Phase-Dependent Language Perception in Neural Oscillations: An Interdisciplinary Study Report

Background

During speech perception, the phase of neural oscillations plays a crucial role in the separation of neural representations and perceptual decisions. However, the specific phase-encoding mechanisms remain unclear. This study aims to reveal how phase encoding is based on the probability of interpreted speech events, particularly the probability of linguistic units. Additionally, the study investigates the region-specific characteristics of this phase encoding. These results suggest that neural oscillations separate linguistic neural representations based on the excitability of neuronal populations.

Source of the Paper

This paper was completed by Sanne Ten Oever, Lorenzo Titone, Noémie Te Rietmolen, and Andrea E. Martin, primarily affiliated with the Max Planck Institute for Psycholinguistics, Donders Centre for Cognitive Neuroimaging, Radboud University, Maastricht University, and the Max Planck Institute for Human Cognitive and Brain Sciences. The paper was published in the 2024 edition of the “Proceedings of the National Academy of Sciences” (PNAS), edited by Richard Aslin of Haskins Laboratories Inc.

Research Process

Subjects and Methods

The study is divided into three parts: psychophysical experiments, magnetoencephalography (MEG) experiments, and computational modeling, involving the following steps:

1. Psychophysical Experiments:

   • Subjects were native Dutch speakers.
   • Four Dutch words were used in the study: “dat” (that), “gat” (hole), “gaat” (go), and “daad” (deed), differing in consonants, vowels, and word frequency.
   • The corresponding IPA (International Phonetic Alphabet) transcriptions are: /xαt/, /dαt/, /xat/, /dat/.

2. Magnetoencephalography (MEG) Experiments:

   • Brain wave activities of subjects were recorded using MEG technology.
   • Subjects were randomly presented with ambiguous words during the experiment, and their brain’s oscillatory phase related to the presented frequency was recorded.
   • Primary experimental areas included the superior temporal gyrus (STG) and middle temporal gyrus (MTG).

3. Computational Modeling:

   • A computational model incorporating speech tracking and content representation was constructed to simulate the results of the psychophysical and MEG experiments.
   • The model included two analytical levels: the consonant and the lexical levels, testing performance in different scenarios by changing the sensitivity of neuronal populations.

Research Results

1. Psychophysical Experiment Results:

   • In response to high-frequency words (such as /dαt/ [cvw] and /γat/ [cvw]), participants reacted more during the high-excitability phase.
   • Specific frequency-phase relationships include: when fitting experimental data with a 6.25Hz sine oscillation, a significant fitting curve was found.

2. Magnetoencephalography (MEG) Experiment Results:

   • The oscillatory phase of the superior temporal gyrus (STG) determined the perception of consonant frequency, while the oscillatory phase of the middle temporal gyrus (MTG) determined the perception of lexical frequency, indicating a double dissociation effect.
   • The average phase difference between different response options and their related words was 0.93π (low-frequency consonants), 0.87π (low-frequency consonants), and 0.002π (high-frequency consonants).

3. Computational Modeling Results:

   • The computational model validated the results of the psychophysical and MEG experiments. High-frequency words or consonants in the model also showed stronger phase dependence.
   • The phase differences generated by the model were consistent with experimental results, but broader validation indicated that phase encoding might involve not only lower frequency bands but also higher frequency bands (like the alpha frequency around 10Hz).

Conclusion

This study elucidates that phase encoding in neural oscillations is based on the probability of linguistic units. Using different frequency units and region-specific characteristics, the study demonstrates how phase provides perceptual category information in language comprehension. This not only helps explain cognitive mechanisms in speech perception but also provides directions for further investigation.

Research Highlights

1. Important Findings: The study demonstrates how the phase of neural oscillations encodes language unit probabilities, thereby affecting word perception.
2. Uniqueness: The double dissociation effect shows the region-specific characteristics of the middle temporal gyrus (MTG) and the superior temporal gyrus (STG).
3. New Methods: Combining psychophysical experiments and computational modeling offers a new perspective, demonstrating the importance of phase encoding in neural activity.

Research Significance

This study enriches our understanding of speech perception and phase encoding in neural oscillations, promoting interdisciplinary research in cognitive neuroscience and linguistics by revealing phase dependence in complex language processing. Future research can further explore the application of phase encoding in dynamic contexts and settings to uncover broader cognitive mechanisms.