EEG Aperiodic Dynamics from Early through Late Childhood: Associations with ADHD, Cognition, and Development

Abstract

Background.

Aperiodic resting electroencephalography (EEG) activity is dynamic, reflecting shifting excitatory:inhibitory (E:I) balance with changing environmental conditions. We examined developmental and cognitive correlates of aperiodic and dynamic aperiodic indices in a cross-sequential cohort of early, middle, and late childhood youth with and without attention deficit hyperactivity disorder (ADHD).

Methods.

Two-hundred eighty-five children ages 2 – 14 years provided resting EEG during high- and low-visual input conditions. Licensed clinical psychologists determined ADHD diagnosis or likelihood (in young children). Linear regressions were estimated to examine associations between aperiodic features and age, ADHD diagnosis, IQ, and experimental condition.

Results.

From early to middle childhood, the aperiodic exponent increased linearly, indicating lower E:I, followed by a decreasing trajectory in late childhood. The aperiodic exponent was greater with high versus low visual input in young children, but this effect reversed with age. The ADHD group had a decreased aperiodic exponent, overall. Dynamic aperiodic activity, i.e. shifts in E:I balance, was associated with IQ.

Conclusions.

The aperiodic exponent and aperiodic dynamics are proxies for age-related cortical maturation and E:I balance, and have distinct associations with ADHD symptoms and cognitive ability.

Significance.

We provide novel evidence that dynamic aperiodic activity is a candidate marker of cortical efficiency in childhood.

Graphical Abstract

1. Introduction

Electroencephalography (EEG) is a non-invasive, temporally sensitive method of measuring cortical electrical fields. EEG signals can be differentiated into periodic and aperiodic activity, with the former reflecting rhythmic, oscillatory potentials, and the latter reflecting spontaneous cortical activity occurring across frequencies (Donoghue, Dominguez, & Voytek, 2020; Donoghue, Haller, et al., 2020). The aperiodic component of the EEG has gained increased attention in recent years due to its associations with development, cognition, and psychiatric diagnoses, including attention deficit hyperactivity disorder (ADHD) (McSweeney et al., 2021; Ostlund et al., 2021; Robertson et al., 2019; Schaworonkow & Voytek, 2021; Virtue-Griffiths et al., 2022; Waschke et al., 2019). The slope of the aperiodic power distribution is quantified by the 1/f^β exponent (“aperiodic exponent”) and the intercept is defined as power at the lowest frequency (“offset”). Steeper versus flatter aperiodic exponents reflect a balance of spontaneous excitatory and inhibitory (E:I) cortical activity, and may relate to integration versus segregation of neural information processing across space and time (Fries, 2005). High power, low-frequency aperiodic activity largely reflects localized inhibitory neuronal circuitries, while lower power, high-frequency signals primarily derive from integrated, excitatory networks (Gao et al., 2017). In previous work, our group has demonstrated that the aperiodic exponent is dynamic, shifting in response to and in anticipation of changing environmental conditions (Arnett, Fearey, et al., 2022; Arnett, Peisch, & Levin, 2022).

Prior studies suggest the aperiodic exponent is developmentally moderated. Cross-sectional and longitudinal investigations have established that the aperiodic exponent decreases from childhood through adulthood in TD samples (Hill et al., 2022; Voytek et al., 2015), and lower exponents are associated with cognitive decline among adults (Finley et al., 2024). Interestingly, one study found that the association between aperiodic exponent and cognition was driven by performance on tests of executive functioning (Finley et al., 2024), suggesting that aperiodic activity reflects higher order cortical processes relevant to ADHD.

