A test of Bayesian observer models of processing in the Eriksen flanker task

Area of Science:

• Cognitive psychology
• Computational neuroscience
• Visual perception

Background:

• The Eriksen flanker task is a common paradigm to study selective attention and interference.
• Two Bayesian observer models (perceptual bias and spatial uncertainty) were proposed to explain flanker task data, particularly the accuracy dip in fast responses.

Purpose of the Study:

• To evaluate the adequacy of existing Bayesian observer models in explaining the full behavioral data of the Eriksen flanker task.
• To contrast these models with a spotlight-diffusion model and a new hybrid model.
• To identify the source of shortcomings in the proposed Bayesian observer models.

Main Methods:

• Fitting two Bayesian observer models (perceptual bias, spatial uncertainty) to empirical data from the Eriksen flanker task.
• Comparing the fit of these models to a spotlight-diffusion model and a novel hybrid model.
• Analyzing the assumptions regarding visual processing in each model.

Main Results:

• Neither of the proposed Bayesian observer models adequately accounted for the complete set of behavioral data from the flanker task.
• A spotlight-diffusion model provided a superior fit to the empirical data compared to the Bayesian observer models.
• The limitations of the Bayesian observer models were attributed to their assumptions about visual processing, not their Bayesian decision framework.

Conclusions:

• The proposed Bayesian observer models do not fully capture the mechanisms underlying performance in the Eriksen flanker task.
• The spotlight-diffusion model offers a more parsimonious and accurate explanation of flanker task behavior.
• Future research should focus on refining models of visual processing rather than solely on Bayesian decision-making principles for explaining attentional interference.