A Bayesian Reformulation of the Extended Drift-Diffusion Model in Perceptual Decision Making

Area of Science:

• Cognitive Neuroscience
• Computational Psychology
• Decision Science

Background:

• Perceptual decision making is often modeled as evidence accumulation to a bound using drift-diffusion models (DDMs).
• A recent Bayesian model offers a direct link between stimulus information and the decision process, contrasting with standard DDMs.

Purpose of the Study:

• To extend the Bayesian model of perceptual decision making by incorporating inter-trial parameter variability.
• To demonstrate the equivalence and utility of the extended Bayesian model (eBM) compared to the extended drift-diffusion model (eDDM).

Main Methods:

• Derivation of parameter distributions for the extended Bayesian model.
• Qualitative comparison of predictions between the eBM and eDDM.
• Application of Bayesian model selection to analyze behavioral data with trial-wise stimulus features.

Main Results:

• The extended Bayesian model (eBM) yields predictions qualitatively equivalent to the extended drift-diffusion model (eDDM).
• Bayesian model selection favored the eBM with inter-trial parameter variability, even with limited data (200 trials/condition).
• The eBM proved effective for analyzing behavioral data constrained by trial-specific stimulus features.

Conclusions:

• The extended Bayesian model provides a novel and promising framework for analyzing perceptual decision making experiments.
• The eBM successfully captures inter-trial parameter variability, offering advantages over traditional models, particularly with limited trial data.
• This approach enhances the understanding of decision-making processes by directly linking stimulus information and parameter variability.