Several studies have reported an association between aperiodic EEG activity and ADHD diagnosis in childhood, suggesting differences in spontaneous cortical activity among this clinical group compared to neurotypical/unaffected controls. During infancy and early childhood, a steeper aperiodic exponent, i.e., lower E:I, has been associated with familial or behavioral risk for ADHD (Karalunas et al., 2022; Robertson et al., 2019); by later childhood, this association appears to reverse, with children and adolescents diagnosed with ADHD showing flatter exponents (increased E:I) than their typically developing (TD) peers (Karalunas et al., 2022; Ostlund et al., 2021). Moreover, reduced signaling of gamma-aminobutyric acid (GABA), the primary inhibitory neurotransmitter in the brain, has been documented in the anterior cingulate as well as subcortical structures in association with ADHD symptoms (Ende et al., 2016; Rivero et al., 2015). Thus, it is possible that the neurobiological mechanisms underlying the association between aperiodic activity and cognition may be shared with those driving the association between aperiodic activity and ADHD diagnosis.

With respect to aperiodic dynamics, evidence for developmental and clinical significance is sparse. We previously reported that 7–11-year-old TD children had a higher resting aperiodic exponent during a low-visual input compared to a high-visual input resting condition. In contrast, children with ADHD showed no change in aperiodic exponent across conditions, suggesting lack of dynamic cortical response (Arnett, Fearey, et al., 2022). A small body of research suggests that children with ADHD have delayed maturation of the aperiodic exponent from 6 – 12 years of age (Dakwar-Kawar et al., 2024). However, no publications have examined developmental trajectories of dynamic aperiodic activity in TD or ADHD samples. Notably, few publications have examined the aperiodic exponent during both high- and low-visual input resting conditions during early childhood, underscoring the need for condition comparisons at this age range (Stanyard et al., 2024). We propose that while aperiodic activity reflects differences in cortical organization and maturation, dynamic aperiodic activity is an indicator of cortical flexibility and response. Thus, both measures are relevant to neurodevelopmental disorders, including pediatric ADHD. In the current study, we characterize developmental trajectories of resting aperiodic activity and dynamics in a cross-sectional sample spanning early, middle, and late childhood, and examine associations with ADHD diagnosis and cognition.

2. Methods

2.1. Procedures

Participants were recruited to one of three studies examining EEG correlates of ADHD symptoms and related disorders in children. The early childhood and middle childhood samples had identical procedures. These participants were recruited from the greater Boston area via flyers, research registries, outreach to pediatric medical providers, and social media. Caregivers completed questionnaires as well as clinical interviews about their child and family member with ADHD (when applicable) either online or in person. The child and a caregiver completed a single research visit at a hospital laboratory, which included approximately 45 minutes of resting EEG and event related potential (ERP) tasks, as well as approximately 90 minutes of cognitive and academic testing. Each study intentionally over-sampled for children who had an immediate family history of ADHD (early childhood sample) or an ADHD diagnosis (middle childhood sample). The third sample included youth in middle-late childhood, recruited from a prior study of childhood ADHD (Arnett, Fearey, et al., 2022; Arnett et al., 2021). All participants from this previous study were invited to complete a second visit consisting of a standardized clinical interview and a 45-minute EEG. The second visit occurred approximately 1 – 3 years after the previous study visit. Clinical interviews were conducted online and/or by phone, and the EEG took place at a university hospital setting. The current study uses data from this second visit.

All procedures were approved by the relevant institutional review boards (Boston Children’s Hospital P00039618 and P00040320; University of Washington CR00005526). For all studies, caregivers provided informed consent and children ages seven or older provided assent prior to study participation. Children completed a 48-hour washout of psychostimulant medications prior to the visit, when applicable. Under the guidance of their medical provider, children taking other psychotropic medications abstained for an appropriate time to ensure elimination from their system prior to the visit.

2.2. Participants

2.2.1. Early Childhood.

133 children ages 30–59 months were enrolled in the early childhood study. Enrollment exclusion criteria included Full Scale IQ < 80, a diagnosis of global developmental delay, prenatal exposure to alcohol, perinatal trauma, diagnosis of autism spectrum disorder (ASD), known genetic syndrome, seizure disorder, or color blindness. Three children were excluded after the study visit due to low IQ or a later diagnosis of autism. For the current analyses, participants were also excluded if they had fewer than 30 seconds of valid EEG data on both resting tasks, after data processing (n=9). The final sample included 121 children (M_age = 4.01, SD = 0.72 years) of whom 73% had an immediate family history of ADHD (i.e., biological parent or sibling). Children were categorized by clinicians as having high or low likelihood for ADHD, regardless of family history (see measures). 55% of children were identified as having high likelihood of meeting diagnostic criteria for ADHD. See Table 1.

