Impact of EEG Reference Schemes on Event-Related Potential Outcomes: A Corollary Discharge Study Using a Talk/Listen Paradigm

doi:10.21203/rs.3.rs-4853183/v1

Download PDF

Research Article

Impact of EEG Reference Schemes on Event-Related Potential Outcomes: A Corollary Discharge Study Using a Talk/Listen Paradigm

https://doi.org/10.21203/rs.3.rs-4853183/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The selection of an appropriate virtual reference schema is pivotal in determining the outcomes of Event-Related Potential (ERP) studies, particularly within the widely utilized Talk/Listen ERP paradigm, which is employed to non-invasively explore the corollary discharge (CD) phenomenon in the speech-auditory system. This research centers on examining the effects of prevalent EEG reference schemas—Linked Mastoids (LM), Common Average Reference (CAR), and Reference Electrode Standardization Technique (REST)—through statistical analysis, Statistical Parametric Scalp Mapping (SPSM), and source localization techniques. Our ANOVA findings indicate significant main effects for both the reference and the experimental condition on the amplitude of N1 ERPs. Depending on the reference used, the polarity and amplitude of the N1 ERPs demonstrate systematic variations: LM is associated with pronounced frontocentral activity, whereas both CAR and REST exhibit patterns of frontocentral and occipitotemporal activity. The significance of SPSM results is confined to regions exhibiting prominent N1 activity for each reference schema. Source analysis provides corroborative evidence more aligned with the SPSM results for CAR than for REST or LM, suggesting that results under CAR are more objective and reliable. Therefore, the CAR reference is recommended for future studies involving Talk/Listen ERP paradigms.

Corollary Discharge

Statistical Parametric Scalp mapping

EEG Reference Schemas

Talk/Listen Paradigm

Auditory N1 Amplitude

Reference Electrode Standardization Technique

Emerging evidence suggests that predictive feed-forward motor sensory communication is key to the optimal functioning of the organism. The hardwiring between the motor and sensory brain regions is quite recursive, facilitating retrograde motor-to-sensory communication besides the usual sensory-to-motor communication (Crapse & Sommer, 2008). This motor-to-sensory signal of the planned motor action is instantiated faithfully in real-time from the generator frontal motor brain regions to the appropriate sensory cortices to modulate the target neurons for responding to the imminent sensory reafference. This neural communication is instrumental in the precise anticipation of the impending sensation resulting from self-generated action and is key to determining the agency of the performed action. This specialized motor-to-sensory neural signaling is referred to as corollary discharge (CD) (Sperry, 1950).

Studies using the Talk/Listen paradigm, where EEG potentials are synchronized to speech sounds during active listening and compared to passive playback, effectively assess CD signalling non-invasively in the speech auditory system. Typically, healthy controls exhibit lower auditory N1 amplitude during the Talk condition relative to the Listen condition, a phenomenon known as speech-induced suppression (SIS) (Whitford, 2019). The modulation by the active CD signaling during speaking has been proposed to induce this SIS phenomenon. Various modifications of this task have studied other psychophysiological factors like agency and ownership (Beño-Ruiz-de-la-Sierra et al., 2023), expectancy (Ford, Gray, et al., 2007), and frequency precision of the CD to the sensory reafference in the speech-auditory system (Heinks-Maldonado et al., 2007). Moreover, dysfunctional CD, as measured by this ERP paradigm, has been proposed as a mechanism underlying various psychiatric conditions, highlighting a deficit in self-monitoring linked to CD signaling dysfunction (Ford et al., 2013).

These findings underscore the utility of this ERP paradigm in measuring CD functionality in healthy and across neuropsychiatric conditions using cost-effective EEG technology. The choice of reference in ERP studies is crucial, as it can significantly influence the interpretation of results within the same experimental framework (Joyce & Rossion, 2005). Prior research has demonstrated that outcomes such as power spectra, EEG coherence, voltage waveform, and statistical parametric scalp mapping (SPSM) can vary with different reference choices (Qin et al., 2010). This underscores the choice of reference schema that more accurately captures the neural phenomenon is critical in drawing meaningful conclusions in ERP studies. The ideal virtual EEG reference is the one that is at zero/neutral potential. Although the ideal EEG reference would be at a zero/neutral potential, the debate over the best reference choice remains unresolved.

Several reference schemes have been adopted to address this issue, including the common average reference (CAR) (Offner, 1950), linked mastoid (LM) (Gevins, 2000), vertex reference (CZ) (Lehmann et al., 1998), and the reference electrode standardization technique (REST) (Yao, 2001).The CAR, the average of the signals from all the channels, is the most commonly employed in ERP paradigms. This is supposed to return unbiased measurements not biased to any specific electrode locations. However, it can be erroneous in the context of sparse spatial sampling not encompassing the entire head surface (Lantz et al., 2003) and might introduce physiological dynamic signals that affect EEG results (Thatcher, 2012; Yao, 2001). REST schema, which calculates a virtual reference point at infinity, has demonstrated superiority in estimating EEG spectrum (Chella et al., 2014, 2017), coherence (Marzetti et al., 2007), and event-related potentials (Tian & Yao, 2013; Yang et al., 2017) in previous studies. Therefore, we conceived this study to explore the reference that best captures the corollary discharge-related neural activity in the context of this widely employed paradigm by objectively comparing N1 ERPs and CD-induced suppression between the conditions measured with CAR, REST, and LM reference schemas.

