Atypical auditory cortical processing in ADHD during early and middle childhood

Abstract

Auditory perception emerges from the dynamic integration of bottom-up sensory encoding with higher-order predictive processes, and disruption of these processes is associated with poor attention and behavior regulation in children. Using an auditory oddball paradigm, we examined event-related potentials reflecting auditory deviance detection (mismatch negativity; MMN) and novelty-related orienting (novelty-P3) in early (n = 115) and middle (n = 71) childhood. Across both age-stratified cohorts, elevated attention-deficit/hyperactivity disorder (ADHD) symptoms or diagnosis were associated with altered MMN responses, whereas evidence for novelty-P3 differences was more limited. In early childhood, elevated ADHD symptoms were associated with larger negative-going MMN amplitudes and altered habituation to repeated standard and novel stimuli. In middle childhood, ADHD was associated with attenuated MMN amplitudes, while novelty-P3 group differences were sensitive to model specification. These results suggest that auditory deviance detection and novelty adaptation may be developmentally informative markers for understanding atypical attention regulation in ADHD-related phenotypes. Together, the findings support a systems-level account in which the integration of ascending sensory input with descending predictive signals supports adaptive responses to changing environments.

1. Introduction

Detection of change in the auditory environment is a fundamental aspect of human perception, enabling individuals to orient attention, evaluate novel or unexpected stimuli, and adapt behavior accordingly (Kral, 2013, Litovsky, 2015). Maturation of the auditory system occurs through a series of precisely timed, experience-dependent stages that form the foundational scaffold for subsequent language acquisition and higher cognitive functions (Chang and Kanold, 2021, Moore and Linthicum Jr., 2007). While thalamocortical afferents and local vertical cortical circuits are formed prenatally, postnatal synaptogenesis, pruning, and the development of corticocortical projections are governed by sensory-evoked activity (Kral, 2013, Moore and Linthicum Jr., 2007). A key developmental effect of experience is the transformation of an initially naive cortical network into one capable of categorization, thereby enabling the formation of auditory objects (i.e., categories) (Griffiths and Warren, 2004).

This process can be conceptualized within the predictive coding framework, which posits a hierarchical organization of the brain in which higher cortical areas generate predictions that are continuously compared with incoming sensory inputs (Friston, 2005). From this perspective, the emergence of auditory object representations reflects not only maturational changes in cortical networks but also the dynamic interplay between ascending sensory encoding and descending predictive mechanisms (Kral, 2013, Chang and Kanold, 2021, Marais and Roche-Labarbe, 2025, Teufel and Fletcher, 2020). Oscillatory dynamics—arising from endogenous rhythms, exogenous entrainment, and rhythmically guided attention—are fundamental to the emergence of auditory object representations (Morillon and Schroeder, 2015, Kayser et al., 2015, Henry et al., 2025, Pranjić et al., 2024). Disruption of these temporally coordinated processes due to genetic, environmental, or sensory deprivation factors can result in aberrant oscillatory circuit formation.

Unsurprisingly, auditory-perceptual and rhythmic timing differences are frequently reported across neurodevelopmental disorders (Lense et al., 2021, Noreika et al., 2013, Slater and Tate, 2018). Likewise, impaired regulation of behavioral attention and goal-directed behavior is a key symptom of several neurodevelopmental disorders, most predominantly attention-deficit/hyperactivity disorder (ADHD) (Paule et al., 2000). A recent meta-analysis confirmed that individuals with ADHD consistently exhibit more pronounced sensory atypicalities across multiple domains—including heightened sensitivity, sensory avoidance, low registration, and sensory seeking—compared to controls (Jurek et al., 2025). Investigation of neurocognitive mechanisms for atypical development of attentional processes is a critical step toward understanding typical development of the same.

Converging evidence indicates that individuals with ADHD have reduced integrity of neural circuits supporting higher-order attentional and broader executive control (Kaiser et al., 2020, Petrovic and Castellanos, 2016, Rao et al., 1997). Within this framework, the brain’s finite neural and cognitive resources engage attentive control to selectively prioritize relevant sensory inputs while suppressing competing, irrelevant signals (Friedman-Hill et al., 2010, Corbetta and Shulman, 2002). Therefore, when interference from irrelevant or salient stimuli exceeds cortical processing capacity, effective regulatory mechanisms are required to filter distractions and sustain goal-directed behavior (Tegelbeckers et al., 2016). Neuroimaging studies support this view, showing that individuals with ADHD exhibit reduced connectivity within prefrontal–parietal networks critical for attentive control, cognitive flexibility, and sensory gating (Albajara Sáenz et al., 2019, Kasper et al., 2012).

Less frequently acknowledged is the hypothesis that individuals with ADHD also show atypical early sensory and deviance-related signaling, reflected in altered activation of primary sensory cortices, atypical sensory-perceptual behaviors, and increased variability in auditory processing speed and timing (Cheng et al., 2016, Puyjarinet et al., 2017, Zhang et al., 2020). Thus, ADHD-related differences in auditory processing may reflect atypicalities at multiple processing stages, including early sensory encoding, automatic deviance detection, novelty-related orienting, and later attentional engagement. Furthermore, little is known about how these processes relate to cortical auditory processing in typically developing children from early to middle childhood, and how neuronal habituation shapes these dynamics. In this context, ADHD-related traits can be modeled dimensionally, as continuous variation in symptoms, or categorically, as clinically meaningful diagnostic or elevated-risk groupings, with each approach offering complementary developmental and clinical insight.

1.1. Cortical indices of auditory deviance detection and novelty-related orienting

At the neurophysiological level, detection of auditory change can be non-invasively measured, beginning with automatic deviance detection, and followed by novelty-related orienting and later attentional engagement (Fogarty et al., 2019). These stages are reliably captured using event-related potentials (ERPs), most commonly through the auditory oddball paradigm involving frequent “standard” tones interspersed with infrequent “deviant” tones (Squires et al., 1975). The passive auditory oddball paradigm is particularly well-suited for studying auditory change detection and attention orienting in young children, as it places minimal cognitive and motor demands on participants (Näätänen et al., 2007, Polich, 2007).

