Abstract
Previous studies have found brain regions to be associated with high-level visual processing. Amongst these, the lateral occipital complex is sensitive to object processing, the parahippocampal place area (PPA), retrosplenial complex (RSC), and the occipital place area (OPA) are sensitive to scene processing such as content and scene categorization. Most research on object modulation in scene processing has focused on computer-generated stimuli with controlled properties. The purpose of this study is to understand to what extent object differentiation in complex, real-world scenes is reflected in the activation patterns of these regions of interest (ROIs). This study utilized data from BOLD5000, an fMRI dataset consisting of 5,000 real-world scene stimuli (Chang et al., 2019). We categorized the images into 24 scene categories that had at least 35 different exemplars (M=57.6). The current analysis included nine outdoor categories and analyzed their object agreement across exemplars. Each image was thoroughly labeled by a consensus of human analyzers to yield an object-scene category matrix. Each matrix was cross-correlated to produce a similarity space defined by object presence. To address how object differentiation affects representation in these ROIs, a representational similarity analysis was performed comparing the similarity matrix defined by objects with a similarity matrix defined by the fMRI response within a given ROI. The results demonstrate a significant main effect of hemisphere across all ROIs, with higher similarity in the right. This was particularly significant within the OPA and trended in the RSC. We plan to incorporate the remaining 15 categories into the current analysis. These results suggest the right hemisphere in general, and in the OPA in specific, is involved in representing objects, at least when observed within real-world scenes. This work proposes a new framework for understanding how category selective regions process the content of scenes.