Age-related alterations in early and late auditory processing in children with attention-deficit/hyperactivity disorder: Evidence from event-related potentials Authors: Jipeng Huang, Shitao Huang, Yuanjun Kong, Manqi Zhou, Han Yang, Danping Hong, Yanling Zhou, Weizhen Yin, Jialiang Guo, Cai-ping Dang, Yan Song Published: 2026 ABSTRACT AbstractThis study investigated age-related differences in auditory detection and attentional processing in school-aged children with attention-deficit/hyperactivity disorder (ADHD) compared to typically developing (TD) peers, focusing on mismatch negativity (MMN), P3a, and P3b components. Participants were children aged 7–12 years with ADHD (N = 80; 12.0% female) and age-matched TD children (N = 80; 26.3% female). All children completed a three-stimulus auditory oddball task with unilateral presentation of standard, nontarget deviant, and target deviant tones while electroencephalography (EEG) was recorded. Compared with the steady increase in MMN amplitudes observed in TD children, children with ADHD showed reduced MMN amplitudes at ages 9–12 years and a U-shaped MMN age-related pattern. In TD children, more negative MMN amplitudes were associated with faster responses on hit trials. In contrast, children with ADHD showed reduced P3a amplitude at ages 7–8 years, and their P3b latency was significantly associated with behavioral performance. Our findings suggest that children with ADHD may rely more on late-stage cognitive processing to complete auditory attention tasks, whereas TD children benefit from more efficient early perceptual and involuntary attention mechanisms. Altered age-related differences across MMN, P3a, and P3b highlight disruptions in the development of auditory attention systems and may represent candidate neural indices that warrant further evaluation in longitudinal and intervention studies.Keywords: Auditory attention, MMN, P3, School-aged children, ADHD Introduction To exhibit successful adaptive behavior, one must be able to concentrate on information that is relevant in a given context while remaining responsive to perturbing events so as not to overlook potentially important information. In a classroom setting, it is crucial for children to be able to focus on the teacher's instructions amidst potential distractions such as whispers or bird noises. The ability to filter out irrelevant information is a key prerequisite for effective learning in school-aged children. Attention-deficit/hyperactivity disorder (ADHD) has become one of the most prevalent disorders diagnosed among U.S. children and adolescents aged 3–17 years (Bitsko et al., 2022) and is characterized by core symptoms of age-inappropriate inattention, hyperactivity, or impulsivity (American Psychiatric Association, 2013). Previous studies have shown that children with ADHD exhibit impaired auditory and visual attention across a range of attention tasks (Lin et al., 2017). The neural mechanisms underlying attention deficits in children with ADHD have been extensively studied in the visual domain (Wang et al., 2016). However, findings regarding auditory attention in children with ADHD remain inconsistent. Some researchers have suggested that children with ADHD may have deficits in auditory attention. For instance, Gomes et al. (2012) reported that children with ADHD showed reduced auditory responses at 100 ms (Ta component) and an absence of the Nd component compared with typically developing (TD) children, indicating impaired early attentional allocation. Fu et al. (2022) observed the absence of the N2ac component, an electrophysiological marker of lateralized auditory attention toward target sounds, in children with ADHD, suggesting deficits in auditory spatial attention. In contrast, Laffere et al. (2021) argued that children with ADHD may struggle with attentional engagement rather than attentional selection, as they showed similar levels of attentional modulation in phase locking and neural phase shifts. However, no published studies have explicitly examined the age-related pattern of auditory attention in children with ADHD. Previous research has shown that the development of brain structures and cognitive functions in children with ADHD differs from that in TD children (Janssen et al., 2017). Compared with TD children, children with ADHD exhibit significantly reduced gray and white matter volumes across widespread frontal, parietal, and occipital regions (Castellanos et al., 2002; Shaw et al., 2007). These interconnected cortical areas integrate information from lower-order sensory systems to generate higher-order percepts that guide attentional and behavioral control. Structural anomalies within this network have been implicated in the pathogenesis of ADHD (Sowell et al., 2003). In addition, previous studies have reported smaller volumes of the amygdala, nucleus accumbens, hippocampus, and caudate in individuals diagnosed with ADHD compared with healthy controls (Ellison-Wright et al., 2008; Frodl and Skokauskas, 2012; Hoogman et al., 2017). Dysfunction in these subcortical structures has been associated with hyperactivity symptoms (Frodl et al., 2010; Shigemune et al., 2010; Sonuga-Barke et al., 2010). In the auditory domain, one well-researched event-related potential (ERP) component is mismatch negativity (MMN), which is elicited when a deviant stimulus, such as a change in duration, frequency, or intensity, occasionally appears in a sequence of standard stimuli (Näätänen et al., 2007). MMN reflects the automatic detection of unpredictable auditory changes during auditory oddball or Go/No-Go tasks. Compared with healthy controls, findings on MMN amplitude in individuals with ADHD have been mixed, with studies reporting both reduced MMN amplitudes in children with ADHD (Rothenberger et al., 2000; Zhang et al., 2020) and comparable MMN amplitudes in children (Gomes et al., 2013), adolescents (Rydkjær et al., 2017), and adults (Le Sommer et al., 2023). Although few studies have explored the MMN age-related patterns in children with ADHD, previous research has revealed nonlinear developmental trajectories in children with risk factors for schizophrenia, in contrast to the linear trajectories observed in TD children (Laurens et al., 2020). Similarly, atypical MMN responses have also been reported in autism spectrum disorder (ASD), further suggesting that abnormal developmental trajectories of auditory change detection may be a transdiagnostic feature across neurodevelopmental conditions (Dunn et al., 