Dynamic encoding of speech sequence probability in human temporal cortex.

Area of Science:

• Neuroscience
• Auditory Perception
• Computational Linguistics

Background:

• Sensory processing integrates stimulus features with context.
• Learned statistical knowledge of sensory environments, like speech sound sequences, tunes auditory perception.
• Listeners' brains are attuned to the statistical regularities of their native language.

Purpose of the Study:

• Investigate the neural encoding of speech sequence statistics in the human brain.
• Determine how the brain integrates acoustic features with learned probabilities of sound sequences.
• Explore the interaction between low-level auditory processing and high-order linguistic knowledge.

Main Methods:

• High-resolution direct cortical recordings from human lateral superior temporal cortex.
• Subjects listened to real English words and pronounceable nonwords with varying sound transition probabilities.
• Analysis of neural responses to acoustic features and contextual sequence statistics.

Main Results:

• Neural responses encoded both preceding and upcoming speech sound probabilities.
• Transition probability modulated neural responses with an initial negative, followed by a positive, deflection.
• Encoding of transition probability differed between real words and nonwords, indicating interaction with linguistic knowledge.

Conclusions:

• Auditory processing of learned stimuli, like speech, integrates physical features with contextual sequential structure.
• Listeners implicitly use phoneme sequence statistics to process spoken input.
• Neural mechanisms link low-level acoustic representations with linguistic information for word recognition.