Abstract
An essential component of visually guided navigation is the ability to perceive features of the environment that afford or constrain movement. For example, in indoor environments, walls limit one’s potential routes, while passageways facilitate movement. Here we attempt to identify the cortical mechanisms that encode such navigational features. Specifically, we test the hypothesis that scene-selective regions of the human brain represent navigational affordances in visual scenes. In an fMRI experiment, subjects viewed images of artificially rendered rooms that had identical geometry as defined by their walls, but varied on the number (one to three) and position (left, right, center) of spatial passageways (i.e., open doorways) connected to them. Thus, the layout of these passageways defined the navigable space in each scene. Several versions of each layout were shown, each with the same set of passageways but different textures on the walls and floors. Furthermore, half the rooms were empty except for the walls and passageways, while the other half included visual clutter in the form of paintings along the walls, which were similar in size and shape to the passageways. Images were presented for 2 seconds while subjects maintained central fixation and performed an unrelated color-discrimination task on two dots overlaid on each scene. Using multivoxel-pattern analysis, we sought to identify representations of navigational layout that were invariant to other visual properties of the images. This analysis revealed consistent and invariant coding of navigational layout in the occipital place area (OPA), a scene-selective region near the transverse occipital sulcus. In this region, scenes elicited similar activation patterns if they had similar navigational layout, independent of changes in texture or the presence of visual clutter (i.e., paintings on the walls). These findings suggest that the OPA encodes representations of navigational affordances that could be used to guide movement through local space.
Meeting abstract presented at VSS 2015