The development of EEG microstates from infancy into middle childhood and their associations with behavior

Abstract

Microstates reflect rapid changes in neural activity with specific scalp topographies. Microstates have been colloquially referred to as the “atoms of thought” and temporal properties of microstates are thought to be associated with mental health and cognition. However, little work has characterized the longitudinal developmental changes and associations with behavior across the infancy and middle childhood period. In this study, we examined developmental changes of microstates from infancy through age 7 years (N = 383 with at least one time point of usable EEG data, 52.2% male, 78.0% white) in a sample of typically developing children. Further, we examined associations of microstates with child temperament and psychopathology symptoms. Results provide evidence for complex age by sex by microstate number interactions for each of the microstate temporal features (occurrence, duration, coverage, and global explained variance [GEV]). Moreover, individual differences in microstates were of relevance to temperament traits. Overall, results point to microstates being an informative marker of brain development.

Humans are constantly processing information from their environment and preparing potential behavioral responses, which requires rapid neural activation and organization. These fast-acting brain activity patterns must be refined across the lifespan to accommodate changing demands. One way to measure and describe fast-acting neural processes is by computing microstates from resting state electroencephalography (EEG) data. Microstates are scalp potentials that remain relatively stable (less than 150 ms) before quickly transitioning to another scalp potential reflecting a different microstate (Michel & Koenig, 2018). Microstates are computed by assessing the global field power, or evenness of electrical activity across the scalp, on a moment-to-moment basis. Microstates are characterized and named by their dipole patterns (e.g., Microstate 1 could be right-frontal to left-posterior, 2 left-frontal to right-posterior, and 3 frontal to occipital). These distinct dipole patterns may be related to functional cognitive networks similar to those identified using fMRI (Britz et al., 2010; Custo et al., 2017; Musso et al., 2010). For a particular dataset, a set of microstates is selected at the group level and then fit to each time series on the individual level. Each moment in an individual’s EEG recording is assigned to the closest matching microstate class, or scalp topography pattern. The time series is then summarized based on the temporal patterns of each microstate’s activation. Therefore, unlike traditional frequency-domain EEG measures that discard temporal information by averaging across the entire recording session, microstates leverage EEG’s high temporal resolution to investigate patterns of temporal changes.

Microstates are characterized by both the spatial topography pattern and specific temporal properties, including occurrence, duration, coverage, and global explained variance (GEV). Occurrence refers to the number of times one microstate occurs in a second. Duration refers to the average time (ms) that a microstate is in one stage before transitioning into a different topography. Coverage is the amount of time a particular microstate is dominant during an EEG recording. GEV is the total variance in the data explained by each microstate class for an individual (Bagdasarov et al., 2024; Khanna et al., 2015). Another microstate feature is transition probability, or the sequence of microstate activations. This feature provides a large array of information (e.g., microstate number * [microstate number −1] transition probabilities). For brevity and because relatively less is known about this feature and how to interpret the findings (e.g., what does it mean to have a higher transition from microstate 1 to 2 as opposed to microstate 3 to 4), this feature is not covered in the present analyses. Historically, microstate topographies or dipole patterns have been named using a numeric or alphabetical classification. The topographies of each microstate are unique, and the number of microstates in a population can be explained best using a set number of microstates, with a 4- to 7-microstate solution being most common. Work with adult and pediatric populations has demonstrated that these features show high levels of internal consistency, even with as few as 60 seconds of EEG data (Bagdasarov et al., 2024). This suggests that microstate analyses may be a reliable tool for understanding the spatiotemporal dynamics of emerging childhood brain and related behavioral development (Bagdasarov et al., 2024).

In neurotypical populations, EEG microstate properties change across the lifespan and likely reflect a necessary neural adaptation to one’s environment (Bagdasarov et al., 2022, 2024; Brown & Gartstein, 2023; Koenig et al., 2002). During infancy, environmental input is limited by nascent sensory-motor abilities (Adolph & Hoch, 2019). As infants gain more autonomy and freedom to explore the world, the demands on the neural system increase; consequently, the brain needs to process a larger volume of information in a more efficient manner (Rachwani et al., 2020). This is exemplified by the duration significantly decreasing in infancy in all microstates observed throughout 2 to 30 months, with the steepest decreases occurring in the first year (Ghosh et al., 2025). Another study of infants in a resting-state task showed an increase in coverage from 6 to 10 months of age for one microstate (topographical pattern left occipital to right frontal orientation) but no developmental differences in the coverage of other microstates (topographical patterns with symmetric occipital to prefrontal orientation and right occipital to left frontal orientation) (Brown & Gartstein, 2023). Therefore, examination of EEG microstates across infancy and middle childhood may shed light on how neural patterns develop and begin to synchronize with one another to form coordinated units. In addition, microstate analysis provides insights into fast-acting neural processes often ignored by averaging or low temporal resolution modalities.

The exact developmental patterns during childhood for each of the microstate features are mixed, with several studies reporting that effects may differ between microstates and are modified by sex assigned at birth (herein referred to as “sex”). Some research has found that each microstate class has its own developmental trajectory, meaning microstate A may increase in a particular feature across development, whereas microstate B may decrease (Koenig et al., 2002). In an eyes-open resting-state paradigm with children aged 4–12 years, age was not associated with four of the microstates, but it was positively associated with occurrence and duration for one of the microstates with a topographical pattern going left to right (Hill et al., 2023). Moreover, the developmental patterns appear to differ between males and females (Bagdasarov et al., 2022; Hill et al., 2023). For instance, among children aged 4–8 years, coverage, occurrence, and GEV for a microstate with a posterior-central maximum decreased with age only in males. In another study of children and adults (age range 6–87 years), the duration for a microstate with central anterior-posterior orientation increased throughout the female lifespan, but this same pattern was not identified in males (Tomescu et al., 2018). These findings suggest both developmental and sex differences in temporal brain dynamics (Bagdasarov et al., 2022). Overall, many aspects of the developmental patterns remain unclear due to both the complex developmental trajectories (sex by microstate by microstate feature interactions) and limitations of prior research, such as cross-sectional designs and narrow age ranges, particularly when brain development is very rapid (e.g., either infants or young children). Therefore, there is a need for longitudinal research to clarify the developmental trajectories of microstates as well as their functional significance in relation to key behavioral outcomes.

