Latent neural dynamics encode temporal context in speech.

Area of Science:

• Neuroscience
• Auditory Processing
• Speech Perception

Background:

• The human auditory cortex encodes acoustic-phonetic features of speech.
• Neural responses capture acoustic amplitude cues related to speech timing.
• Understanding the neural basis of temporal context in speech is essential.

Purpose of the Study:

• To investigate how distributed cortical responses represent temporal context in speech.
• To determine the role of timing cues in speech processing within the auditory cortex.
• To explore the relationship between temporal context and phonetic feature encoding.

Main Methods:

• Utilized direct neural recordings from the human auditory cortex.
• Employed a group reduced rank regression model.
• Analyzed widespread activity over the superior temporal gyrus.

Main Results:

• Demonstrated a low-dimensional latent state representation of temporal context.
• Identified timing cues as capturing more unique variance than phonetic features.
• Observed rotational or cyclical dynamics in latent space related to timing cues.

Conclusions:

• Spatially distributed timing signals in the auditory cortex provide temporal context for phonetic processing.
• These timing signals may bind phonetic features across time to form higher-order representations.
• Temporal context is a critical, low-dimensional component of speech representation in the brain.