Weight illusions explained by efficient coding based on correlated natural statistics

Area of Science:

• Cognitive Neuroscience
• Perception Science
• Computational Neuroscience

Background:

• Human perception often relies on integrating visual and haptic information.
• The Size-Weight Illusion (SWI) contradicts naive Bayesian predictions, where smaller objects of equal weight are perceived as heavier.
• Existing models struggle to fully explain the SWI and related illusions.

Purpose of the Study:

• To provide a normative explanation for the Size-Weight Illusion (SWI) based on efficient coding principles.
• To demonstrate how the joint distribution of object properties in the environment influences perceptual fidelity.
• To quantitatively predict the SWI using Bayesian estimation informed by efficient encoding.

Main Methods:

• Developed an efficient coding model based on the natural statistics of object properties (mass and volume).
• Analyzed human observers' weight estimation precision as a function of object mass and volume.
• Used Bayesian estimation, incorporating efficient encoding's discriminability gradients, to predict perceptual biases.

Main Results:

• Perceptual precision in weight estimation aligns with the natural joint distribution of object mass and volume.
• The SWI is quantitatively predicted by Bayesian inference under efficient encoding constraints.
• The Material-Weight Illusion (MWI) is also explained by this framework, considering visual cues about density.

Conclusions:

• Efficient coding provides a unifying framework for understanding the SWI and MWI.
• Perceptual biases arise from the brain's optimal encoding of statistically prevalent environmental properties.
• This model generalizes to other sensory modalities and perceptual phenomena influenced by natural correlations.