To the best of our knowledge, all the previous studies have largely ignored this well-known bias introduced by the choice of reference while drawing conclusions based on this Talk/Listen paradigm. Historically, the linked mastoid (LM) schema has been predominantly used, and recommended by standard guidelines in the field (Ford et al., 2010). The LM, calculating the arithmetic average of activity from the left and right mastoids and being independent of montage, provides consistent results across different studies, even when varying numbers of ERP recording channels are used. However, the LM approach has faced criticism for shifting outcomes towards the frontocentral regions (Yao, 2001) and diminishing the hemispheric asymmetry apparent in scalp topography, due to reliance on its method of bilateral averaging. Previous studies have nonetheless demonstrated a left-pronounced CD effect in a similar paradigm (Heinks‐Maldonado et al., 2005; Houde et al., 2002). Moreover, the neural activity in the mastoid regions is non-zero, which might introduce its own physiological dynamic signals, mostly from temporoparietal regions, into the ERP findings. Multiple studies have highlighted these limitations of LM in risking erroneous conclusions on ERP results.

To investigate these effects, we utilized ANOVA to analyze differences in N1 amplitude of Talk and Listen conditions across the reference schemas.We also measured the corollary discharge index (CDI) [(N1Talk - N1 Listen)] across the three reference schemes. Additionally, we performed Statistical Parametric Scalp Mapping (SPSM) to map the scalp distribution of statistical differences (paired t-tests) in N1 amplitude between conditions, identifying regions showing significant disparities in neural activity for each reference schema. To corroborate the SPSM findings, we located the source of auditory N1 activity for each condition. We then quantified differences in grandaverage ERPs within the source space between conditions and projected the difference onto the sensor space. This projection, which effectively addresses the forward problem in EEG, was anticipated to align closely with the SPSM results of the reference schema that most accurately captures the CD-related neural activity. The fundamental premise of this approach is that source activity remains consistent irrespective of the reference used for measurement. Consequently, differences in CD-induced source activity, when projected back to the sensor space, should mirror the topographical distribution observed in the SPSM of the reference schema that also delineates the significant statistical differences at sites where N1 activity is most evident. A similar approach has been employed to investigate the most appropriate reference schema to measure the N170 ERP of facial recognition (Wang et al., 2019).

Thirty right-handed healthy subjects (ten females) aged 18 to 35 years (mean age = 27 years) participated in the study. The study was conducted in the facility of the Central Institute of Psychiatry, a tertiary care Psychiatric hospital in eastern India. Subjects were recruited by word of mouth among the hospital staff. All the subjects were recruited after excluding any neurological or psychiatric illness and hearing impairment. The study was approved by the on-site institute ethics committee of the Central Institute of Psychiatry. All the subjects signed an informed consent form before the enrolment.

Procedure

The study participants engaged in a talk-listen paradigm facilitated by E-prime3 (Psychology Software Tools, Inc.) software for the stimulus presentation. During the talk condition, participants were instructed to emit brief (<300 ms), sharp vocalizations of the phoneme 'Ah' at a self-paced interval of approximately every 2-3 seconds for a duration of 180 seconds. The training ensured that participants maintained 'Ah' vocalizations at a consistent sound intensity of 75 to 85 dB, as measured by a handheld decibel meter (model H-M80A) positioned approximately 6 cm from the participants' mouths. This sound intensity was standardized across both talk and listen conditions by equilibrating earphone audio output using the hand-held decibel meter to mitigate variability. Participants were also cautioned against any overt facial or bodily movements that could introduce artifacts into the EEG recordings.

The vocalizations were captured using a microphone (model NT1) connected to the computer managing the stimulus presentation and were immediately played back to the subjects in real-time via audio-technica earphones equipped with foam insert tips and active noise cancellation features (model- ATH-ANC33IS), which were inserted into the ear canals. Following the talk condition, the listen condition commenced, wherein the previously recorded vocalizations were played back passively to the subjects. Throughout both conditions, participants were instructed to maintain their focus on a centrally positioned white cross, displayed against a dark background on a 14-inch monitor.

Offline processing of the audio track:

The recorded vocalizations were digitized and subsequently processed offline utilizing an automated Matlab-based script designed to detect the onset of vocalizations (Ford et al., 2010). The timestamps marking the onset of vocalizations, as identified by this automated routine, underwent further verification using Audacity®, an open-source audio editing software. Erroneous timestamps were meticulously corrected or removed, culminating in the creation of the final event file. This file was then integrated into the EEG recording for subsequent time-locked epoching and averaging of EEG data.

