Recent studies suggest that dementia in the most elderly (90 years of age and above) is only modestly related to Alzheimer's disease pathology. This raises the possibility that other, as yet unknown, disease processes may underlie dementia in this rapidly growing demographic group, and that efforts designed to combat Alzheimer's disease may not be appropriate for treating dementia in very elderly subjects. To study this question more closely, we examined the relationship between neocortical Alzheimer-type brain pathology and dementia in consecutive autopsies from 209 participants in the Baltimore Longitudinal Study of Ageing, a prospective longitudinal cohort study of the effect of ageing on cognition. Almost half of the cohort was older than 90 years of age at death. We found that several measures of neocortical Alzheimer's pathology, including the Consortium to Establish a Registry of Alzheimer's Disease neuritic plaque score and the Braak neurofibrillary tangle score, remained significant predictors of dementia, independent of age. In participants older than 90 years of age, intracranial atherosclerosis emerged as an important predictor of dementia in subjects with low Alzheimer's pathology scores, but did not mitigate the importance or population attributable risk of high Alzheimer's pathology scores on the odds of dementia. There was evidence that the threshold score for neurofibrillary pathology to cause dementia increased in the oldest subjects, but this was offset by an overall increase in neurofibrillary pathology in this age group. We conclude that neocortical Alzheimer's disease pathology remains significantly correlated with dementia, independent of age. In the most elderly, atherosclerosis also emerged as a cause of dementia in subjects with low Alzheimer's pathology scores. We found no evidence for a significant number of elderly subjects having dementia without an apparent cause.