Premotor cortex uses a compositional neural geometry to plan words

Area of Science:

• Neuroscience
• Computational Linguistics
• Speech Motor Control

Background:

• The brain must precisely order words and phonemes for novel speech production.
• Flexible preparation allows for the generation of previously unheard words.

Purpose of the Study:

• To investigate how neural populations in the premotor cortex achieve flexible speech preparation.
• To understand the mechanism of compositional planning for novel word sequences.

Main Methods:

• Recorded intracortical activity from the premotor cortex in two participants of a speech neuroprosthesis clinical trial.
• Participants attempted to speak factorially-balanced phoneme sequences.
• Analyzed neural activity during the preparation phase of speech production.

Main Results:

• Neural activity encoded not only the next phoneme but multiple upcoming phoneme positions, representing whole words.
• Word-level plans were formed by compositionally combining phoneme representations.
• Premotor cortex activity was largely limited to the first word in multi-word utterances, suggesting hierarchical segmentation.

Conclusions:

• The premotor cortex employs a compositional, hierarchically-segmented planning geometry for speech.
• This neural strategy may enable efficient planning of novel spoken sequences.
• Hierarchical segmentation by higher-order features could be a universal neural strategy for language sequence organization.