Data Acquisition and Processing:

Electroencephalogram (EEG) data were captured using a 64-channel cap, incorporating Ag-AgCl electrodes, adhering to the 10-10 system international standard, and interfaced with a Galileo EB Neuro system. Impedance was kept below 5 kΩ. Additional bipolar electrodes were strategically positioned at the outer canthi of both eyes to record the horizontal electrooculogram (HEOG) and above and below the right eye to capture the vertical electrooculogram (VEOG), facilitating the monitoring of eye movements. The EEG data were recorded with an online reference between the Fz and AFz electrodes at a sampling rate of 1024 Hz. The data were imported offline into the EEGLAB (Delorme & Makeig, 2004) and a 1-15 Hz band-pass, Butterworth filter of order 2 was applied before further processing. Subsequent data processing involved an independent component analysis using an extended infomax-based algorithm to segregate the data into statistically independent components. Artefactual components were identified using the Multiple Artefact Rejection Algorithm (MARA) plugin, which is freely available in EEGLAB (Winkler et al., 2011) and subsequently removed. Each data was subsequently offline re-referenced to LM (the average of mastoid electrodes TP9, TP10), CAR (the average of all the channels), and to the REST reference, creating three distinct data processing streams. The REST referencing was done using the freely available plugin available in EEGLAB (Dong et al., 2017). Each data was further segmented into 600m s epochs with a 100ms pre-stimulus interval. Baseline correction was applied to the pre-stimulus baseline of 100ms in all the epochs. All the epochs were screened for amplitude fluctuation, and those exceeding beyond the range of -/+50 microvolt were rejected.

Data Analysis

ERP Analysis:

The N1 peak amplitude was determined by identifying the most negative amplitude relative to the pre-stimulus baseline within a temporal window of 60-150 ms following the vocalization of 'Ah' in both the talk and listen conditions. The N1 amplitude were assessed individually for data referenced to LM, CAR, and REST. The grand average amplitude scalp topography displayed the predominant amplitude negativity centred around the frontocentral area across all reference schemas (as illustrated in Figure 2). Consequently, our analysis concentrated on the amplitude data from this specific region. The electrode locations at Cz and FCz have been consistently utilized as target sites for sampling data in prior investigations examining CD in various neuropsychiatric conditions. For statistical analysis, the N1 Talk and N1 Listen data were computed as the average N1 amplitude from the Cz and FCz channels for each respective condition. Furthermore, the degree of N1 suppression, quantified by the corollary discharge index (CDI), was determined by calculating the difference between the N1 amplitudes during the talk and listen conditions and [CDI = N1Talk - N1Listen]. The resultant CDI values used for statistical comparisons were derived by averaging the CDIs calculated from the Cz and FCz channels.

Statistical analysis:

The primary objective of this investigation was to evaluate the impact of three distinct reference schemes on event-related potential (ERP) data, focusing on identifying the most effective reference scheme for detecting speech-generated corollary discharge during the talking condition. A two-way repeated measures analysis of variance (ANOVA) was employed to assess the influence of the study condition (Talk and Listen) and the reference scheme (LM, REST, and CAR) on the auditory N1 amplitude. The threshold for rejecting the null hypothesis was set at p < 0.05. Subsequently, targeted analyses involved conducting separate paired t-tests between conditions for each offline reference scheme to determine if variations in N1 amplitude persisted irrespective of the reference scheme utilized. We did a separate repeated measures ANOVA to compare the CDI values across the reference schemas, to understand the impact of the schema on this frequently employed measure of corollary discharge induced N1 suppression.

Further analytical measures included a correlational analysis to explore the impact of the selected offline reference scheme on CDI metrics: CDI-LM, CDI-REST, and CDI-CAR. All statistical analyses were performed using IBM SPSS Statistics 20 (SPSS Inc., Chicago, Illinois 60606).

SPSM and Scalp topography:

To obtain a more comprehensive appreciation of the distribution of scalp potential differences across conditions, we further conducted a topographical analysis of scalp potentials by implementing statistical parametric scalp mapping (SPSM) – a statistical technique for elucidating the scalp distribution of significant differences between the condition-specific N1 ERPs. To this end, we performed paired t-tests for the N1 amplitude between the Talk and Listen conditions across all the scalp electrodes and represented all the significant result t-values as a scalp topographic map. The results were marked significant at a p-level of < 0.01 after FDR correction to adjust for the multiple comparisons. This analysis was performed separately for each of the reference schema- CAR, REST, and LM.

Source Analysis:

To further validate the SPSM results, we conducted three sequential analysis steps using the Brainstorm application (Tadel et al., 2019). Firstly, a 3-shell sphere head model (conductivities of the Brain, Scalp, and Skull are 0.33, 0.0042, 0.33, respectively) of each participant was estimated, with a total of 15,002 sources (voxels) distributed around the cerebral cortex in the ICBM 152 head model (a non-linear average of 152 magnetic resonance scans) and the default channel location file in the brainstorm application (Fonov et al., 2011). The parameter of Regularize noise covariance was set to 0.1, and the parameter of signal-to-noise ratio was set to 3.00. Secondly, using these parameters, a source activity reconstruction was performed for the grandaveraged ERPs, estimated with the LM schema, for the Talk and Listen conditions. Subsequently, a 20 ms time range around the N1 peak in the source space was intercepted and averaged for each condition separately. Finally, the averaged source activity differences were projected back to the sensor space to generate the topographic map. The resultant topographic map reflects the source difference activity between the conditions. The source reconstruction and sensor projection were performed by brainstorm in the MATLAB environment.