Table 1.

Sample Characteristics

Early Childhood (N = 121)
	High Likelihood	Low Likelihood	p
N	67	54	---
Age	4.05 (0.70)	3.96 (0.75)	.475
Female	21 (31%)	26 (48%)	.090
IQ	106 (11.70)	110 (9.53)	.074
Hispanic/Latino	7 (10%)	7 (13%)	.567
Race			.485
Black/African-American	2 (3%)	2 (4%)	---
Asian	0 (0%)	2 (4%)	---
Native	0 (0%)	0 (0%)	---
American/Alaskan Native
Native Hawaiian/Other Pacific Islander	0 (0%)	0 (0%)	---
White	56 (84%)	42 (78%)	---
More than one race	9 (13%)	7 (13%)	---
Not Reported	0 (0%)	1 (2%)	---
Primary Caregiver Highest Education			.005
High School or equivalent	2 (3%)	0 (0%)	---
Some College	2 (3%)	0 (0%)	---
2-Year College or Trade School	1 (1%)	0 (0%)	---
4-Year College	9 (13%)	22 (41%)	---
Master’s	36 (54%)	23 (43%)	---
Doctorate (PhD, MD)	17 (25%)	9 (17%)	---
Not Reported	0 (0%)	0 (0%)	---
Middle Childhood (N = 110)
	ADHD	TD	p
N	72	38	---
Age	9.50 (1.35)	9.33 (1.55)	.555
Female	19 (26%)	19 (50%)	.023
IQ	109 (15.30)	118 (10.90)	<.001
Hispanic/Latino	12 (17%)	1 (3%)	.058
Race			.558
Black/African-American	4 (6%)	1 (3%)	---
Asian	0 (0%)	1 (3%)	---
Native	0 (0%)	0 (0%)	---
American/Alaskan Native
Native Hawaiian/Other Pacific Islander	0 (0%)	0 (0%)	---
White	56 (78%)	30 (79%)	---
More than one race	7 (10%)	5 (13%)	---
Not Reported	5 (7%)	1 (3%)	---
Primary Caregiver Highest Education			.079
High School or equivalent	2 (3%)	0 (0%)	---
Some College	8 (11%)	0 (0%)	---
2-year College or Trade School	4 (6%)	0 (0%)	---
4-year College	17 (24%)	8 (21%)	---
Master’s	30 (42%)	18 (47%)	---
Doctoral (PhD, MD)	11 (15%)	12 (32%)	---
Not Reported	0 (0%)	0 (0%)	---
Middle-Late Childhood (N = 54)
	ADHD	TD	p
N	39	15	---
Age	11.50 (1.61)	11.00 (1.50)	.368
Female	12 (31%)	5 (33%)	1.000
IQ	106 (10.40)	117 (10.30)	.001
Hispanic/Latino	3 (8%)	3 (20%)	.334
Race			.158
African-American/Black	1 (3%)	0 (0%)	---
Asian/Pacific Islander	0 (0%)	2 (13%)	---
White	30 (77%)	11 (73%)	---
More than one race	7 (18%)	2 (13%)	---
Not Reported	1 (3%)	0 (0%)	---
Primary Caregiver Highest Education			.494
High School or equivalent	2 (5%)	0 (0%)	---
Some College	2 (5%)	0 (0%)	---
2-year College or Trade School	1 (3%)	1 (7%)	---
4-year College	12 (31%)	6 (40%)	---
Master’s	17 (44%)	4 (27%)	---
Doctorate	5 (13%)	4 (27%)	---
Not Reported	0 (0%)	0 (0%)	---

Notes: Sample comparisons for continuous variables were done using independent samples T-tests with equal variance assumed except when Levene’s test of equality indicated unequal variances. Sex comparisons were completed with chi-square tests. Ethnicity, race, and caregiver education comparisons were done with Fisher’s exact test to account for low cell counts. Demographics questionnaires about race differed in the middle-late childhood sample, which had combined “Asian” and “Pacific Islander”

2.2.2. Middle Childhood.

113 children between the ages of 7 years, 0 months and 11 years, 11 months were enrolled in the middle childhood study. Enrollment exclusion criteria were identical to those for the early childhood study. Three children were excluded for autism or low IQ; all children provided sufficient valid EEG data for at least one resting task. The final sample included in the analyses included 110 children (M_age = 9.44, SD = 1.41 years), of whom 65% had a confirmed ADHD diagnosis.

