Perceptual learning of Vernier discrimination transfers from high to zero noise after double training

Abstract

Perceptual learning is often interpreted as learning of fine stimulus templates. However, we have proposed that perceptual learning is more than template learning, in that more abstract statistical rules may have been learned, so that learning can transfer to stimuli at different precisions. Here we provide new evidence to support this view: Perceptual learning of Vernier discrimination at high noise, which has thresholds approximately 10 times as much as those at zero noise, is initially non-transferrable to zero noise. However, additional exposure to a noise-free Vernier-forming Gabor, which is ineffective alone, not only maximizes zero-noise fine Vernier discrimination, but also further enhances high-noise Vernier performance. Such high-threshold coarse Vernier training cannot impact the fine stimulus template directly. One plausible explanation is that the observers have learned the statistical rules that can apply to standardized input distributions to improve discrimination, regardless of the original precision of these distributions.