More than fulfilled expectations: An electrophysiological investigation of varying cause-effect relationships and schizotypal personality traits as related to the sense of agency

Abstract

The sense of agency (SoA) is central to human experience. The comparator model, contrasting sensory prediction and action feedback, is influential but limited in explaining SoA. We investigated mechanisms beyond the comparator model, focusing on the processing of unpredictable stimuli, perimotor components of SoA, and their relation to schizotypy.

ERPs were recorded from 18 healthy participants engaged in button-pressing tasks while perceiving tones with varying causal relationships with their actions. We investigated the processing of non-causally related tones, contrasted this to causally related tones, and examined perimotor correlates of subjective expectancy and experience of agency.

We confirmed N100 attenuation for self-generated stimuli but found similar effects for expectancy-dependent processing of random tones. SoA also correlated with perimotor ERP components, modulated by schizotypy.

Thus, neural processes preceding actions contribute to the formation of SoA and are associated with schizotypy. Unpredictable events also undergo sensory attenuation, implying additional mechanisms contributing to SoA.

Full text

Consciousness and Cognition 119 (2024) 103667
Available online 29 February 2024
1053-8100/© 2024 Published by Elsevier Inc.
More than fullled expectations: An electrophysiological
investigation of varying cause-effect relationships and schizotypal
personality traits as related to the sense of agency
Nena Luzi
a
, Maria Chiara Piani
b
,
c
, Daniela Hubl
b
, Thomas Koenig
b
,
*
a
Department of Clinical Psychology and Psychotherapy, Institute of Psychology, University of Bern, Fabrikstrasse 8, 3012 Bern, Switzerland
b
Translational Research Center, University Hospital of Psychiatry, University of Bern, Bolligenstrasse 111, 3000 Bern 60, Switzerland
c
Graduate School of Health Sciences, University of Bern, Mittelstrasse 43, 3012 Bern, Switzerland
ARTICLE INFO
Keywords:
N100
Forward model
Comparator model
Sensory attenuation
Sense of agency
Schizotypal personality traits
Readiness potential
ABSTRACT
The sense of agency (SoA) is central to human experience. The comparator model, contrasting
sensory prediction and action feedback, is inuential but limited in explaining SoA. We investi-
gated mechanisms beyond the comparator model, focusing on the processing of unpredictable
stimuli, perimotor components of SoA, and their relation to schizotypy.
ERPs were recorded from 18 healthy participants engaged in button-pressing tasks while
perceiving tones with varying causal relationships with their actions. We investigated the pro-
cessing of non-causally related tones, contrasted this to causally related tones, and examined
perimotor correlates of subjective expectancy and experience of agency.
We conrmed N100 attenuation for self-generated stimuli but found similar effects for
expectancy-dependent processing of random tones. SoA also correlated with perimotor ERP
components, modulated by schizotypy.
Thus, neural processes preceding actions contribute to the formation of SoA and are associated
with schizotypy. Unpredictable events also undergo sensory attenuation, implying additional
mechanisms contributing to SoA.
1. Introduction
1.1. Phenomenology of sense of agency
How we experience ourselves as actors of our actions, and their effects on the environment is an important aspect of the self. From a
phenomenological perspective, the self is seen as “an integrated part of our conscious life, which has an immediate experiential reality”
(Zahavi, 2003, p. 59). Therefore, the sense of agency (SoA), which refers to the experience of the self as an agent of an action, is a basic
pre-reective component of self-experience. To illustrate this, consider the act of ringing a doorbell at a friend’s house. It is unusual for
individuals to question their own role in pressing the doorbell and consequently setting off the ringing sound.
A two-step account of agency presented by Synofzik et al. (2008) distinguishes bottom-up and top-down contributions to the SoA.
* Corresponding author.
E-mail addresses: nena.luzi@bluewin.ch (N. Luzi), maria.piani@upd.unibe.ch (M.C. Piani), daniela.hubl@upd.unibe.ch (D. Hubl), thomas.
koenig@upd.unibe.ch (T. Koenig).
Contents lists available at ScienceDirect
Consciousness and Cognition
journal homepage: www.elsevier.com/locate/yccog
https://doi.org/10.1016/j.concog.2024.103667
Received 5 July 2023; Received in revised form 28 December 2023; Accepted 12 February 2024

Consciousness and Cognition 119 (2024) 103667
2
They argue that SoA consists of the feeling of agency and the judgment of agency. While the former is dened as a non-conceptual, low-
level feeling of being the agent of an action, the latter refers to an explicit judgment of being the agent (Synofzik et al., 2008). How
these two steps contribute to the SoA depends on the task and context (Synofzik et al., 2008). Synofzik et al. (2008) suggest that, in
daily life, we largely rely on the feeling of agency and do not need to make explicit judgments about our agency.
In line with this two-step account, Gallagher distinguishes between two contiguous parts of the SoA. Attribution of agency (also
called reective SoA) refers to the act of attributing a specic action to oneself. From a phenomenological perspective, the attribution or
explicit judgment of agency follows a more basic experience of the self—the experiential or pre-reective SoA (Gallagher & Zahavi,
2008). The latter part of agency is inherent in a so-called minimal self, which refers to the conscious experience of oneself as a direct
subject (Gallagher, 2000). Therefore, the pre-reective SoA does not entail a reective act of consciousness but results from an im-
mediate rst-person experience of being the cause of events. It is dened as “the pre-reective experience or sense that I am the author
of the action” (Gallagher & Zahavi, 2008, p. 161).
1.2. Forward models
One inuential approach to explaining the SoA is based on assumptions of the motor control system. Forward models, categorized
as internal models of the motor control system, propose that individuals employ motor commands or intentions to predict their sensory
consequences, thereby allowing for a distinction between self-generated events (own actions) and non-self-generated events (Wolpert,
1997; Wolpert et al., 1995). Thus, forward models represent causal relationships between actions and their predicted outcomes
(Wolpert, 1997). Before the execution of any motor action, a motor command is generated and an efference copy of the motor command
is produced (von Holst, 1954). This efference copy predicts the sensory outcomes of the action using a forward model mechanism (i.e.,
corollary discharge; Sperry, 1950). Some authors (e.g., Frith et al., 2000) employ the terms “efference copy” and “corollary discharge”
interchangeably. However, we assume that the corollary discharge of a sensory outcome belonging to a motor action occurs after an
efference copy of a motor plan is generated. By comparing the predicted sensory outcomes with the sensory feedback received, it is
possible to distinguish the sensory effects of own movements from those that are not self-caused. The sensory feedback of own
movements is called reafference (von Holst, 1954) and when this is congruent with a matching efference-copy, sensory effects of self-
movements are suppressed. In other words, through reafference processes, the sensory effects of self-movements are canceled out
(Wolpert, 1997). Additionally, forward models can provide information regarding the predicted outcomes that is used before sensory
feedback becomes available to maintain perceptual stability (Miall et al., 1993; Wolpert, 1997).
1.3. The comparator model
In SoA literature, these conclusions drawn from the motor control system are also included in the comparator model (Haggard,
2017; Moore, 2016). The experience of being the agent of a movement and the associated outcomes is elicited when there is a match
between the predicted sensory feedback of a motor action, generated by a forward model, and the actual sensory feedback generated
by the movement (Haggard, 2017). Hence, self-generated sensations can be accurately predicted, owing to the availability of an
efference copy for comparison (Frith et al., 2000). By contrast, externally generated sensations lack an efference copy, rendering such a
comparison unfeasible.
One well-known consequence of the successful prediction of sensory feedback is sensory attenuation. Sensory attenuation assumes
that sensory reafferences of correctly predicted sensory consequences are attenuated. There is evidence that sensory predictions in
healthy individuals lead to perceptual attenuation of self-produced stimulation. For example, research has demonstrated that par-
ticipants rate self-generated tactile stimuli as less tickly and intense than sensations elicited by similar external stimuli (Blakemore
et al., 1999). With the addition of a delay to self-generated stimuli, Blakemore et al. (1999) demonstrated that delayed sensory
feedback led to increased ticklishness. Therefore, they suggested that perceptual attenuation is due to a precise forward model and that
the amount of sensory attenuation depends on the deviation between sensory feedback and the prediction of the model.