Noise covariance:

The noise covariance required for the source reconstruction was estimated based on the resting state EEG data. A 150-second resting state eyes open EEG was recorded for each individual after the completion of the task for the computation of the noise covariance matrix. The dataset was processed in the similar manner and a 1-15 Hz band pass filter was applied. Subsequently all the filtered datasets were concatenated to generate a single dataset to compute the covariance matrix.

ERP Measurements:

For the analysis of N1 amplitude as a dependent variable, a repeated-measure ANOVA incorporating two factors (Reference: CAR, REST, and LM; Condition: Talk and Listen) was executed, demonstrating significant main effects for both reference (F = 22.418, p < .001*, η_p² = .211; Figure 1), condition (F = 54.067, p < .001*, η_p² = .243; Figure 1) and the interaction effect between reference and condition was also significant (F = 5.278, p = .006*, η_p² = .061). Subsequently, three separate paired t-tests were conducted between conditions for each reference schema, with the results summarized in Table 1. Regardless of the reference schema, the N1 amplitude was significantly higher (more negative) in the Listen condition (Figure 1), reflecting the CD-induced suppression (Figures 1 & 2). The grand average N1 ERPs are plotted in Figure 1 for the electrodes TP9, TP10, Cz, and FCz; the reversal of the N100 at the mastoid electrodes measured at the same latency for the REST and CAR references denotes the capture of the signal from the flipped side of the dipole of the auditory N1 generator. The RMANOVA among the CDI values reflects a significant effect of reference on the CDI (F = 63.379, p < 0.001, η_p² = 0.701). The results of the post hoc pairwise comparison have been shown in Table 2: the results denote a significantly higher value for LM (CDI_LM) and the least for CAR (CDI_CAR), with REST (CDI_REST) yielding an intermediate value.

Additionally, we conducted a correlational analysis among the CDI values across the reference schemes. All the CDI values were strongly correlated with each other: CAR_CDI vs LM_CDI, r = .829 ; CAR_CDI vs REST_CDI, r = .931 ; LM_CDI vs REST_CDI, r = .897.

Measurement Pair(N=20)	Mean	Standard Deviation	t	p= 0.05
Listen_CAR	-2.426	1.210	-5.181	<.001*
Talk_CAR	-1.235	0.922	-5.181	<.001*
Listen_REST	-3.650	1.737	-5.982	<.001*
Talk_REST	-1.820	1.422	-5.982	<.001*
Listen_LM	-5.468	2.439	-6.867	<.001*
Talk_LM	-2.571	2.340	-6.867	<.001*

Table 1: Paired t-test: The N1 amplitude is significantly higher (more negative) in the Listen condition across all reference schemas, with significance adjusted for multiple comparisons at a threshold of p = 0.016. The effect size was found to be highest for LM, lowest for CAR, and intermediate for REST.

Pairwise Comparisons
		Mean Difference	Std. Error	Sig. (P=0.05)	95% Confidence Interval for Difference^b
		Mean Difference	Std. Error	Sig. (P=0.05)	Lower Bound	Upper Bound
CDI _CAR	CDI _REST	-.744	.129	<0.001	-1.072	-.415
CDI _REST	CDI _LM	-1.328	.175	<0.001	-1.776	-0.881
CDI _LM	CDI _CAR	2.072	.239	<0.001	1.462	2.681

Table 2: The results of the posthoc Bonferroni test for the CDI across the reference schemas denoting a hierarchical relationship of CDI_LM > CDI_REST > CDI_CAR.

Scalp topography and SPSM:

The scalp topographic map of the Talk and Listen N1 ERP are illustrated in the figure 2 A & B, the waveforms of the ERP in the electrode location TP9, TP10, Cz and FCz are illustrated in the figure 1. The scalp topographic maps feature two distinct patterns. Firstly, the potential values are larger in the Listen condition as compared to the Talk condition reflecting CD induced suppression. Secondly, the voltage distribution pattern of the scalp topography remains similar – a frontocentral negativity and an occipitotemporal positive potential- across reference schemas and conditions. The only difference is a constant shift of voltage induced by different references, which imparts a commensurate variation of colour intensity to the negative and positive regions of the scalp while preserving the spatial distribution pattern (Pascual-Marqui & Lehamann, 1993; Yao et al., 2007). However, the SPSM results between the talk and listen conditions feature distinct differences in the spatial distribution for different references (Figure 2 B). LM indicates a frontocentral distribution, the REST features a comparatively smaller area of frontocentral and left temporo-occipital distribution of significant difference. The CAR demonstrates a frontocentral and bilateral temporo-occipital distribution. Each of the SPSM results features a prominent left-dominant activity, which is more pronounced in the CAR schema.

Source results:

Figure 3 presents the cortical source activity distribution during the Listen and Talk conditions and the source difference activity between the conditions. During both of the conditions, the activation in the upper margin of the posterior superior temporal lobe near Heschl’s gyrus is apparent. The region corresponds to the primary auditory cortex, the location believed to be the generator region of the auditory N1 ERP (Liem et al., 2012).

