Pursuing motion illusions: a realistic oculomotor framework for Bayesian inference.

Area of Science:

• Computational Neuroscience
• Visual Perception
• Oculomotor Systems

Background:

• Estimating object motion is challenged by visual noise and the aperture problem.
• Early visual processing relies on 1D edge cues, while later stages integrate 2D terminator cues for accurate global motion perception.
• Understanding the dynamic integration of these cues is crucial for explaining smooth pursuit eye movements.

Purpose of the Study:

• To propose a computational model for the dynamic integration of 1D and 2D motion information.
• To describe this integration within the context of smooth pursuit eye movements using a Bayesian framework and a model oculomotor plant.
• To quantitatively account for human smooth pursuit responses to visual motion stimuli.

Main Methods:

• Developed a recursive Bayesian framework incorporating edge-related (1D) and terminator-related (2D) motion likelihoods.
• Cascaded the Bayesian framework with a model oculomotor plant to simulate eye velocity responses.
• Tuned the model using human smooth pursuit data for 1D and 2D motion stimuli, stimulus energy, and directional anisotropies.

Main Results:

• The recurrent Bayesian model accurately captures the dynamic transition from 1D to 2D motion cue influence.
• The model's simulated eye velocity responses closely match human smooth pursuit data across varying contrasts and speeds.
• The oculomotor plant component successfully incorporated stimulus energy and directional anisotropy effects.

Conclusions:

• The proposed model provides a robust framework for understanding the dynamic integration of visual motion cues.
• This computational approach successfully explains the interplay between local and global motion information in smooth pursuit.
• The findings offer insights into the neural mechanisms underlying motion perception and eye movement control.