Microstates may be of particular importance to behavioral temperament and psychopathology, given the potential link between microstates, functional brain networks (i.e., fronto-parietal and default mode network), and behavioral outcomes (Brown & Gartstein, 2023; Kelsey et al., 2021; McDevitt & Gartstein, 2025). The analysis of microstates may expand upon our knowledge of these behaviors as it can show how neural processing at a high temporal time scale (on a moment-to-moment basis) contributes to behavior. Few studies have leveraged a longitudinal cohort to examine the link between individual microstate properties and temperament characteristics across early development. To this end, several cross-sectional studies have pointed to microstate features underlying specific components of regulatory capacities, or one’s ability to manage, control, and adapt their emotions, behaviors, and attention (Feldman, 2009), and both externalizing and internalizing psychopathology.

Literature presents mixed associations between externalizing symptoms, surgency, and effortful control temperament behaviors and microstates. One study examining a population of infants showed that duration in a particular microstate was positively associated with key attentional capacities such as sustained attention and initiations of joint attention (Bagdasarov et al., 2025). Another cross-sectional study with infants revealed that surgency temperament behaviors were negatively correlated with occurrence and duration in one microstate, and effortful control temperament behaviors were negatively associated with coverage and occurrence of a different microstate (Brown & Gartstein, 2023). These findings highlight how early microstate development may be important for building blocks of regulatory abilities. However, work with older children and adolescents appears to contradict the infant findings. Specifically, there were fewer occurrences and increased durations for specific microstates in ADHD compared to neurotypical children, suggesting that occurrence and duration may be negatively associated with attention (Luo et al., 2021). This study also showed that microstate properties such as occurrence and coverage for specific microstates were associated with regulatory abilities (i.e., inhibitory control) within the ADHD group.

Associations of microstate characteristics with internalizing and negative affectivity temperament behaviors are also mixed. In infants, negative affectivity temperament behaviors were negatively correlated with coverage and duration of one particular microstate (Brown and Gartstein, 2023). Looking later in development, no association was found between GEV and depressive symptoms in children aged 4–8 years (Bagdasarov et al., 2022), whereas another study found that GEV, coverage, and occurrence of several microstates were linked to anxiety and depressive symptoms in adults (Xue et al., 2024). This discrepancy in results might be explained through differing populations and data collection practices, whereby Bagdasarov et al. conducted their research on a child population during an error-related negativity task, whereas Xue et al. utilized resting state EEG to examine an adult population. Despite growing interest in EEG microstates, little is known about how these spatial and temporal patterns evolve longitudinally across infancy and middle childhood, which is a period of rapid neurodevelopment. Prior research has largely relied on cross-sectional designs and has rarely examined behavioral correlates together with microstate dynamics.

The present study leverages a longitudinal study of resting state electroencephalography data collected at infancy, 3 years, 5 years, and 7 years of age recruited from a database of volunteer participants without any pre identified developmental conditions. We addressed the following questions: 1) How do microstates’ characteristics change across infancy through middle childhood, and 2) Are microstate features associated with concurrent and later temperament characteristics and psychopathology symptoms? We hypothesize that the temporal properties of microstates change across development and that these changes vary by the microstate type and sex (Bagdasarov et al., 2022; Koenig et al., 2002). We hypothesize that the temporal properties of microstates change across development and that these changes vary by the microstate type and sex (Bagdasarov et al., 2022; Koenig et al., 2002). Based on prior work, we hypothesize that age is negatively associated with duration across most microstate patterns and that other features, such as occurrence, coverage, and GEV, will show differing associations (both positive and negative) with age (Ghosh et al., 2025).

Analyses between microstate features with both temperament characteristics and psychopathology were exploratory; thus, we did not have specific a-priori hypotheses about these associations. However, prior work, such as Brown and Gartstein (2023), suggests that, in some microstate classes, occurrence is negatively associated with effortful control and surgency, duration is negatively correlated with negative affectivity and surgency, and coverage is negatively associated with negative affectivity and effortful control (Brown and Gartstein, 2023). Given documented associations of negative affectivity with internalizing behaviors and of surgency and effortful control with externalizing behaviors, we suggest that similar brain-behavior (i.e., microstate-psychopathology) associations may be found for these behaviors (Pérez-Edgar & Guyer, 2014; Oldehinkel, Hartman, de Winter, Veenstra, & Ormel, 2004). Moreover, we hypothesize that these associations are further qualified by age, sex, and microstate class (Bagdasarov et al., 2022; Koenig et al., 2002).

Method

Participants

Participants were recruited during infancy from a registry of families in the Greater Boston region who expressed interest in participating in developmental research. The infancy time period was selected because of the large amount of growth and development in the brain, behavior, and socio-emotional development. Families in this analysis were part of a prospective longitudinal study investigating early emotion processing. Exclusion criteria included known prenatal or perinatal complications, maternal use of medications during pregnancy (e.g., anticonvulsants, antipsychotics, opioids), known vision issues, and known neurological disorders. Data from participants were excluded if, post-enrollment, the child was diagnosed with autism spectrum disorder and/or a genetic/medical condition affecting neurodevelopment. Children were followed at ages 3, 5, and 7 years (M_age infancy = 7.91 months (SD = 2.85), M_age 3 years = 38.13 months (SD = 1.88), M_age 5 years = 63.00 months (SD = 2.54), M_age 7 years = 88.91 months (SD = 3.48 months)). The current analyses included 383 participants (infancy n = 359, 3 years n = 185, 5 years n = 155, 7 years n = 100; note the n’s reflect EEG data retained after preprocessing). Individual participants contributed an average of 2.08 (SD = 1.03) data points of usable EEG data. Sociodemographic information is presented in Table 1. Families were recruited from an urban area, and children in the sample were predominantly non-Hispanic White. Parental education levels and annual household income show that families were mostly from middle-to-high socioeconomic status. By design, not every child who was recruited was followed up with due to budgetary and staffing constraints. Furthermore, the 7-year study visits were disrupted by the COVID-19 pandemic. Correlational analyses were conducted between the number of completed visits and participant characteristics to characterize patterns of missing data. Families with higher incomes were more likely to complete more visits, and children with higher levels of surgency in infancy and at 3 years of age, as well as higher negative affectivity at 3 years of age, were also less likely to remain in the study (see Supplementary Figure 20). The Institutional Review Board at Boston Children’s Hospital approved all study procedures. In line with the Declaration of Helsinki, written informed consent was obtained from parents before participation at each time point, and child assent was obtained when children were 7 years of age. Families received monetary compensation for their time.

Table 1.

