We studied the relationship between infarct size and degree of internal carotid artery stenosis in 227 stroke patients without a source of embolism in the heart, who had a CT-verified territorial brain infarct. We used logistic regression analysis adjusting for differences between groups in co-associated variables such as age, sex, hypertension, diabetes mellitus, and a history of ischaemic heart disease. Ipsilateral carotid stenosis greater than 50% was more strongly associated with large than small infarcts; adjusted odds ration [(a)OR]: 4.56; 95% confidence interval (CI): 1.21-17.2; P = 0.02. For ipsilateral carotid occlusion the association was even stronger--(a)OR: 36.80; 95% CI: 2.54-533; P = 0.007. When large infarcts were compared with infarcts of small and moderate size together the ORs were 2.29, 95% CI 1.14-4.58 and 2.57, 95% CI 1.17-5.67 for carotid stenosis or occlusion, respectively. Our data show a relationship between greater than 50% carotid stenosis or occlusion and large brain infarcts. We suggest that haemodynamic impairment may contribute to infarct size in territorial infarcts of non-cardiac origin.