Speech Rhythms and Multiplexed Oscillatory Sensory Coding in the Human Brain

Joachim Groß, Nienke Hoogenboom, Gregor Thut, Philippe G. Schyns, Stefano Panzeri, Pascal Belin, Simon GarrodPLoS 2013 · DOI

Abstract

Cortical oscillations are likely candidates for segmentation and coding of continuous speech. Here, we monitored continuous speech processing with magnetoencephalography (MEG) to unravel the principles of speech segmentation and coding. We demonstrate that speech entrains the phase of low-frequency (delta, theta) and the amplitude of high-frequency (gamma) oscillations in the auditory cortex. Phase entrainment is stronger in the right and amplitude entrainment is stronger in the left auditory cortex. Furthermore, edges in the speech envelope phase reset auditory cortex oscillations thereby enhancing their entrainment to speech. This mechanism adapts to the changing physical features of the speech envelope and enables efficient, stimulus-specific speech sampling. Finally, we show that within the auditory cortex, coupling between delta, theta, and gamma oscillations increases following speech edges. Importantly, all couplings (i.e., brain-speech and also within the cortex) attenuate for backward-presented speech, suggesting top-down control. We conclude that segmentation and coding of speech relies on a nested hierarchy of entrained cortical oscillations.

Introduction

Speech is a complex, continuous acoustic stream, yet the human brain effortlessly segments it into discrete units like syllables and words. Neuroscientists have long hypothesized that cortical oscillations—rhythmic patterns of neural activity—serve as the "clocks" for this segmentation. However, the exact mechanisms of how different brain frequencies interact to handle varying scales of speech (e.g., slow sentences versus fast phonemes) remained unclear. This study investigates how the brain utilizes multiplexed oscillatory coding to track speech rhythms and whether hemispheric specialization plays a role in this process.

Methodology

The researchers recorded brain activity from 22 healthy participants using Magnetoencephalography (MEG) while they listened to a 7-minute continuous story. To distinguish between passive acoustic processing and active linguistic understanding, a "backwards" version of the story was used as a control.

The technical core of the analysis involved Mutual Information (MI), which captures both linear and non-linear dependencies between the speech signal and neural activity. The analysis pipeline included extracting the speech amplitude envelope, bandpass filtering both speech and MEG signals into specific frequency bands, and applying the Hilbert transform to determine phase and amplitude.

Overview of the Mutual Information analysis pipeline used to relate speech signals to MEG brain activity.

Key Insights and Results

The study revealed a sophisticated "multiplexing" strategy where the brain tracks speech at multiple temporal scales simultaneously.

A Nested Hierarchy of Oscillations

The brain employs a hierarchy of frequencies: Delta (1–3 Hz) tracks prosody and phrasing, Theta (3–7 Hz) tracks syllables, and Gamma (30–90 Hz) tracks phonemes. These processes occur in parallel, providing a multi-layered representation of the acoustic input. Notably, Gamma amplitude tracking adds approximately 23% more information to the signal already captured by theta phase tracking, confirming that the brain utilizes both channels for decoding.

Hemispheric Specialization

A clear "division of labor" was observed between the hemispheres. The right auditory cortex showed stronger entrainment to the phase (timing) of slow rhythms (Delta and Theta). In contrast, the left auditory cortex was more sensitive to the amplitude (intensity) of fast rhythms (Gamma).

Brain maps showing the spatial distribution of mutual information for different frequency bands (Delta, Theta) and cross-frequency coupling (Theta-gamma).

The Role of Acoustic "Edges"

Sudden increases in speech volume, or acoustic "edges," act as a reset mechanism. These edges trigger a "phase reset" in cortical oscillations, re-aligning the brain's internal clock with the speaker. This ensures optimal processing of the subsequent "chunk" of information. Furthermore, coupling between Delta, Theta, and Gamma oscillations increases following these edges.

Top-Down Influence

Tracking and coupling were significantly attenuated for backward-played speech. This suggests that speech entrainment is not merely a passive physical reaction to sound but is modulated by top-down processes related to intelligibility and the brain's attempt to extract meaning.

Discussion and Future Directions

While the use of backward speech highlights the importance of intelligibility, it also introduces unnatural acoustic transitions. Future research could employ vocoded speech or foreign languages to better isolate linguistic components from acoustic features. Additionally, exploring how these hierarchical couplings adapt to varying speaking rates or noisy "cocktail party" environments remains an important next step.

This hierarchical model offers a promising framework for studying language disorders. For instance, in conditions like dyslexia, the "sampling" of speech at specific frequencies (such as the phonemic Gamma level) may be impaired, and this methodology could help pinpoint the underlying neural deficits.

Summary updated

Abstract

Cortical oscillations are likely candidates for segmentation and coding of continuous speech. Here, we monitored continuous speech processing with magnetoencephalography (MEG) to unravel the principles of speech segmentation and coding. We demonstrate that speech entrains the phase of low-frequency (delta, theta) and the amplitude of high-frequency (gamma) oscillations in the auditory cortex. Phase entrainment is stronger in the right and amplitude entrainment is stronger in the left auditory cortex. Furthermore, edges in the speech envelope phase reset auditory cortex oscillations thereby enhancing their entrainment to speech. This mechanism adapts to the changing physical features of the speech envelope and enables efficient, stimulus-specific speech sampling. Finally, we show that within the auditory cortex, coupling between delta, theta, and gamma oscillations increases following speech edges. Importantly, all couplings (i.e., brain-speech and also within the cortex) attenuate for backward-presented speech, suggesting top-down control. We conclude that segmentation and coding of speech relies on a nested hierarchy of entrained cortical oscillations.
BibTeX
@inproceedings{gro2013speech,
  author = {Joachim Groß and Nienke Hoogenboom and Gregor Thut and Philippe G. Schyns and Stefano Panzeri and Pascal Belin and Simon Garrod},
  title = {{Speech Rhythms and Multiplexed Oscillatory Sensory Coding in the Human Brain}},
  booktitle = {PLoS},
  year = {2013},
  doi = {10.1371/journal.pbio.1001752},
}
Figure