ISSN 0439-755X
CN 11-1911/B

Acta Psychologica Sinica ›› 2026, Vol. 58 ›› Issue (11): 2167-2177.doi: 10.3724/SP.J.1041.2026.2167

    Next Articles

Feedback refines visual perception by reducing sensory noise and calibrating perception-action mapping

SUN Qi()   

  1. Zhejiang Key Laboratory of Intelligent Education Technology and Application; School of Psychology, Zhejiang Normal University, Jinhua 321004, China
  • Published:2026-11-25 Online:2026-09-11
  • Contact: SUN Qi, E-mail: sunqi_psy@zjnu.edu.cn

Abstract:

It has been proposed that systematic biases in visual perception can be consistent with a cascaded computational framework of “efficient encoding—Bayesian decoding—perceptual–response mapping.” However, this framework has primarily been developed for open-loop perception under no-feedback conditions. Feedback is a core component of perceptual learning, yet its computational locus within this cascade remains unclear. In the present study, we employed a 2 (feedback: present/absent, between-subjects) × 3 (response range: 80°, 160°, 240°, within-subjects) mixed experimental design, combined with seven alternative Bayesian observer models, to systematically examine how feedback modulates sensory noise ($\kappa $), prior integration weight ($w$), and the mapping scaling factor ($\alpha $). Behavioral results showed that feedback significantly attenuated overestimation bias and reduced estimation variability under wide response ranges. Model comparison identified the best-fitting model, whose parameters revealed that feedback significantly increased sensory encoding precision (larger $\kappa $) and decreased mapping gain (smaller $\alpha $), whereas prior integration weight ($w$) did not change significantly. These findings indicate that feedback optimizes visual perception through a dual mechanism—reducing sensory encoding noise and calibrating perceptual–response mapping gain—rather than by updating prior beliefs. At the computational level, this supports the separability of prior construction and feedback-driven modulation, and extends the open-loop perception framework to closed-loop learning scenarios.

Key words: Visual Perception, Bayesian Observer Model, Efficient Coding, Feedback