Abstract
A subject may group stimulus elements in different ways using, e.g., size, shape, and proximity. Such groupings can be tuned to promote performance for a given task, e.g., look for nearby similarly shaped objects or for horizontally aligned equisized circles. Francis et al. (2017) showed how to tune a neural segmentation circuit to explain various uncrowding effects. The circuit could segment out flankers that did not group with the target, thereby freeing it from crowding effects. Here, we introduce methods to tune neural circuits for perceptual grouping and describe strategies that promote performance in visual search tasks. In the model, grouping occurs when illusory contours connect spatially separated elements. The modified circuit allows for top-down control of a timing parameter that establishes the range of illusory contour formation. Separate top-down control of two size parameters prevents illusory contour formation for elements having too much or too little contour boundaries. We propose that subjects are able to tune these parameters for stimuli and tasks so as to group elements in a way that allows targets to be segmented from distractors. Within an experiment, the parameters can be changed to reflect previous failures to find the target, and strategic control of the parameters within a trial correspond to different search strategies that flexibly change groupings to promote target search. We show that when the model uses particular strategies, simulated results closely match human performance in visual search tasks where perceptual grouping is induced by proximity and shape similarity (Palmer & Beck, 2007) or by the spacing of irrelevant distractors and size similarity (Vickery, 2008). Thus, we show that the model accounts for a variety of grouping effects and indicates which grouping strategies were likely used.