Purpose: To evaluate the 3-dimensional accuracy of dental implant drilling in a computer-assisted navigation (CAN) system using simulated mandible models. Materials and Methods: Eight acrylic resin models were fabricated to simulate human mandibles containing mandibular canal (MC). Computerized tomography (CT) scans were obtained for each model, and the data were transferred to the system for dental implant planning. The models were mounted on a phantom head to simulate surgical situation. The assessment parameters included entry point localization, drill path angulation, and drilling depth, which were directly measured by sectioning of the models. Results: Eighty drill holes were made on the 8 models. The entry point localization showed a mean deviation of 0.43 mm (range, 0 to 2.23 mm; SD, 0.56 mm) from the plan. The angulation showed a mean deviation of 4.0 degrees (range, 0 to 13.6 degrees; SD, 3.5 degrees). The drill aimed at stopping as close to the upper border of the MC as possible without perforating it, and 65% (52) of the drill holes managed to come within 1 mm. Another 5% of the holes stopped 1 to 2 mm above the MC. None of the drill holes stopped more than 2 mm above the MC. However, 30% (24 of 80) of the drill holes perforated the upper border of MC, and the mean depth of perforation was 0.37 mm (range, 0.01 to 1.04 mm; SD, 0.28 mm). Discussion and Conclusion: The CAN system identified the entry location and angulation with mean deviations of 0.43 mm and 4 degrees, respectively. About two thirds of the drillings achieved accuracy within 1 mm above the MC. Thirty percent perforated into the MC, and the maximal depth was 1.04 mm. In the planning stage, the maximal depth of the implant should be at least 1.1 mm above the superior border of MC as a safety margin. (Technical Report) Keywords: computer-assisted navigation, dental implants, image-guided navigation