The mismatch negativity (MMN) component is considered a hallmark of automatic auditory change detection, reflecting the brain’s detection of deviations from predictable auditory patterns (Näätänen et al., 2007, Näätänen et al., 2005, Näätänen et al., 2011). The MMN peaks around 100–250 ms and is calculated by subtracting the average cortical response to standard from deviant trials, with neural generators localized primarily to the supratemporal auditory cortices (Li et al., 2019, Liebenthal et al., 2003). Some studies also support involvement of the prefrontal cortex in the generation of the MMN, suggesting that MMN should not be interpreted as a purely sensory or bottom-up response (Deouell, 2007, Lakatos et al., 2020). Rather, contemporary accounts conceptualize MMN as reflecting partially overlapping mechanisms, including stimulus-specific adaptation or habituation to repeated standards, memory-based comparison, and prediction-error signaling when incoming auditory input violates an established regularity (Winkler, 2007, May and Tiitinen, 2010, Ruiz-Martínez et al., 2022), underscoring that MMN should be interpreted as an index of both sensory adaptation and prediction-based auditory processing. Importantly, the MMN is present from infancy and demonstrates age-related increases in amplitude and decreases in latency, reflecting ongoing maturation of auditory cortex and white matter pathways (Bishop et al., 2011, Morr et al., 2002, Shafer et al., 2000). Meta-analytic and empirical studies provide evidence of reduced MMN amplitude in children with ADHD compared to neurotypical peers (Cheng et al., 2016); however, developmental patterns are less clear and suggest a mixture of both maturational lag and maturational deviance.

The N1 is another key auditory ERP component, which reflects early sensory and perceptual processing of auditory stimuli and is modulated by both stimulus properties and experience (Tomé et al., 2015, Hofmann-Shen et al., 2020). Developmentally, auditory ERPs in infants and young children are dominated by the P1 and N2 components, with the N1-P2 complex emerging in later childhood and adolescence, as demonstrated by increasing amplitudes and decreasing latencies (Tomé et al., 2015, Adibpour et al., 2020). The refinement of the N1 is associated with improved auditory discrimination abilities and parallels the maturation of thalamocortical and corticocortical auditory pathways (Moore and Linthicum Jr., 2007, Fitzroy et al., 2015, Mahajan and McArthur, 2012). While there is temporal and spatial overlap between the N1 and MMN, extensive evidence indicates that the MMN is not simply a byproduct of N1 adaptation but represents a distinct process involving memory-based comparison and prediction error signaling (Näätänen et al., 2011, Takasago et al., 2020).

In contrast, the P300 (or P3 complex) is a broad family of ERP components that are thought to index later cognitive and attentional processes and are modulated by contextual and task-related top-down factors (Escera and Corral, 2007, Ranganath and Rainer, 2003). Its subcomponents (e.g., P3a, novelty-P3, cue-P300) are elicited by a variety of tasks and sensory modalities, each reflecting somewhat distinct cognitive processes, and their amplitudes are typically reduced in individuals with ADHD relative to controls (Kaiser et al., 2020, Johnstone et al., 2013). The novelty-P3 is a later subcomponent of the P3 complex, reliably elicited in passive auditory oddball paradigms that incorporate a third stimulus category of complex, novel sounds, reflecting involuntary orienting, salience processing, and attentional capture (Barry et al., 2020, Watkins et al., 2007). Neural generators of the novelty-P3 are distributed across a fronto-central network, including the frontal cortex and anterior cingulate (Wronka et al., 2012).

Together, these ERP components capture partially dissociable stages of auditory change detection and novelty-related attentional engagement, offering valuable insight into typical and atypical development of perception, attention, and learning (Garrido et al., 2009). However, while mean ERP amplitudes index the aggregate strength of phase-locked neural responses to deviance detection and context updating, their functional significance also depends on the chronological evolution of discrete cortical responses with repeated stimulation. Habituation—a progressive reduction in response to repetitively presented stimuli—is a key marker of the brain’s ability to optimize sensory gain based on stimulus history (Merchie and Gomot, 2023, Thompson and Spencer, 1966). In neurotypical cohorts, repeated exposure to deviant stimuli leads to habituation of neural responses, evidenced by a progressive reduction in ERP amplitudes across trials (Hartkopf et al., 2019). While auditory habituation has not been well studied in ADHD, there is some evidence that habituation to visual stimuli is atypical in this clinical group (Jansiewicz et al., 2004, Massa and O’Desky, 2012). Examination of auditory temporal dynamics in children with and without ADHD may offer novel insight into mechanisms of typical and atypical attention regulation from early through middle childhood.

1.2. Current study

We employed a passive, three-stimulus auditory oddball paradigm to examine auditory deviance detection and novelty-related orienting, as indexed by the MMN and novelty-P3, respectively, in children with typical and atypical attention regulation. The sample included two independent, cross-sectional cohorts spanning early and middle childhood. Our primary analyses tested whether elevated ADHD symptoms or ADHD diagnosis were associated with mean MMN and novelty-P3 amplitudes within each cohort. Secondary analyses examined condition-specific habituation dynamics across repeated trials. Based on developmental models of attention network maturation, we hypothesized that elevated ADHD symptoms in early childhood would be associated primarily with atypical automatic deviance detection and early sensory processing, because frontal attention-control systems are still undergoing substantial maturation at this age, whereas auditory deviance detection mechanisms are already present but continue to be refined through experience-dependent development. In middle childhood, we expected children with ADHD would exhibit differences in both auditory deviance detection and novelty-related orienting, relative to neurotypical peers. We further hypothesized that atypical habituation dynamics would partially explain individual differences in auditory processing in both age groups.

2. Methods

2.1. Participants

2.1.1. Early childhood sample

One hundred twenty-two children ages 2 years, 6 months through 4 years, 11 months completed the auditory oddball paradigm. Participants were recruited from the greater Boston area via flyers, research registries, outreach to pediatric medical providers, and a paid research advertising service. Participants were over-sampled for having an immediate biological family member with a diagnosis of ADHD, which was confirmed through a clinical interview with a licensed clinical psychologist. Because ADHD is highly heritable and formal diagnosis is uncommon and less stable in early childhood, this family-history enrichment strategy was used to increase variability in ADHD-related liability in a very young, understudied age range (Wolraich et al., 2019). Importantly, family-history status was used only as a recruitment-enrichment strategy and was not used to define analytic groups. Exclusion criteria included full-scale IQ (FSIQ) < 80; prenatal exposure to alcohol, tobacco, or other non-prescribed substances; perinatal trauma; diagnosis of autism spectrum disorder; known genetic syndrome; seizure disorder or non-febrile seizure history; color blindness; or inability to abstain from psychotropic medication use (if applicable). Eligibility was confirmed via interview with a primary caregiver and cognitive testing at the research visit.