Demographic information for the included study sample (N = 383)

Variable	Level	Summary
Birth_Weight		3559.41 (679.04)
ageM_Infancy		7.91 (2.85)
ageM_3y		38.13 (1.88)
ageM_5y		63.00 (2.54)
ageM_7y		88.91 (3.48)
Sex	Male	204 (53.3%)
	Female	179 (46.7%)
Ethnicity	NonHispanic	339 (88.5%)
	Hispanic	39 (10.2%)
	Did not respond	5 (1.3%)
Race	White	305 (79.6%)
	More than one Race	48 (12.5%)
	Asian	15 (3.9%)
	Black	9 (2.3%)
	Did not respond	4 (1.0%)
	Pacific Islander	2 (0.5%)
P1_Education	Masters Degree	164 (42.8%)
	Bachelors Degree	114 (29.8%)
	M.D., Ph.D., J.D. or Equivalent	78 (20.4%)
	High School/GED	16 (4.2%)
	Associates Degree	7 (1.8%)
	Did not respond	2 (0.5%)
	8th Grade or Less	1 (0.3%)
	Some High School	1 (0.3%)
P2_Education	Masters Degree	118 (30.8%)
	Bachelors Degree	111 (29.0%)
	M.D., Ph.D., J.D. or Equivalent	97 (25.3%)
	High School/GED	31 (8.1%)
	Associates Degree	19 (5.0%)
	Did not respond	6 (1.6%)
	Some High School	1 (0.3%)
Family_Income	$100,000 and greater	241 (62.9%)
	$75,000 through $99,999	51 (13.3%)
	$50,000 through $74,999	37 (9.7%)
	Did not respond	29 (7.6%)
	$35,000 through $49,999	15 (3.9%)
	$25,000 through $34,999	5 (1.3%)
	$16,000 through $24,999	3 (0.8%)
	Less than $5,000	2 (0.5%)

Measures

Electroencephalography (EEG)

EEG Acquisition.

At infancy and age 3 years, 2 minutes of EEG data were recorded while participants viewed a movie featuring moving toys and sounds (Kelsey et al., 2025; Vincent et al., 2021). At ages 5 and 7 years, 4 minutes of EEG data were recorded as participants viewed silent, slow-moving light animations (Sacks et al., 2025). Scalp EEG was recorded using a 128-channel HydroCel Geodesic Sensor Net (HGSN; Electrical Geodesic Inc.) with Ag/AgCl-coated, carbon-filled electrodes. The net was connected to a NetAmps 300 amplifier (Electrical Geodesic Inc.), and data were recorded at 500 Hz with Cz as the online reference. Impedances were maintained at ≤100 kΩ, which is within recommended guidelines given the high-input impedance capabilities of the system’s amplifier.

Preprocessing.

EEG data were exported from Net Station (Electrical Geodesic Inc.) and processed in MATLAB (R2023a) using the Harvard Automated Preprocessing Pipeline for EEG (HAPPE) version 4.1 (Gabard-Durnam et al., 2018). Data were high-pass filtered at 1 Hz, low-pass filtered at 20 Hz (Nagabhushan Kalburgi et al., 2024), resampled to 250 Hz (Bagdasarov et al., 2024; Bagdasarov et al., 2025), and preprocessed using cleanline noise removal, bad channel rejection, and artifact removal via waveletting. Waveletting entails first performing an ICA (independent component analysis) decomposition of the EEG signal into components, then a wavelet transform and threshold to remove artifacts before translating back to the EEG channel format (Castellanos and Makarov, 2006). Interpolated channels replaced bad channels after artifact removal. Data were re-referenced to the average reference, detrended, and segmented into 2-second epochs (following a similar procedure as Ghosh et al., 2025). Epochs with artifacts were removed using amplitude (min −150 and max 150) and joint probability criteria. Joint probability criteria is the evaluation of the average log power from 1 to 125Hz across all electrodes. Electrodes whose probability falls more than 3 standard deviations from the mean are removed as bad channels.

Exclusion thresholds included recordings with fewer than 30 usable epochs (60 seconds of EEG) and < 80% usable channels. Overall, 37 individuals at specific time points were excluded based on these criteria (infancy n = 30, 3 years n = 3, 5 years n = 4, 7 years n = 0). See Supplementary Figures 1–4 for associations between data quality and microstate features. As a few associations between data quality and microstate features were found, the percentage of segments retained was used as a covariate in subsequent analyses. Descriptive statistics for EEG data quality are in Supplementary Table 51.

Microstate computation.

Microstate analysis was conducted using a Microstate EEGLAB toolbox using all recommended default parameters (Poulsen et al., 2018). First, the global field power (GFP), or the spatial standard deviation of the EEG amplitude across all electrodes (Lehmann & Skrandies, 1980), was computed. The scalp topographies at the GFP maxima were then entered into the modified k-means clustering algorithm (Koenig et al., 2002; Pascual-Marqui et al., 1995), and 3 to 8 microstate prototypes were computed. The following parameters were used: 50 random initializations, 1000 maximum number of iterations, 1000 GFP peaks (4 seconds total of individual time points associated with the local maxima) per subject entered the segmentation, 10 ms minimum distance between peaks, as recommended by Poulsen et al. (2018). GFP peaks represent time points when the topographical EEG maps are most stable and have the strongest signal-to-noise ratio. Using these peaks focuses the clustering on the most meaningful brain activity patterns, rather than on every single time point (which could include a lot of noise). The selection of the number of microstates was based on measures of fit (global explained variance [GEV] and cross-validation [CV] criterion). The microstate solution was chosen based on the default parameters (maximizing GEV and minimizing CV). These microstate classes were then back-fitted to each participant’s EEG data and temporally smoothed using default parameters in the microstate toolbox. To mitigate artifacts, short time frames (shorter than 30 ms) were prohibited from being assigned to different classes (Musaeus et al., 2019; Poulsen et al., 2018). This method reclassifies the labels of those segments to the next best-fitting microstate, as measured by global map dissimilarity (Poulsen et al., 2018). After which, temporal properties (occurrence, duration, coverage, and GEV) for each microstate were computed for each individual.

Temperament (Infancy, 3, 5, and 7 years)

Temperament was assessed using a series of well-established parent-report questionnaires. The Infant Behavior Questionnaire-Short Form (IBQ-SF) was administered at infancy, The Early Child Behavior Questionnaire – Short Form (ECBQ-SF) at age 3 years, and the Child Behavior Questionnaire – Very Short Form (CBQ-VSF) at ages 5 years and 7 years (Gartstein & Rothbart, 2003; Putnam et al., 2006; Putnam et al., 2014; Rothbart et al., 2001). These questionnaires ask parents to rate children’s likelihood to respond in a particular way or display a set of behaviors using an 8-point scale ranging from 0 (never) to 7 (always). Negative affectivity was computed by averaging responses across the subscales, such as Shyness, Sadness, Falling Reactivity, and Fear (note that subscales contributing to the factor scores differ slightly among assessment ages), with higher scores indicating greater negative affectivity. Orienting/regulation and effortful control, called effortful control from this point forth, was computed by averaging responses across the subscales, such as attention focusing, attention shifting, inhibitory control, and low-intensity pleasure, with higher scores indicating greater effortful control. Surgency was computed by averaging subscales such as approach, vocal reactivity, high-intensity pleasure, impulsivity, sociability, smiling and laughter, activity level, and perceptual sensitivity, with higher scores indicating greater surgency. The three factors showed acceptable internal consistency at each time point (Negative affectivity: infancy α = .86, 3 years α = .86, 5 years α = .70, 7 years α = .74; Effortful control: infancy α = .81, 3 years α = .84, 5 years α = .72, 7 years α = .75; Surgency: infancy α = .92, 3 years α = .82, 5 years α = .74, 7 years α =.75).

