September 2019
Volume 19, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2019
Generalization of learning in n-AFC orientation identification
Author Affiliations & Notes
  • Jiajuan Liu
    Department of Cognitive Sciences, University of California, Irvine
  • Barbara A. Dosher
    Department of Cognitive Sciences, University of California, Irvine
  • Zhong-Lin Lu
    Department of Psychology, The Ohio State University
Journal of Vision September 2019, Vol.19, 29a. doi:https://doi.org/10.1167/19.10.29a
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Jiajuan Liu, Barbara A. Dosher, Zhong-Lin Lu; Generalization of learning in n-AFC orientation identification. Journal of Vision 2019;19(10):29a. https://doi.org/10.1167/19.10.29a.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

A large classic literature examined specificity and transfer of perceptual learning, virtually always in two-alternative tasks. Recently, we demonstrated learning in n-alternative identification—a task perhaps more typical of real world visual judgments. However, little is known about specificity or generalization for nAFC tasks. In this study, we examined whether learning in 8-alternative orientation identification was specific to the spatial frequency of the training stimuli. Stimuli were Gabors of 8 orientations (−78° to 79.5° relative to vertical, in 22.5° steps) and five spatial frequencies (0.7 to 2.84 cpd, in half octave steps) embedded in Gaussian external noise (sigma=0.33). Observers identified which of the 8 orientations was displayed on each trial. Learning and transfer were evaluated by comparing performance in a post-test after training to a pre-test of orientation identification in all five spatial frequencies. Four different groups were trained for 5 sessions of 960 trials each either with the lowest, highest, middle, or a mixture of spatial frequencies at 3 Gabor contrasts (0.3 0.6, 1.0). In all four groups, training improved performance on the trained stimuli, including the group trained on the mixture of spatial frequencies (“roving” in the judgment-irrelevant spatial frequency dimension). Some of what was learned generalized to all spatial frequencies for all groups. At the same time, training with a single spatial frequency also showed some specificity, especially for the lowest and highest spatial frequencies, yielding a transfer function across spatial frequency. However, a simulation of the n-AFC integrated reweighting theory without spatial frequency invariant representation (IRT, Dosher et al., 2013; Dosher, Liu & Lu, VSS 2017) predicts more specificity than what was observed in the human data. Example simulations suggest a role for spatial-frequency invariant orientation representations that mediate transfer in a way analogous to the possible role of location-invariant representations in transfer over retinal locations.

Acknowledgement: National Eye Institute Grant # EY–17491 
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×