1.4. Electrophysiology of sense of agency
An important part of the empirical support for these models of the SoA has been based on the study of sensory attenuation using
electrophysiological data. Solid evidence of this attenuated sensation of self-generated stimuli was found in N100 event-related po-
tentials (ERPs). In the auditory-speech domain, several studies have shown that the N100 for self-generated speech is attenuated
compared to playback of one’s own speech or altered or alien auditory feedback (Bühler et al., 2016; Heinks-Maldonado et al., 2005;
Whitford, 2019). Another important example is the non-speech motor-auditory domain. Parallel to the former experiments with
speech, it was found for non-verbal sounds that lower N100 amplitudes were elicited by self-generated sounds than by externally
produced sounds (Baess et al., 2008, 2011; Schafer & Marcus, 1973; Timm et al., 2014). According to ndings by Baess et al. (2011),
sensory attenuation persists even when random sounds are interspersed among self-generated sounds. Given the consistent correla-
tions between the subject’s causal role in events and the sensory attenuation observed in the perception of these events, sensory
attenuation can be considered a reliable indicator of biological processes that represent the subject’s SoA through forward models. It is
also important to note here that there is also evidence for electrophysiological correlates of agency that is not related to the sensory
feedback, such as alterations of the lateralized readiness potential (Ford et al., 2014, Vercillo et al., 2018).
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
3
1.5. Disruption of sense of agency
When it comes to the assessment of the experiential role of the SoA, the importance of these processes for our sense of self becomes
apparent because of the drastic consequences that occur when these mechanisms fail and the SoA is disrupted. Such effects are referred
to as ego disturbances, including symptoms such as depersonalization, alienation, passivity, thought insertion, thought withdrawal
and thought broadcasting. These symptoms are typically found in psychotic disorders. As explained by Moore (2016), specic
problems with sensorimotor prediction are present in schizophrenia. In patients with schizophrenia, an abnormal experience of agency
might be the neuropathological basis for positive symptoms (hallucinations and delusions). Passivity symptoms, which are included in
the positive symptom category, are an excellent example of what happens when SoA is impaired. Mellor (1970) reported the following
statement from a patient with passivity symptoms: “It is my hand and arm which move, and my ngers pick up the pen, but I don’t
control them. What they do is nothing to do with me” (p. 18). Returning to our doorbell example, a person with a distorted SoA may
argue that it was his or her nger moving but would not experience control over the movement and, thus, would not feel responsible
for the ringing.
Evidence for disrupted SoA in individuals with psychotic disorders has been reported by Blakemore et al. (2000). They reported on
patients with auditory hallucinations and passivity experiences who were able to tickle themselves. This indicated that the movement-
associated sensory feedback did not match the predictions in these patients. Therefore, the feedback was unexpected, and no sensory
attenuation occurred. Thus, the sensation was not perceived as internally generated but as coming from an external stimulus and
manifested itself as tickling. Physiologically, there is evidence of a lack of sensory attenuation and thus impaired SoA in patients with
schizophrenia. Concerning N100 attenuation as an indicator of the SoA, several studies reported that patients with schizophrenia and
schizoaffective disorder showed no attenuated N100 for self-generated auditory sensory feedback (Ford et al., 2001, 2014; Heinks-
Maldonado et al., 2007).
1.6. Challenges to existing models
While forward models provide good explanations for many empirical observations such as the relationship between sensory
attenuation, SoA, and important dimensions of ego disturbances, we believe that they may be insufcient for a fully functional account
of the mechanisms underlying SoA. First, patients with ego disturbances frequently have a false-positive SoA; that is, they experience
themselves as causing events when such causation is implausible (or even impossible). This is for example the case when a patient tells
that because they press the light switch, it starts snowing outside. Other examples of disturbance in the SoA outside the psychotic
symptoms can be found in patients with phantom limbs (for a review on phantom limbs see Ramachandran & Hirstein, 1998) or while
dreaming. In these cases, however, it is unclear how a comparator model can account for false-positive experiences of one’s own
agency (see also Carruthers, 2012 for such attempts). Second, if we consider Gallagher’s description of a minimal self, which is not
further reducible (Gallagher, 2000), and if the SoA is a fundamental constituent of this minimal self, this implies that individuals must
possess an understanding of themselves as agents before experiencing the consequences of their actions. Again, such an a priori un-
derstanding of SoA would not be explainable by a comparator model because it would precede, and therefore be independent of, the
eventual sensory consequences of the subject’s actions.
1.7. Current study
In this study, we investigated the features of the SoA that go beyond the comparator model using an electrophysiological approach.
In particular, with regard to the point that the comparator model cannot explain well that there are cases of false-positive SoA, we were
interested in the behavioral and electrophysiological correlates of subjective expectancy and experience of agency in the sensory
processing of events not caused by the subject. Regarding the point that there must be an a priori SoA we were interested in elec-
trophysiological correlates of subjective expectancy and experience of agency during moments when the subject was actively engaged
as an agent, that is, at movement onset. Therefore, we conducted an experiment in which the subjects spontaneously pressed a button
and perceived auditory stimuli that were in varying causal relationships to these button presses. Thus, using this design we were able to
manipulate the causal relationship between action and perception. We expected that this manipulation of the causal relationship
between action and perception would systematically alter subjects’ SoA. This would allow us to study the correlates of SoA a) in stimuli
not caused by the subject and b) at movement onset. In our analysis, we propose the following hypotheses:
The rst part of our work focused on the N100 to investigate whether our experimental manipulation of the SoA was successful, and
to conrm previous ndings. According to the literature, the N100 in healthy subjects is attenuated for self-generated auditory
feedback compared to passive listening to the same stimulus (Baess et al., 2008, 2011). Consequently, our primary hypothesis (H1) was
that self-generated sounds would elicit a smaller amplitude in auditory N100 than passive listening to sounds. Additionally, we hy-
pothesized that an increase in SoA would correspond to a decrease in the N100 amplitude, specically in the case of self-made sounds.
Regarding our second hypothesis (H2), which suggests that the SoA is already represented at the moment of action, there is
supporting evidence from electrophysiological studies. These studies have shown that the expectancy of a stimulus can modulate
premotor brain activity. In a visuomotor paradigm, Vercillo et al. (2018) observed a more negative late readiness potential (RP) where
motor action leads to a visual stimulus, compared with the RP preceding the same motor action alone. Similarly, Ford et al. (2014)
reported a larger lateralized RP preceding button presses that resulted in a tone compared with the lateralized RP preceding button
presses without an associated tone. Notably, they found a relationship between a larger lateralized RP preceding motor action and
greater suppression of the N100 amplitude. Based on these ndings, we assumed that the subjective agency rating and our
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
4
manipulations of the SoA would affect neural activity before and during motor action, rather than solely inuencing sensory feedback.
Our third hypothesis (H3) focused on the processing of sounds that were not caused by the subjects’ button presses. If the concept of
false-positive SoA existed, we anticipated nding correlations between the participants’ expectancy and their rating of agency, and the
processing of these randomly occurring sounds. Interestingly, in an audiomotor task, Moore et al. (2009) demonstrated how priming
with random sounds has an effect on SoA of subsequent sounds, when these are not triggered by the subject. The effect, however, is
dependent on the temporal correlation between the stimuli. This suggests a modulation of SoA though external cues, when internal
motor cues are not available. Thus, we hypothesized that the manipulation of the SoA and explicit subjective agency ratings were
related to the processing of randomly occurring stimuli.
Finally, SoA is a core element of self-experience and the self. Disturbances in SoA have been linked to ego disturbances and various
psychotic symptoms. Taking a dimensional approach to psychotic symptoms (Guloksuz & van Os, 2018), schizotypy is viewed as a
continuously distributed trait among the general population and high levels of schizotypy can be related to disorders within the
schizophrenia spectrum (Claridge & Beech, 1995; Nelson et al., 2013). Since our study focused on healthy individuals, we expected
(H4) variations in the levels of schizotypal personality traits and hypothesized that these traits would be related to the behavioral and
neurophysiological correlates of SoA, as investigated in H1–3.
