Complex Mapping between Neural Response Frequency and Linguistic Units in Natural Speech

Area of Science:

• Neuroscience
• Linguistics
• Signal Processing

Background:

• The brain decodes linguistic units (syllables, words, sentences) from connected speech.
• A hypothesis suggests cortical activity scales match linguistic unit timescales.
• Previous studies used frequency tagging; recent work examines natural speech envelope tracking.

Purpose of the Study:

• To investigate if neural responses to different linguistic units (phones, syllables, words, sentences) can be separated in natural speech.
• To analyze the coherence between speech envelope and idealized neural responses tracking single linguistic unit levels.
• To test the assumption that distinct frequency bands in neural tracking of speech envelope correspond to different linguistic units.

Main Methods:

• Analyzed coherence between the speech envelope and idealized neural responses.
• Considered four linguistic units separately: phones, syllables, words, and sentences.
• Investigated correlations in specific frequency bands (e.g., 3-6 Hz, ~1 Hz).

Main Results:

• Idealized phone-, syllable-, and word-tracking responses correlated with the speech envelope around 3-6 Hz and below ~1 Hz.
• The ~1 Hz correlation was primarily driven by pauses within connected speech.
• Results indicate overlap in neural tracking across different linguistic levels.

Conclusions:

• A simple frequency-domain decomposition of speech envelope-tracking activity is insufficient to separate neural responses to distinct linguistic units in natural speech.
• The assumption that specific frequency bands uniquely represent different linguistic units may not hold for natural speech.
• Further research is needed to understand how the brain segregates linguistic information processed at different timescales.