The application of virtual reality (VR) to the neurosurgical field has been increasing recently, however, the relation between the surgeon and the VR environment is rarely studied. We examined the trajectory of a surgical instrument during manipulation of a virtual object using a video-see-through microscope and a neurosurgical navigator (CANS Navigator) to find better surgeon-microscope interface. A resin cylindrical phantom was produced representing the surgical field, which included two 3 dimensionally arranged small spheres and a virtual 'gate'. The phantom was fixed and set under the microscope with a skull clamp mimicking conditions in an ordinary craniotomy. Firstly, the binocular parallax perception under microscope was examined. Experienced and inexperienced neurosurgeons were asked to learn the position of the virtual 'gate' for 3 minutes. Then, after 5 minutes to point with the navigator probe (suction tube), under various conditions; under the naked eyes, under the microscope, under the navigator without observing the phantom, and under the microscope with picture in picture (PIP) display of the navigational image. The positions of the suction tube were recorded at real time into the navigator for later analysis. Secondly, the task performance in this VR environment was studied by analyzing the trajectory of the suction tube from one sphere to the other sphere passing the virtual 'gate' under various conditions. A significant difference in pointing precision between experienced and inexperienced neurosurgeons was able to be observed only under microscope. This difference was mainly derived from overestimation of the depth of the virtual 'gate' by the inexperienced neurosurgeons. Among the above conditions, pointing under the microscope with PIP was able to be performed the most precisely and the most promptly. This study disclosed the presence of stereoscopic distortion in the microscope. The PIP display of the navigational image in the microscopic view remarkably improved the task performance, which could be accounted for by the correction of the somewhat distorted binocular parallax perception under the neurosurgical microscope by the provision of another visual key.