Purchase this article with an account.
Roger Li, Charlie Ngo, Dennis Levi, Jussi Saarinen; The specificity of learning position discrimination: Noise and stimulus features. Journal of Vision 2007;7(9):467. doi: 10.1167/7.9.467.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Perceptual learning improves position discrimination. Here, we ask whether the learning transfers across (1) positional noise levels and (2) feature orientations. Our stimulus was comprised of two horizontal line segments each made up of 5 discrete Gabor patches (10 cycle/deg). In some runs, binary “positional” noise was added to the right segment by jittering the Gabor patches in vertical direction. The observer's task was to localize the vertical position of the right segment relative to that of the left segment. The experiment was conducted in 5 phases. On day 1, the pre-training session consisted of four stimulus conditions: with or without positional noise, and horizontal or vertical “carrier” orientations. Observers practiced the alignment task with noise from day 2 to 6 with a vertical carrier, and subsequently from day 8 to 12 with a horizontal carrier. Performance for all four conditions was re-evaluated in the two post-training sessions on days 7 and 13. Each session consisted of 1000 responses (13-kilo trials in total). Practicing position discrimination with a vertical carrier in noise improved performance significantly (by 23%). The learning effect transferred completely to the untrained stimulus with zero noise (20% improvement) and partially to the untrained horizontal carrier orientation with noise (13% improvement). Practicing with the horizontal carrier in the second training phase resulted in a further significant improvement (17% - a total of 30% from day 1). Using an efficient reverse correlation technique (Li et al. 2004 & 2006), we reveal a dramatic perceptive field retuning in the first phase training, with no further changes in the second. Improvement in the second phase was due to a reduction in internal noise. Our findings show that learning generalizes across noise levels and feature orientations, however as evidenced by the phase 2 learning, there are two separate mechanisms for learning position discrimination.
This PDF is available to Subscribers Only