BACKGROUND: Prospective population-based studies are important to accurately determine the incidence and characteristics of stroke associated with atrial fibrillation (AF), while avoiding selection bias which may complicate hospital-based studies. METHODS: We investigated AF-associated stroke within the North Dublin Population Stroke Study, a prospective cohort study of stroke/transient ischaemic attack in 294,592 individuals, according to recommended criteria for rigorous stroke epidemiological studies. RESULTS: Of 568 stroke patients ascertained in the first year, 31.2% (177/568) were associated with AF (90.4%, i.e. 160/177 ischaemic infarcts). The crude incidence rate of all AF-associated stroke was 60/100,000 person-years (95% CI = 52-70). Prior stroke was almost twice as common in AF compared to non-AF groups (21.9 vs. 12.8%, p = 0.01). The frequency of AF progressively increased across ischaemic stroke patients stratified by increasing stroke severity (NIHSS 0-4, 29.7%; 5-9, 38.1%; 10-14, 43.8%; >or=15, 53.3%, p < 0.0001). The 90-day trajectory of recovery of AF-associated stroke was identical to that of non-AF stroke, but Rankin scores in AF stroke remained higher at 7, 28 and 90 days (p < 0.001 for all). DISCUSSION: AF-associated stroke occurred in one third of all patients and was associated with a distinct profile of recurrent, severe and disabling stroke. Targeted strategies to increase anticoagulation rates may provide a substantial benefit to prevent severe disabling stroke at a population level.