Figure 4 represents the voltage topographical projection of the source difference between the talk and listen conditions. The result illustrates that the difference between the two stimuli is primarily distributed in the frontocentral region and the right and left occipitotemporal areas, which is reminiscent of the SPSM results of the CAR reference schema.

In this study, we evaluated various offline reference schemas—including CAR, REST, and LM—to identify the most effective method for analyzing CD signaling within the speech-auditory system using a Talk/Listen ERP paradigm. Our findings have consistently replicated the SIS phenomenon in healthy subjects, aligning with existing literature (Whitford, 2019). The feed-forward CD signal-mediated recruitment of inhibitory interneurons in the auditory cortex during speaking is believed to accomplish this N1 suppression (Eliades & Wang, 2008; Nelson et al., 2013; Schneider et al., 2014). Most strikingly, the measurements, regardless of the reference schema, have reliably captured the SIS effect (Table 1).

In the context of ERP analysis, the virtual reference that is most proximal, if not coincident, to the zero/neutral reference point is more likely to capture the most accurate neural signal. The measurements of the Talk/listen paradigm are no exception and our analysis intends to find a solution for the same. This study is pioneering in systematically evaluating how different reference schemas influence the results of this paradigm. Initial observations reveal that the N1 ERPs display frontocentral negativity and occipitotemporal positivity across references and conditions (Figure 2A), with variations in contrast across reference schemas: pronounced for CAR, minimal for LM, and moderate for REST. These differences are vividly captured in the grand average ERP images—LM exhibits significant negativity in frontocentral regions and minimal activity at mastoid regions, while CAR shows balanced positive activity at mastoids concurrent with negative peaks in the frontocentral area (Figure 1). REST presents an intermediary outcome, bridging positive and negative values between CAR and LM. This alludes to the potential presence of dipolar neural generators, with their flipped poles oriented in opposite directions, which is most accurately represented by the results of the CAR schema. The graded distribution of the negative and positive potentials that varies systematically across references indicates the addition of a constant for each reference. Prior intracranial recording studies (Godey et al., 2001, 2001), as well as electrical source localization analysis in EEG and MEG studies (T et al., 2007; Yvert et al., 2001), have consistently localized the source of auditory N1 ERP in and around Heschl’s gyrus in the supratemporal plane on both hemispheres. Our source analysis findings are precisely in agreement, where the N1 ERPs are localized in the same areas across conditions (Figure 3).

The SPSM analysis highlights the distribution of statistically significant amplitude differences between conditions for each reference schema. CAR demonstrates clear separation with frontocentral negativity and bilateral occipito-temporal positivity, aligning with the hypothesis that the auditory N1 dipole orients its negative pole towards the frontocentral and its positive pole towards the occipito-temporal regions, maximizing potential differences and pinpointing condition-specific variations. LM's reference electrodes, positioned over the mastoids, likely overlay the region of positive dipole activity, resulting in lower occipitotemporal potentials while higher negative potentials appear over the extensive frontocentral regions. This represents the sum of the dipole's positive and negative electrical activities. REST presents intermediate results with frontocentral and left occipital-temporal activity differences in SPSM analysis, suggesting transitional values between LM's frontocentral dominance and CAR's dipolar separation. These observations, supported by the ERP amplitude data in Figure 1, indicate the influence of reference schema on neural activity representation. However, these amplitude and topographic variations alone are insufficient to conclusively determine the measurement accuracy of true neural activity

To corroborate the SPSM results that more accurately reflect CD-induced neural activity, we projected the differential source activity from condition-specific grandaverage ERPs into the sensor space, as depicted in Figure 4. This reconstructed source activity is unique and independent of the reference schema used, because the EEG inverse problem yields consistent solutions across different schemas. Consequently, differences in ERPs in the source space align with the activity associated with SIS. Projecting this differential source activity to the sensor space effectively addresses the forward problem of EEG, representing the differential neural activity in the sensor space.Therefore, the SPSM topography that best corresponds with the scalp topography derived from projecting the source difference activity more accurately represents true neural activity. Thus, the reference schema that aligns most closely with this projection should be considered the most accurate for capturing neural activity. Figure 4 demonstrates the source-to-sensor projection of CD-induced source difference, highlighting a frontocentral negativity and occipitotemporal positivity that align closely with the SPSM results of the CAR schema. Moreover, the identical sensor space topography reconstructed from the condition differential activity from the source space computed from all three reference schema further robustly validates the claim that the EEG inverse problem is unique and independent of the reference schema.

The CDI is a proxy marker of CD effectiveness and the most commonly employed parameter to conduct correlational analysis with the clinical variables in previous studies. The LM_CDI calculated over the frontocentral region yields the highest values for the CDI due to an overestimation of N1 amplitudes, due to the summated activity of the opposite poles of the generator dipole. As implicit as it is, the order of the CDI values is CDI_LM > CDI_REST > CDI_CAR (Table 2). Aligning with our prior discussion the CDI_CAR is most likely closer to the true CD activity and should be evaluated in future studies.

