Abstract
Anstis et al. (VSS 2008) demonstrated that weak color signals from afterimages and blurry patterns spread across regions defined by strong luminance borders. While those demonstrations are new, they all depend on principles of visual perception that have previously been identified through work on a two-stimulus afterimage (e.g., Francis & Rothmayer, 2003; Francis & Schoonveld, 2005; VanHorn & Francis, 2007). Almost all of the effects described by Anstis et al. will be explained with computer simulations of a neural model that has already described similar phenomena. The analysis of the new cases did, however, reveal a new prediction of the model. The model predicts that it is possible for a color afterimage to appear for unseen visual colors. As Anstis et al. noted, a color plaid sinusoidal grating made of blue/yellow horizontal components and red/green vertical components appears to look like a blue/yellow horizontal grating when black horizontal lines are superimposed on the plaid. At offset of the image, the model predicts that the black horizontal lines produce vertical orientation after-responses that provide boundaries that constrain color after-responses to flow vertically. The unseen red/green colors produce visible afterimages because they are separated by the vertical boundaries, but the after-responses generated by the previously seen blue/yellow colors spread across each other and cancel out. An empirical study validates the model's prediction.