2.2.3. Middle-Late Childhood.

58 children ages 7 years, 11 months to 14 years, 2 months were enrolled in the middle-late childhood study. Exclusion criteria were identical to the other samples. One child with ADHD was disqualified due to identification of seizures in the EEG; one TD child was excluded due to receiving an ADHD diagnosis since the first study. Two children were disqualified due to no longer meeting diagnostic criteria for ADHD at the second visit. The final sample included 54 children (M_age = 11.35, SD = 1.58 years), of whom 72% had a confirmed diagnosis of ADHD.

2.3. Resting EEG Acquisition

All participants were seated comfortably, approximately 70 cm from a computer monitor on which the visual stimuli were presented. Continuous EEG was recorded using a 128-channel Magstim-EGI HydroCel geodesic sensor net for all samples and digitized with a sampling rate of 1,000 Hz. Data were acquired with Net Station Acquisition software version 4.5.6 (middle-late childhood) or 5.4.2 (early and middle childhood), in combination with a 400 series high-impedance amplifier (Magstim-EGI, Plymouth, MN). The vertex electrode was used as the reference during acquisition and signals were analog filtered (0.1 Hz high pass, 100 Hz elliptical low pass). Impedances were reduced to <50 kΩ before starting the EEG session.

Each participant completed a high visual input resting condition (“lights-on”) and a low visual input condition (“lights-off”). During the lights-on task, the middle and middle-late childhood sample watched small screensaver videos in three blocks and were instructed to look at the screen for the duration of the task. The early childhood sample watched a three-minute video animation created for the purposes of the study. Both samples watched the videos in silence, and segments containing artifactual movement or non-resting behavior (moving, talking, etc.) were marked in real time by the experimenter.

The lights-off task was similarly structured for all three samples. Across four blocks of 30 seconds each, participants sat silently in the EEG acquisition room with the lights off. Participants were instructed to continue looking forward and to keep their eyes open. Trials in which the participant spoke were marked by the experimenter for later removal.

2.4. EEG processing

Continuous EEG data were processed using MATLAB R2021b with the EEGLAB v2021.1 extension and functions. Eye and rim electrodes were excluded, leaving 112 total electrodes for processing. Resting data were isolated and experimenter-marked bad trials were removed prior to filtering. Data were bandpass filtered at 1–100hz after baseline removal, and then the data were downsampled to 500hz. The Cleanline method was used to remove 60hz line noise. Bad channel rejection was done twice, and bad channels were interpolated before average re-referencing. Artifact rejection using window and burst rejection criterion was completed prior to identifying ICA components. MARA was used to automatically detect and subtract artifactual ICA components. Data were segmented and separated by experiment and a fast Fourier transformation was conducted using Welch’s method.

