Articulation Artifacts During Overt Language Production in Event-Related Brain Potentials: Description and Correction

Area of Science:

• Cognitive Neuroscience
• Linguistics
• Biomedical Engineering

Background:

• Overt articulation introduces substantial artifacts into electroencephalogram (EEG) and event-related potential (ERP) recordings.
• These artifacts pose a significant challenge for studying language production using EEG/ERP data.

Purpose of the Study:

• To characterize the properties of articulation-related artifacts in EEG/ERP.
• To introduce and validate a novel artifact correction procedure for overt articulation.

Main Methods:

• Co-recording of ERPs and articulatory trajectories using an electromagnetic articulograph.
• Application of residue iteration decomposition (RIDE) for artifact separation.
• Comparison with independent component analysis (ICA) for artifact removal.

Main Results:

• Articulation artifacts were identified up to 300 ms before voice onset, exhibiting consistent topography across phonemes but varying in time course and amplitude.
• RIDE successfully separated ERPs from artifacts, reducing correlations with articulatory data to near zero.
• ICA was less effective, leaving substantial correlations and attenuating early ERP components.

Conclusions:

• Articulation-related artifacts have distinct temporal and topographic properties.
• RIDE offers a robust method for removing articulation artifacts, significantly improving the quality of ERPs for language production research.
• This approach provides a new perspective for ERP studies involving overt articulation.