Child mental health symptoms

The Infant Toddler Social-Emotional Assessment (ITSEA) and the Child Behavioral Checklist (CBCL) are questionnaires that have been validated for the assessment of emotional and behavioral problems in children (Achenbach & Rescorla, 2001; Achenbach & Rescorla, 2000; Achenback & Rescorla, 2001; Carter, 2013). When the child was 3 years old, parents completed the ITSEA; when the child was 5 years old, parents completed the CBCL 1½ -5; and when children were 7 years old, parents completed the CBCL 6–18. Forms were completed online using the REDCap survey platform (Harris et al., 2019; Harris et al., 2009) or in-person during the laboratory visit. For each of the questionnaires, caregivers were asked to assess how often particular behaviors occurred using a 3-point Likert scale with time frames ranging from not specified to 6 months. T-scores were calculated for two broad-band scales, Internalizing Problems (subscales included: anxious/depressed, somatic complaints, and withdrawn; 3-year α = .82; 5-year α = .82; 7-year α =.82) and Externalizing Problems (subscales included: aggressive behaviors and attention problems; 3-year α = .79; 5-year α = .90; 7-year α = .89), with higher scores indicating greater symptoms.

Transparency and openness

We report data exclusions, all manipulations, and all measures in the study, and we follow JARS (Appelbaum et al., 2018). EEG data were processed using the Harvard Automated Preprocessing Pipeline for EEG (HAPPE) version 4.1 (Gabard-Durnam et al., 2018) and microstates were computed using the Microstate EEGLAB toolbox (Poulsen et al., 2018). Analyses were conducted using R version 4.4.2 (R core team, 2022) using the lme4 package (lmer function) and the reghelper package (simple_slopes function). Figures were made using the corrplot and ggplot2 packages. This study’s design and its analysis were not pre-registered. The data and code that support the findings of this study are available from the corresponding author upon reasonable request. This method was chosen to promote participant confidentiality and because sharing de-identified data on a public platform was not included as part of the consent process.

Analysis Plan

Prior to conducting analyses, outlier detection was performed to identify extreme outliers (Median Absolute Deviation > 3). A total of 1,211 (5.3%) values across all times and microstate variables were identified and removed. This approach has been used in other work (Bagdasarov et al., 2024) and may be especially important in microstate work, as group-based microstate patterns may not be a fit for a particular individual. Sensitivity analyses were conducted with all values included and are presented in the supplementary materials to assess potential influences of this decision. Overall, the results are similar except for the associations between developmental trajectories of microstates and child mental health symptoms, which were all non-significant when outliers were removed.

Analyses were conducted using R version 4.4.2 (R core team, 2022) using the lme4 package (lmer function) and the reghelper package (simple_slopes function). Figures were made using the corrplot package and ggplot2. In the first set of analyses, a series of linear mixed effects models were conducted with each microstate feature (occurrence, duration, coverage, and GEV) as a separate outcome variable. These models used restricted maximum likelihood to account for missing data and included main and interactive fixed effects of age in months, sex, and microstate number, with a random effect of subject. The percentage of EEG segments retained was added as a fixed effect to control for associations due to data quality. Two sets of simple slope analyses were conducted as a follow-up assessment. The first set of simple slopes assessed the association between age and microstate feature separately for each microstate number and sex. The second set of simple slopes assessed the differences between sex for each microstate feature separately for each age group (groupings done at 7, 36, 60, and 84 months) and microstate number. Bonferroni corrected p-values are reported below, and only the associations that survived Bonferroni correction are discussed in the main text.

Next, a series of linear mixed effects models were conducted with the main and interactive fixed effects of age in months, sex, microstate number, and microstate features (occurrence, duration, coverage, and GEV) with a random effect of subject for each of the child behavior outcomes (negative affectivity, effortful control, surgency, internalizing symptoms, and externalizing symptoms). The percentage of EEG segments retained was added as a fixed effect to control for associations due to data quality. Child temperament was assessed at all four time points (infancy, 3 years, 5 years, 7 years). In contrast, child mental health (internalizing and externalizing symptoms) was assessed at three time points (3 years, 5 years, 7 years). Significant interactions in the omnibus models were probed using simple slopes analyses. Specifically, the association between microstate value and behavior is tested for each microstate number and for each age group (groupings done at 7, 36, 60, and 84 months). Bonferroni corrected p-values are reported below, and only the associations that survived Bonferroni correction are discussed in the main text.

Results

Microstate solution

The number of microstates was determined by assessing fit (i.e., maximizing GEV and minimizing CV). Overall, a 7-microstate solution was found to fit the data best (GEV = .93; CV = .00; see Supplementary Table 1 for fit statistics and Figure 1 for topographies of the microstates found). Microstate 1 has a centralized anterior dipole and a diffuse posterior dipole (similar to microstate C). Microstate 2 has an anterior-posterior configuration with an emphasis on the posterior dipole. Microstate 3 has a diffuse anterior dipole and a centralized posterior dipole (similar to microstate E). Microstate 4 has a left lateral posterior to right lateral anterior dipole (similar to microstate A). Microstate 5 has a left lateral posterior to right central anterior dipole. Microstate 6 has a left lateral anterior to right lateral posterior dipole (similar to microstate B). Microstate 7 has a central dipole (similar to microstate D). Correlations among the brain variables of interest are presented in Supplementary Figures 1–4. Summary statistics for all variables are in Supplementary Tables 45–50.

Figure 1. Topographies for each microstate solution for the full group and individual time points.

Note. The top of the topographical maps are the front of the heads and the bottom of the topographical maps are the back of the head. MS = microstate.

