Brain activity reflects the predictability of word sequences in listened continuous speech

Area of Science:

• Neuroscience
• Cognitive Science
• Linguistics

Background:

• Natural speech comprehension relies on predicting upcoming information based on context.
• Understanding the neural mechanisms of predictive processing in speech is crucial for cognitive neuroscience.

Purpose of the Study:

• To investigate the neural underpinnings of predictive processing in continuous natural speech.
• To differentiate the brain's response to acoustic speech features versus contextual word probabilities.

Main Methods:

• Magnetoencephalography (MEG) recorded brain activity in 10 healthy adults listening to an audiobook.
• MEG signals were correlated with acoustic speech envelope and Bayesian word probabilities (N-gram and Unigram).
• Analysis focused on time-lagged correlations to identify predictive temporal patterns.

Main Results:

• Auditory and sensorimotor cortices showed strong coupling to the speech envelope at syllable and prosody rates.
• Word sequence probability influenced brain signals (≤ 2 Hz) in auditory, rolandic, precuneus, occipital, and frontal regions, independent of acoustics.
• Contextual influences on speech prediction were significantly lateralized to the left hemisphere.

Conclusions:

• Brain activity provides a signature for predicting word sequences in continuous speech.
• Deeply-learned knowledge and recent context are dynamically employed for prediction.
• Speech prediction is a left-hemisphere dominant process, integrating acoustic and probabilistic information.