Scene-selective regions play a key role in the perception and recognition of the visual world. However, the principles that govern the topography of these regions have not been fully resolved. In this study, we directly compare the relative importance of low-level image properties and high-level scene category on the organization of scene-selective regions. fMRI responses were measured while 24 participants viewed images of two categories of scene: indoor and natural. Images of scenes were filtered in spatial frequency to generate four stimulus conditions: (1) indoor, high-pass; (2) indoor, low-pass; (3) natural, high-pass; (4) natural, low-pass that were presented in a blocked design. Scene-selective regions were defined in a localizer scan by comparing the response to intact and scrambled scenes. Scene-selective regions included the PPA, RSC and TOS. The patterns of response to each of the four conditions were compared using correlation based MVPA. Our prediction was that if high-level categorization is important, then the pattern of response to conditions that have the same category should be higher than to conditions that contain the same spatial frequency. On the other hand, if low-level image properties are important, the pattern of response to conditions with the same spatial frequency should be higher than to conditions with the same category. Using multiple regression, we found that both image properties and scene category were found to explain a significant proportion of the variance in the patterns of neural response. However, a significantly greater proportion of the variance in neural response was accounted for by the image properties. These results suggest that the topographic organisation of high-level visual regions is tightly coupled to low-level properties of the image.
Meeting abstract presented at VSS 2014