Follow-up analyses examined the microstate solution for each individual time point. In infancy, a six microstate solution was the best-fitting model. The individual infant solution had similar topographies to the group solution, with the exception of the group microstate 5. At ages 3-, 5-, and 7-years, an eight microstate solution was found to fit the data best. The individual 3-year solution had similar topographies to the group solution, with the exception of the group microstate 2. Similarly, the individual 5- and 7-year solutions had similar topographies to the group solution, with the exception of the group microstate 5. In addition, there is an emergence of microstates for the 3-, 5-, and 7-year time points not seen in the group solution, including one microstate with a similar topography to microstate G (see Figure 1 for more details; fit statistics for individual solutions are in Supplementary Table 52). The group-based solution is used in all subsequent analyses.

Developmental trajectories and moderation by sex for microstates.

There were significant age by sex by microstate number interaction effects for each of the microstate features (p < .001; see Supplementary Tables 2–9 and Figures 2–5). The results of the simple slope analyses are detailed below.

Figure 2. Developmental trajectories of occurrence for each of the 7 microstates.

Note, There were significant age by sex by microstate number interaction for occurrence. There were positive associations seen across both sexes between age and occurrence for microstates 1, 3, 4, and 6, and a negative association between age and occurrence for microstate 2 (p < .001). For microstate 5, the association between age and occurrence was negative for females (p < .001) and positive for males (p < .001). For microstate 7, there was a negative association between age and occurrence for females (p < .001) and no association between age and occurrence for males (p = 1.00). There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had smaller occurrences than females for microstates 3, 4, and 6 (p < .001), and males had larger occurrences than females for microstates 5 and 7 (p < .001).

Figure 5. Developmental trajectories of GEV for each of the 7 microstates.

Note, There were significant age by sex by microstate number interaction for GEV. There were positive associations found for both sexes between age and GEV for microstates 1, 3, 4, and 6 (p < .001) and a negative association between age and GEV for microstate 2 (p < .001). For microstate 5, there was a positive association between age and GEV for males (p = .014), and a negative association between age and GEV for females. For microstate 7, there was a negative association between age and GEV for females (p = .028). There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had larger GEVs than females for microstates 5 and 7 (p < .001), and males had smaller GEVs than females for microstates 3, 4, and 6 (p < .001). At 60 and 84 months, males had larger GEVs than females for microstate 1 (p < .028).

Occurrence.

To test for age effects, simple slopes analyses were conducted between age and occurrence for each combination of microstate number (1–7) and sex (male, female). There were positive associations seen across both sexes between age and occurrence for microstates 1, 3, 4, and 6, and a negative association between age and occurrence for microstate 2 (p < .001). For microstate 5, the association between age and occurrence was negative for females (p < .001) and positive for males (p < .001). For microstate 7, there was a negative association between age and occurrence for females (p < .001) and no association between age and occurrence for males (p = 1.00).

To test for sex effects, simple slopes analyses were conducted to assess differences between males and females for occurrence for each microstate and age pairing (simple slope analyses were conducted at the following ages: 7, 36, 60, and 84 months). There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had smaller occurrences than females for microstates 3, 4, and 6 (p < .001), and males had larger occurrences than females for microstates 5 and 7 (p < .001). Findings are summarized in Supplementary Tables 2 and 3 and Figure 2.

Duration.

There were negative associations found for both sexes between age and duration for microstates 1, 2, 3, 5, and 7 (p < .001). For microstate 6, there was a positive association between age and duration for females (p = .014), and no association for males (p = 1.00).

There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had longer durations than females for microstates 5 and 7 (p < .001). At 36, 60, and 84 months, males had shorter durations than females for microstate 4 (p < .001). At 84 months, males had shorter durations than females for microstate 3 (p = .028). Findings are summarized in Supplementary Tables 4 and 5 and Figure 3.

Figure 3. Developmental trajectories of duration for each of the 7 microstates.

Note, There were significant age by sex by microstate number interaction for duration. There were negative associations found for both sexes between age and duration for microstates 1, 2, 3, 5, and 7 (p < .001). For microstate 6, there was a positive association between age and duration for females (p = .014), and no association for males (p = 1.00). There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had longer durations than females for microstates 5 and 7 (p < .001). At 36, 60, and 84 months, males had shorter durations than females for microstate 4 (p < .001). At 84 months, males had shorter durations than females for microstate 3 (p = .028).

Coverage.

There were positive associations for both sexes between age and coverage for microstates 4 and 6, and a negative association between age and coverage for microstates 2 and 7 (p < .001). For microstate 1, there was a positive association between age and coverage for males (p < .001). For microstate 3, there was a positive association between age and coverage for females (p < .001). For microstate 5, there was a negative association between age and coverage for females (p < .001).

There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had larger coverages than females for microstates 5 and 7 (p < .001), and males had smaller coverages than females for microstates 4 and 6 (p < .001). At 60 and 84 months, males had smaller coverages than females for microstate 3 (p < .001), and males had larger coverages than females for microstate 1. Findings are summarized in Supplementary Tables 6 and 7 and Figure 4.

Figure 4. Developmental trajectories of coverage for each of the 7 microstates.

Note, There were significant age by sex by microstate number interaction for coverage. There were positive associations for both sexes between age and coverage for microstates 4 and 6, and a negative association between age and coverage for microstates 2 and 7 (p < .001). For microstate 1, there was a positive association between age and coverage for males (p < .001). For microstate 3, there was a positive association between age and coverage for females (p < .001). For microstate 5, there was a negative association between age and coverage for females (p < .001). There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had larger coverages than females for microstates 5 and 7 (p < .001), and males had smaller coverages than females for microstates 4 and 6 (p < .001). At 60 and 84 months, males had smaller coverages than females for microstate 3 (p < .001), and males had larger coverages than females for microstate 1.

GEV.

There were positive associations found for both sexes between age and GEV for microstates 1, 3, 4, and 6 (p < .001) and a negative association between age and GEV for microstate 2 (p < .001). For microstate 5, there was a positive association between age and GEV for males (p = .014), and a negative association between age and GEV for females. For microstate 7, there was a negative association between age and GEV for females (p = .028).

There were no differences between males and females at 7 months. At 36, 60, and 84 months, males had larger GEVs than females for microstates 5 and 7 (p < .001), and males had smaller GEVs than females for microstates 3, 4, and 6 (p < .001). At 60 and 84 months, males had larger GEVs than females for microstate 1 (p < .028). Findings are summarized in Supplementary Tables 8 and 9 and Figure 5.

Overall, the findings indicate that there are significant changes occurring for each of the microstate features and individual microstate numbers across development. Moreover, the developmental pattern appears to differ between males and females.

Associations between developmental trajectories of microstates and child temperament.

Child temperament.