2.1.2. Middle childhood sample

Eighty-one children ages 7 years, 0 months through 11 years, 11 months completed the auditory oddball paradigm. Participants were recruited from the greater Seattle area via flyers, social media, research registries, and outreach to pediatric clinical providers. Participants were over-sampled for a diagnosis of ADHD, any presentation. Exclusion criteria included FSIQ < 80; prenatal exposure to alcohol, tobacco, or other non-prescribed substances; perinatal trauma; diagnosis of autism spectrum disorder; known genetic syndrome; seizure disorder or non-febrile seizure history; color blindness; or inability to abstain from psychotropic medication use (if applicable). Typically developing (TD) children did not have an immediate family member with ADHD, a criterion intended to minimize the likelihood of unrecognized ADHD-related liability or attention dysregulation in the smaller control group. Eligibility was confirmed via interview with a primary caregiver and cognitive testing at the research visit. Six participants were excluded from the analysis due to a lack of clinical consensus for an ADHD diagnosis.

2.2. Procedures

Study procedures were approved by the Institutional Review Boards at Boston Children's Hospital (early childhood sample) and the University of Washington (middle childhood sample). Caregivers provided informed consent and children provided assent prior to study activities. Eligible participants attended a single research visit at a hospital laboratory, which included approximately 45 min of resting EEG and ERP tasks, as well as approximately 90 min of neuropsychological testing. FSIQ scores were obtained using the Wechsler Preschool and Primary Scale of Intelligence, Fourth Edition (WPPSI-IV) (Wechsler, 2012) for the early childhood participants and the Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II) (Wechsler, 2011) for middle childhood participants. FSIQ is calculated as an age-standard score with M = 100, SD = 15. Because middle childhood groups differed in cognitive ability, FSIQ was included in primary models to evaluate whether ADHD-related ERP effects were present above and beyond this group difference. Given that IQ may also be clinically intertwined with ADHD-related developmental burden, parallel sensitivity models excluding FSIQ were conducted for major outcomes and are reported in the Supplementary Material. For children currently prescribed ADHD medication and for whom washout was medically appropriate, stimulant medications were withheld for 48 h prior to the research visit, whereas non-stimulant ADHD medications were withheld for a duration determined by the child’s prescribing physician.

2.3. ADHD categorization

2.3.1. Early childhood sample

Trained clinical research staff (i.e., a licensed clinical psychologist or clinical psychology trainee under their supervision) assigned global ADHD symptom ratings for each participant on two occasions. The first rating was assigned following a 30–60-minute, semi-structured clinical interview conducted with a primary caregiver. During this interview, caregivers were asked to report on their child’s attention, behavior, and emotion regulation at home and at daycare/preschool (when applicable), as well as the level of functioning and presence of formal and informal supports across settings. The second rating was assigned by the research clinician who completed neuropsychological testing during the in-person research visit. Child behaviors, such as activity level, compliance, persistence, and response to redirection during testing, were considered. Both ratings were made on a 1–5 scale, where 1 = ADHD diagnosis highly unlikely and 5 = ADHD diagnosis highly likely. Children who received a rating of 4 or 5 on either the post-interview or the post-visit rating were classified in the current analyses as having elevated ADHD symptoms, whereas those with ratings of 3 or lower were characterized as typically developing (i.e., low symptoms). ADHD diagnoses are typically not assigned until a child reaches school age; as such, we used the designation "elevated" for the younger sample. Thus, the early-childhood analytic groups were defined by comprehensive clinical ratings derived from caregiver interview and observed child behavior during the baseline EEG visit, not by family-history status alone. Further details on this categorization are reported in Arnett et al. (2024) (Arnett et al., 2024). Caregivers also completed standardized questionnaires normed for age and sex, including a parent report on the Behavior Assessment System for Children Third Edition (BASC−3) Preschool Version (Reynolds et al., 2011). The BASC−3 assesses children's behaviors on a 4-point Likert scale ranging from never to almost always. BASC−3 Attention Problems and Hyperactivity subscales are scored as age- and sex-standardized T scores (M = 50, SD = 10), with higher scores indicating greater levels of psychopathology.

Following the baseline visit, caregivers and teachers completed remote questionnaires at 6, 12, and 18 months, and caregivers completed a standardized diagnostic interview via video call at 18 months. These follow-up diagnostic data support the clinical relevance of the baseline symptom-based classification, as clinician ratings were significantly associated with later diagnostic status, Fisher’s exact p < .001. Because follow-up data were collected after EEG acquisition and were not available for 24.4% of participants, analytic groups were defined using clinician ratings described above. Two children (1.7%) were currently prescribed non-stimulant ADHD medication, and 11 children (9.6%) had at least one additional neurodevelopmental concern, including speech/language-related difficulties (n = 7), motor coordination difficulties (n = 3), and behavioral concerns (n = 2).

2.3.2. Middle childhood sample

ADHD diagnoses were confirmed by a licensed clinical psychologist on the study team, using information gleaned from clinical interview with a primary caregiver, standardized questionnaires, and direct observation of the child during the research visit. Other possible contributors to the symptoms were also assessed, including posttraumatic stress disorder, learning disorders, anxiety, depression, and underlying medical conditions. ADHD symptom severity was assessed using the caregiver-report Strengths and Weaknesses of ADHD Symptoms and Normal Behavior questionnaire (SWAN) (Swanson et al., 2012). This measure employs a balanced 7-point Likert format and positively-valenced questions to capture the full spectrum of abilities across the 18 DSM−5 ADHD symptoms (American Psychiatric Association, 2022). Caregivers rated their child’s ability to regulate attention, activity level, and impulses relative to same-age peers, considering behavior when off medication if applicable. Responses were numerically recoded from + 3 (“far below”) to –3 (“far above”) and averaged to yield continuous scores for inattentive (items 1–9) and hyperactive/impulsive symptoms (items 10–18). Higher (positive) values reflected greater ADHD symptom severity, consistent with prior publications (Pranjić et al., 2025a). Because BASC−3 and SWAN scores were derived from different instruments, age ranges, and scoring conventions, symptom scores were not treated as directly commensurable across cohorts. Among the 31 children prescribed ADHD medication, 28 were prescribed stimulants and 3 non-stimulants. Fifteen children (21.1%) had at least one additional neurodevelopmental concern, most commonly reading/writing-related difficulties (n = 8), speech/sensory-related difficulties (n = 4), oppositional/behavioral concerns (n = 2), or other/unspecified concerns (n = 2). Because medication subgroups were small, inferential analyses by medication status or class were not conducted.

