August 2023
Volume 23, Issue 9
Open Access
Vision Sciences Society Annual Meeting Abstract  |   August 2023
Visual perceptual learning of natural and Portilla & Simoncelli images occurs in a significantly different manner than visual perceptual learning of unnatural images
Author Affiliations & Notes
  • Kazuhisa Shibata
  • Daiki Ogawa
    Nagoya University
  • Yuka Sasaki
    Brown University
  • Takeo Watanabe
    Brown University
  • Footnotes
    Acknowledgements  This work is supported by JSPS KAKENHI Grant Number 19H01041, 20H05715 (to KS), JST Moonshot R&D JPMJMS2013 (to KS), NIH R01EY027841, R01EY019466 (to TW), R01 EY031705 (to YS), and United States - Israel Binational Science Foundation BSF2016058 (to TW).
Journal of Vision August 2023, Vol.23, 5071. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Kazuhisa Shibata, Daiki Ogawa, Yuka Sasaki, Takeo Watanabe; Visual perceptual learning of natural and Portilla & Simoncelli images occurs in a significantly different manner than visual perceptual learning of unnatural images. Journal of Vision 2023;23(9):5071.

      Download citation file:

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

  • Supplements

Task-irrelevant perceptual learning (TIPL) of orientation in a Gabor patch occurs when it is subthreshold. Results of several studies consistently indicated that it is because subthreshold signals are not subject to attentional suppression. However, recently we found that TIPL of suprathreshold natural images or Portilla & Simoncelli (PS) images consisting of lower- and higher-order statistics from natural images occurred, whereas TIPL of suprathreshold images only consisting of lower-order statistics did not occur. These results raise the possibility that the orientation information in higher order statistics of natural images is first implicit and is not subject to attentional suppression but later is reconstructed to be explicit, leading to TIPL of PS and natural images. Here, to partially test this possibility, we examined whether the orientation of natural images is retrieved only from the higher-order statistics in natural images. First, we measured performances of orientation discrimination of each of three types of images (natural, PS, and lower-order statistics) with a visual mask that disrupts lower-order statistics signals, with the SOA from an image to the mask varied between 40ms and 180ms. We found that the mask did not disrupt the performance on the natural and PS images, whereas the performance on the lower-order statistics images was nearly at a chance level with shorter SOAs. Next, we examined whether the orientation information can be extracted only from higher-order statistics in images. We combined broad bands orientation power with higher-order statistics in natural images to create new stimuli with no dominant orientation power. While the Fourier analysis showed the new images had no peak orientation that matched the natural images, participants still responded accordingly to the orientation in the natural images, with significantly higher accuracy than a chance level. These results demonstrate that higher-order visual statistics restore orientation information and support the above possibility.


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.