2008; Sapey-Triomphe et al., 2025). This raises the possibility that children with ADHD may also show atypical age-related patterns in auditory detection, which could provide a potential explanation for the contradictory conclusions in the MMN literature. The design of the MMN paradigm allows for the division of the P3 into P3a and P3b subcomponents. In the three-stimulus paradigm, P3a is elicited by an infrequent distractor stimulus that is randomly inserted into the target or standard sequence, to which the participant does not need to respond. P3a is assumed to reflect an involuntary shift in attention toward task-irrelevant deviant sounds (Kushnerenko et al., 2002; Shestakova et al., 2003). P3b refers to the classical P3, which is elicited by a task-relevant infrequent stimulus (Polich and Criado, 2006). P3b is thought to be associated with the contextual updating of working memory (Kok, 2001), which reflects the attentional processes involved in updating the stimulus representation following P3a (Polich, 2007). Similarly, research findings for both P3a and P3b have yielded contradictory results, reporting both attenuated amplitudes (Johnstone et al., 2013; Zhang et al., 2020) and comparable amplitudes (Oja et al., 2016; Rydkjær et al., 2017) in children with ADHD relative to TD children. Consequently, it is important to investigate the age-related patterns in these neurophysiological markers, which may be closely related to the development of auditory attention in children with ADHD. In the present study, we employed a three-stimulus auditory oddball paradigm with monaural stimulation in which children were required to discriminate the side of the target deviant sound by pressing a button (Gamble and Woldorff, 2015). We aimed to investigate (1) whether attentional problems in children with ADHD are related to the MMN, P3a, or P3b components in school-aged children and, if so, (2) whether children with ADHD exhibit TD-like or atypical age-related patterns in the MMN, P3a, or P3b components. Methods Participants From January 2021 to September 2022, a total of 239 Chinese children were recruited for this study. All participants and their primary caregivers, typically mothers, underwent semi-structured diagnostic interviews conducted by a qualified psychiatrist via the Kiddie-SADS-Lifetime Version (K-SADS-PL; Kaufman et al., 1997) to preliminarily assess the presence of ADHD symptoms. Furthermore, children with ADHD met the full criteria of the DSM-5 (Diagnostic and Statistical Manual of Mental Disorders, 5th Edition), whereas TD children did not. The severity of symptoms was measured using parent-reported scores on the SNAP-IV (Swanson, Nolan, and Pelham, Version IV Rating Scale). All participants fulfilled the specified inclusion criteria stated for this study: (a) normal hearing; (b) normal or corrected-to-normal vision; (c) no comorbidities such as schizophrenia, mood disorders, autism spectrum disorders, or epilepsy; (d) no history of head trauma with loss of consciousness; (e) no history of organic diseases, neurological illness, or other severe diseases; (f) right-handedness; (g) a normal full-scale IQ (FSIQ >80 on the Chinese WISC-IV; Wechsler, 2003); and (h) being able to complete the EEG task. However, 48 children were excluded based on these criteria: (1) children with comorbidities (n = 10, 4 ADHD); (2) those with a total score of less than 80 on the Chinese WISC-IV full-scale IQ test or those who did not complete the test (n = 7, 5 ADHD); (3) children who were unable to control drowsiness during the EEG test or had difficulty distinguishing left from right, resulting in high EEG noise (n = 11, 9 ADHD); (4) those with less than 50.0% accuracy on the auditory task (n = 9, 6 ADHD); (5) those with fewer than 60 trials for unilateral deviant (target or nontarget) stimuli (n = 11, 7 ADHD). After examining the age distribution of the remaining 191 children, we eliminated individuals who were either older than 12 years or younger than 7 years. To achieve age matching, we utilized ERP data obtained from 160 participants. The participants were subsequently categorized into three distinct age groups (7–8 years, 9–10 years, and 11–12 years; see Table 1) based on school age. Furthermore, there were no significant sex differences (χ2 (5, n = 160) = 6.062, p = .295) across the six age groups (including the ADHD and TD groups). All children with ADHD were required to abstain from medication for at least 24 h prior to EEG recording, ensuring no acute pharmacological effects during testing. Within the ADHD group, 10 participants were medication-naïve, whereas 70 had a history of pharmacological treatment. Reported medications included methylphenidate, atomoxetine, and other commonly prescribed ADHD medications. Table 1Demographic data of the age groups (N = 160).ADHD groupTD groupχ2/tdp7–8 yearsN2422Age in years7.68 ± 0.587.55 ± 0.460.89440.381FSIQ111.83 ± 14.44112.05 ± 14.71−0.05440.961Sex (male: female)18:616:60.0310.8619–10 yearsN3432Age in years9.57 ± 0.589.40 ± 0.481.25640.217FSIQ110.32 ± 13.59113.88 ± 9.80−1.21640.230Sex (male: female)30:425:71.2110.27111–12 yearsN2226Age in years11.40 ± 0.6111.34 ± 0.430.35460.732FSIQ111.09 ± 11.98111.77 ± 14.89−0.17460.864Sex (male: female)20:218:8NANA0.084Note. Data are presented as the mean ± standard deviation (SD), unless otherwise indicated. ADHD = attention-deficit/hyperactivity disorder; TD = typical developing; FSIQ = full-scale intelligence quotient (standardized scores); NA = not applicable. The sex difference within the 11–12-year age group was confirmed via Fisher's exact test. Demographic data of the age groups (N = 160). Note. Data are presented as the mean ± standard deviation (SD), unless otherwise indicated. ADHD = attention-deficit/hyperactivity disorder; TD = typical developing; FSIQ = full-scale intelligence quotient (standardized scores); NA = not applicable. The sex difference within the 11–12-year age group was confirmed via Fisher's exact test. Written informed consent was obtained from all children and their parents/guardians in accordance with the Declaration of Helsinki, and this study was approved by the Ethics Committee of the Affiliated Brain Hospital of Guangzhou Medical University. Experimental stimuli and procedure An active three-stimulus auditory oddball task adapted from Gamble and Woldorff (2015) was implemented through E-Prime software (Version 2.0, Psychology Software Tools, Inc., USA). The task is illustrated in Fig. 1. Three pure-tone