| Visual perception begins at the retina, with cluttered two-dimensional projections of the three-dimensional world. In the early twentieth century, Gestalt psychologists observed that we tend to organize this clutter through a process of figure-ground segregation---i.e. by identifying those regions of the retinal images that are object-related (figures) for further processing, and relegating other regions to the background. Recent neurophysiological evidence indicates that this process is carried out in the primate visual cortex through border ownership coding.;Border ownership coding involves neurons whose firing rates indicate the relationship of boundaries to figural regions in a retinotopic representation of a scene. This type of response selectivity, which has been observed in areas V1, V2, and V4, requires a mechanism through which edge-sensitive neurons can rapidly integrate contextual information that lies far outside of their classical receptive fields. Prior studies have invoked two anatomical mechanisms to explain how the necessary context information might be conveyed to such neurons: horizontal-fiber projections from neighboring neurons in the same area of visual cortex and inter-areal (i.e. white-matter) projections from neurons in other areas of visual cortex. The exact contributions made by these two mechanisms are still poorly understood, however.;The goal of my dissertation research has been to help elucidate the role that inter- and intra-areal circuits play in facilitating border ownership coding Toward this end, my research progressed in two distinct stages. In Chapter 2, I present a computational neural model that investigates the role of inter-areal feedback and contour grouping in shaping the responses of border ownership selective neurons. In Chapters 3 and 4, I then develop a computational/functional anatomy study of the constraints on context integration via intra-areal circuits. My analysis in these latter two chapters offers a novel, quantitative look at the cortical extent of the context integration required for border ownership coding. Overall, my results indicate that intra-areal circuits, alone, would not be sufficient to account for the full spatial extent and speed of context integration that has been observed experimentally for border ownership neurons. |
