Decoding the temporal dynamics of spoken word and nonword processing from EEG.

Area of Science:

• Neuroscience
• Cognitive Science
• Psycholinguistics

Background:

• Spoken word recognition is crucial for communication, relying on implicit activation and competition of word candidates.
• Existing methods struggle to capture the neural dynamics of this competition at a millisecond-by-millisecond level.
• Understanding these dynamics is vital for neuroscience of language and for studying hearing, language, and cognitive disorders.

Purpose of the Study:

• To develop and validate a machine learning-based approach using electroencephalography (EEG) to decode spoken word recognition.
• To analyze the temporal dynamics of neural competition between target words and acoustic competitors.
• To assess the robustness and reliability of this paradigm for future clinical applications.

Main Methods:

• Applied machine learning techniques to standard EEG signals to decode heard words on a trial-by-trial basis.
• Analyzed temporal patterns of confusion between target words and similar-sounding competitors.
• Investigated robustness across different EEG systems, channel counts, and trial numbers.

Main Results:

• The decoding model initially showed confusion between target words and competitors, mirroring psycholinguistic findings.
• Neural suppression of competing word candidates was observed around 500 milliseconds post-stimulus onset.
• Results demonstrated robustness across variations in EEG setup and were reliable within individuals.

Conclusions:

• The developed EEG-based paradigm offers a powerful and simple method to assess neural dynamics of speech decoding in real-time.
• This approach has significant potential for understanding lexical development and aiding in the diagnosis of various clinical disorders.
• The findings provide a millisecond-level neural correlate of spoken word recognition competition and suppression.