The natural history of aggressive intracranial dural arteriovenous fistulae (ICDAVF) is unknown. Despite this, the recently proposed classification scheme of Borden et al (Borden(*)) has the potential to predict aggressive lesion behavior after presentation for any lesion, but has so far been untested. In addition, they discuss a new but logical treatment strategy for aggressive ICDAVF based on the elimination of retrograde leptomeningeal venous drainage (RLVD). Our similar philosophy and substantial experience with these lesions, provides a unique opportunity to test these hypotheses. A cohort of 46 Borden(*) grade II and III ICDAVF was selected from a series of 102 ICDAVF seen at a single institution between 1984 and 1995. Patients with these lesions, presumed to have an aggressive course were all offered treatment. Conservative therapy was chosen by 14 (30%) patients, 22 (47%) had surgery, and 20 (43%) had embolisation either as sole treatment or prior to surgery. During the follow-up period (249 lesion months) for the conservatively treated group, Jour (29%) patients died. Excluding presentation, these patients were observed to have interval rates of intracranial hemorrhage (ICH), non haemorrhagic neurological deficit (NHND), and mortality, of 19.2%, 10.9%, and 19.3% / lesion year respectively. The 11 patients who had embolisation alone were followed for a total of 344 months after treatment. All nine patients who had lesion obliteration, or subtotal obliteration with elimination of RLVD, as confirmed by angiography, experienced improvement or complete clinical recovery. Two patients had subtotal obliteration without elimination of RLVD. One died from interval ICH and the other experienced a delayed NHND. Twenty-five surgical operations were performed on 23 ICDAVF in 22 patients. Resection of the ICDAVF was performed in 9 patients, and 16 patients were treated with surgical disconnection alone. Complications occurred in 3/9 (33%) patients who had their lesions resected and none of the disconnected group. Failure to achieve angiographic obliteration of RLVD in 2 patients treated with resection was associated with an adverse outcome in both cases (death, and interval NHND). All 16 (100%) of the disconnected group were shown to have undergone angiographic obliteration with excellent clinical outcome. Untreated, Borden(*) grade II and III ICDAVF have a poor natural history. Also, persistence of RLVD after inadequate treatment results in adverse outcomes. Embolisation usually improves the safety of surgical access and may lead to obliteration on its own in some cases. For the aggressive ICDAVF, surgery is required in most cases, and our data confirm that surgical disconnection alone results in cure of all Borden(*) grade III ICDAVF, and in grade II lesions, ifnot cure, conversion to a benign grade I lesion.