A Neural Circuit for Perceptual Grouping, Segmentation, and Selection

Abstract

Francis et al. (2017) showed that a neural circuit for grouping and segmentation explains a variety of visual uncrowding effects. When a central target did not group with flankers, the circuit could segment out the flankers, thereby freeing the target from crowding effects. A limitation of that model is that fixed parameter values strictly determine the range of perceptual grouping. We describe a revised grouping circuit that allows for top-down control of grouping range via modification of a single timing parameter. Different values can result in different groupings of elements in a scene. Depending on how the grouping circuit connects visual elements, the segmentation process may select individual items, all items, or a subset of items. Based on the behavior of this revised model, we propose that various empirical tasks designed to measure grouping track how easily observers can segment and select relevant parts of a visual scene. Model simulations that quantify ease of grouping closely match reported subjective ratings of target distinguishability (Manassi et al., 2012). Such judgments are well explained by the segmentation process. Flankers larger than the target, for example, are not grouped with the target for a wide span of top-down range control values. In contrast, flankers equal in size to the target cannot form separate groups: either all or none of the elements group together. The same model also accounts for objective measures of grouping. Palmer and Beck (2007) varied spacing between items and had subjects locate a matching pair of adjacent items. We show that challenging conditions for this task occur when a narrow set of range control parameters are required. This set varies as a function of spacing between items. Overall, we show that the model provides a novel characterization of grouping effects that is applicable to a wide variety of situations.