Dementia associated with cerebrovascular disease is common. It has been reported that ∼30% of elderly patients who survive stroke develop delayed dementia (post-stroke dementia), with most cases being diagnosed as vascular dementia. The pathological substrates associated with post-stroke or vascular dementia are poorly understood, particularly those associated with executive dysfunction. Three separate yet interconnecting circuits control executive function within the frontal lobe involving the dorsolateral prefrontal cortex, anterior cingulate cortex and the orbitofrontal cortex. We used stereological methods, along with immunohistological and related cell morphometric analysis, to examine densities and volumes of pyramidal neurons of the dorsolateral prefrontal cortex, anterior cingulate cortex and orbitofrontal cortex in the frontal lobe from a total of 90 elderly subjects (age range 71–98 years). Post-mortem brain tissues from post-stroke dementia and post-stroke patients with no dementia were derived from our prospective Cognitive Function After Stroke study. We also examined, in parallel, samples from ageing controls and similar age subjects pathologically diagnosed with Alzheimer’s disease, mixed Alzheimer’s disease and vascular dementia, and vascular dementia. We found pyramidal cell volumes in layers III and V in the dorsolateral prefrontal cortex of post-stroke and vascular dementia and, of mixed and Alzheimer’s disease subjects to be reduced by 30–40% compared to post-stroke patients with no dementia and controls. There were no significant changes in neuronal volumes in either the anterior cingulate or orbitofrontal cortices. Remarkably, pyramidal neurons within the orbitofrontal cortex were also found to be smaller in size when compared to those in the other two neocortical regions. To relate the cell changes to cognitive function, we noted significant correlations between neuronal volumes and total CAMCOG, orientation and memory scores and clinical dementia ratings. Total estimated neuronal densities were not significantly changed between patients with post-stroke dementia and post-stroke patients with no dementia groups or ageing controls in any of the three frontal regions. In further morphometric analysis of the dorsolateral prefrontal cortex, we showed that neither diffuse cerebral atrophy nor neocortical thickness explained the selective neuronal volume effects. We also noted that neurofilament protein SMI31 immunoreactivity was increased in post-stroke and vascular dementia compared with post-stroke patients with no dementia and correlated with decreased neuronal volumes in subjects with post-stroke dementia and vascular dementia. Our findings suggest selective regional pyramidal cell atrophy in the dorsolateral prefrontal cortex—rather than neuronal density changes per se—are associated with dementia and executive dysfunction in post-stroke dementia and vascular dementia. The changes in dorsolateral prefrontal cortex pyramidal cells were not associated with neurofibrillary pathology suggesting there is a vascular basis for the observed highly selective neuronal atrophy.