Abstract
Anderson & Winawer (2005, Nature) presented a dramatic lightness illusion, where identical texture patches appear either black or white depending on background properties. Albert (2007, VR) argues that simple well-known visual mechanisms detecting local occlusion account for the illusion, determining which stimulus areas are perceived in plain view. Here, we present modeling and experimental evidence that a local occlusion explanation fails to account for the illusion (see also Poirier, submitted). Indeed, modeling shows that local occlusion theories predict a complete reversal of the effect at lower figure contrasts, even within the range used by Anderson & Winawer. However, this predicted reversal was absent in both the original data and in our replication (N=7). Moreover, repeating the experiment using thresholded textures (thus emphasizing local occlusion cues) did produce the expected reversal, thus providing evidence that our modeling did capture the essence of local occlusion. This further reinforces our conclusion that theories based on local occlusion cues fail to account for the Anderson-Winawer illusion. We propose a simple model to account for these effects, as well as other related phenomena, e.g. the White effect (White, 1979, 1981, Perception), and simultaneous contrast (Heinemann, 1955, JEP; Horeman, 1963, VR).
This research was supported by a CIHR grant awarded to Martin Arguin, Frédéric Gosselin, and Dan Bub.