2.5. Measures

2.5.1. ADHD/Likelihood Categorization

2.5.1.1. Early Childhood.

Caregivers of early childhood participants completed a remote clinical interview with a licensed clinical psychologist (AA or VP) or a psychology trainee under their supervision. The interview lasted approximately 45 minutes and focused on the child’s attention, activity level, emotion regulation, and associated functional impairment (e.g., difficulties learning; peer challenges). Following the interview, the clinician provided a rating on a scale of 1 to 5, with 1 = no symptoms and very low likelihood of ADHD, 3 = some symptoms of ADHD without functional impairment, and 5 = many symptoms with accompanying impairment and high likelihood of ADHD. The child was rated on the same scale a second time by a clinician who completed the neuropsychological evaluation during the in-person research visit, with the rating assigned based solely on the child’s behavioral presentation. For the current analyses, children given a rating of 4 or 5 at either time point were categorized as high likelihood and those with ratings less than 4 were characterized as typically developing. We examined sensitivity and specificity of this cutoff by comparing these ratings with results of a standardized diagnostic interview (K-SADS; Kaufman et al., 2016) completed with caregivers 18 months following the research visit. At the time of these analyses, 27 participants had completed the K-SADS, with 12 (44%) meeting criteria for ADHD AS defined by the Diagnostic and Statistical Manual for Mental Disorders, 5^th Edition (DSM-5; American Psychiatric Association & American Psychiatric Association, 2013). Among the subsample of children who had completed both assessments, the ADHD likelihood ratings showed adequate sensitivity (83%) and specificity (73%) for K-SADS ADHD diagnosis at the 18-month follow-up.

2.5.1.2. Middle Childhood.

ADHD diagnoses in the middle childhood sample were confirmed by a licensed clinical psychologist using DSM-5 criteria. Caregivers completed an unstructured clinical interview with a licensed psychologist or psychology trainee prior to the research visit to confirm the presence of ADHD symptoms across settings and associated impairment. Alternative explanations for symptoms, such as posttraumatic stress disorder, specific learning disorders, anxiety, depression, and medical diagnoses were also evaluated.

2.5.1.3. Middle-Late Childhood.

Caregivers of the middle-late childhood sample completed a standardized clinical interview using the computerized version of the Kiddie Schedule of Affective Disorders and Schizophrenia (KSADS-COMP) (Townsend et al., 2019). When responses required clarification, a licensed clinical psychologist conducted an additional interview with the caregiver via telephone or video call to confirm or rule out a DSM-5 ADHD diagnosis.

2.5.2. IQ

2.5.2.1. Early Childhood.

Children in the early childhood cohort completed the core subtests of the Wechsler Preschool and Primary Scale of Intelligence – Fourth Edition (WPPSI-IV; Wechsler, 2012), from which a full-scale IQ score was derived. One participant had a significant speech language impairment and their performance on the WPPSI-IV was deemed invalid. As they had completed IQ testing using the Differential Abilities Scale, Second Edition (DAS-II; Elliott et al., 2007) with a clinical psychologist only 7 months prior, the participant’s nonverbal reasoning score on the DAS-II was used as a proxy for IQ. The WPPSI-IV IQ and DAS-II nonverbal reasoning scores have previously been shown to be highly correlated (r = .61 – .75)(Wechsler, 2012).

2.5.2.2. Middle and Middle-Late Childhood.

Participants in the middle and middle-late childhood cohorts completed the two-subtest version of the Wechsler Abbreviated Scale of Intelligence – Second Edition (Wechsler, 2011) to derive an abbreviated IQ score. The middle-late childhood sample completed this testing as part of their participation in the previous study; thus, their scores had been recorded 1–3 years prior to the EEG. However, given the stability of IQ at this age range (For review, see Schneider et al., 2014), the scores were interpreted as valid. The WASI-II and the WPPSI-IV, while measuring similar constructs, have not been tested for test-retest reliability to our knowledge. However, in comparison to a comprehensive childhood intelligence test, the WASI-II and the WPPSI-IV were both highly correlated with the Wechsler Intelligence Scale for Children – Fourth Edition (WISC-IV), with corrected r coefficients of .84 and .85, respectively (Wechsler, 2011; Wechsler, 2012).

2.5.3. Aperiodic Exponent

The fitting oscillations & one over f (FOOOF) tool (Donoghue, Haller, et al., 2020) version 1.0.0 was used to extract the aperiodic 1/f exponent from the power distribution. Peak width limits were set from 2 to 12, with 8 max peaks at 0.5 minimum peak height. Peak threshold was set to 2.0 with a fixed aperiodic mode, and the model was fit from 1 to 50 Hz. Following examination of topographic plots (Figure 1) and review of prior literature (Stanyard et al., 2024), it was decided to calculate the average aperiodic exponent across the full scalp for analyses.