A series of linear mixed effects models were conducted to test the main and interactive effects of age in months, sex, microstate number, and microstate features (occurrence, duration, coverage, and GEV) on child temperament (negative affectivity, effortful control, and surgency). Significant effects with microstate number and microstate feature value are reported below. Across all analyses, no significant interactive effects between sex and microstate number or sex and microstate feature were found.

Occurrence and child temperament.

There was a significant age by microstate number by occurrence interaction for negative affectivity and surgency (p < .001; see Supplementary Tables 10 and 14 and Supplementary Figures 5 and 7). There was a significant age by occurrence interaction for effortful control (p < .001; see Supplementary Table 12 and Supplementary Figure 6). Simple slopes analyses were conducted for each microstate and age pairing. At 7 and 36 months, there was a negative association between occurrence and negative affectivity for microstate 1 (p < .001). At 60 months, there was a negative association between occurrence and negative affectivity for microstate 2 (p = .028). At 7 and 36 months, there was a negative association between occurrence and effortful control (p < .044). At 7 months, there was a positive association between occurrence and surgency for microstates 1, 3, 4, and 6 (p < .012). At 36 months, there was a positive association between occurrence and surgency for microstates 1 and 2 (p < .040). At 60 months, there was a positive association between occurrence and surgency for microstate 2 (p = .004), and a negative association between occurrence and surgency for microstate 3 (p = .040). At 84 months, there was a positive association between occurrence and surgency for microstates 2, 5, and 7 (p < .044) and a negative association between occurrence and surgency for microstates 3, 4, and 6 (p < .020). The results are presented in Supplementary Tables 11, 13, and 15.

Duration and child temperament.

There was a significant microstate number by age by duration interaction for negative affectivity (p = .017). There was a significant microstate by duration interaction (p = .020) and age by duration interaction (p < .001) for effortful control and surgency (see Supplementary Tables 16, 18, and 20 and Supplementary Figure 8–10). Simple slopes analyses were conducted for each microstate and age pairing. At 7 months, there was a positive association between duration and negative affectivity for microstates 2, 3, 4, 5, 6, and 7 (p < .004). At 36 months, there was a positive association between duration and negative affectivity for microstates 3, 4, 5, and 6 (p < .004). At 60 months, there was a negative association between duration and negative affectivity for microstate 2 (p = .028), and a positive association between duration and negative affectivity for microstates 4 and 6 (p < .004). At 84 months, there was a negative association between duration and negative affectivity for microstate 2 (p = .004), and a positive association between duration and negative affectivity for microstate 4 (p = .012). At 7 months, there was a positive association between duration and effortful control for microstates 3, 5, and 7 (p < .004). At 36 months, there was a positive association between duration and effortful control for microstates 3 and 4 (p < .008). At 60 months, there was a negative association between duration and effortful control for microstate 2 (p = .004) and a positive association between duration and effortful control for microstate 4 (p = .028). At 84 months, there was a negative association between duration and effortful control for microstate 2 (p < .001). At 7 months, there was a negative association between duration and surgency for microstates 1, 2, 3, 4, 5, 6, and 7 (p < .028). At 36 months, there was a negative association between duration and surgency for microstates 3, 4, 5, and 6 (p < .016). At 60 months, there was a negative association between duration and surgency for microstate 6 (p = .020). At 84 months, there was a positive association between duration and surgency for microstate 2 (p = .012). The results are presented in Supplementary Tables 17, 19, and 21.

Coverage and child temperament.

There was a significant age by microstate by coverage interaction for negative affectivity, effortful control, and surgency (p < .001; see Supplementary Tables 22, 24, and 26 and Supplementary Figure 11–13). Simple slopes analyses were conducted for each microstate and age pairing. At 60 and 84 months, there was a negative association between coverage and negative affectivity for microstate 2 (p < .028). At 60 and 84 months, there was a negative association between coverage and effortful control for microstate 2 (p < .028). At 7 months of age, there was a positive association between coverage and surgency for microstate 1 (p < .001) and a negative association between coverage and surgency for microstate 5 (p < .001). At 36 months, there was a positive association between coverage and surgency for microstate 1 (p < .001). At 60 months, there was a positive association between coverage and surgency for microstate 2 (p = .028). At 84 months, there was a positive association between coverage and surgency for microstate 2 (p < .001) and a negative association between coverage and surgency for microstates 3 and 6 (p < .028). The results are presented in Supplementary Tables 23, 25, and 27.

GEV and child temperament.

There was a significant age by microstate by GEV interaction and age by sex by GEV interaction for negative affectivity (p < .031; see Supplementary Table 28 and Supplementary Figure 14). There was also a significant age by microstate by GEV interaction for effortful control and surgency (p < .010; see Supplementary Table 30 and 32 and Supplementary Figures 15 and 16). Simple slopes analyses were conducted for each microstate and age pairing. At 7 months, there was a negative association between GEV and negative affectivity for microstates 1, 3, and 4 (p < .028). At 36 months, there was a negative association between GEV and negative affectivity for microstates 1 and 7 (p < .001). At 7 months, there was a negative association between GEV and effortful control for microstate 1 (p = .028). At 60 months, there was a negative association between GEV and effortful control for microstate 2 (p < .001). At 84 months, there was a negative association between GEV and effortful control for microstate 2 (p = .028) and a positive association between GEV and effortful control for microstate 3 (p < .001). At 7 months, there was a positive association between GEV and surgency for microstate 1 (p < .001). At 36 months, there was a positive association between GEV and surgency for microstate 1 (p = .028). At 84 months, there was a negative association between GEV and surgency for microstate 6 (p = .028). The results are presented in Supplementary Tables 29, 31, and 33.

Associations between developmental trajectories of microstates and child mental health symptoms.

Child mental health symptoms.

A series of linear mixed effects models were conducted to test the main and interactive effects of age in months, sex, microstate number, and microstate features (occurrence, duration, coverage, and GEV) on child mental health symptoms (internalizing and externalizing symptoms). Significant effects with microstate number and microstate feature value are reported below. Across all analyses, no significant interactive effects between sex and microstate number or sex and microstate feature were found.

Occurrence and child mental health symptoms.

There were no significant main or interactive effects between occurrence and either internalizing or externalizing symptoms.

Duration and child mental health symptoms.

There were no significant main or interactive effects between duration and internalizing symptoms. There was a significant age by microstate number by duration interaction for externalizing symptoms (p = .019; see Supplementary Table 37 and Supplementary Figure 17). Simple slopes analyses were conducted for each microstate and age pairing, but none survived Bonferroni multiple comparisons correction.

Coverage and child mental health symptoms.