stimuli, with frequencies of 200 Hz, 600 Hz, and 1000 Hz and durations of 200 ms, were created using Praat software (Version 5.3.64, Paul Boersma and David Weenink, Phonetics Sciences Department, University of Amsterdam, Netherlands). The stimuli were presented at approximately 65 dB via headphones (HP GH10 Wired Gaming Headphones). Illustration of the stimulus sequence and an example scene from the experimental setup. The upper panel depicts the three stimulus types: target deviants (16.3%, 240 trials), nontarget deviants (16.3%, 240 trials), and standard tones (67.4%, 992 trials). The two deviant tones (200 and 1000 Hz) were counterbalanced across participants, and all stimuli were presented monaurally to either the left or right ear. The lower panel shows a segment of the trial sequence. Each trial lasted 1200 ms, consisting of a 200 ms tone followed by a 1000 ms interstimulus interval. Children were instructed to fixate on a central point and to press the “2” or “3” key with the right hand to indicate whether a target deviant was presented on the left or right side, respectively. All other stimuli were to be ignored. The task consisted of eight blocks with a total of 1472 trials. In each trial, a single stimulus lasted 200 ms, with an interstimulus interval (ISI) of 1000 ms, resulting in a total trial duration of 1200 ms, and was pseudorandomly presented monaurally via headphones (either the right or the left side). Therefore, the trials consisted of three types of stimuli: target deviants (16.3% probability, 240 trials divided equally between sides), nontarget deviants (16.3% probability, 240 trials divided equally between sides), and nontarget standards (67.4% frequency, 992 trials divided equally between sides). The two tones (200 Hz and 1000 Hz) were counterbalanced between subjects for trials with target and nontarget deviants. For trials with nontarget standards, the other tone (600 Hz) was used. Trials were presented in a pseudorandom order in accordance with the following criteria: (1) no more than four stimuli were presented consecutively on the same side; (2) no more than three identical stimuli were presented in succession; (3) nontarget standard stimuli were interspersed with target and nontarget deviants after every 1–3 nontarget standard stimuli; and (4) each block began and ended with nontarget standards. The participants were provided with a break of approximately 2 min between blocks. After practice, the task took approximately 45 min. Participants were instructed to sit comfortably approximately 70 cm away from the computer screen in a dimly lit, sound-attenuated, and electrically shielded room. Unlike the classical passive MMN paradigm, the present task was an active auditory oddball paradigm. The participants were instructed to accurately and quickly identify the side of the target deviants, using either the index or middle finger of their right hand to press either the “2” or “3” button on the keyboard, respectively, when the target deviants were presented on the left or right side. Additionally, the probability of target deviants and nontarget deviants was identical (16.3%). Therefore, the task required continuous engagement, and equal emphasis was placed on the accuracy and speed of the participants' responses. The behavioral indicators included the mean reaction time across all trials (RTAll), the mean reaction time for hit trials (RTH), the mean reaction time for false alarm trials (RTFA), the RTAll coefficient of variability (RTCV), the hit rate, and the false alarm rate. RTCV is the standard deviation of RTAll divided by the mean RTAll, reflecting the variability in response times. EEG acquisition and pre-processing EEG data were collected with a 64- or 128-channel network (HydroCel Geodesic Sensor Net; Electrical Geodesics, Eugene, OR) using Net Station EEG software. The impedance of all electrodes was kept under 50 kΩ during data acquisition. During recording, all electrodes were referenced to the Cz electrode, and data were sampled at a rate of 1000 Hz. The data were preprocessed offline using bespoke scripts implemented with EEGLAB functions (Delorme and Makeig, 2004) in MATLAB (The MathWorks, Inc., Natick, MA, USA). Twenty-nine TD children (21 boys, 8 girls) were originally recorded using the 128-channel sensor net, and only homologous 64-channel electrodes were retained for preprocessing and analysis. For all datasets, four facial electrodes (Nos. 61, 62, 63, and 64 in the 64-channel EGI system) were excluded (see Supplementary Fig. S1). The data were downsampled to 250 Hz and filtered using the default FIR filter in EEGLAB, with a bandpass range of 1–30 Hz. After electrodes with consistently poor signal quality were interpolated, the data were re-referenced to the average signal across all EEG electrodes. The continuous EEG data were then visually inspected to remove irregular amplifier saturation or movement artifacts. Independent component analysis (ICA) was conducted using the “runica” function with the extended infomax algorithm in EEGLAB (Jung et al., 2000; Lee et al., 1999). Prior to ICA decomposition, principal component analysis (PCA) reduction to 20 dimensions was applied as a preprocessing step to improve computational stability and reduce redundancy in the data, consistent with prior pediatric EEG preprocessing practices (Artoni et al., 2018; Lee et al., 1999). ICA components associated with ocular artifacts were subsequently identified and removed through visual inspection. The continuous time series were divided into six epoch datasets spanning from −200 to 800 ms for the analysis of MMN, P3a, or P3b. Additionally, trials were eliminated when the voltage difference between F9 or F10 (No. 17 or No. 1 in the 64-channel EGI system) exceeded ±50 μV at any point during the epoch, to eliminate horizontal eye movements from the data. Furthermore, baseline correction (−200 to 0 ms) was applied relative to stimulus onset. After preprocessing, no significant differences were found between the ADHD and TD groups in the remaining trial numbers of nontarget deviants (217 ± 30 vs. 220 ± 22, p = .498), nontarget standards (885 ± 128 vs. 900 ± 98, p = .430), or target deviants (194 ± 33 vs. 199 ± 24, p = .237). ERP analysis of MMN, P3a and P3b Prior to the statistical analyses, ERP component quantification procedures were determined based on inspection of grand-average waveforms and scalp topographies collapsed across all participants within the canonical time windows of the MMN, P3a, and P3b