Time course of EEG complexity reflects attentional engagement during listening to speech in noise

Eqlimi, Ehsan; Bockstael, Annelies; Schönwiesner, Marc; Talsma, Durk; Botteldooren, Dick · 2023 · Crossref

DOI: 10.1101/2023.07.11.548528

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the temporal dynamics of attentional engagement and memory encoding during speech perception in noisy environments. Motivated by the challenge auditory noise poses to information encoding, the research aims to quantify neural activity associated with "learning from speech" using electroencephalography (EEG). Unlike previous studies focusing on speaker selection in cocktail-party scenarios, this work examines an ecologically valid setting where participants attend to speech amidst background noise. The authors hypothesize that EEG microstate properties and complexity metrics will differ significantly between conditions requiring focused speech attention versus those involving background noise attention or disregard. The experimental design involved 23 healthy adult participants who underwent three sequential listening tasks while 64-channel EEG data were recorded. In the first task (Lecture Attended, LA), participants listened to five-minute English lectures presented with various background noises (highway, traffic, babble) or pink noise, instructed to retain content for a subsequent exam. The second and third tasks involved listening to three-minute fragments of the same background noises alone, either with focused attention (Background Attended, BA) or intentional disregard (Background Unattended, BUA). EEG preprocessing included artifact removal and synchronization with auditory stimuli. The analysis focused on the initial three minutes of each task to ensure consistency. Researchers employed microstate analysis to identify seven canonical topographical classes and used Recurrence Quantification Analysis (RQA) to quantify the complexity of microstate transitions. Additionally, time-frequency analysis calculated the alpha-to-theta power ratio. Generalized Additive Mixed Modeling (GAMM) was utilized to model non-linear, time-varying changes in these metrics across conditions. The results revealed distinct neural signatures for speech-in-noise processing. Specifically, directing attention to speech resulted in increased complexity during microstate transitions and slower microstate recurrence compared to tasks without speech. A two-stage time course was observed for both microstate complexity and the alpha-to-theta power ratio. In early epochs, these metrics exhibited lower levels, gradually increasing to reach a steady state in later epochs. This temporal pattern suggests that the initial stage of processing is driven primarily by sensory processes and information gathering, while the subsequent stage involves higher-level cognitive engagement, including mnemonic binding and memory encoding. These findings provide evidence that EEG complexity and spectral power ratios serve as quantifiable indicators of attentional engagement and cognitive load during speech perception. The study highlights the brain's dynamic reorganization in response to task demands, distinguishing between sensory-driven early processing and cognitively driven later stages. This understanding contributes to the development of cognitively-controlled hearing solutions and improved communication interfaces, particularly for individuals in noisy environments or with hearing impairments. By linking specific neural dynamics to learning outcomes, the research offers a framework for assessing the quality of information encoding beyond traditional behavioral measures.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices