Sequential sampling models with variable boundaries and non-normal noise: A comparison of six models.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Decision Science

Background:

• The diffusion model, a prominent response-time model, assumes decision-making involves continuous evidence accumulation via a Wiener diffusion process.
• Standard diffusion model analyses rely on specific assumptions about decision thresholds and the nature of evidence accumulation.
• Investigating these assumptions is crucial for refining models of human decision-making.

Purpose of the Study:

• To examine if decision thresholds adjust during decision-making.
• To compare the explanatory power of diffusion models against Lévy-flight models with random jumps.
• To assess accumulator models' fit to empirical data under varying task difficulties and response constraints.

Main Methods:

• Compared six accumulator models, including diffusion and Lévy-flight variants.
• Analyzed data from a number-letter classification task across four conditions (speed vs. accuracy, simple vs. multiple stimuli).
• Evaluated model fits to response times and accuracy.

Main Results:

• Found minimal evidence for collapsing decision boundaries across experimental conditions.
• Demonstrated that Lévy-flight models, incorporating heavy-tailed noise distributions, provided a superior fit to the experimental data compared to the standard Wiener diffusion model.
• Identified potential limitations of the standard diffusion model in capturing complex decision dynamics.

Conclusions:

• Decision thresholds appear relatively stable during evidence accumulation.
• Lévy-flight models offer a more accurate account of decision-making processes, particularly when incorporating random jumps.
• Findings suggest a need to reconsider the standard diffusion model's assumptions in favor of models that allow for more complex noise structures.