High-level neural structures constrain visual behavior

Abstract

Visual cognition is notoriously limited: only a finite amount of information can be fully processed at a given instant. What is the source of these limitations? Here, we suggest that the organization of higher-level visual cortex into content-specific channels constrains information processing across the visual system. Each channel is primarily involved in representing one particular type of visual content (e.g. faces, cars, certain types of shapes, etc.). Furthermore, each channel has a finite processing capacity/bandwidth and is limited in the amount of information it can process. When multiple items are simultaneously presented across space, or quickly in time, the extent to which those items activate overlapping channels will constrain the amount of information that can be successfully processed. To examine this, we used brain/behavior correlations in which we directly compared behavioral performance on a perceptual task with the amount of overlap amongst the neural channels used to support the items from the behavioral task. In Chapter 1, we found that the amount of information that could be encoded on a change detection task was correlated with the amount of channel overlap within occipitotemporal cortex, but not early visual regions such as V1-V3. In Chapter 2, we extend this finding by showing that the amount of information that could reach visual awareness in a masking paradigm was also predicted by overlap amongst occipitotemporal, as well as occipitoparietal channels, but once again not in V1-V3. Finally, in Chapter 3, we sought to identify which particular channels were the most behaviorally relevant and found that virtually any part of higher-level visual cortex (e.g. across occipitotemporal cortex, within category selective regions, within the least active voxels, amongst a random sample of voxels, etc.) was significantly correlated with behavioral performance. Together, these results suggest that visual cognition is limited by a set of neural channels that extend across the majority of higher-level visual cortex. These findings have direct implications on many prominent models of visual cognition, specifically those focused on perceptual limitations, and help clarify the large-scale representational structure in higher-level visual cortex.