Characterizing multi‐word speech production using event‐related potentials

Abstract

Event‐related potentials (ERPs) derived from electroencephalography (EEG) have proven useful for understanding linguistic processes during language perception and production. Words are commonly produced in sequences, yet most ERP studies have used single‐word experimental designs. Single‐word designs reduce potential ERP overlap in word sequence production. However, word sequence production engages brain mechanisms in different ways than single word production. In particular, speech monitoring and planning mechanisms are more engaged than for single words since several words must be produced in a short period of time. This study evaluates the feasibility of recording ERP components in the context of word sequence production, and whether separate components could be isolated for each word. Scalp EEG data were acquired, while participants recited word sequences from memory at a regular pace, using a tongue‐twister paradigm. The results revealed fronto‐central error‐related negativity, previously associated with speech monitoring, which could be distinguished for each word. Its peak amplitude was sensitive to Cycle and Phonological Similarity. However, an effect of sequential production was also observable on baseline measures, indicating baseline shifts throughout the word sequence due to concurrent sustained medial‐frontal EEG activity. We also report a late left anterior negativity (LLAN), associated with verbal response planning and execution, onsetting around 100 ms before the first word in each cycle and sustained throughout the rest of the cycle. This work underlines the importance of considering the contribution of transient and sustained EEG activity on ERPs, and provides evidence that ERPs can be used to study sequential word production.

Speaker-induced suppression in EEG during a naturalistic reading and listening task

“…21 participants (11F, age 24.4±3.9) were recruited from The University of Texas at Austin. This is in line with sample sizes of recent EEG studies of speech production (Ries et al 2021; Zhao and Rudzicz 2015; Goregliad Fjaellingsdal et al 2020). All participants were native speakers of English with typical hearing as assessed through pure tone audiometry and a speech-in-noise hearing test (QuickSIN, Interacoustics).…”

“…While we utilized methods to correct for EMG artifact that have been previously demonstrated in the literature to be successful (Ries et al 2021; Vos et al 2010; Chen et al 2019), there is no definitive way to rule out residual EMG given the lack of ground truth in the sources that contribute to the electroencephalogram. As a result, we further explored the influence of EMG artifact on model performance by fitting linear encoding models that included normalized EMG activity recorded from auxiliary facial electrodes in tandem with the EEG as a regressor.…”

“…The selected ICA components were next removed from the unfiltered data. After ICA, data were filtered at 0.16Hz and corrected for EMG artifact via blind source separation algorithm based on Canonical Correlation Analysis (CCA, (De Clercq et al 2006)), a technique that has been previously demonstrated to correct for EMG artifact in speech production EEG tasks (Vos et al 2010; Ries et al 2021; Riès et al 2013). In line with these studies, CCA was performed in two passes: first, a 30-second window to remove tonic muscle activity; second, a 2-second window to remove rapid bursts of EMG associated with speech production.…”

Encoding category-level and context-specific phonological information at different stages: An EEG study of Mandarin third-tone sandhi word production

“…In particular, the T2 prime elicited more negative amplitude in the left anterior and central scalp regions. Based on the time window and channel locations, the effect of the T2 prime in the current study is probably linked to a left-lateralized slow-rising negative frontal component documented in previous EEG research on speech production (Riès et al, 2013(Riès et al, , 2021, which is termed as the left-lateralized anterior negativity (LLAN) and thought to be associated with response preparation, motor planning and execution (Riès et al, 2021). Due to the pitch contour similarity between the T2 prime and the T3 sandhi variant (Nixon et al, 2015), the T2 prime probably brought benefits to the phonetic encoding or motor preparation stage, and the surface form was more likely to be encoded during this stage.…”

Greater Social Competence Is Associated With Higher Interpersonal Neural Synchrony in Adolescents With Autism

“…The minimum sufficient data quantity for hyperscanning studies is yet to be established. In the meantime, novel EEG signal cleaning procedures for addressing speech-related motor artifacts are starting to emerge (e.g., Riès et al, 2021 ) and may support higher data retention rates in social interaction paradigms. Future work capitalizing on these developments as well as on ecological validity of real-time, face-to-face reciprocal social interactions and high temporal resolution of EEG will also need to investigate more fine-grained moment-to-moment associations between various social actions (e.g., eye contact, positive affect, head nods) and the neural coordination of activity between the interacting brains to more precisely characterize the mechanisms underlying social behaviors, the contributions of sex differences, and identify specific factors that facilitate or impede social synchrony.…”