Abstract
Discerning occluding objects from their backgrounds is an automatic and fundamental process of vision, called figure-ground perception. The compulsory nature of this process and its profound impact on perception can be observed in many simple 2D displays, such as Rubin’s vase/faces illusion. Subsets of neurons in V1, V2, and V4 of the rhesus macaque have a neuronal correlate of this phenomenon. When stimulated with the contour of a figure, these neurons show an elevated firing rate when the figure is on one side of their receptive field (the "preferred" side) versus the other side. Thus, a local piece of a figure’s border has the side of the figure encoded by neurons with opposing side preferences (border-ownership coding). About half of the V2 neurons and a smaller proportion of V1 neurons exhibit border-ownership coding even in absence of depth cues (such as stereo disparity or dynamic occlusion). So far, such selectivity has only been demonstrated with displays of simple geometrical shapes (e.g., squares). Here we studied neurons with images of natural scenes in order to evaluate the role of border-ownership under more normal visual conditions. We record from V1 and V2 neurons of a rhesus macaque using microelectrodes. While the subject maintains fixation, we align occlusion-based contours within natural scenes (and manipulations thereof that we use to control for various factors) to the receptive field of the recorded neuron. We show that natural scenes do indeed elicit a preference for the side-of-object in a subset of V1/V2 neurons and that there is a significant agreement with the side-of-figure preference elicited by single squares. Furthermore, we show that the global context rather than information within the classical receptive field elicits most of this side-of-object related difference.
Meeting abstract presented at VSS 2012