We studied a synatophysin-like immunoreactivity in the hippocampal formation of patients with definite Alzheimer's disease, multi-infarct dementia, patients with no evidence of clinical dementia with neuropathological findings fulfilling the criteria of possible Alzheimer's disease, and age-matched nondemented controls. Possible Alzheimer's disease cases were of special interest because they were considered to represent early Alzheimer's disease. We also studied the spatial relationship of synaptophysin-like immunopositivity with amyloid-beta-protein immunopositive senile plaques and anti-paired helical filament immunopositive degenerating neurons locally as well as considering the intrinsic circuits in the hippocampal formation. The synaptophysin-like immunoreactivity was decreased in the hippocampus and the entorhinal cortex in patients with definite and possible Alzheimer's disease but not in multi-infarct dementia patients compared to controls. Equal loss of synapses in possible and definite Alzheimer's disease patients supports the hypothesis that synaptic loss is an early phenomenon in Alzheimer's disease. Unchanged synaptophysin-like immunopositivity in patients with multi-infarct dementia suggests that the loss of synapses is centrally involved in the pathogenesis of Alzheimer's disease and not dementia per se. There was no spatial correlation between loss of synapses and amyloid-beta-protein positive senile plaques. Moreover, we could not find a strict spatial relationship between senile plaques and degenerating neurons. Our results do not support the amyloid cascade hypothesis of Alzheimer's disease that local accumulation of amyloid-beta-protein leads to the loss of synapses.