A Bayesian model of distance perception from ocular convergence.

Area of Science:

• Visual neuroscience
• Perception science
• Computational modeling

Background:

• Ocular convergence is a key cue for absolute distance perception.
• Humans systematically underestimate object distances, especially for farther objects.
• This distance underestimation challenges the reliability of vergence as a cue.

Purpose of the Study:

• To explain the systematic underestimation of distance despite accurate ocular convergence.
• To propose a probabilistic geometric model for distance perception.
• To investigate the role of noise in the vergence signal.

Main Methods:

• Developed a probabilistic geometric model.
• Modeled the visual system's estimation of the most likely distance based on noisy vergence signals.
• Formulated a likelihood function considering the generative relationship between distance and convergence.

Main Results:

• The model successfully explains human distance underestimation.
• The required noise level in the vergence signal aligns with experimental measurements.
• The formulation of the likelihood function is critical for explaining the observed bias.

Conclusions:

• Distance underestimation is a consequence of the visual system's probabilistic inference from noisy sensory data.
• Vergence remains a viable, albeit noisy, cue for absolute distance perception.
• The study highlights the importance of generative models in understanding perceptual biases.