Figure 1.

Topographic plots of the aperiodic exponent by sample (from left: Early Childhood, Middle Childhood, and Middle-Late Childhood) and resting experiment (top: Lights Off; bottom: Lights On). Topographic plots are averaged over ADHD and TD groups.

2.5.4. Aperiodic Dynamics

Aperiodic dynamic values were computed as the difference in the aperiodic exponent between the lights-on and lights-off experiments, divided by the exponent during the lights-off exponent. Thus, higher dynamic scores indicated a relatively steeper aperiodic exponent in the lights-on compared to lights-off experiment.

2.6. Analytic Plan

Analyses were completed in RStudio version 2023.09.1. One participant from the middle childhood cohort was missing an IQ score. Two early childhood participants had insufficient lights-on data (i.e., < 30 seconds after processing). There were insufficient or missing lights-off data for 35 (n=20 early, n=10 middle, and n=5 middle-late childhood) participants in this dataset. Continuous variables of interest were examined for normal distributions and found to have skew < |2.0| and kurtosis < |4.0|. The early, middle, and middle-late childhood samples were combined for all analyses. Linear and quadratic age values were centered prior to analyses.

The first set of analyses aimed to characterize cross-sectional developmental, diagnostic and cognitive associations with the aperiodic exponent. Multilinear models (MLM) were estimated with aperiodic exponent as the dependent variable, growth parameters as independent variables, and a random intercept for individual. Experiment and sex were included as covariates. First, we determined the optimal shape of growth by comparing regression coefficients and model fit statistics for linear and quadratic models (Supplementary Materials). Next, we added main effects of ADHD and IQ. Lastly, interaction terms were added to test for diagnostic and cognitive moderators of aperiodic development.

The second set of analyses focused on developmental, diagnostic and cognitive correlates of aperiodic dynamics, i.e., change in aperiodic activity between lights-off and lights-on resting conditions. A series of linear models were estimated in the same progression as the MLM described above.

3. Results

3.1. Aperiodic Exponent

The quadratic growth model fit the aperiodic exponent best. The aperiodic exponent was characterized by a linear increase that decelerated after approximately age 8 years (Table 2; Figure 2a). On average, the aperiodic exponent was significantly higher during the lights-on compared to lights-off experiment; the sex effect was not significant. The main effects model revealed that children with or at high likelihood for ADHD had a lower aperiodic exponent compared to TD children on average; there was no main effect of IQ. The combined fixed and random effects in this model explained the majority of variance in the aperiodic exponent (conditional R² = 0.77). The interaction model did not reveal any interactions between the growth parameters and either ADHD or IQ (ps > .303), nor did it explain additional variance (conditional R² = 0.77).

Table 2.

Multilevel Linear Models

Aperiodic Exponent
	Unstandardized B	SE	p	${\eta }_{\text{p}}^2$
Age	0.003	0.001	.007	0.03
Age2	−0.00002	0.000006	.013	0.02
Experiment	0.049	0.006	<.001	0.20
Sex	0.018	0.016	.267	0.004
ADHD	−0.036	0.017	.036	0.02
IQ	0.0002	0.0006	.783	0.0003
Aperiodic Dynamics
	Unstandardized B	SE	p	${\eta }_{\text{p}}^2$
Age	−0.0006	0.0001	<.001	0.12
Sex	0.005	0.009	.579	0.002
ADHD	0.008	0.009	.402	0.01
IQ	−0.001	0.0003	.002	0.04

Notes: Statistically significant results at p < .01 are indicated in bold.

Figure 2.

A) The aperiodic exponent followed a negative quadratic trajectory from early to late childhood for both ADHD (red) and TD (black) participants during both experiments. B) The dynamic aperiodic exponent decreased from early to late childhood among both ADHD and TD children, indicating increasing exponent during the lights off relative to lights on experiment with age. C) The dynamic aperiodic exponent was negatively correlated with age-standardized IQ across the full sample.