2. Material and methods
2.1. Participants
Eighteen healthy individuals took part in the study (10 women, 8 men; age range =18–39 years, M =22.89, SD =4.68). Seventeen
undergraduate psychology students participated as part of their curriculum in the Faculty of Psychology at the University of Bern. One
participant volunteered to participate in the study. All the participants had normal or corrected-to-normal vision and spoke standard
German or Swiss German. Normal hearing was assessed using the Whispered Voice Test (Pirozzo et al., 2003). Individuals with
neurological or psychiatric diagnoses or with diagnosed rst-degree relatives were excluded. None of the participants had been in a
substance-induced state of intoxication for at least seven days before the recording session. The participants were not paid for their
participation, but study credits were granted to the undergraduate psychology students. Informed consent was obtained from all
participants before the start of the study. The study was conducted at the vonRoll University Center in Bern, was approved by the Ethics
Commission of the University of Bern and met all the requirements of the Declaration of Helsinki. Participants provided demographic
data and were asked to complete the German version of the Schizotypal Personality Questionnaire (SPQ-G; Klein et al., 1997).
2.2. Electrophysiological recordings
Electrical brain activity was measured using a 64-channel ActiCap system (Brain Products GmbH, Gilching, Germany; plus one
reference and one ground electrode) mounted in an electrode cap according to the international 10–20 system. Electroencephalog-
raphy (EEG) recordings were obtained using BrainAmp DC ampliers and the BrainVision Recorder (Brain Products GmbH, Gilching,
Germany). Impedances of the electrodes were kept below 20 kΩ, and FCz was used as the recording reference. The data was sampled
with 500 Hz. All EEG recordings were conducted in an electrically and acoustically shielded, dimly lit cabin. To preserve the EEG data
quality, the participants were instructed to move and blink as little as possible during the recording, placed their chin on a chin rest,
and looked at a xation cross at the center of the screen during the blocks.
2.2.1. Resting EEG
A resting EEG was recorded to optimize the preprocessing of the experimental EEG data, both with open eyes, closed eyes, and
during eye movements. The participants then performed a trial of the experimental task without EEG recording, before starting the
experimental protocol.
2.3. Experimental protocol
2.3.1. Control tasks
After practicing, the EEG recording was started, and the participants completed two control tasks. In the rst control task, par-
ticipants listened to sounds (beep-only; 0.3 s, 440 Hz (A), and 80 dB). The interval between these sounds varied randomly from 3 to 5 s.
In the second control task, the participants were instructed to press a button approximately every 3–5 s in a randomized manner. The
control tasks consisted of 30 trials each (30 sounds and 30 button presses).
2.3.2. Experimental task
In the experimental task, the participants were instructed to press a button with the index nger of their dominant hand every 3–5 s
in a randomized manner. No predictable pattern should occur within button presses. By pressing the button, participants could
eventually trigger sounds (self-made sound =0.3 s, A, 80 dB) with varying delay and probability. Identical sounds could also occur that
were not caused by participants (random sound =0.3 s, A, 80 dB). All sounds were presented in blocks of 10 button presses. Within each
block, the probability and delay of self-made sounds were approximately xed. The self-made sounds had a delay of either 200, 500, or
800 ms, with a random variance of up to 500 ms for each delay. The blockwise probabilities of the self-made sounds were 0.25, 0.50,
0.75, or 1.00. The probability of random sounds was inversely proportional to the block’s probability of the self-made sounds such that
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
5
the sum of the two probabilities was always one. The timing of the random sounds was adapted to match the participants’ behavior
based on the timing of the average interval of the last 10 button presses. To assess the subjective experience of SoA, participants were
instructed to promptly rate the question “Did you cause the sounds?” after every block on a 7-point Likert scale (1 =I do not agree at all.
to 7 =I fully agree.). After the rating, the participants could take a self-paced pause before the next block.
The combination of all delays and probabilities yielded twelve different block types. Each block type was presented three times,
yielding 36 blocks. The blocks were presented in random order. On average, the participants took approximately 2 h to complete the
entire experiment.
During the experimental task, the participants viewed a computer monitor (21.26 in, 103 cm viewing distance) with a xation
cross. The instructions, presentation of sounds, and collection of responses were implemented using Psychopy (Peirce et al., 2019).
See Fig. 1 for a visual representation of the task design.
2.3.3. Data preprocessing
The EEG data were preprocessed using the Brain Vision Analyzer 2.2 (Brain Products GmbH, Gilching, Germany). First, in the
resting EEG data, channels with poor or absent EEG signals were interpolated using spherical splines when necessary. After bandpass
ltering of the data (1–20 Hz, Notch Filter 50 Hz), we performed an Independent Component Analysis (ICA). The resulting ICA
components were visually inspected to identify eye movement and heartbeat artifacts and to construct a spatial lter matrix that
eliminated the identied components for each participant. Next, we applied this spatial lter to the temporally unltered task data of
each participant after interpolating bad channels. Finally, we manually identied and excluded any remaining artifacts. The data were
then re-referenced to the average reference and bandpass-ltered from 0.3 to 7 Hz, with an additional 50 Hz Notch Filter to suppress
line noise that may have leaked through the bandpass lter. Finally, continuous EEG recordings were segmented based on the markers
into epochs from −500 to 200 ms for motor-evoked potentials (MEPs) and from 0 to 500 ms for auditory-evoked potentials (AEPs).
2.4. Analysis of behavioral data
We analyzed the effects of delay, probability, and SPQ-G scores on subjective agency ratings with a linear mixed model using the R
package lme4 (Bates et al., 2015). We included delay, probability, SPQ-G scores, and associated interactions as xed-effect factors and
participants as random-effect factors.
2.5. Analysis of evoked potentials
An overview of the analysis strategies employed to test our different hypothesis is shown the Fig. S2 of the supplementary material.
2.5.1. Evoked potentials
Grand mean AEPs were computed separately for each sound type (beep-only, self-made sound, and random sound) and each
subject. For the button press, we calculated the MEP grand mean for each subject. The average included trials per subject were 27.9
(SD =1.8) for the beep-only condition, 217.2 (SD =12) for the self-made sound condition, 103.1 (SD =46.9) for the random sound
condition and 374.7 (SD =15.0) for the button presses. Finally, the grand mean MEP and AEP for each sound type across all subjects
were computed.
2.5.2. Microstate analysis
We performed microstate analyses to dene the time windows for the ERP components and thereby reduce the number of statistical
tests (Koenig & Pascual-Marqui, 2009). These analyses were conducted using the grand mean MEP and AEP across all subjects. Mi-
crostates are short time windows of quasi-stable topographic congurations (topographies) that are assumed to reect different
stimulus processing steps (Habermann et al., 2018). The identication of microstates as periods of quasi-stable topographies justies
averaging within the time window of a given microstate. Given the assumption that the source conguration remains highly stable
Fig. 1. Experimental Task Design. Note: s =seconds.
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
6
during this timeframe, it is reasonable to regard all data within this period as a single component and collapse them (Koenig & Pascual-
Marqui, 2009). Consequently, every microstate indicates spatially different neural activity.
An important parameter in the microstate analysis is the number of microstates. As our main focus was on robustly identifying the
perimotor MEP and early post-stimulus AEP components, we generated microstate models of the data using various numbers of
clusters. We then carefully selected the nal microstate model for the MEP and AEP, which consistently captured these components,
regardless of the addition or removal of a microstate. In other words, we ensured that the selected model remained highly similar to the
models obtained with one additional or one fewer microstate class.
2.5.3. Covariance analysis
To analyze the effects of our experimental manipulations and subjective agency ratings on event-related EEG potentials, we
conducted regression analyses at the single-trial level. Therefore, we created the predictor delay, probability, and subjective agency rating
for each trial and subject according to the experimental design and subject’s ratings. We then conducted a covariance analysis of the
single-trial MEPs and AEPs with these three predictors by computing the weighted averages of the single-trial data, where the
weighting of these averages stemmed from the predictors (Koenig et al., 2008). For each subject, this yielded the scalp elds of the
sources that linearly covaried with the given predictors. The obtained covariance maps were averaged over time for each microstate
component. Computing such covariance maps is mathematically equivalent to computing difference maps between category-wise
individual mean ERP map series, with the exception that the experimental variable is assumed to be a linear instead of a categori-
cal predictor of the variance in the ERPs and that, therefore, estimates of slopes instead of estimates of differences are obtained. For
visualization, specic ERP traces for the stratied predictors were computed and presented in the supplementary material (Fig. S3).
2.5.4. Null-hypothesis testing of the covariance maps using the Topographic Consistency test
Next, to establish the signicance of the above covariance analyses, we tested the across-subject mean covariance maps against the
null hypothesis that these mean maps would not be systematically different from an all-zero, “at” map. In order to avoid problems of
multiple testing across electrodes, we employed a technique known as Topographic Consistency Test (TCT) (Koenig & Melie-García,
2010). As a measure of spatial consistency across subjects, the TCT uses the Global Field Power (GFP) of the grand mean MEPs, AEPs,
and covariances. To statistically test the GFP of our experimental data, the TCT drew instances of GFP under the null hypothesis by
computing the GFP values after randomly shufing the channel assignments of the data per subject before averaging across subjects.
