Moving beyond category-selectivity: What can fMRI tell us about large-scale interactions in vision?

Visual perception is commonly viewed as a stimulus-driven process, whereby neural representations of increasing complexity are hierarchically assembled from primary sensory areas through category-selective regions to high-level association areas. Vision provides a great opportunity to study cortical mechanisms of perception, as the ordered hierarchical organization has been amply demonstrated and modeled formally in many computational models. Despite their success, however, computational models rarely perform as well as the biological system, and often fail to take account of the highly interactive nature of cortical networks - involving interactions between different processing pathways as well as across different levels of the hierarchy.

In the current talk, I will present a series of neuroimaging studies, which demonstrate how representations in dedicated brain regions in visual cortex emerge from interactions with large-scale networks, exemplifying both functional and neuroanatomical constraints. Specifically, I will describe recent investigations of object- and scene-selective cortex that reveal (1) the large impact that top-down factors, such as experience and task demands have on the neural representations of visual objects and (2) how the distinction between object and scene representations can be accounted for by the patterns of connectivity within and across the ventral and dorsal visual processing pathways.