Compatibility of a Competition Model for Explaining Eye Fixation Durations During Free Viewing.

Area of Science:

• Computational neuroscience
• Cognitive neuroscience
• Vision science

Background:

• Eye fixation durations (EFDs) are typically around 250 ms, but simple statistics like mean and standard deviation fail to capture their distribution.
• The exGaussian distribution has been used to model EFD histograms, providing a better fit than basic statistical measures.

Purpose of the Study:

• To introduce and evaluate a novel competition model (C model) for explaining EFD histograms.
• To compare the explanatory power of the C model against the established exGaussian model for EFD distributions.

Main Methods:

• Adjusted both the C model and the exGaussian model to EFD data from a large open database (179,473 eye fixations).
• The C model's parameters were interpreted computationally: parameter A relates to saccade refractory periods (sigmoid equation), and parameter ps relates to network competition (geometric probability density function).

Main Results:

• The C model, alongside the exGaussian model, effectively explains EFD distributions.
• The C model's parameters provide insights into the neural mechanisms underlying saccade generation and fixation.
• The model's dynamics explain visual scene scanning behavior, including initial foveal stability and subsequent broad exploration.

Conclusions:

• Competition between neural networks is a fundamental organizational principle facilitating decision-making in perception and action.
• The C model offers a conceptual framework for understanding the interplay between saccadic and fixation networks in visual processing.
• The study provides computational insights into how the brain manages visual exploration and information processing during scene viewing.