The signicance of the TCT was then dened as the percentage of cases in which the GFP under the null hypothesis was equal to or
larger than the GFP of the experimental data and yielded evidence of consistent activation of neural sources across subjects (Koenig &
Melie-García, 2010). For our data, TCTs were conducted for the components obtained from the microstate analysis and by drawing
5000 samples for the null hypothesis.
2.5.5. TANOVA in microstates
In addition to examining consistency across subjects, we were also interested in eventual topographic differences in ERP and
covariance maps as a function of experimental conditions and predictors. To test such map differences, we used a randomization
procedure called Topographic Analysis of Variance (TANOVA). In general, in TANOVA, condition-wise grand means are computed
across subjects and compared using a single overall index of topographic difference between the conditional mean maps (Koenig &
Pascual-Marqui, 2009). To test the index of the map difference for signicance, it was compared with the distribution of the same index
under the null hypothesis. To compute this, the TANOVA uses randomization statistics. Therefore, the topographies of the subjects
were randomly shufed between conditions and averaged across the subjects for each condition. The index of the map difference was
then recomputed using the randomized data. This randomization procedure was repeated 5000 times to obtain a proper estimate of the
null hypothesis distribution (Habermann et al., 2018). The signicance of the differences between the conditions was then estimated
by comparing the distribution of the index under the null hypotheses and the index obtained from the observed data. For our nal
analysis, we included SPQ-G as a between-subjects factor. In this case, TANOVA used the GFP of the covariance map for the SPQ-G
across subjects as an index of the size of the effect. For the randomization procedure, the subjects’ covariance maps and their SPQ-
G scores were randomly swapped, and GFP was computed for these data. This procedure was repeated 5000 times to estimate the
distribution of GFP under the null hypothesis. The observed GFP was then compared to the GFP distribution under the null hypothesis.
We conducted TANOVAs for MEPs, AEPs, and covariance maps within the components resulting from the microstate analysis, as
outlined below. We conducted these TANOVAs on the covariance data after normalizing all data to the unit GFP. This normalization
eliminated quantitative differences between the maps owing to scaling differences in the predictors (Habermann et al., 2018).
TCT and TANOVA analyses were computed in the MATLAB-based program Ragu (Koenig et al., 2011).
2.5.6. Testing of H1 (N100)
According to previous literature ndings on auditory N100 attenuation for self-generated sounds, we expected an attenuated N100
for self-made sounds compared to that for beep-only. Additionally, we anticipated that the covariance map of probability, delay, and
subjective agency rating with the AEP N100 would yield inverted maps compared with the N100. This is because an increasing
probability and a decreasing delay for self-made sounds should produce a stronger attenuation and therefore show an inverted N100 on
the covariance map. Similarly, concerning subjective agency rating, we anticipated an inverted covariance map for the N100
component. This expectation arose from the understanding that higher subjective ratings would indicate greater condence in the
experience of being the cause of the sounds.
To test the rst assumption, we conducted a one-factorial repeated measures TANOVA across subjects comparing the sound types,
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
7
beep-only and self-made sound, for the N100 mean AEP. To investigate the second assumption regarding whether the N100 AEP of the
self-made sound was inuenced by experimental manipulation and the subjective agency rating, we conducted across-subjects TCTs
for the covariance maps: First, we performed within-subject covariance analyses with the predictors probability, delay, and subjective
agency rating to extract the effect of these predictors on the single-trial N100 AEPs of the self-made sounds for each subject. We then
computed across-subjects TCTs to investigate whether the scalp eld maps representing the neurophysiological correlates of variations
in the predictors were stable across subjects.
2.5.7. Testing of H2 (Perimotor Component)
The second hypothesis predicted that our experimental manipulations (delay and probability) and the subjective experience of
agency would systematically inuence perimotor evoked potentials. Similar to the procedure fo H1, we rst performed within subject
covariance analyses with the predictors probability, delay, and subjective agency rating to extract the effect of these predictors on the
single-trial MEPs of the button presses. We when conducted a TCT to investigate whether the thereby obtained scalp eld maps
representing the neurophysiological correlates of variations in delay, probability, and subjective agency rating in the MEPs were
consistent across subjects.
Finally, to determine whether our predictors were associated with different regions of the brain, we conducted a one-factorial,
repeated measures TANOVAs across subjects of the covariance maps in the perimotor component, with the predictor type as a
within-subject factor.
2.5.8. Testing of H3 (Random Sound)
Our third hypothesis predicted that experimental manipulations and subjective agency ratings would systematically inuence the
N100 AEP of the random sounds. Because we investigated the N100 of the self-made sounds, we focused on the same time window for
random sounds. We conducted a TCT to test whether the correlates of the experimental manipulations were consistent across subjects
in the N100 AEP of the random sounds.
Furthermore, we wanted to examine whether different brain regions were associated with delay, probability, and subjective agency
judgment for random sounds compared with those for self-made sounds. Therefore, we conducted a two-factorial TANOVA of the
covariance maps with the two within-subject factors of predictor type (delay, probability, and subjective agency rating) and sound type
(random sounds and self-made sounds) for the N100 AEP.
2.5.9. Testing of H4 (Sense of agency and SPQ)
The fourth hypothesis predicted that schizotypal personality traits would be related to SoA. To investigate the neuronal correlates
of subjects’ SPQ-G scores on subjective agency ratings, we included the SPQ-G sum score as a between-subjects factor in the analysis of
the ERP components, where the previous hypotheses provided substantial evidence of SoA effects. Thus, we conducted a two-factorial
TANOVA with predictors (delay, probability, and subjective agency rating) as the within-subjects variable and the SPQ-G score as the
between-subjects variable.
Fig. 2. Subjective Sense of Agency Rating for the Different Experimental Manipulations. Note. Subjective rating scale: 1 =I do not agree at all to 7 =I fully
agree. Error bars show the standard deviation.
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
8
3. Results
3.1. Behavioral data
The SPQ-G scores ranged from 0 to 31 (Mean =10.72, Standard deviation =8.90, Median =9.50, see also Fig S2 of the supple-
mentary material). The linear mixed model that aimed to explain the variance in the subjective ratings of agency across blocks and
subjects by systematic inuences of the factors sound delay, sound probability and individual SPQ-G score revealed a signicant effect
of sound probability (β =5.97, SE =0.82, t =7.29, p <0.001, Cohen’s d =1.72). The xed effects of delay, SPQ-G, and their in-
teractions were not signicant (all p >0.41, all Cohen’s ds <0.2). The mean ratings as functions of probability and delay are shown in
Fig. 2.
3.2. Segmentation of the components
The selected MEP microstate model (Fig. 3) consisted of three components and yielded an early microstate from −482 to −64 ms
(premotor microstate), a second microstate from −64 to 64 ms (perimotor microstate), and a late microstate from 64 to 200 ms
(postmotor microstate). These components were very similar to the two and four microstate models; these models only added or
omitted components outside of the time periods of interest.
For the AEP, the selected microstate model (Fig. 4) included ve microstates: an early pre-N100 microstate from 0 to 49 ms, a
second microstate from 49 to 127 ms (N100 AEP), a third microstate from 127 to 281 ms (P200 AEP), a fourth microstate from 281 to
397 ms (early P300 AEP), and a late microstate from 397 to 500 ms (late P300 AEP). Alternative microstate models yielded similar
results for the rst four components.
3.3. H1: N100 conrmation
The conguration of the N100 AEP maps displayed central negativity and temporoparietal positivity for the self-made sounds and
beep-only (Fig. 5). Signicant differences between the two sound types (p =0.018) were indicated by TANOVA. The t-map contrasting
the self-made against random beep-evoked N100 showed a pattern of central positivity (t
max
at Fz =3.886) and temporoparietal
negativity (t
min
at TP7 =-3.664). These observations are jointly compatible with sources in the left and right auditory cortices
(N¨
a¨
at¨
anen & Picton, 1987) and indicate an attenuated N100 AEP, and thus assumingly less activity in the auditory cortices, for the self-
made sound compared with that for the beep-only.