2.4. EEG acquisition

The experiment was conducted in a sound-attenuated room. Participants were seated approximately 70 cm from a computer monitor and were instructed to sit still and attend to the videos. Continuous EEG was recorded using a 128-channel Magstim-EGI HydroCel Geodesic Sensor Net and digitized at a sampling rate of 1000 Hz. Data were acquired with Net Station Acquisition software version 5.4.2 for the early childhood sample and version 4.5.6 for the middle childhood sample, both integrated with a 400 series high-impedance amplifier (Magstim-EGI, Plymouth, MN). Although different acquisition systems were used across samples, all data were collected using the same processing parameters and task characteristics. During acquisition, EEG signals were referenced to the vertex electrode and analog filtered (0.1 Hz high-pass, 100 Hz elliptical low-pass). Electrode impedances were reduced to < 50 kΩ prior to recording. Participant behavior was continuously monitored via a camera mounted above the screen. Epochs of 1.5 s were flagged for exclusion if the child exhibited signs of inattention (e.g., gaze aversion or excessive blinking), and these segments were subsequently removed during offline preprocessing.

2.5. Auditory oddball paradigm

Participants watched a silent video while auditory stimuli were presented without any required behavioral response (Fig. 1). Auditory stimuli included randomly presented frequent tones (n = 320, 80%), infrequent tones (n = 40, 10%), and novel sounds (n = 40, 10%). Frequent and infrequent tones were sinusoidal stimuli of either 1000 or 1500 Hz, counterbalanced between subjects. Novel stimuli (e.g., chimes, cricket chirps, drums, whooshing noise) were identical to those used by Hudac et al. (2018) Tone stimuli had near-instantaneous onset envelopes, with rise times < 1 ms. Sounds were presented for 200 ms with random inter-onset intervals of 900–1100 ms. A total of 400 stimuli were presented in a pseudorandom order with the constraint that no two deviants were presented consecutively. The first four standard stimuli were excluded from the analysis to account for adjustment effects. Thus, analyses targeted a total of 316 frequent, 40 infrequent, and 40 novel tones. Sounds were presented at 65 dB at the child’s midline from speakers. Stimuli were displayed using E-Prime 3.0 software for the early childhood sample and 2.0 software for the middle childhood sample (Psychology Software Tools, Inc., Pittsburgh, PA), and the video subtended a visual angle of 14.6° by 28.9°.

Fig. 1.

Auditory oddball paradigm and EEG electrode layout. Participants watched a silent video while auditory stimuli were presented without any required behavioral response. Standard tones occurred on 80% of trials, deviant tones on 10% of trials, and novel environmental sounds on 10% of trials, with stimuli presented at 900–1000 ms intervals. The EEG electrode layout shows the frontal-midline electrode cluster comprising Fz and neighboring electrodes (Fz, E4, E5, E10, E12, E16, E18, and E19), highlighted in purple, used for MMN and novelty-P3 analyses.

2.6. EEG preprocessing

EEG data were processed using MATLAB R2019b with the EEGLAB v2021.1 extension and functions, and the Harvard Automated Processing Pipeline for Electroencephalography (HAPPE) version 3.3 (Gabard-Durnam et al., 2018). First, we applied a 60 Hz CleanLine noise filter with 120 Hz harmonic frequency reduction and then resampled the EEG data to 500 Hz. Data were bandpass filtered at 1–50 Hz, after which bad channels were detected and removed, and artifacts were subsequently identified and eliminated using HAPPE v3.3 wavelet thresholding (soft rule). This filter range was selected to reduce slow drift and high-frequency artifact in high-motion pediatric EEG data and is consistent with HAPPE-based preprocessing approaches for developmental samples. To reduce artifacts, eye electrodes and 14 rim channels were excluded from analyses. Data were then segmented into 1200 ms time windows from −300 to 900 ms around the stimulus onset, with a baseline correction at −200–0 ms. Bad channel data within segments were interpolated before segment rejection based on amplitude over or under ± 150, and similarity criteria. Finally, bad channels marked prior to wavelet thresholding were interpolated before average re-referencing at the end of preprocessing. Seven participants from the early childhood group and four from the middle childhood group were excluded from analyses due to an unsatisfactory signal quality caused by excessive motion artifacts, as determined using the HAPPE Processing Report metrics (e.g., if the percentage of good channels was below 50%) (Gabard-Durnam et al., 2018).

2.7. ERP components

Based on prior literature, expected fronto-midline topography, and group-blind inspection of waveforms collapsed across groups, we selected time windows and locations for the MMN and the novelty-P3 components. Considering maturational changes across development, we computed grand-averaged waveforms for the early and middle childhood groups separately to ensure that the selected time windows were appropriate across ages. Within each cohort, time-window selection was performed using pooled waveforms collapsed across diagnostic/symptom group to avoid selecting windows based on group differences. For the early childhood group, MMN amplitudes were computed from 100 to 250 ms and novelty-P3 from 200 to 400 ms. For the middle childhood group, MMN amplitudes were computed from 70 to 200 ms and novelty-P3 from 300 to 420 ms. MMN and novelty-P3 waveforms were computed for each participant as a difference wave (i.e., deviant – standard and novel – standard). The same time windows were used for habituation analyses, with standard, deviant, and novel stimuli condition amplitudes computed separately. Statistical analyses were then performed on the unweighted average waveform across the frontal-midline cluster comprising Fz and neighboring electrodes (Fz, E4, E5, E10, E12, E16, E18, E19) known to show the most pronounced MMN and novelty-P3 responses in children (Fig. 1) (Chang et al., 2021). In line with prior research, we used wide time windows to account for the individual variability in ERP latencies in children.