components (Kappenman and Luck, 2016). The same electrode clusters and quantification procedures were applied consistently across all groups and age bands. MMN and P3a components were calculated from the difference waveforms of nontarget minus standard stimuli, whereas the P3b component was calculated from the difference waveform of target minus standard stimuli. ERP amplitudes were quantified as mean amplitudes within a fixed ±70 ms window centered on the average peak latency of each component within each age subgroup of the ADHD and TD groups to accommodate age-related variability in component latency and morphology. For MMN, amplitudes were quantified within the 100–300 ms latency range across seven centroparietal electrodes (Cz, C1, C2, C3, C4, CP1, CP2; Fig. 2B). For P3a, amplitudes were quantified within the 250–450 ms latency range across eight frontocentral and centroparietal electrodes (FCz, FC1, FC2, Cz, C1, C2, CP1, CP2; Fig. 3B). For P3b, amplitudes were quantified within the 400–800 ms latency range across six posterior electrodes (Pz, P1, P2, POz, PO3, PO4; Fig. 4B). Peak latency for MMN and P3a was identified as the time point at which the difference waveform reached its maximum amplitude for each participant. The onset latency of the P3b component was defined as the time point at which the difference waveform reached 50% of its peak amplitude for each participant. (A) Grand-averaged MMN difference waveforms (nontarget deviant minus standard) at electrodes Cz, C1, C2, C3, C4, CP1, and CP2 for three age groups (7–8, 9–10, and 11–12 years) in children with ADHD and TD children. The shaded region indicates the 170–240 ms analysis window, which captures the MMN peak across age groups and diagnostic groups. (B) Topographical distributions of MMN amplitudes within the 170–240 ms window corresponding to the shaded area in Panel A. (C) Mean MMN amplitudes extracted from a ± 70 ms window centered on the grand-average peak latency for each age group in both diagnostic groups (red = ADHD; blue = TD). (D) Scatterplot showing the association between MMN amplitude and age. Standardized residuals (SR) of MMN amplitude were computed after statistically controlling for IQ and sex. *p < .05. (A) Grand-averaged P3a difference waveforms (nontarget deviant minus standard) at electrodes FCz, FC1, FC2, Cz, C1, C2, CP1, and CP2 for three age groups (7–8, 9–10, and 11–12 years) in children with ADHD and TD children. The shaded region indicates the 300–370 ms analysis window, which captures the P3a peak across age groups and diagnostic groups. (B) Topographical distributions of P3a amplitudes within the 300–370 ms window corresponding to the shaded area in Panel A. (C) Mean P3a amplitudes extracted from a ± 70 ms window centered on the grand-average peak latency for each age group in both diagnostic groups (red = ADHD; blue = TD). (D) Scatterplot showing the association between P3a amplitude and age. SRs of P3a amplitude were computed after statistically controlling for IQ and sex. *p < .05; **p < .01; ***p < .001. (A) Grand-averaged P3b difference waveforms (target deviant minus standard) at electrodes Pz, P1, P2, POz, PO3, and PO4 for three age groups (7–8, 9–10, and 11–12 years) in children with ADHD and TD children. The shaded region indicates the 400–800 ms analysis window, which captures the P3b peak across age groups and diagnostic groups. (B) Topographical distributions of P3b amplitudes within the 400–800 ms window corresponding to the shaded area in Panel A. (C) Mean P3b onset latencies, measured as the mean of latency at which each participant's difference waveform first reached 50% of the individual peak amplitude, for each age group in both diagnostic groups (red = ADHD; blue = TD). (D) Scatterplot showing the association between P3b onset latency and age. SRs of P3b onset latency were computed after statistically controlling for IQ and sex. *p < .05, ***p < .001. Statistical analysis Behavioral, symptom, and ERP data were analyzed using ANCOVAs. SNAP-IV scores were analyzed across age groups using separate one-way ANCOVAs. Behavioral and ERP measures were analyzed using 2 (group: ADHD, TD) × 3 (age: 7–8, 9–10, 11–12 years) ANCOVAs, with IQ and sex as covariates. Significant interactions were followed by Bonferroni-corrected simple-effects analyses to control for multiple comparisons. To assess age-related trends, correlation analyses between ERP measures and age (treated as a continuous variable) were performed separately for each group. To further characterize age-related patterns that were not adequately captured by linear associations, exploratory regression analyses comparing linear, quadratic, and cubic models were conducted to examine the observed age-related variation. Exploratory partial correlation analyses were also conducted to examine associations between ERP parameters and behavioral measures, controlling for age, IQ, and sex. To account for multiple comparisons, false discovery rate (FDR) correction was applied across all correlation analyses. Statistical analyses were performed using EEGLAB, MATLAB 2020b, and SPSS 26.0. The significance level was set at p = .05 (adjusted for multiple comparisons where applicable). Results Symptom and behavior Children with ADHD exhibited significantly more severe ADHD symptoms (ps < 0.001, see Supplementary Table S1) than TD children did. With respect to behavior, the main effect of group was significant. Compared with TD children, children with ADHD showed a greater RTCV (F (1,152) = 5.72, p = .018, ηp2 = 0.036) and a lower hit rate (F (1,152) = 5.09, p = .025, ηp2 = 0.032). There was also a significant main effect of age, with older age groups performing better than younger age groups on RTAll, RTH, RTCV and hit rate in both ADHD and TD groups (ps < 0.001; see Table 2). Table 2ANCOVAs of task performance.ADHD groupTD groupADHD vs TDFpηp2Age(years)①7–8②9–10③11–12①7–8②9–10③11–12RTAllMean715.73676.00647.00717.37716.28651.93Age×Groupn.s.(SD)(66.06)(79.22)(74.60)(55.21)(46.41)(67.01)n.s.Age12.68<0.001***0.143① > ③***, ② > ③**Groupn.s.RTHMean740.55696.07652.60732.96727.89654.96Age×Groupn.s.(SD)(74.63)(88.92)(79.70)(64.98)(48.59)(70.10)n.s.Age16.13<0.001***0.175① > ③***, ② > ③***Groupn.s.RTFAMean528.85489.69531.39548.88542.91563.25Age×Groupn.s.(SD)(93.50)(130.38)(173.26)(135.43)(132.13)(142.16)n.s.Agen.s.n.s.n.s.Groupn.s.RTCVMean0.270.270.240.260.240.23Age×Groupn.s.(SD)(0.05)(0.07)(0.04)(0.05)(0.04)(0.04)ADHD>TD*Age4.260.016*0.053① > ③*Group5.720.018*0.036Hit RateMean0.780.820.900.840.850.92Age×Groupn.s.