Holistic Object Representation in Human Visual Cortex

J. Zhang,Y. Song,J. Liu,Y. Xu

DOI: https://doi.org/10.1167/12.9.271

2012-01-01

Journal of Vision

Abstract:Object parts often interact to form holistic representations. Here we used fMRI and multi-voxel pattern analysis to study the neural underpinnings of holistic object representation for faces and cars. We used a block design and extracted fMRI response pattern from predefined regions of interest (ROI) in lateral and ventral visual cortex for object parts either presented alone, together in the correct configuration or in a scrambled configuration. We first trained our classifier to distinguish between the patterns evoked by the intact and the scrambled object images using support vector machine. We then averaged the patterns evoked by the individual parts presented alone to create a synthesized pattern and asked the trained classifier to classify this synthesized pattern. If holistic object representation is not coded in an ROI, because the synthesized image looks as dissimilar to the intact as it does to the scrambled object, the synthesized pattern should be classified at chance as either an intact or a scrambled object. However, if holistic object representation is coded in an ROI, because only the intact object pattern contains such information, the synthesized pattern would be more likely to be classified as a scrambled than as an intact object. For faces, only the right fusiform face area (FFA) exhibited such response bias; and for cars, only the right lateral occipital (LO) area did so. All other object areas examined showed no classification bias. These results indicate holistic face processing in the right FFA and holistic car processing in the right LO. Together they suggest that right visual object areas play important roles in representing holistic object information, with different right visual object areas representing such information for different categories of objects. Meeting abstract presented at VSS 2012