Abstract
The visual system groups parts of an object together and segments different objects from each other and the background. We challenge the view that grouping is done in parallel and hypothesize that attention spreads over the area of an object to group the object together and to segment it from other objects and the background. We distinguish three models that can explain how the brain achieves perceptual grouping and make a prediction on the time-course of image-segmentation: 1) the eccentricity model, which predicts that the speed of attentional spread depends on the Euclidean distance between points; 2) the pixel-by-pixel model, which predicts that attention spreads at the same speed in all directions within the object; 3) the growth-cone model, which takes into consideration the size of receptive fields in visual areas and predicts that attention spreads faster over homogeneous areas. To test the models, we investigate whether the reaction time (RT) pattern is in accordance with the model prediction for various stimulus sets that vary in their level of complexity. Participants indicate whether two cues are on the same or on two different objects. Eighty-eight participants were randomly assigned to one of the following conditions: natural images, detailed cartoons, cartoon outlines, and scrambled cartoons. Regression analysis of the RT-data shows that the explained variance is highest for the growth-cone model. This demonstrates that attention spreads over the object, thereby serially grouping the different object-parts together and that this attentional spread is faster if it spreads over homogeneous areas of the object and slower on small or narrow part of the object. The superior performance of the growth-cone model is very consistent for the various image conditions, showing that our brain probably uses similar strategies to perceptually group simplified as well as complex objects together into one single object.
Meeting abstract presented at VSS 2012