McCollough Effect: A neural network model based on source separation

International audienceMcCollough effects (MEs) are a group of visual contingent aftereffects that involve colour and contour. These effects have been the subject of a large body of literature concerning their properties and theoretical accounts, but the mechanisms underlying the ME have never been fully clarified. We make the assumption that a general adaptive neural process tending to maintain independent dimensions in visual perception could account for the ME. The proposed neural network model generating the ME, though of minimal complexity, can reproduce various detailed experimental results (such as the tilt effect contingent to colour) and above all it accounts for the distinctive long temporal persistence of this aftereffect