Abstract
We examined the time-course of perceptual learning of face- and texture-identification, and the robustness of that learning across years. On the first session (baseline), observers performed a total of 840 trials of 10AFC face- or texture-identification in three levels of external noise. Method of constant stimuli was used to display the images at one of seven different contrasts at each level of external noise. Proportion correct and identification thresholds were measured. Observers returned the next day to complete the second session under the same experimental conditions. Perceptual learning was indicated by a significant increase in proportion correct relative to baseline in both tasks. The time course of learning revealed substantial within-session gains during the first session followed by some overnight improvement, with performance continuing to improve during the second session. Observers then returned to perform a third session about 12–18 months later, and were tested in two conditions. One condition used the same stimuli that were used during the first two sessions. The other condition used novel stimuli. Even after more than a year, for both faces and textures, there was almost complete retention of perceptual learning. However, some differences between faces and textures were observed. Identification of previously-viewed faces was slightly less accurate than the second session performed over a year prior, but was still considerably greater than baseline. Moreover, there was no evidence that learning transferred to novel faces. Identification of previously-viewed textures was virtually identical to that measured during the second session over a year earlier, and learning partially generalized to the novel textures. Clearly, perceptual learning of faces and textures endures for years, and the specificity of learning depends on stimulus type.
This research was supported by NSERC operating grants 105494 and 42133 and the Canada Research Chairs Programme (ABS & PJB), and an NSERC PGS-D fellowship (ZH)