Object. Surgery to monitor and resect epileptogenic foci may be undertaken in 2 stages, providing an opportunity to use skull-fixated fiducials implanted during the first stage to improve the accuracy of cortical resection during the second stage. This study compared the intrinsic accuracy of skin-based and skull-fixated fiducial markers in registering frameless stereotaxy during pediatric epilepsy surgery. To the authors' knowledge, these modalities of registration have not previously been directly compared in this population. Methods. The authors undertook a retrospective review of pediatric patients who underwent resection of epileptogenic foci in 2 stages with frameless stereotactic assistance, performed by a single surgeon at Oregon Health & Science University. For the first stage (subdural grid implantation), 9 skin fiducial markers were used to register anatomical data in a frameless stereotactic station. Intraoperatively, four 3-mm screws were placed circumferentially around the craniotomy. Postoperatively, thin-slice brain MR and CT images were obtained and fused. For the second stage, the 4 screws were used as fiducial markers to register the stereotactic anatomical data. For both stages, accuracy (difference in millimeters from zero of the manual fiducial registration compared with the computer model) was determined using navigation software. The intrinsic accuracy of these 2 methods of fiducial registration was compared using a paired Student t-test. Results. Between 2004 and 2009, 40 pediatric patients with epilepsy underwent frameless stereotactic surgical procedures. Fourteen patients who had 2-stage procedures using skin-based and skull-fixated registration with complete accuracy data were included in this retrospective review. Mean registration error was significantly lower using skull-fixated fiducials (1.35 mm, 95% CI 1.09-1.60 mm) than using skin-based fiducials (1.85 mm, 95% CI 1.56-2.13 mm; p = 0.0016). Conclusions. A significantly higher degree of accuracy was achieved using 4 skull-fixated fiducials compared with using 9 skin-based fiducials. This simple and accurate method for registering frameless stereotactic anatomical data does not involve the potential time, expense, discomfort, and morbidity of extraoperative skull-fixated fiducial placement. The method described in this paper could also be extrapolated to other planned 2-stage cranial surgical procedures such as combined skull base approaches.