Within the framework of cognitive rehabilitation using virtual reality (VR), one of the major challenges is to study beforehand the effectiveness of the virtual-real transfer of learning and to define cognitive aids. The aim of this experiment was to verify if, after learning spatial knowledge (i.e., a route) in VR, performances can be transferred to reality, then maintained in real time, and improved with the aid of an active navigation (i.e., using a joystick). Ninety student volunteers from the University of Bordeaux 2 (45 men and 45 women) participated in the experiment. The virtual environment (VE) used for learning was a replica of an area of Bordeaux. The factors tested were retention delay (Immediate vs. 48 hours) and type of navigation (Passive virtual vs. Active virtual vs. Real), using three recall tasks: wayfinding, freehand sketch and photograph classification. Our results showed that the virtual-real transfer was not degraded by a retention delay of 48 hours and that active navigation allowed performances to be optimized.