Abstract
Perceptual learning of 10AFC texture identification is stimulus-specific: After a fixed set of ten textures is learned, accuracy drops substantially when those textures are rotated 180 degrees, reversed in polarity, or when a novel set of textures is presented. Two questions arise: 1) Can perceptual learning occur without any repetition of items during training? 2) Does exposure to a more variable training set increase transfer of learning? We trained three groups of observers in a 10AFC texture identification task on two days (420 trials/day). Method of constant stimuli was used to present the textures at multiple contrasts and noise levels. The Fixed group viewed a fixed set of 10 textures throughout training (840 exposures per texture set). The Unlimited group viewed 840 novel sets of 10 textures (1 exposure per texture set). The Switch group viewed a fixed set of 10 textures on Day 1, and a different fixed set on Day 2 (420 exposures per texture set). In all groups, transfer of learning was tested on Days 3 and 4 by using a fixed set of textures in both sessions and having half the observers from each group switch to a novel set on Day 4. Response time and accuracy were measured. Results: 1) Gradual, but significant, learning occurred even when each trial comprised novel stimuli. 2) Greater transfer of learning from Day 3 to Day 4 was found in the Unlimited and Switch groups. Specifically, the Unlimited and Switch groups were affected less by, and recovered faster from, a change in textures viewed on Day 4. We conclude that the amount of stimulus-specificity can vary: increasing stimulus variability during practice may induce the use of strategies that increase generalization of learning to new stimuli.