The epidemiology and natural history of cerebral arteriovenous malformations (AVMs) remains incompletely elucidated. Several factors are responsible. With regard to the incidence and prevalence of AVMs, the results of prior studies have suffered because of the retrospective design, the use of nonspecific ICD-9 codes, and a focus on small genetically isolated populations. Recent data from the New York Islands AVM Hemorrhage Study, an ongoing, prospective, population-based survey determining the incidence of AVM-related hemorrhage and the associated rates of morbidity and mortality in a zip code-defined population of 10 million people, suggests that the AVM detection rate is 1.21/100,000 person-years (95% confidence interval [CI] 1.02-1.42) and the incidence of AVM-hemorrhage is 0.42/100,000 person-years (95% CI 0.32-0.55). Contemporaneous data from the Northern Manhattan Stroke Study, a prospective, longitudinal population-based study of nearly 150,000 patients in which the focus is to define the incidence of stroke, suggest the crude incidence for first-ever AVM-related hemorrhage to be 0.55/100,000 person-years (95% CI 0.11-1.61). Efforts are ongoing to study the natural history of both ruptured and unruptured AVMs in these datasets to examine the relevance of prior studies of patients selected for conservative follow up in Finland. In addition, data are being gathered to determine whether risk factors for future hemorrhage, which have previously been established in small case series, are valid when applied to whole populations. Together, these data should help inform therapeutic decisionmaking.