Fig. 3. MEP Microstates. Note. Below, the GFP (y-axis) for the time windows of the MEP microstates in ms (x-axis) is shown. Above, the MEP maps
are shown for the three different time windows (blue: premotor from −482 to −64 ms; orange: perimotor from −64 to 64 ms; yellow: postmotor
from 64 to 200 ms). The background color of the maps corresponds to the coloring of the time windows. МV: microvolts; GFP =Global Field Power;
MEP =motor evoked potential; ms: milliseconds. (For interpretation of the references to color in this gure legend, the reader is referred to the web
version of this article.)
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
9
The TCT showed signicant covariations in the AEP of the self-made sounds and every predictor (all p <0.001). Maps with frontal
positivity for rating covariance and frontocentral positivity for probability covariance indicated an inverted N100 as the corresponding
predictor increased (Fig. 6). However, contrary to the effect obtained for increased probability and rating, for decreased delay the
covariance map showed lateralized frontal negativity and temporoparietal positivity (Fig. 7).
To further explore whether the covariance maps of the predictors differed, we performed a TANOVA. The covariance maps of the
three predictors differed signicantly (p =0.00020). To investigate this, pairwise TANOVAs were employed. As expected from Fig. 6,
we found that the covariance map of the predictor delay differed from that obtained for probability (p =0.0004) and for rating (p =
0.0002), whereas the latter two were not signicantly different (p =0.0908).
3.4. H2: Perimotor component
The TCT revealed a signicant covariation between delay (p =0.0022), probability (p =0.001), and the perimotor component
(Fig. 1). A marginal effect was observed for rating (p =0.067). Maps for probability and rating indicated stronger frontal negativity and
positivity in the temporal area with increasing values of predictors (Fig. 1) and were similar in conguration to the map of the per-
imotor microstate (Fig. 7). TANOVA yielded no signicant differences between the covariance maps of the distinct predictors (p =
0.3814).
3.5. H3: Random sound
The N100 of the random sounds showed a map conguration of central negativity and temporoparietal positivity (Fig. 8).
Supporting our assumption that the random sounds N100 were affected by our experimental manipulations, TCT revealed sig-
nicant covariations in the N100 AEP with delay (p =0.002) and probability (p =0.0152) (Fig. 9). For the subjective agency rating, the
TCT also yielded a signicant covariation with the N100 AEP of random sounds (p =0.0408) (Fig. 9).
To investigate whether different brain regions were connected to the three predictors for self-made sounds compared with those for
random sounds, we conducted a two-factorial TANOVA of the covariance maps, with sound type and predictor as within-subject
factors. The TANOVA yielded a signicant main effect of predictor (p =0.0002) and a signicant interaction between sound type
and predictor (p =0.0094). The main effect of the sound type (p =0.185) was not signicant. To further investigate the signicant
interaction effect of the sound type and predictor, paired TANOVAs were performed. A TANOVA comparing the covariance maps of
delay obtained in the self-made and random sounds yielded no signicant effect (p =0.2216). A comparison of the probability co-
variations in the same two sound types showed a marginally signicant difference (p =0.0542), with a pattern of occipital positivity
(t
max
=3.875 at O1) and left hemispheric frontal negativity (t
min
=-3.160 at F7) (Fig. 4). Comparing the covariance maps of the two
sound types with the subjective agency rating, TANOVA revealed a signicant difference (p =0.0274), with a t-map showing parietal
Fig. 4. AEP Microstates. Note. Below, the GFP (y-axis) for the time windows of the AEP microstates in ms (x-axis) is shown. Above, the AEP maps are
shown for the ve different time windows (blue: pre-N100 from 0 to 49 ms; orange: N100 from 49 to 127 ms; yellow: P200 from 127 to 281 ms;
purple: early P300 from 281 to 397 ms; green: late P300 from 397 to 500 ms). The background color of the maps corresponds to the coloring of the
time windows. МV: microvolts; AEP =auditory evoked potential; GFP =Global Field Power; ms: milliseconds. (For interpretation of the references
to color in this gure legend, the reader is referred to the web version of this article.)
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
10
and occipital positivity (t
max
=3.239 at PO10) and left hemispheric frontal negativity (t
min
=-4.333 at AF7) (Fig. 10).
3.6. H4: Sense of agency and SPQ-G
The criterion for further investigation of the relationship between the SPQ-G and SoA was that the effect of the subjective agency
Fig. 5. N100 ERP Mean Maps and Difference Map. Note. A: Mean maps of the N100 of the self-made sound and beep-only and the difference map
(Diff). The maps are shown with contour lines in steps of 0.5 µV. The p-value indicates the signicance of the difference map. ERP =event related
potential. B: Traces of the same ERPs at channel FCz, The suppression of the N100 in the self-made condition is clearly visible. µV: microvolts; ms:
milliseconds. The experimental condition explained 16.6 % of the individual variance between condition.
Fig. 6. N100 Covariance Sound t-Maps. Note. Covariance maps of the N100 of the self-made sound with the three predictors. The maps are shown as
electrodewise single-sample t-values with contour lines in steps of 1 t. P-values indicate the signicance of the TCT. TCT =Topographic Consistency
Test. The grand mean covariance map of the N100 explained 42.5 % (delay predictor), 40.7 % (probability predictor), and 48.6 % (Rating predictor)
of the variance of the individual covariance maps of the self-made sounds. µV: microvolts; AU: arbitrary units; p: p-value; prob: probability;
s: seconds.
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
11
rating or probability should be signicant in the relevant components. As mentioned in the section concerning the perimotor
component, the TCT showed signicant covariation between probability and delay with the MEP of the perimotor microstate. For the
subjective agency rating, covariation with the MEP of this microstate was marginally signicant. Therefore, we further explored the
relationship between the SPQ-G and SoA in this perimotor component using covariance maps of the three predictors as dependent
variables. The two-factorial TANOVA showed no signicant main effect of predictor (p =0.3722) and SPQ-G score (p =0.2366) but a
Fig. 7. Perimotor Covariance t-Maps. Note. Covariance maps of the perimotor component with the three predictors. The maps are shown as elec-
trodewise single-sample t-values with contour lines in steps of 1 t. P-values indicate the signicance of the TCT. TCT =Topographic Consistency
Test. The grand mean covariance map of the perimotor component explained 30.8 % (delay predictor), 31.9 % (probability predictor), and 28.6 %
(Rating predictor) of the variance of the individual covariance maps of the motor-evoked potential. µV: microvolts; AU: arbitrary units; p: p-value;
prob: probability; s: seconds.
Fig. 8. N100 ERP Sound Mean Map. Note. Mean map of the N100 for random sounds. The map is shown as electrodewise single-sample t-values with
contour lines in steps of 0.3 µV. ERP =event related potential. µV: microvolts.
Fig. 9. N100 Covariance Sound t-Maps. Note. Covariance maps of the N100 of random sounds with the three predictors. The maps are shown as
electrodewise single-sample t-values with contour lines in steps of 1 t. P-values indicate the signicance of the TCT. TCT =Topographic Consistency
Test. The grand mean covariance map of the N100 explained 32.9 % (delay predictor), 29.4 % (probability predictor), and 30 % (Rating predictor)
of the variance of the individual covariance maps of the random sounds. µV: microvolts; AU: arbitrary units; p: p-value; prob: probability; s: seconds.
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
12
Fig. 10. N100 Covariance Sound t-Maps and Difference Maps. Note. Covariance maps (left and center) and difference maps (right) of the N100 of the
self-made and random sounds with the two predictors probability and rating. The maps are shown as electrodewise single-sample t-values with
contour lines in steps of 1 t. P-values indicate the signicance of the difference maps. In the comparison of the probability covariance maps between
random and self-made sounds (upper row), the factor sound type explained 10.48 % of the variance. In the comparison of the rating covariance
maps between random and self-made sounds (lower row), the factor sound type explained 12.30 % of the variance. µV: microvolts; AU: arbitrary
units; Diff =difference map; p: p-value.
Fig. 11. Perimotor Covariance SPQ-G Rating t-Map. Note. Covariance map of the perimotor component with rating and SPQ-G. The map is shown as
electrodewise single-sample t-values with contour lines in steps of 0.5 t. P-values indicate the signicance of the TCT. The SQP-G score explained
10.24 % of the variance of the individual perimotor ERP component. µV: microvolts; AU: arbitrary units; p: p-value; TCT =Topographic Consis-
tency Test.
