Figure 5.

Perceptual learning depends on perceptual constancy. Pairs of squares were shown on each trial, and the observer’s task was to discover which ones were members of some category (see text). Perceived lightness depends on ratios of the local luminance (within a square) to its surround (Wallach, 1948). Bottom, center: When the invariant determining the classification was identical lightness of the two squares (accomplished by having the same local luminance for each and placing both on the same background), learning readily occurred. Acquisition curves are shown form 5 observers (bottom right). Top, center: When the identical local luminance squares were shown on a variable backgrounds, it caused differences in perceived lightness. In this condition, only the local luminances of the squares (a sensory invariant, decorrelated from perceived lightness) could be used to learn the classification. As in the other experiments involving size and motion, learning did not occur for any observer under this condition. These results suggest that perceptual learning works through constancy-based perceptual representations. (After Garrigan & Kellman, 2008).