To further address potential analytic flexibility in time-window selection, we conducted group-blind, ROI-constrained temporal cluster-permutation sensitivity analyses within broad literature-guided latency ranges over the same frontocentral ROI used in the primary ERP models. Analyses were collapsed across group and tested whether difference waves differed from zero in the expected direction: negative clusters for MMN (deviant-minus-standard) and positive clusters for novelty-P3 (novel-minus-standard). Broader search windows were 80–300 ms for early-childhood MMN, 180–450 ms for early-childhood novelty-P3, 50–250 ms for middle-childhood MMN, and 250–500 ms for middle-childhood novelty-P3. Adjacent significant time points were grouped into temporal clusters using a cluster-forming threshold of p < .05, with cluster mass defined as the sum of absolute t-values. Cluster-level significance was evaluated using participant-level sign-flipping with 5000 permutations. Analyses were implemented in R following established nonparametric cluster-based approaches for EEG/MEG data (Meyer et al., 2021, Maris and Oostenveld, 2007). These analyses provided support for the selected component windows (see Figure S1).

2.8. Statistical analyses

Statistical tests were performed in R Studio version 4.4.2. Separate analyses were conducted for the early- and middle-childhood samples because the data were collected on slightly different EEG systems, which precluded the use of a single combined model and prevented direct statistical comparisons across age groups. In addition, the two cohorts differed in age range, ADHD characterization, family-history enrichment, and symptom instruments; therefore, analyses were designed to test cohort-specific associations rather than direct developmental continuity across cohorts. Primary analyses examined whether group status was associated with mean ERP component amplitudes within each cohort. We computed a series of multiple linear regression models to examine the effect of group (elevated ADHD/ADHD vs. TD) on ERP component amplitudes (MMN and novelty-P3). In each model, age and full-scale IQ (FSIQ) were included as covariates to control for maturational effects and established IQ differences. FSIQ was included because we aimed to evaluate whether ADHD-related ERP effects were present above and beyond this group difference. Parallel sensitivity models excluding FSIQ were also conducted and are reported in the Supplementary Material. For primary ERP amplitude analyses, Benjamini-Hochberg false-discovery-rate correction was applied within each cohort across the two primary group effects (MMN and novelty-P3). The group × age interaction term was included when main effects of both group and age were observed, but additive models were retained as the primary models for interpretability. The alpha level was fixed at p = .05 (two-tailed) for all data analyses. Effect sizes were estimated using partial eta squared (η²p) for each fixed effect.

Secondary analyses examined condition-specific habituation dynamics across trials. To examine habituation effects, we computed single-trial mean amplitudes for each condition within time windows corresponding to the MMN (standard and deviant trials) and novelty-P3 (standard and novel trials), and estimated individual slopes using trial number as the predictor. These slopes, reflecting each participant’s rate of habituation, were analyzed using linear mixed-effects models with random intercepts and slopes by participant. To assess potential nonlinearity, we compared the fits of linear and quadratic models using likelihood ratio tests and AIC values. Linear models provided better fits and were thus used to test a group × trial interaction, controlling for age and IQ. Real trial numbers were used to preserve the positions of valid trials after artifact rejection. All continuous predictors were z-scored to improve convergence, reduce multicollinearity, and stabilize variance estimates. FDR correction was applied separately within MMN-window and novelty-P3-window model families, across the corresponding standard/deviant or standard/novel group × trial interaction terms.

Additionally, for models that yielded nonsignificant or imprecise effects, we conducted Bayesian analyses using the BayesFactor package (v0.9.12) to quantify how compatible the data were with the presence or absence of an effect. For non-hierarchical ERP amplitude models, Bayes factors were computed by comparing the full model including group, age, and FSIQ with a reduced covariate-only model including age and FSIQ but omitting group. Thus, BF₁₀ quantified evidence that adding group improved model fit beyond covariates, whereas BF₀₁ quantified evidence against an incremental group effect after accounting for covariates. For hierarchical and multilevel models, we used brms (v2.21) with weakly informative priors (Normal[0, 1]) to ensure conservative estimation and model stability. All models were fit with four Markov chains (4000 iterations per chain, 2000 warm-up). Convergence was verified using R̂ < 1.01 and visual inspection of posterior traces. Model identification was supported by centering continuous predictors and using appropriate random-effects structures. Reduced models omitted the group-related term(s) of interest, and BF₀₁ values are reported in the direction of support for the reduced model. Thus, for both simple and multilevel models, the Bayes factors we report are expressed in the direction of support for the null.

3. Results

3.1. Sample demographics

The final sample included 115 early childhood participants (elevated ADHD symptoms = 64, low symptoms/TD = 51) and 71 middle childhood participants (ADHD = 53, TD = 18). Within each age group, the clinical groups did not significantly differ on age, gender, race, and ethnicity distributions; nor did they differ in the percentage of trials retained after artifact rejection (see Table 1); however, children with ADHD (middle childhood) had significantly lower full-scale IQ than their TD peers.

Table 1.

Demographic characteristics of the early and middle childhood samples.

	Early Childhood (N = 115)			Middle Childhood (N = 71)
Variable	Elevated ADHD	TD	Difference	ADHD	TD	Difference
N	64	51	−	53	18	−
Age, M ± SD Years Months	4.04 ± 0.71 48.48 ± 8.54	3.98 ± 0.75 47.76 ± 8.94	t(113) = 0.43, p = .661	9.67 ± 1.44	9.03 ± 1.12	t(69) = 1.70, p = .092
Sex, % Female Male	31.2% 68.8%	49% 51%	χ2(1) = 3.05, p = .080	20.8% 79.2%	27.8% 72.2%	χ2(1) = 0.08, p = .772
Race, %
Asian Black/African-American Multiracial Native American/Indian White Prefer not to answer	0% 1.6% 14.1% 0% 84.3% 0%	3.9% 3.9% 15.7% 0% 74.5% 2%	p = .336	0% 3.9% 33.3% 2% 58.8% 2%	0% 0% 11.8% 0% 88.2% 0%	p = .282
Ethnicity, %
Hispanic or Latino Not Hispanic or Latino Prefer not to answer	9.4% 89.1% 1.6%	11.8% 82.2% 0%	p = .869	15.7% 82.4% 2%	17.6% 82.4% 0%	p = .833
FSIQ, M ± SD	107.1 ± 10.53	110.47 ± 9.59	t(111) = 1.76, p = .080	108.23 ± 12.58	117.33 ± 10.63	t(69) = 2.75, p= .007
ADHD-IA, M ± SD (BASC−3 / SWAN)	59.41 ± 10.19	48.18 ± 9.97	t(113) = 5.92, p< .001	1.48 ± 0.89	−0.66 ± 0.68	t(69) = 9.32, p< .001
ADHD-HI, M ± SD (BASC−3 / SWAN)	63.81 ± 11.32	51.12 ± 8.71	t(113) = 6.59, p< .001	1.3 ± 0.75	−0.72 ± 0.81	t(69) = 9.71, p< .001
Trials retained, %	96.10%	98.05%	t(117) = 1.53, p = .128	97.02%	95.31%	t(79) = 0.56, p = .575