3.2. Dynamic Aperiodic Exponent

The dynamic aperiodic exponent was best described by a linear decrease from early to late childhood (Figure 2b). There was no effect of sex on aperiodic dynamics. The regression revealed no effect of ADHD status, but a negative association between aperiodic dynamics and IQ (Figure 2c). This regression explained a significant amount of variance in the dynamic exponent (adjusted R² = 0.15, p < .001). The interaction terms were not statistically significant (ps > .271), nor did they explain additional variance (adjusted R² = .15, p < .001).

4. Discussion

In the current study, we investigated developmental, behavioral and cognitive correlates of the aperiodic exponent and aperiodic dynamics in a cross-sectional sample of children ages two through 14 years. Diagnosis and likelihood of ADHD were uniquely associated with a flatter aperiodic exponent across all ages, while IQ was uniquely associated with aperiodic dynamics. Both cortical measures had unique developmental trajectories, with the exponent characterized by a negative quadratic slope, and dynamics by negative linear change (indicating higher aperiodic exponent during the lights-off compared to lights-on condition, with greater age).

Aperiodic EEG activity is an aggregate measure of excitatory and inhibitory signaling in cortical networks. The aperiodic exponent, or slope of the distribution of non-oscillatory EEG power over frequencies, has previously been linked to excitatory/inhibitory balance in rodents and human populations (Gao et al., 2017; Mamiya et al., 2021). Specifically, Gao and colleagues (2017) used electrocorticography and application of anesthesia to demonstrate that a steeper 1/f slope (i.e., higher exponent) is associated with greater inhibitory relative to excitatory activity. They reasoned that fast-decaying GABA signaling is consistent with higher power at low frequencies, while increased power in the high frequency range reflects excitatory circuitries with dense glutamate inputs. Additional research suggests that slow-frequency inhibitory neuronal activity derives largely from local circuitries and excitatory activity originates with globally integrated networks (Hindriks & van Putten, 2012). Thus, the aperiodic exponent and its dynamics are candidate proxies for cortical efficiency related to an optimal balance of locally versus globally integrated functional connectivity(Deery et al., 2023).

Consistent with prior publications (Dakwar-Kawar et al., 2024; Hill et al., 2022; McSweeney et al., 2021; Stanyard et al., 2024; Voytek et al., 2015), the aperiodic exponent became steeper across early childhood, followed by a flattening from middle to late childhood. This trajectory indicates increased inhibitory neural population signaling across this developmental period, likely coinciding with specialization of cortical networks (Johnson, 2001). In contrast, the decrease in aperiodic exponent after age eight is suggestive of maturation of excitatory, long-range cortical connections. Children with or at high likelihood for ADHD and TDs showed comparable development of the aperiodic exponent, but the high-likelihood/ADHD group had reduced aperiodic exponent across all ages. This aligns with previous reports on school age (Arnett, Fearey, et al., 2022; Chen et al., 2024), and adolescent samples (Karalunas et al., 2022; Ostlund et al., 2021); for an exception, see (Dakwar-Kawar et al., 2024). These findings suggest that ADHD in childhood and adolescence is associated with reduced inhibitory neurotransmitter signaling (Ende et al., 2016; Mamiya et al., 2021). However, it remains unknown whether this is due to differences in neurotransmitter availability and/or cortical organization, or a tendency for children with ADHD symptoms to remain more cognitively alert than their peers in resting conditions. Moreover, the specificity of association between aperiodic exponent and ADHD (but not IQ) suggests that reduced inhibitory signaling may be most prominent in cortical networks involved in self-regulation, e.g., the ventral and dorsal attention networks and frontal-striatal circuitries (Cortese & Coghill, 2018).

On the other hand, our results did not align with two studies in which a family history of ADHD in one-month-old infants (Karalunas et al., 2022) and an ADHD diagnosis in 3–7 year-old children (Robertson et al., 2019) were associated with a steeper aperiodic exponent. One possible explanation is that the Robertson et al. (2019) study focused on only a subset of midline electrodes, while our analyses estimated whole-scalp exponents, a decision we made after visual examination of the topographic plots. Additionally, our sample did not include children who were five or six years old at the time of testing, and there is a possibility that aperiodic activity during this developmental stage is quite different from the age ranges we did capture (Delorme, 2023).