(SD)(0.10)(0.14)(0.08)(0.08)(0.07)(0.05)ADHD ③***, ② > ③***Group5.090.025*0.032False RateMean0.0070.0140.0050.0050.0040.004Age×Groupn.s.(SD)(0.009)(0.038)(0.006)(0.005)(0.005)(0.005)n.s.Agen.s.n.s.n.s.Groupn.s.Note: *p < .05; **p < .01; ***p < .001; n.s. = not significant. P-values were adjusted using the Bonferroni–Holm method. RTAll = mean reaction time across all responses; RTH = mean reaction time for hits; RTFA = mean reaction time for false alarms; RTCV = coefficient of variation of RT; Hit Rate = proportion of targets correctly identified (hits / [hits + misses]); False Rate = proportion of nontargets incorrectly identified as targets (false alarms / [false alarms + correct rejections]); ADHD = attention-deficit/hyperactivity disorder; TD = typically developing. ANCOVAs of task performance. Note: *p < .05; **p < .01; ***p < .001; n.s. = not significant. P-values were adjusted using the Bonferroni–Holm method. RTAll = mean reaction time across all responses; RTH = mean reaction time for hits; RTFA = mean reaction time for false alarms; RTCV = coefficient of variation of RT; Hit Rate = proportion of targets correctly identified (hits / [hits + misses]); False Rate = proportion of nontargets incorrectly identified as targets (false alarms / [false alarms + correct rejections]); ADHD = attention-deficit/hyperactivity disorder; TD = typically developing. To further examine whether task engagement changed differently across groups over the course of the experiment, supplementary block-wise analyses of behavioral performance were conducted. Although mild block-related changes were observed for RTAll, RTH, and hit rate, no significant Group × Block interactions were found for these measures, suggesting that changes in behavioral performance across blocks were comparable between the ADHD and TD groups, with no evidence of greater decline in task engagement in children with ADHD. Detailed results are provided in the Supplementary Fig. S5. MMN elicited by nontarget-minus-standard sound For the MMN component (Fig. 2A; for original waveforms, see Supplementary Fig. S2), the analysis revealed a significant main effect of group (F (1,152) = 3.93, p = .049, ηp2 = 0.025) and a significant Group × Age interaction (F (2,152) = 3.07, p = .049, ηp2= 0.039) for amplitude, suggesting that the MMN amplitude in children with ADHD was significantly lower than that in TD children (−0.55 vs. –0.79 μV). Further simple effects analyses revealed that, compared with TD children, the MMN amplitudes were significantly smaller in children with ADHD in the 9–10-year-old group (−0.34 vs. –0.74 μV, p = .031) and 11–12-year-old group (−0.57 vs. –1.08 μV, p = .021), but not in the 7–8-year-old group (−0.73 vs. –0.54 μV, p = .380). On the other hand, TD children aged 7–8 years exhibited significantly smaller MMN amplitudes than those aged 11–12 years (−0.54 vs. –1.08 μV, p = .031; see Fig. 2C). Regarding the correlation analyses, a significant correlation between MMN amplitude and age was observed in TD children (r = −0.26, p = .032) but not in children with ADHD (r = 0.04, p = .756), indicating that MMN amplitudes increased gradually with age in TD children. Further regression analysis revealed that, in children with ADHD, MMN amplitude exhibited a U-shaped age-related pattern across the continuous age distribution (R2 = 0.08, adjusted R2 = 0.05, p = .015; Fig. 2D), with an inflection point at 9.73 years of age (95% CI [8.71, 10.75]). The quadratic coefficient of age was significantly negative (β = −0.08, 95% CI [−0.144, −0.016]). The specific regression results of each model are presented in Supplementary Table S2. The fitting results of the linear and quadratic terms of the MMN amplitude with age in children with ADHD are presented in Supplementary Fig. S3. We also conducted several sensitivity analyses to evaluate the robustness of the quadratic MMN pattern. Together with the model comparison results, these analyses support the quadratic model as the preferred and most parsimonious description of the observed age-related MMN pattern (Supplementary Fig. S4). For MMN latency, neither the main effects of age (F (2,152) = 0.30, p = .738, ηp2 = 0.004) nor group (F (1,152) = 0.45, p = .502, ηp2 = 0.003), nor the interaction effect (F (2,152) = 2.65, p = .074, ηp2 = 0.034) were significant. P3a elicited by nontarget-minus-standard sound For the P3a component (Fig. 3A; for original waveforms, see Supplementary Fig. S2), there were no significant main effects of age (F (2,152) = 2.20, p = .114, ηp2 = 0.028) or group (F (1,152) = 0.01, p = .933, ηp2 < 0.001) on amplitude. However, the Age × Group interaction was significant (F (2,152) = 4.33, p = .015, ηp2 = 0.054). Simple effects analyses revealed significantly smaller P3a amplitudes in children with ADHD than in TD children in the 7–8-year-old group (1.07 vs. 1.77 μV, p = .026). In TD children, the P3a amplitudes of 7–8-year-olds were significantly greater than those of 9–10-year-olds (1.77 vs. 0.99 μV, p = .023) and 11–12-year-olds (1.77 vs. 0.85 μV, p = .011; Fig. 3C). No other significant differences were observed between any of the age groups. Negative correlations were observed between P3a amplitude and age in TD children (r = −0.33, p = .009) but not in children with ADHD (r = −0.08, p = .571). In contrast to MMN, the P3a amplitudes gradually decreased with age in TD children. Further regression analysis revealed that the P3a amplitude in children with ADHD did not present a significant nonlinear age-related pattern across the age continuum (R2 = 0.03, adjusted R2 = 0.01, p = .143; Fig. 3D). For P3a latency, neither the main effects of age (F (2,152) = 0.93, p = .395, ηp2 = 0.012) nor group (F (1,152) = 0.55, p = .459, ηp2 = 0.004), nor their interaction were significant (F (2,152) = 2.87, p = .060, ηp2 = 0.036). P3b elicited by target-minus-standard sound For the P3b component (Fig. 4A; for original waveforms, see Supplementary Fig. S2), there were no significant main effects of group (F (1,152) = 0.60, p = .442, ηp2 = 0.004) or age (F (2,152) = 0.91, p = .405, ηp2 = 0.012), nor was the Age × Group interaction significant (F (2,152) = 0.42, p = .660, ηp2 = 0.005) on amplitude. The correlation between P3b amplitude and age did not reach significance in either the TD or ADHD group (ps > 0.10). The analysis of P3b latency revealed that both the main effect of group (F (1,152) = 1.33, p = .250, ηp2 = 0.009) and the Group × Age interaction (F (2,152) = 0.44, p = .645, ηp2 = 