Purchase this article with an account.
Minjung Kim, Jason Gold, Richard Murray; Classification images reveal that local grouping within lighting frameworks drives the argyle illusion. Journal of Vision 2015;15(12):632. doi: 10.1167/15.12.632.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The argyle illusion (Adelson, 1993) is a brightness illusion in which lines, triangles, and diamonds are arranged to resemble patterns on knitted garments. The standard explanation is that, due to shapes forming x-junctions, the visual system perceives strips of dark and light transparent filters over the argyle. Since diamonds under the two different filters have the same luminance, the visual system infers that the diamonds must be of different reflectances, thus perceiving a brightness difference. Using classification images, a technique which reveals image components that influence perceptual decisions, we examined whether the x-junction explanation was correct, i.e., whether x-junctions would be the most influential parts of the classification image. We chose two test diamonds, one from each of the filtered regions, and set their luminances to observers’ PSEs. Other argyle elements were corrupted with white, Gaussian noise (SD=.18). Observers (n=3, 10,200 trials each) chose which test diamond appeared brighter (2AFC). Surprisingly, classification images did not show that x-junctions were the most influential stimulus elements; rather, local contrast better explained observer choices. We also found that diamonds immediately neighbouring the test diamonds contributed to the illusion, but only if the neighbours were under the same filters as the test diamonds. This suggests a localized grouping effect, where neighbours sharing the same lighting framework moderate perceived brightness of the test diamonds. We conclude that x-junctions do not play a decisive role in the argyle illusion, and that instead, stimulus patches are compared to local grouped elements. Our experiments point towards the importance of lighting frameworks for brightness perception, and demonstrate classification images as a general technique for examining mid-level lighting effects, such as shading and transparency.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only