Note. ADHD = attention-deficit/hyperactivity disorder; BASC−3 = Behavior Assessment System for Children Third Edition, Preschool Version (reported for the early childhood sample); FSIQ = Full Scale IQ; HI = Hyperactivity/Impulsivity symptoms; IA = Inattention symptoms; M = mean; SD = standard deviation; SWAN = Strengths and Weaknesses of ADHD Symptoms and Normal Behavior (reported for the middle childhood sample); TD = typically developing controls. Higher BASC−3/SWAN ADHD scores indicate greater symptom severity. Because BASC−3 and SWAN scores were derived from different instruments, age ranges, and scoring conventions, symptom scores should be interpreted within cohort and are not directly comparable across cohorts. Statistically significant results at p < 0.05 are indicated in bold.

3.2. Early childhood

Auditory Deviance Detection and Novelty-Related Orienting. The primary model predicting MMN mean amplitude was statistically significant: F(3, 109) = 4.25, R²_adj = 0.08, p = .007 (Fig. 2). The elevated ADHD symptom group showed larger negative-going MMN amplitudes relative to the low symptom/TD group (B = −0.72, SE = 0.27, p = .010, pFDR =.021, η²p = 0.06), after controlling for age (p = .065) and FSIQ (p = .299). Sensitivity models excluding FSIQ yielded the same pattern of findings (Table S1). In contrast, the primary model predicting novelty-P3 amplitude was not statistically significant (F(3, 109) = 0.25, R²_adj = −0.02, p = .858), and group was not associated with novelty-P3 amplitude (B = −0.07, SE = 0.37, p = .850). A complementary Bayesian model comparison provided moderate evidence against an incremental group effect beyond covariates (BF₀₁ = 5.05).

Fig. 2.

Auditory deviance detection and novelty-related orienting in the early-childhood sample. Children with elevated ADHD symptoms are displayed in purple and low-symptom/TD children are displayed in gray. For all group comparisons, the TD/low-symptom group served as the reference group; therefore, negative group coefficients indicate more negative amplitudes in the elevated ADHD symptom group relative to TD children. ERP amplitudes were extracted from a frontal-midline cluster comprising Fz, E4, E5, E10, E12, E16, E18, and E19. For MMN, more negative values reflect larger negative-going MMN amplitudes. (a) ERP waveforms at frontal-midline channels representing stimulus-locked, grand-averaged data by condition. The MMN difference wave was computed as deviant minus standard and quantified from 100 to 250 ms. More negative MMN amplitudes reflect larger negative-going MMN responses. Topographies highlight fronto-central distributions. (b) The novelty-P3 difference wave was computed as novel minus standard and quantified from 200 to 400 ms, with fronto-central distributions. Group-level differences in MMN and novelty-P3 amplitudes are visualized using raincloud plots, which simultaneously depict individual datapoints, distributional density, and summary statistics. Panels (c) and (d) illustrate condition-specific single-trial amplitudes across trials. These analyses used time windows selected to correspond to the MMN and novelty-P3 components, respectively, but were conducted on standard, deviant, and novel trial amplitudes separately; difference waves were not used for the temporal-dynamics analyses. Shaded areas indicate the 95% confidence interval. Asterisks indicate statistical significance (p < .05) and reflect either a main effect of group or a group × trial interaction. Full model outputs are reported in the Supplementary Material.

Habituation Effects. The mixed effects model with N1 standard trial amplitudes showed a significant main effect of trial number (B = −0.14, SE = 0.06, p = .018), and a nominal group × trial interaction (B = 0.20, SE = 0.09, p = .026, pFDR =.051), such that TD children showed stronger habituation (i.e., amplitudes became less negative across trials) than children with elevated ADHD. For deviant N1 trials, group (B = 0.69, SE = 0.26, p = .010) and age (B = −0.37, SE = 0.12, p = .004) were significant predictors, indicating more negative amplitudes among older children and those with elevated ADHD symptoms. However, we found no credible evidence for habituation across trials (B = 0.01, SE = 0.15, p = .943).

Finally, the model revealed a significant group × trial interaction for novel trials in the P3 window (B = −0.52, SE = 0.21, p = .016, pFDR =.032), showing that TD children exhibited greater attentional habituation to novel stimuli than those with elevated ADHD. For standard P3 trials, we found no credible evidence of group or group × trial effects, and a Bayesian model comparison indicated strong evidence for the null model (BF₀₁ = 11.7). See Table S3 for full model specifications, including sensitivity models.

3.3. Middle childhood

Auditory Deviance Detection and Novelty-Related Orienting. The primary model explaining MMN mean amplitude was statistically significant: F(3, 67) = 4.35, R²_adj = 0.13, p = .007 (Fig. 3). TD children had larger negative-going MMN amplitudes relative to children with ADHD (B = −0.94, SE = 0.35, p = .010, pFDR =.019, η²p = 0.10), after controlling for age (p = .003) and FSIQ (p = .064). Sensitivity models excluding FSIQ yielded a similar pattern of findings (Table S2). In contrast, we found little credible evidence that any predictors were associated with novelty-P3 amplitude (F(3, 67) = 0.91, R²_adj = −0.00, p = .438). A complementary Bayesian analysis provided only weak evidence for the null model (BF₀₁ = 1.40).

Fig. 3.

Auditory deviance detection and novelty-related orienting in the middle-childhood sample. Children with ADHD are displayed in blue and TD children are displayed in gray. For all group comparisons, the ADHD group served as the reference group; therefore, negative group coefficients indicate lower/more negative amplitudes in TD children relative to children with ADHD. ERP amplitudes were extracted from a frontal-midline cluster comprising Fz, E4, E5, E10, E12, E16, E18, and E19. For MMN, more negative values reflect larger negative-going MMN amplitudes. (a) ERP waveforms at frontal-midline channels representing stimulus-locked, grand-averaged data by condition. The MMN difference wave was computed as deviant minus standard and quantified from 70 to 200 ms. More negative MMN amplitudes reflect larger negative-going MMN responses. Topographies highlight fronto-central distributions. (b) The novelty-P3 difference wave was computed as novel minus standard and quantified from 300 to 420 ms, with fronto-central distributions. Group-level differences in MMN and novelty-P3 amplitudes are visualized using raincloud plots, which simultaneously depict individual datapoints, distributional density, and summary statistics. Panels (c) and (d) illustrate condition-specific single-trial amplitudes across trials. These analyses used time windows selected to correspond to the MMN and novelty-P3 components, respectively, but were conducted on standard, deviant, and novel trial amplitudes separately; difference waves were not used for the temporal-dynamics analyses. Shaded areas indicate the 95% confidence interval. Asterisks indicate statistical significance (p < .05) and reflect a main effect of group. Full model outputs are reported in the Supplementary Material.