0.006) were not significant. However, a significant main effect of age was observed (F (2,152) = 9.76, p < .001, ηp2 = 0.114). Post hoc analyses revealed that the onset latency of 7–8-year-old children (494 vs. 408 ms, p < .001) and 9–10-year-old children (454 vs. 408 ms, p = .032; Fig. 4C) was significantly longer than that of 11–12-year-old children. Significant negative correlations were observed between P3b latency and age in both TD children (r = −0.42, p = .001) and children with ADHD (r = −0.29, p = .022; Fig. 4D), indicating that P3b onset latency decreased with age in both groups. Correlation analysis between behavioral data and ERPs Partial correlation analyses were conducted to examine the relationships between ERP parameters and task performance, controlling for age, IQ, and sex. In addition, when testing the association between ERP measures and RTH, RTFA was included as an additional covariate to control for its potential influence. Similarly, RTH was controlled for when examining the correlation between ERP measures and RTFA. FDR correction was applied across these correlation analyses. For the MMN component, a significant negative correlation was observed between MMN amplitude and RTH in TD children (r = −0.30, p = .033; Fig. 5), indicating that more negative MMN amplitudes were associated with faster hit responses. However, this association was not observed in children with ADHD (r = 0.07, p = .653). With respect to the P3b component, P3b latency was significantly correlated with behavioral performance measures in children with ADHD, including RTH (r = 0.44, p < .001), RTCV (r = 0.24, p = .026), hit rate (r = −0.40, p < .001), and false alarm rate (r = 0.34, p = .003; Fig. 5). These results indicate that shorter P3b latencies were associated with better behavioral performance in the ADHD group. However, no such effects were observed in TD children (ps > 0.10, see Supplementary Table S3 for the full correlation results). No other correlations between ERP measures and behavioral performance remained significant after FDR correction. Associations between ERP components and task performance. Scatterplots showing the associations between MMN amplitude and RTH, and between P3b latency and behavioral performance measures. SRs of MMN amplitude and RTH were computed while controlling for IQ, sex, age, and RTFA. Associations between P3b latency and behavioral performance included correlations with RTH, RTCV, and hit rate. SRs of P3b latency and RTH were computed while controlling for IQ, sex, age, and RTFA, whereas SRs of P3b latency with RTCV and hit rate were computed while controlling for IQ, sex, and age. Red dots indicate individual data from children with ADHD; blue dots indicate individual data from TD children. To further examine the unique contributions of different ERP components to behavioral performance, hierarchical regression analyses were additionally conducted separately for the ADHD and TD groups. Age and IQ were entered in the first step, followed by MMN amplitude, P3a amplitude, and P3b latency in the second step. For children with ADHD, the ERP variables significantly improved the prediction of RTH (ΔR2 = 0.14, p = .002), with P3b latency emerging as the only significant predictor (β = 0.41, p < .001). In contrast, for TD children, the ERP variables showed a marginal improvement in model fit (ΔR2 = 0.08, p = .053), and MMN amplitude was the only significant predictor of RTH (β = 0.27, p = .010). Similar patterns were also observed when RTAll was used as the behavioral outcome measure (see Table 3). Table 3Hierarchical regression analyses predicting RTAll and RTH in ADHD and TD children.ADHDTDDependentStepFactorβtpModel R2ΔR2βtpModel R2ΔR2RTAll1Age (years)−0.31−2.960.0040.21–−0.31−2.750.0070.19–IQ0.100.890.3750.030.250.8022Age (years)−0.32−3.130.0030.320.12**−0.32−2.700.0090.260.10*IQ0.080.810.4180.030.270.791MMN amplitude−0.04−0.360.7200.292.790.007P3a amplitude0.030.340.734−0.12−1.150.256P3b latency0.373.50<0.0010.100.870.387RTH1Age (years)−0.35−3.390.0010.24–−0.37−3.140.0020.19–IQ0.070.650.5170.060.530.5982Age (years)−0.34−3.51<0.0010.380.14**−0.35−2.890.0050.270.08†IQ0.050.530.6000.040.410.683MMN amplitude−0.04−0.420.6790.272.630.010P3a amplitude0.020.170.866−0.06−0.580.564P3b latency0.414.00<0.0010.080.770.446Note: †p < .10, *p < .05, **p < .01; β = standardized regression coefficient. Age and IQ were entered in Step 1. MMN amplitude, P3a amplitude, and P3b latency were entered in Step 2. RTAll = overall reaction time. RTH = hit reaction time; ADHD = attention-deficit/hyperactivity disorder; TD = typically developing. Hierarchical regression analyses predicting RTAll and RTH in ADHD and TD children. Note: †p < .10, *p < .05, **p < .01; β = standardized regression coefficient. Age and IQ were entered in Step 1. MMN amplitude, P3a amplitude, and P3b latency were entered in Step 2. RTAll = overall reaction time. RTH = hit reaction time; ADHD = attention-deficit/hyperactivity disorder; TD = typically developing. Discussion In the current study, we examined whether attentional difficulties in school-aged children with ADHD were associated with alterations in auditory MMN, P3a, and P3b during an active auditory detection task. Analyses of behavioral performance alongside simultaneous EEG recordings revealed clear differences between children with ADHD and their TD peers. TD children showed age-related patterns characterized by increasing MMN amplitudes and decreasing P3a amplitudes from ages 7 to 12. Compared with TD children, children with ADHD showed attenuated MMN amplitudes at ages 9–12 years and attenuated P3a amplitudes at ages 7–8 years. Notably, the ADHD group exhibited a U-shaped age-related pattern of MMN amplitude across ages 7–12. Both groups displayed similar age-related patterns in P3b, marked by decreasing latencies during childhood. Moreover, TD children showed significant correlations between MMN or P3a amplitudes and reaction time, whereas children with ADHD showed a significant association between P3b latency and task performance. The three-stimulus auditory oddball paradigm was used to test the ability to detect and discriminate salient auditory stimuli. The higher RTCV and the lower hit rate observed in the ADHD group are consistent with previous evidence of increased response variability and impaired inhibitory control in children with ADHD (Tamm et al., 2012). Furthermore, the behavioral results indicate continuous improvement in cognitive performance with increasing age in