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
13
nearly signicant effect for the interaction between SPQ-G score and predictor (p =0.0512). To further examine the effect of the SPQ-G
on the covariation of each predictor with the MEP of the perimotor component, we performed TANOVAs for each predictor using the
SPQ-G as a between-subjects variable. There was no signicant main effect of the SPQ-G score in the TANOVA, including delay (p =
0.66840), a marginal main effect with probability (p =0.0924), and a nearly signicant main effect in the TANOVA including rating (p
=0.0508) (Fig. 11). The covariance map of the perimotor component with rating and SPQ-G obtained after the reversion of its polarity
was similar to the perimotor ERP microstate topography (Fig. 3) and the topography of the perimotor covariance maps of ratings
(Fig. 7) but lacked lateralization. TANOVA with sound type, predictor, and SPQ-G score of the N100 covariance maps yielded no
signicant effects (all p >0.17) that included the SPQ-G score. Therefore, no further post hoc tests were performed.
4. Discussion
We investigated whether there are aspects of SoA that go beyond the comparator model and explored how these aspects can be
measured using electrophysiological methods. Our ndings conrmed previous evidence supporting the comparator model but also
provided evidence for neurobiological representations of the SoA that are challenging to elucidate using the comparator model.
Specically, we identied an early motor-based neuronal correlate of pre-reective SoA and discovered that schizotypal personality
traits affect this correlation. Furthermore, our study showed that neural processing of environmental stimuli can be modulated by
manipulating the SoA.
Regarding the conrmatory aspect of our ndings, we rst observed that the behavioral data provided evidence supporting the
successful implementation of our experimental manipulations. Explicit agency rating could be predicted based on the probability of
self-generated stimuli. The delay in the self-made sound and the schizotypal personality traits of the participants did not predict
explicit agency rating. Regarding delay, this is in line with the conclusion of a recent review indicating that “delay does not impair the
perception of causal relations between one’s action and feedback” (Wen, 2019, p. 6). Therefore, by varying the probability, we were
able to systematically inuence the authorship experience. Contrary to our hypothesis 4, there was no signicant association between
the SPQ-R scores and the subjects’ agency ratings. While negative ndings typically allow for more than one logically possible
explanation, we speculate here that there may have been other and large sources of variance in the subjects’ mean ratings of agency
that obscured such a putative effect.
In the rst step of our analysis of the neurophysiological data, we aimed to replicate N100 suppression for self-generated sensory
stimuli. Our results successfully demonstrated a reduction in the N100 component of self-generated sounds compared with that of
randomly occurring sounds. These ndings are in line with those of previous research in the motor-auditory eld by Baess et al.,
(2008,2011) and Timm et al. (2014). Therefore, our ndings provide evidence for the forward model (Frith et al., 2000; Wolpert,
1997), showing that the predictions generated by the forward model when acting lead to a reduction in the neuronal response through
the predicted self-generated sound.
Similarly, the relationship between N100 and the probability of self-generated sounds indicated that a more predictable situation
was associated with a stronger attenuation of self-generated sounds. In other words, the covariation map of probability yielded an
inverted N100, similar to the difference map, indicating a sensory attenuation effect. As expected, this was reected in the electro-
physiological correlate of subjective experience. Specically, a similar map for the covariation of N100 with subjective agency ratings
showed an attenuated N100 of self-generated sounds with an increasing subjective experience of agency. This suggests that the N100
suppression effect is correlated with explicit SoA. This is contrary to the existing ndings of Kühn et al. (2011), which indicated that
agency judgment was not reected by N100.
One possible explanation for these inconsistent ndings could be different experimental designs. In our study, we asked partici-
pants to make ratings on their explicit SoA, while we systematically modulated a context that strongly affected their expectations of
agency. In contrast, the participants in Kühn et al. (2011) made agency judgments after each button press for ambiguous stimuli
presented in an ambiguous but constant context. This may have caused the participants (contrary to our case) to respond according to
their judgment rather than their feeling of agency.
Another difference that could have led to the difference in the ndings could be that Kühn et al. (2011) had a setup in which all
button presses were followed by a tone, and no tones occurred without the button press. Thus, their data were insensitive to the motor
contribution to the explicit SoA. Our ndings showed a similarity between the difference map of self-made and beep-only and the
covariation map of the N100 AEP with subjective agency ratings. Based on these observations, we believe that motor action
contributed to explicit SoA, as assessed in our experimental setup. In simpler terms, if we assume that the only difference between the
beep-only and self-made sounds is motor prediction and action, and the N100 attenuation reects this, our covariance mapping results
strongly suggest a contribution of motor preparation to the later explicit SoA.
Although we found an effect of delay in the N100, the spatial distribution of this effect was different from the distribution of the
other covariates, and it did not resemble the suppression of the N100 observed when comparing self-made sound with beep-only N100
maps. In addition, the delay did not affect SoA in the behavioral data; therefore, we assumed that the identied representation of the
delay in N100 was unrelated to SoA. Therefore, we do not discuss this nding further.
After conrming the previous ndings, we turn to the core topics of our study. We assumed that the SoA emerges before and during
the action, and should thus be apparent in the perimotor microstate component. Our results conrmed that subjective ratings of the
SoA were associated with neuronal activity in the perimotor component. This nding is important because the assumptions of the
comparator model suggest that the experience of agency arises from a comparison between the predicted state, involving predictions
about consequences and changes to the motor system, and the actual state (Moore, 2016).
In other words, the comparator model suggests an SoA that emerges post hoc. In contrast, our ndings showed that an a priori
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
14
neuronal correlate of explicit SoA was present in the EEG before and during motor activity. Our data also suggest that this effect can be
considered as an increase in RP, since the obtained covariance maps closely resemble the perimotor ERP microstate topography. The
current nding is consistent with evidence from previous studies on motor-auditory and visuomotor processes, such as the work by
Ford et al. (2014) and Vercillo et al. (2018). These studies demonstrated greater (lateralized) RP preceding actions that resulted in
feedback than before actions that were not followed by any stimuli. This led us to draw the following several conclusions. First, the pre-
reective SoA elicited in our paradigm is represented by the perimotor-evoked potential, indicating that it is not conned to the
processing of the stimulus, as outlined by Synofzik et al. (2008). Second, the explicit rating of agency, as collected in our study, must
have been related to pre-reective SoA. This nding is consistent with the two-step approach explained by Synofzik et al. (2008) and
the phenomenological perspective of the self. From this perspective, the pre-reective SoA is a basic component of the self that
contributes to the reective SoA (Gallagher, 2000; Gallagher & Zahavi, 2008). Third, our ndings support the notion that the func-
tional scope of the RP extends beyond motor planning and preparation. It appears to be sensitive to the participant’s expectations
regarding the outcomes or consequences of the motor act (Hughes & Waszak, 2011; for a review of recent literature on the RP see
Schurger et al., 2021). Lastly, our manipulation of the SoA by varying the probability of eliciting tones was represented in the peri-
motor component. Our interpretation of this nding was that the uctuating expectancy of the self-made sound inuenced the sub-
jective experience of the SoA even before the action took place, resulting in a more or less distorted SoA.
In addition, we investigated the experience of agency in externally generated events that were not predictable to participants. Thus,
we were interested in understanding how the manipulation of delay, probability, and subjective SoA inuenced the sensory processing
of random sounds. Our results demonstrated that alterations in expectations related to SoA were reected in the neural processing of
these random stimuli. The observed inverted topography of the N100, resulting from the covariation of random sound and probability,
indicated that there was greater sensory attenuation of random sounds when the probability of self-made sounds was high, and when
participants reported more agency overall. Therefore, we concluded that the attenuation of the N100 was not selective for self-
generated sounds when the stimuli were more predictable based on button presses. In addition to previous results of Baess et al.
(2011), which showed evidence for the selective attenuation of self-made sounds compared to computer-generated sounds occurring
under the same experimental conditions, our results suggest that the attenuation of random and self-made sounds was dependent on
the context. This was supported by the observation that the topographies of the contrast between the covariance maps of the self-made
and random sounds were rather orthogonal to the N100 topography (Fig. 4). Moreover, the covariance maps of self-made and random
sounds were consistently different, indicating that they originated from distinct neuronal sources. These ndings are in line with the
comparator model, which assumes that self-generated and therefore predictable stimuli are processed differently from unpredictable
stimuli from the environment (Haggard, 2017). However, contrary to the comparator model, we showed sensory attenuation for
random sounds, which suggests that no exact matching from the prediction and sensory feedback is necessary to attenuate sensory
stimuli.
