A Bayesian framework for crowding effect

In crowding, neighboring distractors impair the visual perception of a presented ta get. We study influences by the configuration of distractors on the bias to perceive the orientation of a target. Our results show that: (a) when distractors are similar to each other but different from target, crowding is decreased; (b) when distractors form a subjective contour, crowding is also reduced. These results illustrate that crowding is weak whenever the target stands out from the context and strong when the target is grouped into the context as a part of a global percept. In addition, we show how a Bayesian model, based on the principle of spatial resolution of attention that is modulated by the large size of receptive fields, can account for the behavioral data.