Introduction

While the prominence of brain alpha oscillations continues, ongoing advances diversify our understanding of their functional roles, moving away from a unified perspective. This diversity is particularly pronounced in auditory contexts. Auditory alpha’s existence, EEG measurability, and its potential dependence on visual alpha collectively pose three layers of uncertainty. Moreover, an alternative viewpoint transcends sensory distinctions, introducing a dimension that extends beyond the confines of specific modalities. In 1999, a pivotal finding [3] emerges: upper alpha selectively synchronizes during maximal episodic short-term memory demands, challenging the conventional expectation of alpha desynchronization during mental activity at that time. One perspective on interpreting this type of finding involves inhibitory processes, confirmed through controlled experiments such as intracranial EEG and animal studies (e.g., [2]). Alpha oscillations act as a neural ’brake,’ increasing power coincides with reduced neuronal firing rate. The interpretation and evidence are grounded in segregation, suggesting that when a brain region is deemed task-irrelevant, the brain exhibits an increase in alpha activity, signaling subsequent inhibition in that region (sim- ply, alpha synchronization causes net inhibition). However, whether this relationship is a cause or consequence remains a subject of debate [4].

Alpha oscillation research has also focused on distinguishing lower and upper frequencies—crucial for identifying individual alpha peaks. This focus aids in recognizing their manifestation in diverse brain regions during cognitive processes. Moreover, numerous studies emphasize alpha oscillations’ role in parsing sensory information into (discrete) events, especially for visual stimuli (references in [5]). Therefore, alongside power, both frequency and phase are crucial in tasks including those requiring perceptual parsing and temporal binding. Furthermore, in alpha oscillations EEG research, the emphasis is on parsing the brain’s response to repeated brief stimuli into distinct components, including phase-locked (evoked), non- phase-locked (induced), and ongoing characteristics. This provides an understanding of short-term information processing, yet a gap remains in comprehending more prolonged behaviors, including the brain’s engagement in achieving specific goals with continuous and complex information. To facilitate further exploration and address this gap, in this study, conducted with HD-EEG, we investigate alpha oscillation modulations during attentive listening to 5-minute speeches with and without multitalker background noise. We also explore alpha oscillations’ predictive role in memory and retention performance.

Methods

The analyzed data are part of our extensive study on neural markers for attentive/inattentive listening in noisy environments using single-trial EEG [1]. Involving 23 English-speaking listeners, the study comprised four 5-minute speeches without background noise and three speeches with multitalker background noise, covering various topics. Simultaneously, the 64-channel EEG data were continuously recorded during attentive listening. All seven speeches addressed different facets of Belgian culture, with participants having minimal prior knowledge. After an intermediate interval (45 minutes), participants underwent a written exam, which assessed their scores on all seven speeches presented. After preprocessing, three features within alpha oscillations were extracted for each EEG fragment related to each of the seven speeches: alpha peak frequency, alpha peak power, and the long-range temporal correlations (LRTC) of alpha. To examine the impact of multitalker noise on alpha oscillations and assess the linear relationship with exam scores, we utilized linear mixed-effect modeling, accounting for variations in participant topics. Additionally, employing time-frequency analysis, we extracted the time-varying alpha oscillations over the 5-minute speeches. We used a generalized additive mixed model (GAMM) to compare the time course across different speech topics and assess the impact of background noise. Finally, we employed nonparametric cluster-based permutation testing across channels to assess channel influence on alpha manifestation.

Results

Although the behavioral analysis reveals a pronounced decline in exam scores when background noise is present compared to the absence of background noise, there was no significant difference in alpha peak power and frequency between conditions. However, LRTC of alpha oscillations exhibited a significant increase in speech with multitalker noise. Interestingly, this increase was accompanied by a negative correlation between alpha LRTC and exam scores. Moreover, we observed increased synchronization activation of alpha peak frequency in temporal lobe regions, contrasting with higher alpha peak power activity predominantly in occipital lobe regions and comparatively less in temporal versus central and frontal areas. This suggests modulation in both task-relevant and irrelevant cortical areas. Although none of these variations reached significance when comparing speech alone versus speech with noise using nonparametric tests, LRTC of alpha oscillations revealed significant changes. Specifically, for speech with additional background noise, LRTC increased in occipital and temporal regions and decreased in the frontal region. In theory, increased LRTC in occipital and temporal regions could facilitate suppression of task-irrelevant background audio and visual input. Conversely, decreased LRTC in frontal executive control areas could facilitate task-vigilance. Finally, when considering the average across channels, there was an overall increase in LRTC in the noisy condition. Analyzing the modulated alpha changes over time, measured by LRTC, inspired us to examine the evolving alpha power pattern between conditions [7]. Although no significant difference was found in average alpha power, temporal progression analysis using GAMM uncovered a significant increase in noisy conditions in the second part of the trend (after the first minute) [6].

Conclusion and Discussion

The presented results support the role of alpha oscillations in top-down mechanisms when attending and listening to speech. Although the mean alpha peak power and frequency roles have been demonstrated in numerous studies on selective auditory attention and working memory tasks, in our study, we did not observe any influence. Instead, we find that the temporal dynamics of alpha are much more critical, especially in the presence of a distractor in the background, leading to an increase in the alpha LRTC. Increased alpha LRTC, signifying persistent temporal patterns in neural activity, reflects heightened cognitive effort for tasks like focused attention. Simultaneously, it may indicate the quality of speech information encoding, as suggested by the inverse relationship between alpha and behavioral exam scores on information retention. The two-stage trend of alpha temporal dynamics exhibits a high slope in the first minute, stabilizing afterward. In the second part, there is an increase in alpha, particularly when multitalker noise is present. Our interpretation is that the first stage involves sensory processes and information gathering, while the second stage reflects increased effort for higher-level cognitive engagement, including mnemonic binding and memory encoding.

References

1 E. Eqlimi. Exploring neural markers modulated by learning from speech in environmental noise using single-trial EEG. PhD thesis, Ghent University, 2022.

2 S. Haegens, V. N ́acher, R. Luna, R. Romo, and O. Jensen. α-oscillations in the monkey sen- sorimotor network influence discrimination performance by rhythmical inhibition of neuronal spiking. Proceedings of the National Academy of Sciences, 108(48):19377–19382, 2011.

3 W. Klimesch, M. Doppelmayr, J. Schwaiger, P. Auinger, and T. Winkler. Paradoxical’alpha synchronization in a memory task. Cognitive Brain Research, 7(4):493–501, 1999.

4 C. Peylo, Y. Hilla, and P. Sauseng. Cause or consequence? alpha oscillations in visuospatial attention. Trends in Neurosciences, 44(9):705–713, 2021.

5 R. VanRullen. Perceptual cycles. Trends in cognitive sciences, 20(10):723–735, 2016.

6 E. Eqlimi et al. Time course of EEG complexity reflects attentional engagement during listening to speech in noise. European Journal of Neuroscience 58.9 (2023): 4043-4069.

7 E. Eqlimi et al. EEG correlates of learning from speech presented in environmental noise. Frontiers in Psychology 11 (2020): 542295.