To assess whether a rigorous clinical classification, based on computerised tomography, of patients with cerebral ischaemia would identify subgroups at higher or lower risk with respect to cigarette smoking habits, a case-control study was carried out on 422 cases of first-episode cerebral ischaemia matched for age and sex with 422 community-based neighbourhood controls. Patients with ischaemic stroke due to extracranial or intracranial vascular disease were at higher risk from smoking than has previously been reported for stroke (relative risk 5.7, 95% confidence interval 2.8, 12.0) whereas those with stroke due to cardiac emboli had no excess risk associated with smoking (relative risk 0.4 [0.1, 1.8]). After cessation of smoking, the relative risk declined gradually over 10 years, at the end of which time a significant risk was still evident. This finding may imply that the risk incurred by smoking is due mainly to atheroma formation, rather than transient haematological effects. Exposure to smoking by a spouse was an independent risk factor for the whole group of cerebral ischaemia patients (relative risk 1.7 [1.1, 2.6]), but this was not so for smoking by either parent (relative risk 1.2 [0.8, 1.8]). These findings suggest that smoking is a more potent risk factor for the most common form of ischaemic stroke than has previously been appreciated. The persistent nature of the risk even after cessation of smoking and the possible risk associated with passive exposure strengthens public health arguments against smoking.