both the TD and ADHD groups (Tau and Peterson, 2010). MMN has traditionally been interpreted as an automatic neural response to unpredictable auditory changes, typically emerging around 200 ms over the fronto-parietal scalp regions (Luck and Kappenman, 2011; Marco-Pallarés et al., 2005). However, because the present study employed an active three-stimulus auditory oddball task requiring target discrimination and behavioral responses, MMN may also be influenced by attentional set and task context (Kong et al., 2025; Sussman et al., 2003). Therefore, the observed MMN responses likely reflect early deviance-related auditory processing under active task demands rather than purely automatic auditory change detection. In the previous ERP studies, the age-related changes in MMN amplitude have been mixed, with some studies reporting reduced MMN around 11 years (Rothenberger et al., 2000) and others reporting unchanged MMN at preschool age (Zhang et al., 2020) and early school age (7–10 years; Gomes et al., 2013). Our results further revealed reduced MMN amplitudes in ADHD only in the two older groups (9–10 and 11–12 years), but not in the youngest group (7–8 years), suggesting that atypical auditory change detection may become more pronounced during late childhood (Cheng et al., 2016). Importantly, we observed a U-shaped age-related pattern of MMN amplitude in the ADHD group, which contrasts with the gradual increase observed in TD children. Studies using fMRI have shown that MMN generators may originate from the frontal and parietal cortices (Molholm et al., 2005), and delayed maturation of frontal and parietal cortical thickness has been reported in children with ADHD (Shaw et al., 2009). Therefore, one possible interpretation is that the U-shaped age-related pattern in our study, with an inflection point at approximately 10 years, might reflect delayed maturation of early auditory deviance processing in children with ADHD, which could contribute to the divergence in MMN amplitudes between the two groups during the later critical period of 9–12 years. Consistent with a previous finding in adults (Toyomaki et al., 2008), we found that more negative MMN amplitudes were associated with faster responses in TD children, indicating that more mature processing of nontarget deviants was linked to enhanced attentional capture of target deviants. However, this relationship was absent in children with ADHD. The P3a component is thought to index an early attentional orienting process involving frontal mechanisms of working memory updating (Polich, 2012). Previous findings regarding P3a in children with ADHD have been inconclusive. Reduced P3a amplitudes have been reported in preschoolers (Zhang et al., 2020) and in younger school-age children aged 8–10 years (Gumenyuk et al., 2005) with ADHD, whereas comparable amplitudes were observed in older children (Oja et al., 2016) and adolescents (Rydkjær et al., 2017). In our study, children with ADHD showed significantly smaller P3a amplitudes than TD children only in the youngest group (7–8 years), but not in the two older groups (9–10 and 11–12 years). Moreover, a linear decrease in P3a amplitude was found across age in TD children, which was absent in children with ADHD. A similar linear decrease in P3a amplitude has also been observed in healthy adolescents aged 10–18 years (Mahajan and McArthur, 2015), suggesting a normative age-related refinement of frontal attention networks. Although P3a amplitude was not significantly correlated with behavioral performance in the present study, previous research has suggested that P3a reflects not only exogenous attentional orienting but may also be associated with inhibitory control processes (Cowley, 2018; Saarikivi et al., 2016). Given that the P3a in the current study was elicited by task-irrelevant nontarget stimuli, the age-related decrease in P3a amplitude observed in TD children may reflect reduced attentional orienting toward irrelevant deviant stimuli, possibly indicating improved inhibitory control efficiency with age. However, children with ADHD aged 7–8 years exhibited both lower P3a amplitudes and poorer behavioral performance compared with their TD peers. This finding aligns with evidence from schizophrenia research showing that attenuated P3a responses are linked to fronto-subcortical dysfunction (Fisher et al., 2014; Kiang et al., 2009; Light et al., 2015; Takahashi et al., 2013). One possible interpretation is that the reduced P3a amplitude in young children with ADHD reflects pronounced immaturity or inefficiency within cortical and subcortical attention control circuits (Castellanos et al., 2002; Hoogman et al., 2017; Shaw et al., 2007; Sowell et al., 2003). Although comparable MMN amplitudes indicate a similar level of auditory deviance processing between groups at ages 7–8 years, the diminished P3a in children with ADHD indicates that their poorer behavioral performance may be more strongly associated with less efficient attentional control rather than sensory detection. Recent studies have suggested that P3b reflects not only the updating of contextual information in working memory (Polich, 2007, Polich, 2012), but also the working memory-guided target identification process (Rac-Lubashevsky and Kessler, 2019). Previous auditory oddball studies using two-stimulus paradigms have consistently reported reduced P3b amplitude and prolonged P3b latency in school-aged children with ADHD (Janssen et al., 2016; Schall et al., 2024; Tsai et al., 2012). In contrast, the present study using a three-stimulus auditory oddball paradigm revealed only a significant age-related decrease in P3b latency in both the ADHD and TD groups, suggesting that task-related working memory updating processes may be relatively comparable between children with ADHD and TD children under the current experimental task. Moreover, although P3b latency has been shown to correlate with reaction time obtained from separate cognitive tests (e.g., CPT and Stroop) in typical school-aged children (Boucher et al., 2010), our results indicated that shorter P3b latencies were associated with faster target responses in children with ADHD, rather than in TD children. The inconsistent findings regarding P3b may be attributed to task characteristics. Unlike previous studies, the present study used a three-stimulus auditory oddball paradigm with monaural stimulation. Under this task context, the MMN and P3a results have indicated atypical early auditory processing in children with