Robust encoding of scene anticipation during human spatial navigation.

In a familiar city, people can recall scene views (e.g., a particular street corner scene) they could encounter again in the future. Complex objects with multiple features are represented by multiple neural units (channels) in the brain, but when anticipating a scene view, the kind of feature that is assigned to a specific channel is unknown. Here, we studied neural encoding of scene view anticipation during spatial navigation, using a novel data-driven analysis to evaluate encoding channels. Our encoding models, based on functional magnetic resonance imaging (fMRI) activity, provided channel error correction via redundant channel assignments that reflected the navigation environment. We also found that our encoding models strongly reflected brain activity in the inferior parietal gyrus and precuneus, and that details of future scenes were locally represented in the superior prefrontal gyrus and temporal pole. Furthermore, a decoder associated with the encoding models accurately predicted future scene views in both passive and active navigation. These results suggest that the human brain uses scene anticipation, mediated especially by parietal and medial prefrontal cortical areas, as a robust and effective navigation processing.