Perceptual biases are inconsistent with Bayesian encoding of speed in the human visual system

Area of Science:

• Neuroscience
• Visual Perception
• Computational Neuroscience

Background:

• Bayesian models are increasingly proposed for brain function and sensory processing.
• Previous research suggested Bayesian accounts for encoding object speed, contrasting with mechanistic models.

Purpose of the Study:

• To investigate competing models of human visual speed encoding.
• To assess the role of Bayesian processes versus temporal channel activity in speed perception.

Main Methods:

• Measured perceived speed of low-contrast patterns at low (2.5 cd m(-2)) and high (25 cd m(-2)) luminance.
• Analyzed perceptual biases and discrimination thresholds across different speeds and luminance levels.

Main Results:

• Perceptual bias varied with speed and luminance: underestimation at slow speeds, overestimation at fast speeds.
• The speed at which bias reversed differed between low and high luminance conditions.
• Discrimination thresholds remained similar across luminance levels.
• Findings align with temporal channel models but contradict Bayesian models.

Conclusions:

• Bayesian processes do not adequately explain speed encoding in the human visual system.
• Relative activity within broadly tuned temporal channels better predicts observed speed perception phenomena.
• Challenges the prevailing Bayesian framework for visual speed encoding.