Stroke mortality and morbidity remain high despite downward trends in incidence and case fatality. Population-based longitudinal studies which include collection of risk factor data are required for a better understanding of stroke aetiology. From a representative cohort of men from South Wales and South-west England, followed up for a median of 17 y, details of possible cerebrovascular events were collected from questionnaires, hospital admission data, general practitioner records, death certificates, radiology records and post-mortem reports. Radiology records, and strokes and transient ischaemic attacks were independently validated. There were 433 strokes and 163 transient ischaemic attacks identified during follow-up. Of these, 333 were the first ever in a lifetime strokes of which 139 were definite ischaemic, 20 were haemorrhagic and 168 were probable ischaemic strokes. The crude incidence rate for stroke was 445 (95% confidence interval 398-493) per 100 000 person years. The age-standardised rates for 10 y age-bands were: 45-54 y 91 (10-172); 55-64 y 351 (269-432) and 65-74 y 855 (669-1040). The 30 d case-fatality rate was 21.0% (70/333) for all strokes and 19.2% (60/312) for ischaemic strokes. For transient ischaemic attacks the age-standardised incidence rates for the same 10 y age bands were 92 (4-179), 111 (64-157), and 273 (167-80), respectively. These rates for stroke transient ischaemic attack are likely to be accurate given the high ascertainment of events in this representative population of middle-aged men. Such studies, reporting reliable measures of cerebrovascular events, are important for measuring burden of disease, and for analysis of risk factor associations to help improve understanding of stroke aetiology and inform preventive efforts.