ADHD. Consequently, one possible interpretation is that, because early auditory processing may be less efficient, children with ADHD may rely more strongly on later target-related cognitive processes, such as working memory updating and controlled target evaluation, to achieve successful task performance. Hierarchical regression analyses were consistent with this interpretation, showing that P3b latency uniquely predicted behavioral performance in children with ADHD, whereas MMN amplitude was more strongly associated with behavioral performance in TD children. Recent evidence has also shown that ADHD symptom severity is strongly linked to working memory deficits (Yang et al., 2024), which is consistent with the potential involvement of late-stage cognitive processes in compensating for earlier processing inefficiencies. In contrast, TD children seem to rely more efficiently on early-stage processing (MMN) to guide behavioral responses, which may explain the weaker association between P3b latency and behavioral performance in this group. Overall, our findings suggest distinct age-related differences in auditory attention in children with ADHD from 7 to 12 years. At ages 7–8, children with ADHD exhibited reduced P3a amplitudes and poorer behavioral performance, a pattern that may be consistent with less efficient attentional orienting and control processes. By ages 10–12, behavioral performance appeared to improve, whereas MMN showed an atypical age-related pattern relative to TD children, which might reflect differences in the maturation of auditory change detection mechanisms. Within the predictive coding framework, MMN reflects both the generation of sensory predictions and the detection of deviations from these predictions (Fong et al., 2020). Thus, a more mature MMN indicates greater integration of bottom-up sensory encoding with top-down model updating. One possible interpretation of our findings within this framework is that MMN amplitudes in TD children, even when elicited by nontarget stimuli, remaine significantly associated with behavioral performance, suggesting that efficient early-stage auditory processing may contribute to stable task performance. In contrast, children with ADHD exhibited weaker early prediction-error signaling and disrupted orienting responses, which may be associated with a relatively greater involvement of late-stage cognitive processes in supporting task performance. However, several limitations should be acknowledged. First, the recorded ERP responses may be influenced by sensorimotor processes related to target detection because we used an active three-stimulus auditory oddball paradigm. It remains unclear whether the age-related differences observed here would generalize to passive auditory oddball paradigms that minimize task-related motor demands. Second, the current sample included only school-aged children and was based on a cross-sectional design. Consequently, the present findings can only characterize age-related differences across children of different ages and cannot directly capture within-individual developmental changes. Longitudinal studies spanning a broader age range are needed to determine whether the age-related patterns observed here correspond to true developmental trajectories of auditory MMN and P3 components in children with ADHD. Finally, although ERP measures provide valuable temporal information, complementary approaches such as time-frequency analyses or neural decoding methods may reveal more comprehensive characteristics of auditory predictive processing and attentional control in children. Conclusions This study examined the age-related differences in auditory MMN and P3 components in school-aged children with ADHD. We revealed that age-related neural patterns in ADHD diverge substantially from those in TD children, highlighting persistent difficulties in auditory attention. Specifically, children with ADHD showed reduced MMN amplitudes at ages 9–12 years, reduced P3a amplitudes at ages 7–8 years, and an atypical U-shaped age-related pattern of MMN. These differences might reflect delayed maturation of early auditory prediction processes and less efficient attentional orienting and control processes during early childhood. Moreover, P3b latency was associated with behavioral performance in children with ADHD, whereas MMN amplitudes were associated with behavioral responses in TD children. This dissociation may reflect differential contributions of early- and late-stage processing to task performance between the two groups, with late-stage cognitive processing appearing to play a relatively greater role in children with ADHD. Together, these findings provide electrophysiological evidence for atypical auditory attention in children with ADHD and offer novel insights into atypical cognitive developmental impairments in this population. CRediT authorship contribution statement Jipeng Huang: Writing – original draft, Visualization, Investigation, Formal analysis, Data curation. Shitao Huang: Writing – original draft, Investigation, Formal analysis, Data curation. Yuanjun Kong: Supervision, Methodology, Investigation. Manqi Zhou: Writing – review & editing, Visualization, Investigation. Han Yang: Investigation, Data curation. Danping Hong: Investigation. Yanling Zhou: Investigation. Weizhen Yin: Investigation. Jialiang Guo: Supervision, Methodology, Conceptualization. Cai-ping Dang: Supervision, Funding acquisition. Yan Song: Writing – review & editing, Supervision, Funding acquisition, Conceptualization. Funding This work was supported by the Brain Science and Brain-like Intelligence Technology-National Science and Technology Major Project (2021ZD0200500), the National Natural Science Foundation of China (No. 32271094), the Guangzhou Higher Education Teaching Quality and Teaching Reform Project (2023XWSJJD002), and the 2023 Guangdong Provincial Project on Teaching Quality and Educational Reform (01–408-2401067). 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. Acknowledgments We thank all the children and their parents who participated in this study. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.nicl.2026.104043. Contributor Information Cai-ping Dang, Email: dcp619@163.com. Yan Song, Email: songyan@bnu.edu.cn. Supplementary data Supplementary material. Data availability Data will be made available on request. References Associated Data Supplementary Materials Supplementary material. Data Availability Statement Data will be made available on request.