Default mode network connectivity and temperament are associated with internalizing symptoms in early childhood

Abstract

This study investigates neural and temperament precursors to internalizing psychopathology. Increased default mode network (DMN) functional connectivity and decreased frontoparietal network (FPN) connectivity are established neural signatures of internalizing disorders in adults, but it is unknown if these patterns represent early-emerging vulnerability. In a community sample of typically developing children aged 4–6 years, we examined the concurrent associations of resting-state functional connectivity and temperament with internalizing symptoms. Of the 78 children retained following fNIRS quality control, 66 were included in the analytic sample (age range = 48–83 months; 38 female, 28 male; 78% White; predominantly high SES), We hypothesized that increased DMN connectivity, decreased FPN connectivity, and temperament traits of high negative affect and low surgency would be associated with higher internalizing problems. Functional connectivity was measured using functional near-infrared spectroscopy (fNIRS), a child-friendly neuroimaging technique. Internalizing symptoms and temperament were assessed via parent report using the Child Behavior Checklist and Child Behavior Questionnaire. A multiple linear regression revealed that greater DMN connectivity, higher negative affect, and lower surgency were significantly and independently associated with internalizing symptoms. FPN connectivity was not significantly associated with internalizing symptoms. These findings suggest that a key neural signature of adult depression—DMN hyperconnectivity—is associated with subclinical symptoms in early childhood, supporting its role as a primary vulnerability marker. The absence of an FPN association may reflect the protracted maturation of this cognitive control network, though longitudinal data are required to test whether neural risk factors for internalizing disorders emerge on distinct developmental timetables.

1. Introduction

The terms “internalizing” and “externalizing” are widely used to describe 2 broad-band groupings of behavioral, emotional, and social problems (Achenbach, 1966, Achenbach et al., 2016). Internalizing disorders, which are internally directed negative behaviors, thoughts and emotions that primarily encompass depression and anxiety, constitute a profound and escalating public health challenge (Achenbach, 1978, Solmi et al., 2022; C. H. Miller et al., 2015). Now recognized as the leading cause of morbidity and disability among adolescents, these conditions impose immense social, academic, and personal costs (C. H. Miller et al., 2015). A critical feature of internalizing psychopathology is its early developmental origin. In children as young as 24–36 months old, clinically significant levels of internalizing psychopathology can be detected and subclinical levels of internalizing symptoms are similarly present in toddlers and preschoolers (Egger and Angold, 2006, Luby et al., 2003, Sterba et al., 2007, Zhou and Buss, 2021). Symptoms can be reliably identified in children as young as toddlers, and these early difficulties demonstrate significant stability over time (Tandon et al., 2009). However, findings regarding the developmental stability of internalizing symptoms are mixed. Previous evidence suggests that dimensions of internalizing temperament change from infancy to childhood and continue to evolve into adulthood (Thomas et al., 2019, Zorzetti et al., 2019). Yet other studies have reported stability of internalizing symptoms across development (Bayer et al., 2010, McElroy et al., 2018, Murray et al., 2020) and multiple large scale studies report linear, stable developmental trajectories of internalizing symptoms (Bayer et al., 2010, Flouri et al., 2018, Katsantonis and Symonds, 2023, McElroy et al., 2018, Murray et al., 2020). Notably, preschoolers identified with an internalizing disorder are three times more likely than their peers to exhibit similar psychopathology eight years later (Tandon et al., 2009). Thus, early childhood is a crucial period for understanding etiological pathways.