There were no significant main or interactive effects between coverage and internalizing symptoms. There was a significant age by microstate number by coverage interaction for externalizing symptoms (p = .014; see Supplementary Table 40 and Supplementary Figure 18). Simple slopes analyses were conducted for each microstate and age pairing, but none survived Bonferroni multiple comparisons correction.

GEV and child mental health symptoms.

There were no significant main or interactive effects between GEV and internalizing symptoms. There was a significant age by microstate number by GEV interaction for externalizing symptoms (p = .023; see Supplementary Table 43 and Supplementary Figure 19). Simple slopes analyses were conducted for each microstate and age pairing, but none survived Bonferroni multiple comparisons correction.

Discussion

Microstate analysis of EEG data provides insights into moment-to-moment neural activity often ignored with other measures and modalities. In the present study, the developmental changes in temporal properties of neural activity were assessed in a large longitudinal sample using EEG microstates. Associations with early temperament and psychopathology were also examined. For analyses including all individuals at all time points, a 7-microstate solution was found to fit the data best. When assessing the developmental trajectories of the microstates, there were significant age by sex by microstate interactions found for each microstate feature (occurrence, duration, coverage, and GEV). Further, microstate patterns were associated with child behavioral temperament and externalizing symptoms.

When assessing microstate features (Occurrence, Duration, Coverage, and GEV), there were significant age by sex by microstate number interactions pointing to a complex developmental trajectory. Duration is characterized by the stability of a microstate based on the average time the microstate was present. There is some evidence to support that this feature is indicative of relative excitation to inhibitory synapse activity, important for optimal regulation of neural functions (Ahmad et al., 2022; Brown & Gartstein, 2023). In line with our hypothesis, the majority of microstates decreased in duration for both sexes, with the exceptions of microstate 4 (where age was not significantly associated with duration in either sex) and microstate 6 (where age was positively associated with duration in females but not males). These results expand upon prior findings, which showed a decrease in duration across the first two years of life (Ghosh et al., 2025), by demonstrating that this pattern continues throughout early and middle childhood. These findings also lend support that as children mature, they can more efficiently switch between brain states (Koenig et al., 2002).

One possible reason why the developmental trajectory of occurrence, duration, coverage, and GEV for particular microstates is flat or positive is that it reflects an emergence of these microstates. Therefore, individual analyses were conducted at each time point to better understand microstate class emergence. Here, classes with no or positive associations between age and duration (microstate 4 [A] and microstate 6 [B]) were present across all individual solutions. Therefore, it is less likely to be an emergence but rather a continual fine tuning. However, other microstates emerged after the infancy timepoint, such as one microstate similar in topography to microstate G. This developmental pattern is interesting to consider in the context of what is known about the development of functional brain networks. Many higher-order brain networks are thought to rapidly develop during the infancy period and be refined more slowly throughout the childhood period (Chen et al., 2021; Kelsey et al., 2025). Therefore, one may not expect a large increase from 3 to 7 years of age. This change raises the possibility of EEG microstates being sensitive to components of brain coordination that differ from functional MRI connectivity measures. Other findings, such as GEV and Coverage increasing across age for certain microstates and decreasing across age for other microstates, may be in line with the network segregation findings from functional MRI. Thus supporting the notion that continued refinement and increased segregation between networks (e.g., frontoparietal and default mode networks) continue throughout childhood (Richardson et al., 2018). Future work using synchronous EEG and fMRI will help to shed light on the similarities and differences between the two sets of measures.

The developmental patterns were further qualified by sex effects with significant differences between males and females emerging by 3 years of age. Several previous studies, including EEG studies of early to middle childhood, have found sex effects (Bagdasarov et al., 2022; Hill et al., 2023). For example, prior work with 4- to 8-year-old children found males had lower GEVs than females (Bagdasarov et al., 2022). Taken together, microstate features appeared to differ between males and females in the current analyses, pointing to differences in brain development between males and females. Continued consideration of sex effects is needed in future work and further discussion of the implications of the sex differences is reviewed below in the context of the behavioral findings.

To begin to understand the behavioral function of microstate properties, analyses were conducted to test the associations with child temperament. Here, similar complex age by microstate interactions were found for negative affectivity, effortful control, and surgency. In partial alignment with our hypothesis, based on Brown and Gartstein (2023), that occurrence would be negatively associated with effortful control and surgency, occurrence was negatively correlated with effortful control at the younger ages, whereas occurrence was positively associated with surgency for several microstates. Similarly, in partial alignment with our hypothesis that duration is negatively correlated with negative affectivity and surgency, we found that duration was negatively associated with surgency and positively associated with negative affectivity. Furthermore, in line with our hypothesis, coverage was negatively associated with negative affectivity and effortful control. Here, there are several key differences between the present study and the Brown and Gartstein study that are important to note when considering differences in the results. The present study had a much wider age range, the paradigm did not feature music, and there was a larger microstate solution (7 vs 4).

We also found that duration was positively associated with effortful control. This is in line with other work with infants, which has found that the duration of sustained attention (a component of effortful control) is also associated with an increased duration of a microstate with fronto-central spatial topography (Bagdasarov et al., 2024). However, the present results are in contrast to work with older children, showing elevated duration times in certain microstates for ADHD compared to neurotypical children (Luo et al., 2021). Similar paradoxical findings have been found for other EEG features such as aperiodic slope, where larger exponents are found in infants with an elevated likelihood for ADHD (i.e., a first-degree relative) and smaller exponents are found in adolescents with ADHD (Karalunas et al., 2022; Ostlund et al., 2022; Ostlund et al., 2021).

There were also a few individual microstate topographies that were consistently linked to temperamental traits. Microstates 2, 3, and 4 were associated with effortful control, and microstate 1 was associated with surgency. Several microstates have been linked to particular functional brain network patterns, such as the default mode and fronto-parietal networks. Both of these networks are thought to be involved in regulation and emotion processing (Kelsey et al., 2021; Allison et al., 2025). Therefore, these microstates may be reflective of functional network properties important for behavioral temperament.

To assess if similar brain patterns underlie both temperament and psychopathology, analyses were conducted to assess relations with internalizing and externalizing symptoms. Here, only age by microstate number by microstate feature interactions were found for externalizing symptoms, and no significant interactions were found for internalizing symptoms. It is interesting to consider why there would be an association between microstate features and negative affectivity and not internalizing symptoms, given the close tie between aspects of negative affectivity (e.g., fear and behavioral inhibition) and internalizing symptoms (Pérez-Edgar & Guyer, 2014; Rankin Williams et al., 2009). One possible reason why associations were found only for temperament is that these analyses, unlike the psychopathology analyses, included both EEG data and behavioral assessments during the infancy time period.