To date, almost all the studies have employed the LM to draw conclusions in the framework of this paradigm. The strategy is to match the results with the previous findings without addressing the concern of the bias introduced due to the selection of the reference schema. The tendency to draw biased conclusions from different reference selections in ERP studies is well-documented, yet there remains a lack of systematic investigation into this issue within this widely used paradigm. The LM approach is known to shift the results to the frontocentral location and obscures the scalp representation of occipitotemporal sources by distorting the scalp topography representation. Our results not only highlight these methodological flaws but also demonstrate the superior capability of the CAR schema in mitigating these distortions. Furthermore, the deficient CD has been a robust finding in psychosis. Yet, there have been inconsistencies in finding a correlation between N1 suppression and the psychopathological variables, with some showing a correlation (Bose et al., 2019; Heinks-Maldonado et al., 2007) and others not (Ford et al., 2013). Thus, based on our findings, we advocate for the adoption of the CAR schema in future studies, as it more accurately reflects CD-related neural activity. However, we cautiously recommend that the superiority of the CAR over LM be thoroughly evaluated in clinical scenarios to evaluate its efficacy in discerning CD anomalies in various neuropsychiatric conditions.

The outcomes of this study, conducted with a 64-channel EEG system, elucidate several pivotal insights: (1) Across all reference schemas, the amplitude of the N1 ERP component is consistently elevated in the listen condition relative to the talk condition, indicative of active CD-induced cortical suppression during speech production. (2) Scalp topography and SPSM analyses demonstrate frontocentral negativity and occipitotemporal positivity, which is most apparent in the schema, which accurately reflects neural activity from dipolar generators located in the supratemporal plane responsible for auditory N1 generation. (3) Source localization analysis pinpointed the auditory N1 activity for both conditions to Heschl’s gyrus within the supratemporal plane, corroborating prior research findings. (4) The projection of differential source activity into the sensor space corresponds closely with the SPSM results for the CAR schema, substantiating the assertion that the CAR schema more precisely captures the neural dynamics associated with SIS while illustrating the potential for the LM schema to distort and redistribute these values towards the frontocentral regions. Future investigations should aim to confirm these findings through the application of a high-density EEG montage (128/256 channels), which may enhance the detection accuracy of neural signals, particularly within the CAR schema.

CD Corollary Discharge

EC Efference Copy

ERP Event-Related Potential

EEG Electroencephalographic

MEG Magnetoencephalographic

LM Linked Mastoids

CAR Common Average Reference

REST Reference Electrode Standardization Technique

CDI Corollary Discharge Index

HEOG Horizontal Electrooculogram

VEOG Vertical Electrooculogram

MARA Multiple Artifact Rejection Algorithm

SIS Speech Induced Suppression

SPSM Statistical Parametric Scalp Mapping

Data availability statement

The data analyzed in this study is subject to the following licenses/restrictions: The datasets (including clinical data) generated during and/or analyzed during the current study are available from corresponding authors on a reasonable request.

Ethics statement

The studies involving human participants were reviewed and approved by the Institute Ethics Committee of the Central Institute of Psychiatry (CIP)(IEC/CIP/2022-23/1747). The participants provided their written informed consent to participate in this study.

Author contributions

Subham Samantaray: Conceptualization; participated in the investigation; curated data; performed formal analysis; developed methodology; contributed to software programming; validated research findings; and was involved in writing both the original draft and subsequent revisions. Muralidharan Kesavan: Contributed to the conceptualization; helped develop methodology; and assisted in the writing and review process and funding acquisition. Nishant Goyal: Engaged in the conceptualization; assisted in developing methodology; and contributed to the writing and review process; Funding acquisition. Venkat Subramanian Ganesan: Participated in the conceptualization; investigation, and methodology development and was involved in the writing and review process and funding acquisition. Anushree Bose: Conceptualization; developed methodology; contributed to software programming; validated the research; and was involved in writing both the original draft and subsequent revisions. Sreeraj V S: Contributed to the conceptualization, helped develop methodology, and participated in the writing and review process. Umesh S: Engaged in the conceptualization, assisted in developing methodology, and contributed to the writing and review process. Manul Das: Helped develop methodology; and participated in the writing and review process. Justin Raj: Participated in the investigation, provided technical assistance, and contributed to the writing and review process.

Acknowledgments

This work was supported by the grant from the Department of Biotechnology (DBT) - Wellcome Trust India Alliance (IA/CRC/19/ 1/610005) to Venkatsubramanian Ganesan, Muralidharan Kesavan and Nishant Goyal. Umesh S, Subham Samantaray, Justin Raj and Manul Das acknowledge personal support from the Department of Biotechnology (DBT) - Wellcome Trust India Alliance (IA/CRC/19/ 1/610005). Anushree Bose acknowledges the support of the Department of Biotechnology (DBT)-Wellcome Trust India Alliance Early Career Fellowship grant (IA/CPHE/19/1/504591).

Sreeraj V S: acknowledges support of the India-Korea joint program cooperation of science and technology by the National Research Foundation (NRF) Korea (2020K1A3A1A68093469), the Ministry of Science and ICT (MSIT) Korea, and the Department of Biotechnology (India) (DBT/IC-12031(22)-ICD-DBT).

