This study examines how spatial memory acquired from navigation is used to perform a survey task involving pointing. Participants learned a route through a virtual city by walking it multiple times in one direction on an omnidirectional treadmill. After learning, they were teleported to several locations along the route, self-localized and pointed to multiple other locations along the route. Pointing was done away from or towards the current location. Preliminary data show that participants were faster in pointing away. This suggests that pointing was based on an incremental process rather than an all-at-once process which is consistent with mentally walking through a cognitive map or constructing a mental model of currently non-visible areas of the city. On average participants pointed faster to targets located further down the route towards the end than to targets located route upwards towards the start. Analysis of individual performance showed that more participants than expected by chance showed such an effect of target direction also in their pointing accuracy. The direction of this effect differed between participants. These direction biases suggest that at least some participants encoded the environmental space by multiple interconnected locations and used this representation also for pointing.