Cortical Tracking of Surprisal during Continuous Speech Comprehension

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Linguistics

Background:

• Speech comprehension involves processing continuous acoustic signals for meaning.
• Previous research examined neural responses to word predictability and precision in single sentences.
• Naturalistic speech involves continuous variations in word predictability and precision.

Purpose of the Study:

• To investigate how cortical activity tracks word surprisal in continuous natural speech.
• To determine if this neural tracking is modulated by the precision of word predictions.
• To explore the neurobiological underpinnings of natural speech comprehension.

Main Methods:

• Quantified word surprisal and precision in naturalistic speech using a deep neural network.
• Recorded electroencephalography (EEG) responses from human volunteers during auditory story comprehension.
• Analyzed the relationship between quantified speech features and cortical activity across different frequency bands.

Main Results:

• Significant cortical tracking of word surprisal was observed in low frequencies (delta band) and high frequencies (beta and gamma bands).
• The neural tracking of surprisal was modulated by the precision of word predictions.
• Demonstrated frequency-specific neural correlates of surprisal and precision in continuous speech.

Conclusions:

• Cortical activity dynamically tracks linguistic surprisal during natural speech comprehension.
• Prediction precision plays a crucial role in modulating neural responses to surprisal.
• These findings provide a foundation for further research into the neurobiology of real-time speech processing.