Spatial Representations and Spatial Updating

This chapter discusses the basic properties of spatial representations for navigation, spatial reasoning, and object/scene recognition. Contrary to the traditional models of allocentric cognitive maps, recent findings suggest that spatial representations for navigation, real world object and scene recognition, and spatial reasoning are primarily egocentric, and these egocentric representations are updated as the viewer moves. Much of the evidence for allocentric representations, such as the novel shortcut ability, the place and HD cells in rodents, and findings in spatial reasoning tasks, is shown to be either insufficient or based on the wrong assumptions. Moreover, spatial representations of environments at adjacent levels of a "hierarchy" cannot be accessed at the same time, and spatial updating in one environment does not automatically result in updated orientation in another environment, suggesting that representations of these environments are fragmented in nature, rather than integrated hierarchical networks. The fragmentation of the representations of navigational space is consistent with, and may be a direct consequence of, the egocentric nature of the spatial representations and is difficult to explain by allocentric cognitive maps. According to the egocentric updating model, the amount of computation increases as the number of targets increases. Thus, a direct consequence of such a system is that only a subset of the targets may be updated at a time due to limitations in the capacity of the updating process. Accordingly, one may update targets in one environment but not the others, and people switch the environment they update when they navigate from one environment to another (Wang & Brockmole, in press). In contrast, both a single, comprehensive cognitive map and an interconnected, hierarchical network predict that knowing one's orientation in one environment would also specify one's orientation in another environment, and thus have difficulty explaining these findings. In summary, recent findings suggest that navigation, spatial reasoning, and object/scene recognition are primarily based on egocentric representations that are updated as the animal moves. The updating process may have limited computational capacity, and does not apply to all environments simultaneously. Thus, spatial representations learned through navigation are fragmented in nature, rather than integrated hierarchical networks.