this occurs might then correlate with the degree of dementia. We used 3 dimensional triple immunofluorescent confocal microscopy to examine the fine structural relationships between Abeta deposits and neurites in their vicinity. In non-demented elderly, Abeta deposits were porous structures with numerous normal appearing processes coursing through them. In AD, dendrites within Abeta deposits, compared with dendrites in the surrounding neuropil, were likely to have decreased SMI32 immunoreactivity and increased Alz-50 immunoreactivity. We found that the degree to which Abeta deposits disrupt the neuropil, as assessed by local loss of SMI32 immunoreactivity, correlates closely with the amount of neuronal loss and with duration of dementia. These observations support the hypothesis that a subset of Abeta deposits contribute directly to neural system failure in AD.