Purchase this article with an account.
Barbara A. Doshe, Zhong-Lin Lu; Perceptual learning in first- and second-order letter identification. Journal of Vision 2004;4(8):296. doi: 10.1167/4.8.296.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
It has been proposed that there exist two distinct visual information processing systems (Cavanaugh, 1989, Chubb & Sperling, 1989; Lu & Sperling, 1995). The first-order system operates on luminance modulation based on the output of linear filters. The second-order system operates on modulation of features, such as texture, in the stimulus. A typical second-order system consists of a first stage of linear filters, rectification, and a second stage of filters (Sutter & Graham, 1992; Wilson, Ferrera, & Yo, 1992). Perceptual learning in first- and second-order processing may reflect learning at different stages of visual processing. In these experiments, we compared perceptual learning in a first-order and in a second-order version of a letter identification task. Observers discriminated a K from its mirror reversal embedded in one of eight levels of external Gaussian pixel noise. In the first-order task, the letter was shown in black pixels; in the second-order task, the letter appeared as a pattern of black and white pixels. Four observers practiced the first-order task for 5 days (> 6000 trials) yet exhibited no systematic reductions in contrast thresholds, although there was some significant day-to-day variation. In contrast, four observers practicing the second-order task for 5 days primarily showed substantial learning in low noise tests, although there may have been significant but small improvements in high noise tests as well. Practice beyond 5 days in the second-order task, tested in a subset of observers, continued to lead to improvement in performance. The large improvement in perceptual learning in low noise for the second-order task reflects enhancement of first-stage channels in second-order processing, while the lack of substantial perceptual learning in this case of first-order processing suggests only minor improvements in first-order stages of processing.
This PDF is available to Subscribers Only