The sensitivity and specificity of Hachinski's Ischaemic Score (IS) in the diagnosis of the vascular aetiology of dementia was studied in a series of 32 demented patients, dementia of the Alzheimer type (16), multi-infarct dementia (7), mixed dementia (6), Pick's disease (3), with neuropathological diagnosis as the point of reference. The IS distinguished between primary degenerative dementia and multi-infarct or mixed dementia. As single features of the IS "a positive history of stroke" and "a fluctuating course" showed differing prevalences in the latter two diagnostic categories. The IS labelled 21% of patients with primary degenerative dementia as having a vascular aetiology. The uncritical application of the IS to large samples in epidemiological studies may cause incorrect labelling of a significant proportion of patients with primary degenerative dementia as vascular dementia. These results are based on observations of long-term inpatients and depend on neuropathological criteria. While the definite diagnosis of DAT by threshold criteria concerning plaque and tangle counts is well established, neither clinical nor pathological evidence of stroke necessarily means that cerebrovascular disease has anything to do with a patient's dementia.