Habituation Effects. We found no credible evidence for main effects of group or group × trial interaction for either standard or deviant N1 trials (BF₀₁ = 10.8 and 2.2, respectively, supporting the null; Table S4). However, older age was associated with more negative amplitudes to deviant tones (B = −0.56, SE = 0.18, p = .002), suggesting increased neural sensitivity to auditory change with development.

Within the P3 time window, no credible group or group × trial effects were observed for standard or novel P3 trial amplitudes (BF₀₁ = 10.5 and 1.5, respectively, supporting the null). However, there was a main effect of trial number for novel stimulus trials (B = −0.54, SE = 0.17, p = .003), indicating habituation of the novelty-P3 response across repeated presentations in both groups. Because the Bayes factor for novel trials provided only weak evidence for the null (BF₀₁ = 1.5), we conducted an exploratory post-hoc model excluding FSIQ to better understand the group pattern observed in Fig. 3d. In this model, the TD group exhibited reduced mean P3 amplitude to novel stimuli relative to the ADHD group (B = −1.28, SE = 0.58, p = .033). Sensitivity models excluding FSIQ are reported in Supplementary Table S4.

4. Discussion

Auditory deviance detection and novelty-related orienting are key components of auditory perception and may provide insight into how children with ADHD process predictable versus unexpected sensory events. In pediatric ADHD, difficulties with attention and behavior regulation may arise from atypical early sensory encoding, altered automatic deviance detection, inefficient novelty-related orienting, or difficulty adapting neural responses to repeated stimulation. The present findings provide age-stratified, cross-sectional evidence that auditory ERP responses differ in children with elevated ADHD symptoms or ADHD diagnosis. These patterns suggest alterations in auditory deviance detection processing, while the early childhood habituation analyses additionally point to atypical temporal adaptation to repeated standard and novel stimuli. It is important to reiterate that the early- and middle-childhood samples were analyzed separately given differences in EEG acquisition systems, and the cohorts also differed in clinical ascertainment and ADHD characterization. This precluded a single combined model, and therefore any cross-sample differences should be considered descriptive rather than reflecting formally tested developmental contrasts.

In early childhood, children with elevated ADHD symptoms showed larger negative-going MMN amplitudes relative to low-symptom/TD children. The habituation analysis further showed atypical temporal adaptation to repeated standard and novel stimuli, however no credible habituation effect was observed for deviant trials, suggesting that differences in auditory deviance detection may be influenced by differential temporal dynamics in responses to repeated standards. An alternative, and potentially complementary, interpretation is that elevated ADHD symptoms in early childhood are associated with heightened sensory responsiveness or atypical deviance detection. Accordingly, the early-childhood MMN finding may reflect a combination of altered responses to repeated standard input and enhanced or dysregulated responses to deviant auditory events.

These findings are compatible with, but do not directly establish, broader neurodevelopmental accounts of atypical sensory gating, predictive processing, and experience-dependent cortical refinement. Synaptic pruning, an activity-dependent developmental process that occurs as a result of learning and sensory experience, is a plausible mechanism that may contribute to sensory gating and auditory change detection, as seen in both animal models and human studies of neurodevelopmental disorders (Faust et al., 2021, Chaudhury et al., 2016). Aberrant connectivity arising from disruptions in synaptic pruning could modulate the brain’s ability to adaptively suppress responses to repetitive, non-salient stimuli, potentially contributing to atypical temporal dynamics in responses to repeated auditory input (Blok et al., 2022). However, the current data do not provide a direct measure of synaptic pruning, therefore, these mechanisms remain interpretive frameworks rather than direct empirical conclusions.

In middle childhood, by contrast, the ADHD group showed attenuated (i.e., more positive) MMN relative to TD children, a pattern aligned with meta-analytic findings of reduced MMN in school-aged ADHD samples and with accounts of atypical auditory sensory memory and automatic deviance detection (Cheng et al., 2016). Unlike the early childhood group, this difference was not clearly associated with habituation effects for either standard- or deviant-trial amplitudes. The cross-sample comparison suggested typical development is characterized by age-related enhancement of MMN amplitude, consistent with maturation of sensory and higher cortical networks (Morr et al., 2002, Shafer et al., 2000). Notably, the directionality of MMN effects differed across cohorts: children with elevated ADHD symptoms showed larger negative-going MMN in early childhood, whereas children with ADHD showed attenuated MMN in middle childhood. However, this pattern should not be interpreted as a statistically demonstrated reversal across development. Rather, it raises the provisional hypothesis that ADHD-related auditory deviance responses may differ across developmental periods (Figure S2).

One possible explanation for this age-stratified pattern is that the relationship between MMN and overlapping early auditory components changes with maturation. The N1-P2 complex emerges and becomes more robust across later childhood and adolescence, whereas auditory ERPs in younger children are more strongly shaped by broader P1/N2 morphology (Tomé et al., 2015, Adibpour et al., 2020). Thus, apparent differences in MMN amplitude across early and middle childhood may partly reflect developmental changes in the morphology and maturation of overlapping auditory ERP components, including the N1. Within the middle childhood ADHD sample, MMN also increased linearly with age; thus, the group difference at this developmental stage could be consistent with a maturational lag in children with ADHD. However, given that MMN alterations have been reported in adults with ADHD (Hsieh et al., 2022), a pure maturational lag account is unlikely to be sufficient and instead may point to persistent atypicality in auditory deviance-related processing. Therefore, the precise mechanisms by which auditory deviance detection and novelty-related orienting contribute to behavior may vary over the lifetime, particularly with the development and integration of complementary and compensatory cortical networks.