We also thank Dr Arun Sasidharan, Department of Neurophysiology, NIMHANS, for guidance on analysis and all the participants in our research.

Conflict of interest

The authors declare that the research was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest.

Narayanaswamy, J. C., Kumar, D., & Venkatasubramanian, G. (2019). Effect of fronto-temporal transcranial direct current stimulation on corollary discharge in schizophrenia: A randomized, double-blind, sham-controlled mediation analysis study. Schizophrenia Research, 204, 411–412. https://doi.org/10.1016/j.schres.2018.07.040Chella, F., D’Andrea, A., Basti, A., Pizzella, V., & Marzetti, L. (2017). Non-linear Analysis of Scalp EEG by Using Bispectra: The Effect of the Reference Choice. Frontiers in Neuroscience, 11, 262. https://doi.org/10.3389/fnins.2017.00262
Chella, F., Marzetti, L., Pizzella, V., Zappasodi, F., & Nolte, G. (2014). Third order spectral analysis robust to mixing artifacts for mapping cross-frequency interactions in EEG/MEG. NeuroImage, 91, 146–161. https://doi.org/10.1016/j.neuroimage.2013.12.064
Crapse, T. B., & Sommer, M. A. (2008). Corollary discharge across the animal kingdom. Nature Reviews. Neuroscience, 9(8), 587–600. https://doi.org/10.1038/nrn2457
Dong, L., Li, F., Liu, Q., Wen, X., Lai, Y., Xu, P., & Yao, D. (2017). MATLAB Toolboxes for Reference Electrode Standardization Technique (REST) of Scalp EEG. Frontiers in Neuroscience, 11, 601. https://doi.org/10.3389/fnins.2017.00601
Fonov, V., Evans, A. C., Botteron, K., Almli, C. R., McKinstry, R. C., & Collins, D. L. (2011). Unbiased average age-appropriate atlases for pediatric studies. NeuroImage, 54(1), 313–327. https://doi.org/10.1016/j.neuroimage.2010.07.033
Ford, J. M., Mathalon, D. H., Roach, B. J., Keedy, S. K., Reilly, J. L., Gershon, E. S., & Sweeney, J. A. (2013). Neurophysiological Evidence of Corollary Discharge Function During Vocalization in Psychotic Patients and Their Nonpsychotic First-Degree Relatives. Schizophrenia Bulletin, 39(6), 1272–1280. https://doi.org/10.1093/schbul/sbs129
Ford, J. M., Roach, B. J., & Mathalon, D. H. (2010). Assessing corollary discharge in humans using noninvasive neurophysiological methods. Nature Protocols, 5(6), Article 6. https://doi.org/10.1038/nprot.2010.67
Gevins, A. (2000). Neurophysiological Measures of Working Memory and Individual Differences in Cognitive Ability and Cognitive Style. Cerebral Cortex, 10(9), 829–839. https://doi.org/10.1093/cercor/10.9.829
Godey, B., Schwartz, D., de Graaf, J. B., Chauvel, P., & Liégeois-Chauvel, C. (2001). Neuromagnetic source localization of auditory evoked fields and intracerebral evoked potentials: A comparison of data in the same patients. Clinical Neurophysiology: Official Journal of the International Federation of Clinical Neurophysiology, 112(10), 1850–1859. https://doi.org/10.1016/s1388-2457(01)00636-8
Heinks‐Maldonado, T. H., Mathalon, D. H., Gray, M., & Ford, J. M. (2005). Fine‐tuning of auditory cortex during speech production. Psychophysiology, 42(2), 180–190. https://doi.org/10.1111/j.1469-8986.2005.00272.x
Heinks-Maldonado, T. H., Mathalon, D. H., Houde, J. F., Gray, M., Faustman, W. O., & Ford, J. M. (2007). Relationship of Imprecise Corollary Discharge in Schizophrenia to Auditory Hallucinations. Archives of General Psychiatry, 64(3), 286. https://doi.org/10.1001/archpsyc.64.3.286
Houde, J. F., Nagarajan, S. S., Sekihara, K., & Merzenich, M. M. (2002). Modulation of the auditory cortex during speech: An MEG study. Journal of Cognitive Neuroscience, 14(8), 1125–1138. https://doi.org/10.1162/089892902760807140
Joyce, C., & Rossion, B. (2005). The face-sensitive N170 and VPP components manifest the same brain processes: The effect of reference electrode site. Clinical Neurophysiology, 116(11), 2613–2631. https://doi.org/10.1016/j.clinph.2005.07.005
Lantz, G., Grave De Peralta, R., Spinelli, L., Seeck, M., & Michel, C. M. (2003). Epileptic source localization with high density EEG: How many electrodes are needed? Clinical Neurophysiology, 114(1), 63–69. https://doi.org/10.1016/S1388-2457(02)00337-1
Lehmann, D., Strik, W. K., Henggeler, B., Koenig, T., & Koukkou, M. (1998). Brain electric microstates and momentary conscious mind states as building blocks of spontaneous thinking: I. Visual imagery and abstract thoughts. International Journal of Psychophysiology: Official Journal of the International Organization of Psychophysiology, 29(1), 1–11. https://doi.org/10.1016/s0167-8760(97)00098-6
