Decoding Speech Perception utilizing EEG-Based Partial Information Decomposition and Hierarchical Drift Diffusion Modeling

Abstract

This study investigates the neural and cognitive mechanisms underlying speech perception and decision-making under varying task difficulties. By combining Partial Information Decomposition (PID) and Hierarchical Drift Diffusion Modeling (HDDM), we explored how unique, synergistic, and redundant information encoded in neural activity influences key decision-making parameters such as drift rate, decision boundary, and non-decision time. EEG data from 22 participants were analyzed during a speech perception task involving rhyming and non-rhyming stimuli, categorized into easy and hard conditions. PID analysis revealed task-specific differences in neural encoding across Delta (4 Hz), Theta (7 Hz), and Broadband (10 Hz) frequency bands. Hard tasks elicited higher cognitive load, with increased unique and synergistic information in Delta and Theta bands, especially in fronto-central and parietal regions. Theta waves played a prominent role in easier tasks, facilitating efficient processing of unique and synergistic information. HDDM analysis showed that drift rate increased with higher unique or synergistic neural information, indicating faster evidence accumulation in congruent trials. Non-decision times were shorter during trials with high synergy, reflecting faster sensory encoding, while incongruent trials with higher redundancy led to longer non-decision times. Decision boundary parameters remained consistent across conditions, suggesting stable decision thresholds. Additionally, choice switching significantly impacted reaction times and non-decision times, with faster responses and lower non-decision times observed during correct switches, emphasizing the neural basis of adaptive behavior. These findings highlight the dynamic interplay between neural encoding, decision-making, and cognitive load. By integrating PID and HDDM, this study provides novel insights into how the brain processes sensory and motor information to adapt decision-making strategies during speech perception. Keywords: Speech Perception, Decision-Making, Neural Encoding, Partial Information Decomposition (PID), Hierarchical Drift Diffusion Modeling (HDDM), Cognitive Load, Unique Information, Synergistic Information, Redundant Information, Evidence Accumulation