Dynamical analysis of Bayesian inference models for the Eriksen task.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Psychology

Background:

• The Eriksen task is a key paradigm for studying selective attention and response control under conflicting sensory input.
• Previous models include neural networks and normative Bayesian accounts.
• Understanding the dynamics of these models is crucial for explaining cognitive processes.

Purpose of the Study:

• To analyze the dynamics of Bayesian models of the Eriksen task.
• To simplify complex nonlinear models into linear, decoupled systems for analytical solutions.
• To connect Bayesian updating with drift-diffusion processes.

Main Methods:

• Analysis of simplified, linear, decoupled dynamical systems derived from nonlinear Bayesian models.
• Derivation of analytical solutions for posterior probabilities and psychometric functions.
• Comparison with numerical simulations and experimental data.
• Investigation of continuum limits to link with drift-diffusion models.

Main Results:

• Analytical solutions were obtained for simplified models, describing parameter dependencies.
• The simplified models showed good agreement with original models and experimental data.
• Bayesian updating was shown to be closely related to drift-diffusion processes in continuum limits.

Conclusions:

• Simplified dynamical systems provide valuable insights into complex Bayesian models of attention.
• The connection between Bayesian updating and drift-diffusion processes offers a bridge between normative and implementational levels of analysis.
• This work enhances understanding of how neural systems perform Bayesian computations for cognitive tasks.