Cognitive function was examined in 227 patients three months after admission to hospital for ischaemic stroke, and in 240 stroke-free controls, using 17 scored items that assessed memory, orientation, verbal skills, visuospatial ability, abstract reasoning, and attentional skills. After adjusting for demographic factors with standardised residual scores in all subjects, the fifth percentile was used for controls as the criterion for failure on each item. The mean (SD) number of failed items was 3.4 (3.6) for patients with stroke and 0.8 (1.3) for controls (p < 0.001). Cognitive impairment, defined as failure on any four or more items, occurred in 35.2% of patients with stroke and 3.8% of controls (p < 0.001). Cognitive domains most likely to be defective in stroke compared with control subjects were memory, orientation, language, and attention. Among patients with stroke, cognitive impairment was most frequently associated with major cortical syndromes and with infarctions in the left anterior and posterior cerebral artery territories. Functional impairment was greater with cognitive impairment, and dependent living after discharge either at home or nursing home was more likely (55.0% with, v 32.7% without cognitive impairment, p = 0.001). In a logistic model examining the risks related to dependent living after stroke, cognitive impairment was a significant independent correlate (odds ratio, OR = 2.4), after adjusting for age (OR = 5.2, 80 + v 60-70 years) and physical impairment (OR = 3.7, Barthel index < or = 40 v > 40). It is concluded that cognitive impairment occurs frequently after stroke, commonly involving memory, orientation, language, and attention. The presence of cognitive impairment in patients with strike has important functional consequences, independent of the effects of physical impairment. Studies of stroke outcome and intervention should take into account both cognitive and physical impairments.