This study investigated the extent to which people can develop a global representation of local environments through across-boundary navigation. Participants learned objects' locations in two misaligned rectangular rooms in an immersive virtual environment. After learning, they adopted a local view in one room and judged directions of objects within the room; the views in two consecutive trials were from different rooms and locally or globally consistent (priming task). In some experiments, participants learned locations of five buildings before learning the objects in the rooms. In testing, after the priming task, they pointed to the buildings while adopting local views inside the rooms (across-boundary pointing task). Participants' estimated global headings were calculated from their pointing responses. The results showed that the priming effect from the globally consistent views occurred when participants learned the buildings and then locomoted between the rooms through a simple path. Consistent with the global priming effect, the means of participants' estimated global headings were accurate. In contrast, there was only the priming effect from the locally consistent views when participants did not learn the buildings before learning the objects inside the rooms or when participants were teleported between the rooms after learning the buildings. These results suggest that people can develop global representations of local environments through across-boundary navigation while traveling a simple path, provided that there are prior global representations.