A cohort of 469 hospital-referred patients with transient ischaemic attacks (TIA) of the brain (66%) or eye (34%) due to presumed atheromatous thromboembolism, lipohyalinosis or cardiogenic embolism, without prior stroke, was assembled between 1976-86. Follow up was prospective and complete until the patients death or the end of 1986. During a mean period of follow up of 4.1 years there were 82 deaths (58 vascular, 24 non-vascular), 63 first-ever strokes and 58 patients with coronary events. A coronary event accounted for 51% of deaths whilst stroke was the cause in 12%. The average risk of death over the first five years after TIA was 4.5% per year. The risk of stroke was 6.6% in the first year and 3.4% per year on average over the first five years. Stroke occurred in the same vascular territory as the initial TIA in about two-thirds of cases, and was of lacunar type in one fifth of these strokes. The average risk of a coronary event over the first five years after TIA was 3.1% per year, similar to that of stroke. However, the risk of a coronary event, and also death, was fairly constant each year after a TIA, in contrast to the risk of stroke which was highest in the first year. The average risk of stroke, myocardial infarction or vascular death over the first five years after TIA was 6.5% per year and the average risk of stroke, myocardial infarction or death from any cause was 7.5% per year. The prognosis of this cohort of hospital-referred TIA patients was better than that of TIA patients in the same community who presented to the Oxfordshire Community Stroke Project (OCSP), and reflected the impact of referral bias. The hospital-referred patients were younger, assessed at a later date after their last TIA, and comprised a greater proportion of patients who had had a TIA of the eye (amaurosis fugax), which had a better prognosis than TIA of the brain. Knowledge of the prognosis of different populations of TIA patients not only enhances understanding and interpretation of previous studies but is also required for optimal patient management and the planning of treatment trials.