Technical Program

Paper Detail

Paper: PS-2A.41
Session: Poster Session 2A
Location: Symphony/Overture
Session Time: Friday, September 7, 17:15 - 19:15
Presentation Time:Friday, September 7, 17:15 - 19:15
Presentation: Poster
Paper Title: How are the statistics of object co-occurrence represented in human visual cortex?
Manuscript:  Click here to view manuscript
Authors: Michael Bonner, Russell Epstein, University of Pennsylvania, United States
Abstract: We set out to identify cortical representations of contextual associations between objects. First we used a large database of densely labeled images to model the natural statistics of object co-occurrence. We then applied this co-occurrence model to fMRI data collected while subjects viewed images of isolated, individual objects. Using voxelwise encoding models, we tested whether object co-occurrence statistics could be used to predict the stimulus-evoked responses of visual cortex. We found that multiple regions of high-level visual cortex reliably encode the statistical associations between objects in their natural contexts, with prominent effects in scene-selective visual regions.