Serial reproduction reveals the geometry of visuospatial representations

Abstract

A primary function of human vision is to encode and recall spatial information about visual scenes. We developed an experimental paradigm that reveals the structure of human spatial memory priors in unprecedented detail. We ran a series of 85 large-scale online experiments with 9,202 participants that paint an intricate picture of these priors. Our results suggest a way to understand visuospatial representations as reflecting the efficient allocation of coding resources. In a radical departure from traditional theory, we introduce a model that reinterprets spatial memory priors as reflecting an optimal allocation of perceptual resources. We validate the predictions of the model experimentally by showing that perceptual biases are correlated with variations in discrimination accuracy.

Document Details

Document Type
Pub Defense Publication
Publication Date
Mar 26, 2021
Source ID
10.1073/pnas.2012938118

Entities

People

  • Jordan W. Suchow
  • Nori Jacoby
  • Thomas L. Griffiths
  • Thomas Langlois

Organizations

  • Columbia University
  • Max Planck Institute for Empirical Aesthetics
  • National Science Foundation
  • Princeton University
  • Stevens Institute of Technology

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Computational Modeling and Simulation
  • Vision Science/Vision Psychology/Cognitive Neuroscience.