For the delay manipulation, no effect on sensory attenuation occurred, as the topography of the covariance map with delay (Fig. 3)
was very different from that of an inverted N100 map. We suggest that this occurred because the delay had no detectable effect on the
occurrence of random sounds and thus did not affect the attenuation of the N100 of random sounds. Overall, the processing of random
events should be investigated in more detail when studying SoA.
Finally, we hypothesized that SoA would be related to schizotypal personality traits. Our results showed no effect of SPQ-G score on
N100 AEP. This nding is in line with the results of a previous study by Oestreich et al. (2015), which demonstrated no difference in
N100 attenuation between individuals with high or low schizotypal personality trait scores when motor-evoked tones were delayed by
50 ms or more.
Our ndings revealed that higher SPQ-G scores impacted the covariance maps of the ratings within the perimotor ERP microstate.
Interestingly, this effect exhibited an opposite pattern compared to that of the topography of the perimotor ERP microstate itself and
the topography of the covariance of this microstate ERP with rating.
We interpret this nding as a decrease in the effect of agency rating on the evoked potential in people with high SPQ-G scores. In
other words, this effect, which can be interpreted as a positive correlation between RP and SoA, suggests that the representation of SoA
in RP is less pronounced in individuals with high SPQ-G scores. Considering that SoA is an a priori experience and an important
component of the minimal self (Gallagher, 2000, 2012), we argue that this a priori and pre-reective sense contributes less to the
experience of authorship in people with higher schizotypal personality traits. Following the idea proposed by Ford et al. (2014)
regarding impaired efference copy mechanisms in individuals with disorders within the schizophrenia spectrum, we conclude that
individuals with higher levels of schizotypal traits exhibit reduced reliance on the efference copy in their explicit sense of agency.
Our study has some limitations. First, the sample size is small, which may affect the generalizability of our results to larger pop-
ulations. At the same time, our results showed internal consistency. The covariance maps of rating and probability closely resembled
the (independently obtained) contrast between self-caused sounds and the beep-only ERP (Figs. 5 and 6), and the within-subject effects
of probability and rating in the perimotor ERP component (Fig. 7) conrmed the between-subject effect of the SPQ-G in the same
component (Fig. 11). Therefore, it is both important and promising to validate our ndings by conducting further investigations with
larger sample sizes. We recommend implementing a preselection process to include individuals with either high or low schizotypal
personality trait scores. This approach allows for the exploration of SoA at both extremes of this dimensionally distributed personality
trait. Another limitation is the relatively young mean age of the sample. This could limit the generalizability of our ndings to other age
groups, as brain maturation processes may be incomplete in younger participants. However, it is important to note that schizophrenia
typically occurs within this age range and that clinical EEG studies in this eld are often carried out with younger samples. Therefore,
the relatively young age of our participants may have enabled the comparability of our results and strengthened the relevance of our
ndings.
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
15
When it comes to motor-evoked potentials, handedness is an important confounder. Our already small sample included 3 left-
handers, which makes it problematic to systematically assess or exclude the effects of handedness. Future studies should, therefore,
aim to design their study in a way that the effect of handedness can be systematically assessed.
Another limitation of our study is the fact that our analyses have not taken into account that the subjects were likely to update their
expectancies within each block. To account for this, one would have to introduce the within-block trial number and/or beep number as
an additional factor to our analysis. This would have entailed an even more complex set of analyses and results than what we already
had. In conjunction with the limited sample size and the novelty already introduced by our, to some degree, unconventional analysis
strategies, we thought this would overload the paper and refrained from conducting such further analyses.
5. Conclusions
In conclusion, our ndings suggest that the brain processing before and during an action contributes to the experience of agency.
This does not put into question the important and empirically well-supported feedback mechanisms contained in the comparator
model but suggests that relying solely on post hoc matching of predictions and actual states may not be sufcient to fully capture SoA.
Furthermore, our study highlights the signicance of expectancy and contextual factors in the processing of random events, which
offers a more comprehensive understanding of the complex nature of the authorship experience.Our ndings suggest the potential use
of schizotypal personality traits to investigate divergent agency experiences in the general population. Finally, we propose that our
experimental design will yield important additional ndings on the complex neurobiology of the SoA when applied to a population of
patients on the schizophrenia spectrum.
6. Funding statement
This research did not receive any specic grant from funding agencies in the public, commercial, or not-for-prot sectors.
CRediT authorship contribution statement
Nena Luzi: Conceptualization, Data curation, Writing-original draft, Visualization, Investigation, Formal Analysis, Writing-review
& editing, Visualization. Maria Chiara Piani: Writing-original draft, Investigation, Writing-review & editing, Visualization Daniela
Hubl: Writing-original draft, Supervision Thomas Koenig: Conceptualization, Writing-review & editing, Visualization, Validation,
Formal Analysis, Methodology, Supervision
Declaration of competing interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to
inuence the work reported in this paper.
Data availability
Data will be made available on request.
Appendix A. Supplementary material
Supplementary data to this article can be found online at https://doi.org/10.1016/j.concog.2024.103667.
References
Baess, P., Horv´
ath, J., Jacobsen, T., & Schr¨
oger, E. (2011). Selective suppression of self-initiated sounds in an auditory stream: An ERP study. Psychophysiology, 48(9),
1276–1283. https://doi.org/10.1111/j.1469-8986.2011.01196.x
Baess, P., Jacobsen, T., & Schr¨
oger, E. (2008). Suppression of the auditory N1 event-related potential component with unpredictable self-initiated tones: Evidence for
internal forward models with dynamic stimulation. International Journal of Psychophysiology, 70(2), 137–143. https://doi.org/10.1016/j.ijpsycho.2008.06.005
Bates, D., M¨
achler, M., Bolker, B. M., & Walker, S. C. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1–48. https://doi.
org/10.18637/jss.v067.i01
Blakemore, S.-J., Frith, C. D., & Wolpert, D. M. (1999). Spatio-temporal prediction modulates the perception of self-produced stimuli. Journal of Cognitive Neuroscience,
11(5), 551–559. https://doi.org/10.1162/089892999563607
Blakemore, S.-J., Smith, J., Steel, R., Johnstone, E. C., & Frith, C. D. (2000). The perception of self-produced sensory stimuli in patients with auditory hallucinations
and passivity experiences: Evidence for a breakdown in self-monitoring. Psychological Medicine, 30(5), 1131–1139. https://doi.org/10.1017/
S0033291799002676
Bühler, T., Kindler, J., Schneider, R. C., Strik, W., Dierks, T., Hubl, D., & Koenig, T. (2016). Disturbances of agency and ownership in schizophrenia: An auditory verbal
event related potentials study. Brain Topography, 29, 716–727. https://doi.org/10.1007/S10548-016-0495-1/FIGURES/4
Carruthers, G. (2012). The case for the comparator model as an explanation of the sense of agency and its breakdowns. Consciousness and Cognition, 21(1), 30–45.
https://doi.org/10.1016/j.concog.2010.08.005
N. Luzi et al.

Consciousness and Cognition 119 (2024) 103667
16
Claridge, G., & Beech, T. (1995). Fully and quasi-dimensional constructions of schizotypy. In A. Raine, T. Lencz, & S. A. Mednick (Eds.), Schizotypal Personality (pp.
192–216). Cambridge University Press. 10.1017/cbo9780511759031.010.
Ford, J. M., Mathalon, D. H., Heinks, T., Kalba, S., Faustman, W. O., & Roth, W. T. (2001). Neurophysiological evidence of corollary discharge dysfunction in
schizophrenia. The American Journal of Psychiatry, 158(12), 2069–2071. https://doi.org/10.1176/appi.ajp.158.12.2069
Ford, J. M., Palzes, V. A., Roach, B. J., & Mathalon, D. H. (2014). Did I do that? Abnormal predictive processes in schizophrenia when button pressing to deliver a tone.
Schizophrenia Bulletin, 40(4), 804–812. https://doi.org/10.1093/schbul/sbt072
Frith, C. D., Blakemore, S.-J., & Wolpert, D. M. (2000). Abnormalities in the awareness and control of action. Philosophical Transactions of the Royal Society of London.