Liem, F., Zaehle, T., Burkhard, A., Jäncke, L., & Meyer, M. (2012). Cortical thickness of supratemporal plane predicts auditory N1 amplitude. NeuroReport, 23(17), 1026–1030. https://doi.org/10.1097/WNR.0b013e32835abc5c
Marzetti, L., Nolte, G., Perrucci, M. G., Romani, G. L., & Del Gratta, C. (2007). The use of standardized infinity reference in EEG coherency studies. NeuroImage, 36(1), 48–63. https://doi.org/10.1016/j.neuroimage.2007.02.034
Offner, F. F. (1950). The EEG as potential mapping: The value of the average monopolar reference. Electroencephalography and Clinical Neurophysiology, 2(2), 213–214. https://doi.org/10.1016/0013-4694(50)90040-x
Pascual-Marqui, R. D., & Lehamann, D. (1993). Topographic maps, source localization inference, and the reference electrode: Comments on a paper by Desmedt et al. Electroencephalography and Clinical Neurophysiology/Evoked Potentials Section, 88(6), 532–533. https://doi.org/10.1016/0168-5597(93)90043-O
Qin, Y., Xu, P., & Yao, D. (2010). A comparative study of different references for EEG default mode network: The use of the infinity reference. Clinical Neurophysiology: Official Journal of the International Federation of Clinical Neurophysiology, 121(12), 1981–1991. https://doi.org/10.1016/j.clinph.2010.03.056
Sperry, R. W. (1950). Neural basis of the spontaneous optokinetic response produced by visual inversion. Journal of Comparative and Physiological Psychology, 43(6), 482–489. https://doi.org/10.1037/h0055479
T, Z., L, J., & M, M. (2007). Electrical brain imaging evidences left auditory cortex involvement in speech and non-speech discrimination based on temporal features. Behavioral and Brain Functions : BBF, 3. https://doi.org/10.1186/1744-9081-3-63
Tadel, F., Bock, E., Niso, G., Mosher, J. C., Cousineau, M., Pantazis, D., Leahy, R. M., & Baillet, S. (2019). MEG/EEG Group Analysis With Brainstorm. Frontiers in Neuroscience, 13. https://www.frontiersin.org/articles/10.3389/fnins.2019.00076
Thatcher, R. W. (2012). Coherence, phase differences, phase shift, and phase lock in EEG/ERP analyses. Developmental Neuropsychology, 37(6), 476–496. https://doi.org/10.1080/87565641.2011.619241
Tian, Y., & Yao, D. (2013). Why do we need to use a zero reference? Reference influences on the ERPs of audiovisual effects. Psychophysiology, 50(12), 1282–1290. https://doi.org/10.1111/psyp.12130
Wang, Y., Huang, H., Yang, H., Xu, J., Mo, S., Lai, H., Wu, T., & Zhang, J. (2019). Influence of EEG References on N170 Component in Human Facial Recognition. Frontiers in Neuroscience, 13, 705. https://doi.org/10.3389/fnins.2019.00705
Yang, P., Fan, C., Wang, M., & Li, L. (2017). A Comparative Study of Average, Linked Mastoid, and REST References for ERP Components Acquired during fMRI. Frontiers in Neuroscience, 11, 247. https://doi.org/10.3389/fnins.2017.00247
Yao, D. (2001). A method to standardize a reference of scalp EEG recordings to a point at infinity. Physiological Measurement, 22(4), 693–711. https://doi.org/10.1088/0967-3334/22/4/305
Yao, D., Wang, L., Arendt-Nielsen, L., & N Chen, A. C. (2007). The effect of reference choices on the spatio-temporal analysis of brain evoked potentials: The use of infinite reference. Computers in Biology and Medicine, 37(11), 1529–1538. https://doi.org/10.1016/j.compbiomed.2007.02.002
Yvert, B., Crouzeix, A., Bertrand, O., Seither-Preisler, A., & Pantev, C. (2001). Multiple supratemporal sources of magnetic and electric auditory evoked middle latency components in humans. Cerebral Cortex (New York, N.Y.: 1991), 11(5), 411–423. https://doi.org/10.1093/cercor/11.5.411

No competing interests reported.

Download PDF

Reviews received at journal
03 Sep, 2024
Reviewers agreed at journal
10 Aug, 2024
Reviewers invited by journal
05 Aug, 2024
Editor assigned by journal
05 Aug, 2024
Submission checks completed at journal
05 Aug, 2024
First submitted to journal
03 Aug, 2024

You are reading this latest preprint version

Impact of EEG Reference Schemes on Event-Related Potential Outcomes: A Corollary Discharge Study Using a Talk/Listen Paradigm

Status:

Version 1

Abstract

Figures

INTRODUCTION

MATERIAL AND METHOD

RESULTS

DISCUSSION

CONCLUSION

Abbreviations

Declarations

References

Additional Declarations

Status:

Version 1