Nonagenarians and centenarians represent a quickly growing age group worldwide. In parallel, the prevalence of dementia increases substantially, but how to define dementia in this oldest-old age segment remains unclear. Although the idea that the risk of Alzheimer's disease (AD) decreases after age 90 has now been questioned, the oldest-old still represent a population relatively resistant to degenerative brain processes. Brain aging is characterised by the formation of neurofibrillary tangles (NFTs) and senile plaques (SPs) as well as neuronal and synaptic loss in both cognitively intact individuals and patients with AD. In nondemented cases NFTs are usually restricted to the hippocampal formation, whereas the progressive involvement of the association areas in the temporal neocortex parallels the development of overt clinical signs of dementia. In contrast, there is little correlation between the quantitative distribution of SP and AD severity. The pattern of lesion distribution and neuronal loss changes in extreme aging relative to the younger-old. In contrast to younger cases where dementia is mainly related to severe NFT formation within adjacent components of the medial and inferior aspects of the temporal cortex, oldest-old individuals display a preferential involvement of the anterior part of the CA1 field of the hippocampus whereas the inferior temporal and frontal association areas are relatively spared. This pattern suggests that both the extent of NFT development in the hippocampus as well as a displacement of subregional NFT distribution within the Cornu ammonis (CA) fields may be key determinants of dementia in the very old. Cortical association areas are relatively preserved. The progression of NFT formation across increasing cognitive impairment was significantly slower in nonagenarians and centenarians compared to younger cases in the CA1 field and entorhinal cortex. The total amount of amyloid and the neuronal loss in these regions were also significantly lower than those reported in younger AD cases. Overall, there is evidence that pathological substrates of cognitive deterioration in the oldest-old are different from those observed in the younger-old. Microvascular parameters such as mean capillary diameters may be key factors to consider for the prediction of cognitive decline in the oldest-old. Neuropathological particularities of the oldest-old may be related to "longevity-enabling" genes although little or nothing is known in this promising field of future research.