Purchase this article with an account.
Jason M Gold; Dynamic classification images reveal the effects of perceptual learning in a hyperacuity task. Journal of Vision 2003;3(9):162. doi: 10.1167/3.9.162.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: Performance in hyperacuity tasks often improves with practice. One possible mechanism for this effect is an improvement in the spatial tuning of observers' templates. Here, I use response classification to measure trial-by-trial changes in observers' templates with practice in a vernier acuity task. Methods: 3 observers discriminated between 2 vertical line segments in which the entire top half was shifted by 1 pixel (∼20′) to the left or right. On each trial, a vernier stimulus was chosen randomly and presented in high contrast Gaussian white noise. Each observer participated in 1000 trials/day over the course of 10 days. Contrast thresholds were measured during each session with a staircase that maintained ∼71% correct performance throughout the session. Results & Conclusions: Practice reduced thresholds by a factor of ∼1.5 over the course of the experiment. A series of N classification images was computed using the noise shown on trials N through N+2,000 (where N ranged from 1 to 8,000). The result was a classification movie that dynamically revealed the changes that took place in an observer's linear template over time. When viewed in rapid succession, the series of images showed a stable template slowly emerging from a background of noise. This effect was quantified by cross-correlating each human classification image with the classification image for a model observer that a) assumed both the bottom and top halves of the vernier stimuli were shifting by 1 pixel in opposite directions; and b) was subject to a modest amount of spatial uncertainty. The correlation between the human and model classification images increased systematically by a factor of ∼1.5 over the course of the experiment. I am currently using double-pass response consistency and spatial jitter to measure any reductions in multiplicative internal noise and/or spatial uncertainty that may have contributed to the changes that took place in the classification images over time.
This PDF is available to Subscribers Only