This is the first study to examine developmental trajectories of dynamic aperiodic activity across childhood. We quantified dynamic activity as the relative increase in aperiodic exponent during the lights-on compared to lights-off experiment. A multi-method study of adults using EEG, functional magnetic resonance imaging (fMRI), and positron emission tomography (PET) found evidence that inhibitory signaling is increased during resting state (Rajkumar et al., 2021). Specifically, GABA, but not glutamate, receptor availability was increased in the default mode network (DMN) during rest, compared to sensorimotor networks and average gray matter. In our sample, the dynamic aperiodic measure was positive in early childhood, indicating a steeper exponent (i.e., reduced excitatory signaling) during the lights-on compared to lights-off resting condition. However, the dynamic measure became increasingly negative from middle to later childhood, indicating that emphasis on localized, inhibitory signaling during low visual input conditions emerges over development. This result is largely consistent with results from our previous study, where we found that TD children between the ages of 7–11 showed a steeper aperiodic exponent during the lights-off compared to lights-on condition; i.e., a more negative dynamic value (Arnett, Fearey, et al., 2022). We also found consistency with a recently published synthesis that reported steeper aperiodic exponents were during eyes-closed compared to eyes-open conditions in middle to late childhood (Stanyard et al., 2024). Unlike our prior publication, we did not find evidence for overall reduced aperiodic dynamics in children with or at high likelihood for ADHD; rather, the ADHD group likewise showed a more negative dynamic value over development. Although visual inspection of Figure 2 suggests that this maturational decline was somewhat delayed in the ADHD group, the effect did not reach statistical significance in our cross-sectional sample.

Interestingly, higher IQ was associated with a more negative dynamic aperiodic value, over and above the effects of age and ADHD. Consistent with this, our prior work showed that during event related potential tasks, the aperiodic exponent was steeper prior to correct versus incorrect responses among 7–11-year-olds, regardless of ADHD diagnosis (Arnett, Peisch, & Levin, 2022). Altogether, relatively greater engagement of inhibitory circuitries during resting and low-input conditions, and greater engagement of excitatory circuitries with increased cognitive demand, appear to support flexible, efficient cortical information processing. This process matures as children get older and appears to relate to higher intelligence, but is independent of the overall higher E:I ratio associated with ADHD.

There are limitations to our study that we hope to address in future research. First, the current analyses were done on a cross-sectional sample; thus, trajectories of development can only be inferred and may not reflect true individual growth patterns. Possibly, individual differences in developmental trajectories of aperiodic activity could explain phenotypic heterogeneity in ADHD. Additionally, the middle-late childhood sample differed from the other two samples in sex distribution between ADHD and TD groups. Although sex effects were not found for either aperiodic or dynamic aperiodic activity, the difference in sex distribution at this age range could have influenced results. Third, classification of high-likelihood for ADHD in the early childhood sample was based on information gathered prior to the typical age of diagnosis. This approach is most likely to result in false negatives for children who will eventually present as predominantly inattentive and/or whose symptoms emerge in later childhood. Lastly, our sample did not include five- and six-year-olds. Additional data at these ages may have altered our growth trajectory estimates.

5. Conclusions

A growing body of research supports aperiodic resting EEG activity as a putative marker of neurodevelopment, cognition, and behavioral and psychiatric symptoms. The current study expands on this literature by demonstrating that aperiodic dynamics, i.e., within-person variability in aperiodic exponent across stimulus conditions, has a unique developmental trajectory and is associated with general intelligence. Future investigations should model developmental trajectories of aperiodic features within longitudinal samples and examine momentary associations between aperiodic dynamics and attention.

Highlights.

1. The aperiodic exponent is dynamic, responding to changing environmental conditions.

2. The aperiodic exponent and dynamic aperiodic activity are developmentally moderated.

3. Flatter aperiodic exponent is associated with pediatric ADHD, while aperiodic dynamics relate to intelligence in childhood.

Data Statement

Research data used in this study may be made available upon reasonable request to the authors.