Gradient echo T2*-weighted MRI has high sensitivity in detecting cerebral microbleeds, which appear as small dot-like hypointense lesions. Microbleeds are strongly associated with intracerebral haemorrhage, hypertension, lacunar stroke and ischaemic small vessel disease, and have generated interest as a marker of bleeding-prone microangiopathy. Microbleeds have generally been considered to be clinically silent; however, since they are located in widespread cortical and basal ganglia regions and are histologically characterized by tissue damage, we hypothesized that they would cause cognitive dysfunction. We therefore studied patients with microbleeds (n = 25) and a non-microbleed control group (n = 30) matched for age, gender and intelligence quotient. To avoid the confounding effects of coexisting cerebrovascular disease, the groups were also matched for the extent of MRI-visible white matter changes of presumed ischaemic origin, location of cortical strokes, and for the proportion of patients with different stroke subtypes (including lacunar stroke). A battery of neuropsychological tests was used to assess current intellectual function, verbal and visual memory, naming and perceptual skills, speed and attention and executive function. Microbleeds were most common in the basal ganglia but were also found in frontal, parieto-occipital, temporal and infratentorial regions. There was a striking difference between the groups in the prevalence of executive dysfunction, which was present in 60% of microbleed patients compared with 30% of non-microbleed patients (P = 0.03). Logistic regression confirmed that microbleeds (but not white matter changes) were an independent predictor of executive impairment (adjusted odds ratio = 1.32, 95% confidence interval 1.01-1.70, P = 0.04). Patients with executive dysfunction had more microbleeds in the frontal region (mean count 1.54 versus 0.03; P = 0.002) and in the basal ganglia (mean 1.17 versus 0.32; P = 0.048). There was a modest correlation between the number of microbleeds and the number of cognitive domains impaired (r = 0.44, P = 0.03). This study provides novel evidence that microbleeds are associated with cognitive dysfunction, independent of the extent of white matter changes of presumed ischaemic origin, or the presence of ischaemic stroke. The striking effect of microbleeds on executive dysfunction is likely to result from associated tissue damage in the frontal lobes and basal ganglia. These findings have implications for the diagnosis of stroke patients with cognitive impairment, and for the appropriate use of antihypertensive and antiplatelet treatments in these patients.