Although the present study focuses on phase-locked ERP responses, event-related spectral perturbations may provide complementary information about non-phase-locked oscillatory dynamics, including theta-band engagement and alpha-band modulation of distractor suppression, which have been implicated in ADHD-related attentional regulation (Michelini et al., 2022, Arjona et al., 2023). Time-frequency analyses may, therefore, be especially useful for distinguishing early cognitive engagement from later regulatory processes that are not fully captured by mean ERP amplitudes.

Our findings complement neuroimaging and neurophysiological studies wherein children and adolescents with ADHD have exhibited atypical sensory processing and functional immaturity in sensory cortical networks (Jurek et al., 2025, Marcos-Vidal et al., 2018, Carmona et al., 2015). For example, functional connectivity analyses have shown increased local and reduced long-range connectivity within sensory and attentional networks among 7–17-year-old children with ADHD, suggesting a bias toward local sensory processing with weaker feed-forward integration into cognitive cortical streams (Marcos-Vidal et al., 2018, Carmona et al., 2015). Likewise, our observed neurophysiological differences are consistent with prior behavioral research demonstrating core atypicalities in auditory attention and temporal processing in children with poor attention regulation (Puyjarinet et al., 2017, Laffere et al., 2021). For instance, Laffere et al. (2021) showed that children with ADHD have difficulty accurately attending to and synchronizing neural activity with target sound streams, reflecting altered attentional engagement during active listening. Similarly, Puyjarinet et al. (2017) reported that both children and adults with ADHD struggle with basic temporal processing, such as moving to the beat and detecting rhythmic deviations, indicating deficits in internal timing mechanisms. Our ERP results, therefore, fit within a broader framework in which children with ADHD show altered coupling between early auditory encoding and later attentional selection.

For the novelty-P3, we found little credible evidence for group differences in mean difference-wave amplitude in either sample, and the complementary Bayesian analysis in middle childhood (BF₀₁ = 1.38) suggested that the data were only weakly more compatible with the null than with an effect, thus indicating that the evidence for a difference was limited. The most consistent novelty-related finding was in early childhood, where children with elevated ADHD symptoms showed altered habituation to novel stimuli. This pattern remained significant after FDR correction, suggesting that children with elevated ADHD symptoms differed in how neural responses to novel sounds changed across repeated presentations. This pattern is compatible with less efficient adaptation to novel auditory events, even though mean novelty-P3 amplitude was not significantly different between groups (Karakas, 2022). In middle childhood, novel-trial amplitudes decreased across repeated presentations across groups, suggesting that ADHD-related differences in novelty-response adaptation were stronger in the early-childhood cohort. Moreover, our exploratory sensitivity model excluding FSIQ yielded a nominal group effect for novel-trial amplitudes in middle childhood, indicating that evidence for group differences was limited and sensitive to covariate specification. Because FSIQ differed between groups and may reflect clinically meaningful developmental burden rather than a purely nuisance variable, we retained FSIQ-adjusted models as primary but report models without FSIQ as sensitivity analyses.

The absence of significant differences in habituation to standard P3 trials in both samples suggests that the group effect was specific to novel stimuli, consistent with the view that ADHD-related cortical inefficiency is most pronounced in the context of novelty detection and involuntary attentional capture (Tegelbeckers et al., 2016, Hilger et al., 2020, Arnett et al., 2023). In other words, this pattern suggests that ADHD-related habituation differences were not generalized across all repetitive auditory input. However, because habituation analyses were conducted on condition-specific amplitudes rather than MMN or novelty-P3 difference waves, these findings should be interpreted as changes in responses to specific stimulus classes rather than as direct evidence of habituation of the difference-wave components themselves. These findings are broadly consistent with predictive-processing accounts in which attention regulation depends on the weighting of expected and unexpected sensory input (Gonzalez-Gadea et al., 2015), (Tegelbeckers et al., 2015). Converging evidence from fMRI also points to altered expectancy and salience processing in ADHD, including greater reliance on sensory input and inadequate suppression of task-irrelevant novelty (Kaiser et al., 2020, Rao et al., 1997), (Gonzalez-Gadea et al., 2015). Such frameworks offer useful hypotheses for future studies but should be considered provisional in the context of the present findings.

4.1. Limitations and future directions

Several limitations should be noted. First, the cross-sectional design prevents us from tracking within-child changes in auditory deviance detection, novelty-related orienting, or temporal adaptation. Longitudinal studies are needed to determine whether the age-stratified patterns observed here reflect maturational lag, developmental deviance, compensatory processes, or cohort-specific differences. Second, the early-childhood group was defined by elevated ADHD-related symptoms, whereas the middle-childhood group was defined by confirmed ADHD diagnosis. However, available 18-month follow-up diagnostic data supported the clinical relevance of the early-childhood clinician ratings, suggesting that the elevated-symptom group captured meaningful ADHD-related liability. Additionally, the middle-childhood TD group was smaller than the ADHD group, which may have reduced precision of TD estimates and limited power to detect smaller group differences, particularly for novelty-P3 outcomes. Finally, although the present analyses used clinically interpretable categorical groupings, future longitudinal work using harmonized dimensional symptom measures across development will be important for determining whether auditory ERP differences track continuous ADHD-related traits, categorical clinical status, or both.

Together, these findings support the broader view that atypical attention regulation in ADHD-related phenotypes may involve altered processing of predictable and unexpected sensory input, rather than impairments in higher-order control alone. Although the present design cannot distinguish maturational lag from developmental deviance, it identifies auditory deviance detection and novelty adaptation as developmentally sensitive processes that warrant longitudinal study. Clarifying how these neural responses mature, persist, or reorganize across childhood may help refine developmental models of ADHD, identify early neurophysiological markers of atypical attention regulation, and inform future precision-medicine approaches (Pranjić et al., 2025b).

Author contributions

M.P. performed the analyses, interpreted the results, and drafted the manuscript. J.L. assisted with manuscript preparation. A.B.A. conceived and designed the study, supervised the project, contributed to data interpretation, and revised the manuscript. All authors reviewed and approved the final version of the manuscript.

CRediT authorship contribution statement

Anne B. Arnett: Writing – review & editing, Writing – original draft, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Data curation, Conceptualization. Marija Pranjic: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Formal analysis, Conceptualization. Jirapat Likitlersuang: Writing – review & editing, Writing – original draft.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material

Data availability

Analysis code and documentation are publicly available on OSF at https://osf.io/k8xjh (DOI: 10.17605/OSF.IO/K8XJH). De-identified EEG datasets are stored in a controlled-access OSF component and are available from the corresponding author upon reasonable request.