Bayesian Semiparametric Longitudinal Drift-Diffusion Mixed Models for Tone Learning in Adults.

Area of Science:

• Cognitive Neuroscience
• Computational Linguistics
• Psycholinguistics

Background:

• Adults learning nonnative speech tones offers insights into experience-dependent brain plasticity.
• Traditional studies use longitudinal experiments and multi-category decision-making paradigms.
• Drift-diffusion models are frequently used to simulate neural mechanisms in decision-making.

Purpose of the Study:

• Develop a novel Bayesian semiparametric inverse Gaussian drift-diffusion mixed model.
• Adapt the model for multi-alternative decision-making in longitudinal settings.
• Analyze how learning affects biologically interpretable model parameters.

Main Methods:

• Developed a novel Bayesian semiparametric inverse Gaussian drift-diffusion mixed model.
• Designed a Markov chain Monte Carlo algorithm for posterior computation.
• Evaluated performance using synthetic experiments and a longitudinal tone learning study.

Main Results:

• The novel model successfully captures longitudinal tone learning dynamics.
• Biologically interpretable parameters evolved with learning and varied by performance.
• Differences were observed in tone learning across input-response combinations.

Conclusions:

• The developed model provides new insights into adult nonnative speech learning.
• It elucidates the mechanisms of experience-dependent brain plasticity in auditory learning.
• The method offers a robust framework for analyzing longitudinal decision-making data.