The population of oldest old, or people aged 85 and older, is growing rapidly. A better understanding of dementia in this population is thus of increasing national and global importance. In this review, we describe the major epidemiological studies, prevalence, clinical presentation, neuropathological and imaging features, risk factors, and treatment of dementia in the oldest old. Prevalence estimates for dementia among those aged 85+ ranges from 18 to 38%. The most common clinical syndromes are Alzheimer's dementia, vascular dementia, and mixed dementia from multiple etiologies. The rate of progression appears to be slower than in the younger old. Single neuropathological entities such as Alzheimer's dementia and Lewy body pathology appear to have declining relevance to cognitive decline, while mixed pathology with Alzheimer's disease, vascular disease (especially cortical microinfarcts), and hippocampal sclerosis appear to have increasing relevance. Neuroimaging data are sparse. Risk factors for dementia in the oldest old include a low level of education, poor mid-life general health, low level of physical activity, depression, and delirium, whereas apolipoprotein E genotype, late-life hypertension, hyperlipidemia, and elevated peripheral inflammatory markers appear to have less relevance. Treatment approaches require further study, but the oldest old may be more prone to negative side effects compared with younger patients and targeted therapies may be less efficacious since single pathologies are less frequent. We also highlight the limitations and challenges of research in this area, including the difficulty of defining functional decline, a necessary component for a dementia diagnosis, the lack of normative neuropsychological data, and other shortcomings inherent in existing diagnostic criteria. In summary, our understanding of dementia in the oldest old has advanced dramatically in recent years, but more research is needed, particularly among varied racial, ethnic, and socioeconomic groups, and with respect to biomarkers such as neuroimaging, modifiable risk factors, and therapy.