Series B, Biological Sciences, 355(1404), 1771–1788. https://doi.org/10.1098/rstb.2000.0734
Gallagher, S. (2000). Philosophical conceptions of the self: Implications for cognitive science. Trends in Cognitive Sciences, 4(1), 14–21. https://doi.org/10.1016/
s1364-6613(99)01417-5
Gallagher, S. (2012). Multiple aspects in the sense of agency. New Ideas in Psychology, 30(1), 15–31. https://doi.org/10.1016/j.newideapsych.2010.03.003
Gallagher, S., & Zahavi, D. (2008). The phenomenological mind. Routledge.
Guloksuz, S., & van Os, J. (2018). The slow death of the concept of schizophrenia and the painful birth of the psychosis spectrum. Psychological Medicine, 48(2),
229–244. https://doi.org/10.1017/S0033291717001775
Habermann, M., Weusmann, D., Stein, M., & Koenig, T. (2018). A student’s guide to randomization statistics for multichannel event-related potentials using Ragu.
Frontiers in Neuroscience, 12. https://doi.org/10.3389/fnins.2018.00355
Haggard, P. (2017). Sense of agency in the human brain. Nature Reviews Neuroscience, 18(4), 197–208. https://doi.org/10.1038/nrn.2017.14
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), 286–296. https://doi.org/10.1001/archpsyc.64.3.286
Hughes, G., & Waszak, F. (2011). ERP correlates of action effect prediction and visual sensory attenuation in voluntary action. NeuroImage, 56(3), 1632–1640. https://
doi.org/10.1016/j.neuroimage.2011.02.057
Klein, C., Burghard, A., & Jahn, T. (1997). Erfassung der schizotypen Pers¨
onlichkeit nach DSM-III-R. Diagnostica, 43(4), 347–369. https://doi.org/10.1037/t10727-
000
Koenig, T., Kottlow, M., Stein, M., & Melie-García, L. (2011). Ragu: A free tool for the analysis of EEG and MEG event-related scalp eld data using global
randomization statistics. Computational Intelligence and Neuroscience, 2011, Article 938925. https://doi.org/10.1155/2011/938925
Koenig, T., & Melie-García, L. (2010). A method to determine the presence of averaged event-related elds using randomization tests. Brain Topography, 23(3),
233–242. https://doi.org/10.1007/s10548-010-0142-1
Koenig, T., Melie-García, L., Stein, M., Strik, W., & Lehmann, C. (2008). Establishing correlations of scalp eld maps with other experimental variables using
covariance analysis and resampling methods. Clinical Neurophysiology, 119(6), 1262–1270. https://doi.org/10.1016/j.clinph.2007.12.023
Koenig, T., & Pascual-Marqui, R. D. (2009). Multichannel frequency and time-frequency analysis. In C. M. Michel, T. Koenig, D. Brandeis, L. R. R. Gianotti, &
J. Wackermann (Eds.), Electrical Neuroimaging (pp. 145–168). Cambridge University Press. https://doi.org/10.1017/CBO9780511596889.008.
Kühn, S., Nenchev, I., Haggard, P., Brass, M., Gallinat, J., & Voss, M. (2011). Whodunnit? Electrophysiological correlates of agency judgements. PLoS One, 6(12),
Article e28657. 10.1371/journal.pone.0028657.
Mellor, C. S. (1970). First rank symptoms of schizophrenia. I. The frequency in schizophrenics on admission to hospital. II. Differences between individual rst rank
symptoms. The British Journal of Psychiatry, 117(536), 15–23. https://doi.org/10.1192/s0007125000192116
Miall, R. C., Weir, D. J., Wolpert, D. M., & Stein, J. F. (1993). Is the cerebellum a Smith Predictor? Journal of Motor Behavior, 25(3), 203–216. https://doi.org/10.1080/
00222895.1993.9942050
Moore, J. W., Wegner, D. M., & Haggard, P. (2009). Modulating the sense of agency with external cues. Consciousness and Cognition, 18(4), 1056–1064. https://doi.
org/10.1016/j.concog.2009.05.004
Moore, J. W. (2016). What is the sense of agency and why does it matter? Frontiers in Psychology, 7. Article, 1272. https://doi.org/10.3389/fpsyg.2016.01272
N¨
a¨
at¨
anen, R., & Picton, T. (1987 Jul). The N1 wave of the human electric and magnetic response to sound: A review and an analysis of the component structure.
Psychophysiology., 24(4), 375–425. https://doi.org/10.1111/j.1469-8986.1987.tb00311.x. PMID: 3615753.
Nelson, M. T., Seal, M. L., Pantelis, C., & Phillips, L. J. (2013). Evidence of a dimensional relationship between schizotypy and schizophrenia: A systematic review.
Neuroscience and Biobehavioral Reviews, 37(3), 317–327. https://doi.org/10.1016/j.neubiorev.2013.01.004
Oestreich, L. K. L., Mifsud, N. G., Ford, J. M., Roach, B. J., Mathalon, D. H., & Whitford, T. J. (2015). Cortical suppression to delayed self-initiated auditory stimuli in
schizotypy: Neurophysiological evidence for a continuum of psychosis. Clinical EEG and Neuroscience, 47(1), 1–8. https://doi.org/10.1177/1550059415581708
Peirce, J., Gray, J. R., Simpson, S., MacAskill, M., H¨
ochenberger, R., Sogo, H., Kastman, E., & Lindeløv, J. K. (2019). PsychoPy2: Experiments in behavior made easy.
Behavior Research Methods, 51(1), 195–203. https://doi.org/10.3758/s13428-018-01193-y
Pirozzo, S., Papinczak, T., & Glasziou, P. (2003). Whispered voice test for screening for hearing impairment in adults and children: Systematic review. BMJ (Clinical
Research Ed.), 327(7421), 1–5. https://doi.org/10.1136/bmj.327.7421.967
Ramachandran, V. S., & Hirstein, W. (1998). The perception of phantom limbs. The D. O. Hebb lecture. Brain, 121(9), 1603–1630. https://doi.org/10.1093/brain/
121.9.1603
Schafer, E. W. P., & Marcus, M. M. (1973). Self-stimulation alters human sensory brain responses. Science, 181(4095), 175–177. https://www.jstor.org/stable/
1736870.
Schurger, A., Hu, P., Pak, J., & Roskies, A. L. (2021). What is the readiness potential? Trends in Cognitive Sciences, 25(7), 558–570. https://doi.org/10.1016/j.
tics.2021.04.001
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
Synofzik, M., Vosgerau, G., & Newen, A. (2008). Beyond the comparator model: A multifactorial two-step account of agency. Consciousness and Cognition, 17(1),
219–239. https://doi.org/10.1016/j.concog.2007.03.010
Timm, J., SanMiguel, I., Keil, J., Schr¨
oger, E., & Sch¨
onwiesner, M. (2014). Motor intention determines sensory attenuation of brain responses to self-initiated sounds.
Journal of Cognitive Neuroscience, 26(7), 1481–1489. https://doi.org/10.1162/jocn_a_00552
Vercillo, T., O’Neil, S., & Jiang, F. (2018). Action–effect contingency modulates the readiness potential. NeuroImage, 183, 273–279. https://doi.org/10.1016/j.
neuroimage.2018.08.028
von Holst, E. (1954). Relations between the central nervous system and the peripheral organs. The British Journal of Animal Behaviour, 2(3), 89–94. https://doi.org/
10.1016/S0950-5601(54)80044-X
Wen, W. (2019). Does delay in feedback diminish sense of agency? A review. Consciousness and Cognition, 73, Article 102759. https://doi.org/10.1016/j.
concog.2019.05.007
Whitford, T. J. (2019). Speaking-induced suppression of the auditory cortex in humans and its relevance to schizophrenia. Biological Psychiatry: Cognitive Neuroscience
and Neuroimaging, 4(9), 791–804. https://doi.org/10.1016/j.bpsc.2019.05.011
Wolpert, D. M. (1997). Computational approaches to motor control. Trends in Cognitive Sciences, 1(6), 209–216. https://doi.org/10.1016/S1364-6613(97)01070-X
Wolpert, D. M., Ghahramani, Z., & Jordan, M. I. (1995). An internal model for sensorimotor integration. Science, 269(5232), 1880–1882. https://doi.org/10.1126/
science.7569931
Zahavi, D. (2003). Phenomenology of self. In T. Kircher, & A. David (Eds.), The Self in Neuroscience and Psychiatry (pp. 56–75). Cambridge University Press.
N. Luzi et al.