Despite the moderation by sex found for the developmental trajectories of the microstate features, there were no significant sex by microstate interactions found for behavior (with the singular exception of GEV and negative affectivity). This is particularly interesting to consider in light of sex differences found across aspects of psychopathology (Rapee et al., 2019; Zahn-Waxler et al., 2008). Here, it is important to acknowledge that this sample is a general population sample and thus the number of children meeting clinical criteria thresholds for psychopathology is relatively low compared to clinical samples. Therefore, it may be that sex by microstate feature interactions are detected in cases with greater variance in psychopathology or temperament traits. It is also possible that the differences in brain temporal properties between males and females develop early in childhood, but the differential impact of this developmental pattern on relevant behaviors becomes apparent only later in development. In particular, diverging rates of internalizing versus externalizing problems by sex may be a byproduct of increases in vulnerabilities to mental health challenges. Future work should examine these constructs through middle childhood and adolescence to explore these possibilities.

The overall topographic patterns in the present study resemble patterns found in other work with children and adults (Al Zoubi et al., 2019; Bagdasarov et al., 2024; Bagdasarov et al., 2022; Luo et al., 2023). One possible interpretation of this is that the topographic patterns and global networks are in place from a young age, but the temporal properties change with development (Bagdasarov et al., 2022; Muetzel et al., 2016). Despite similarities in topographic features between studies, there are also inherent differences across studies, with microstate solutions typically ranging from 4 to 7 microstates (Adibpour et al., 2025; Bagdasarov et al., 2024; Brown & Gartstein, 2023). The large number of solutions was consistent in the group solution and individual time point solution, making the large age range an unlikely reason for the large number of microstate classes. Related to this idea, the main analysis used the microstate solution was computed based on all time points, i.e., ages, together. This decision was made to allow for the comparison of microstate features across development. However, it is possible that some patterns for specific microstates would differ if we used the individualized solutions. Future work should continue to use both individual and group solutions to better understand how this decision impacts the interpretation of microstate work.

It is also possible that the measurement context impacts the microstate solution. In the present study, the children viewed either moving toys or light displays in their resting-state task. Whereas other work has examined microstates under sleep (7 microstate solution) or eyes closed (5 microstate solution) conditions (Adibpour et al., 2025; Hill et al., 2023). One study of young children (4–8 years of age) computed microstate profiles across two conditions and found that fewer microstates emerged during an eyes-open resting state task as opposed to an error-related negativity task (Bagdasarov et al., 2024). Another study assessing microstate profiles across measurement contexts in infancy found that topographies were similar across contexts, but the temporal properties differed (McDevitt & Gartstein, 2025). Measurement conditions may also be of particular relevance for associations between microstates and behavior. This raises the possibility that unique microstates emerge during relevant contexts (e.g., when the subject is experiencing stress or viewing highly emotional stimuli) and these microstates are of particular relevance to a particular behavior or psychological outcome (e.g., internalizing symptoms). Therefore, it will be important for future work to continue to consider the measurement context when comparing work across studies and closely examine the state versus trait properties of microstates.

When assessing the patterns of microstates found in the present study, it is important to acknowledge the naming schemes used. The present study chose to label microstates using numbers as opposed to lettering conventions (e.g., microstate A, right-frontal to left-posterior). This decision was made as many of the meta-analytic databases are based on adult data, and more work is needed to understand the structure and function of microstates in pediatric populations (Bagdasarov et al., 2022). However, this decision makes comparisons across studies more difficult. Researchers should consider submitting pediatric data to meta-analytic efforts to inform microstate naming schemes.

This study should also be interpreted within the context of its strengths and limitations. Strengths include its relatively large longitudinal sample, spanning infancy through 7 years of age. Limitations include the relatively homogeneous, middle-to-high socioeconomic status study sample recruited from an urban area. Furthermore, study retention was associated with some sociodemographic and behavioral features of interest (e.g., income, surgency, and negative affectivity), which may have restricted the range of temperament and socio-economic status represented and potentially weakened the power to detect associations at particular time points. Future work should investigate associations in a sample that has larger variability in temperament and socio-economic status across age groups.

Therefore, future studies should examine whether the results hold when assessing more diverse samples, including those with higher rates of clinically relevant mental health problems. In addition, longitudinal studies that extend into late childhood and early adolescence are important to explore how developmental changes may continue to unfold in these periods. It is also possible that non-neurophysiological changes affected by development and sex (e.g., skull thickness) impacted some aspects of the EEG measurements and microstate computations. Multimodal research (e.g., MRI and EEG) will be critical to disentangling the impact of such non-neurological contributions.

We also acknowledge the limitations associated with the change in stimuli across developmental periods. The change in stimulus presentation from the infancy and 3-year time points to the 5- and 7-year time points may have introduced confounds related to changes in the auditory and visual presentations. Prior work has shown that differences in stimuli and tasks during the EEG recording session are associated with differences in microstate features (McDevitt & Gartstein, 2025). Although the stimuli differed in audio and visual content, they were deliberately chosen to be age-appropriate and neutral (i.e., slow-moving, non-arousing) to minimize potential confounds and enhance data quality at each age. This approach aligns with established methodological practices in longitudinal neurophysiological analyses of similar age groups (Sacks et al., 2025; Kelsey et al., 2025). It is also possible that the change in recording length from 2 to 4 minutes impacted the reliability of the microstate features. However, a study that tested the impact of recording length found that a high level of internal consistency was achieved with only 1 minute of data (Bagdasarov et al., 2024). Moreover, the observed microstate differences were not limited to the period corresponding to the change in stimulus between 3 and 5 years, making it unlikely that the stimulus change alone accounted for the developmental patterns.

It is also important to acknowledge that our main analyses removed outliers. This approach has been used in other work (Bagdasarov et al., 2024) and may be especially important in microstate research, as group-based microstate patterns may not be a fit for a particular individual. Sensitivity analyses with the full set of data were conducted, and overall, the developmental patterns were robust to outliers. However, the associations between developmental trajectories of microstates and child mental health symptoms appeared to be sensitive to outlier removal. Therefore, additional caution should be taken when interpreting these findings.

Overall, the results of this study revealed developmental changes in the temporal dynamics of microstates across infancy and middle childhood. This work provides insights into normative trajectories of microstate features and helpful groundwork for research exploring trajectories in other populations. Moreover, some microstate patterns and temporal dynamics were associated with aspects of behavior, pointing to the importance of brain coordination and temporal properties.

Public significance statement:

This work characterizes normative trajectories of microstate features and their associations with behavior. Therefore, this work provides groundwork for research exploring trajectories in other populations and highlights the importance of brain coordination and temporal properties.