Despite their prevalence and long-term impact, internalizing disorders are often characterized by quiet, internal distress—such as sadness, worry, and social withdrawal—which can make them difficult for parents, caregivers, and educators to detect (Rubin and Mills, 1991, Tandon et al., 2009). Unlike externalizing problems, which are maladaptive behaviors directed toward one's environment and externally oriented, the "quiet" suffering of an internalizing child may go unnoticed, delaying or preventing access to necessary support (Achenbach et al., 2016, Rubin and Mills, 1991, Yang et al., 2022). This diagnostic challenge, coupled with the high rates of treatment non-response in adolescence (Rubin & Mills, 1991), makes early childhood an important period in which to characterize the correlates of emerging internalizing symptoms. By examining the neurodevelopmental correlates of internalizing symptoms in preschool-aged children, we aim to advance our understanding of the developmental processes underlying early vulnerability to internalizing psychopathology and determine whether neural signatures associated with adult internalizing disorders are already related to variation in subclinical symptoms during an earlier developmental period. Additionally, examining externalizing symptoms alongside internalizing would elucidate whether distinct neural signatures are differentially associated with internalizing versus externalizing dimensions of psychopathology or reflect broader risk for childhood behavioral and emotional problems.

In the search for biological markers of risk, neuroimaging research in adults with major depressive disorder (MDD) has consistently converged on a model of dysregulated large-scale brain networks (Kaiser et al., 2015, Tooley et al., 2022). Two networks are central to this model: the Default Mode Network (DMN) and the Frontoparietal Network (FPN). The DMN, comprising regions such as the medial prefrontal cortex and posterior cingulate cortex, is most active during periods of rest and is implicated in internally oriented, self-referential cognitive processes (Buckner and Schacter., 2008, Kaiser et al., 2015). In adults with MDD, the DMN is frequently found to be hyperconnected, a within network state of excessive functional integration or greater connectivity that is thought to be the neural substrate for rumination—the persistent, repetitive focus on negative thoughts and feelings that is a core symptom of depression (Yan et al., 2019). DMN hyperconnectivity is associated with internalizing symptoms in non-clinical samples as well (Azarias et al., 2025, Luo et al., 2016). In contrast, the FPN (also known as the central executive network), which includes the dorsolateral prefrontal cortex and posterior parietal cortex, is critical for "top-down" cognitive control, goal-directed attention, and the regulation of emotion (Menon, 2011, Ochsner and Gross, 2005). This network is often found to be hypoconnected, or in a state of decreased within network connectivity, in MDD, reflecting a deficit in the neural systems that support adaptive, flexible behavior and emotional modulation (Kaiser et al., 2015, Schultz et al., 2018). Previous work in children also implicates the FPN in internalizing symptoms as left amygdala connections to FPN regions has been linked to withdrawn/depressed symptomology (Thomas et al., 2019). Collectively, these findings support a model of internalizing symptoms characterized by increased within-network connectivity of the DMN and decreased connectivity of the FPN (Kaiser et al., 2015).

Parallel to neurobiological investigations, developmental psychopathology has long established temperament as a primary risk factor for mental illness (Muris and Ollendick, 2005, Nigg, 2017, Rubin and Mills, 1991). Temperament refers to early-appearing, biologically-based individual differences in emotional reactivity and self-regulation that form the foundation of later personality (Muris and Ollendick, 2005, Nigg, 2017). Two core temperament dimensions are robustly implicated in the developmental pathways to internalizing disorders. First, high negative affectivity (also termed emotionality or neuroticism), a constitutional predisposition to experience negative emotions such as fear, sadness, and frustration, is a powerful predictor of both anxiety and depression (Muris and Ollendick, 2005, Nigg, 2017). Second, deficits in regulatory capacity, particularly low effortful control, are strongly linked to psychopathology (Nigg, 2017, Tandon et al., 2009). Effortful control is the ability to inhibit a dominant or prepotent response in favor of a subdominant one, a capacity essential for managing emotional impulses and focusing attention (Nigg, 2017, Tandon et al., 2009). The related construct of Surgency, which encompasses traits of positive emotionality, activity level, and approach behaviors, is also relevant; low surgency often manifests as shyness and behavioral inhibition, key features of an internalizing profile (Gartstein et al., 2012). It is widely theorized that the combination of high reactivity (negative affect) and low regulation (effortful control) confers the greatest vulnerability to psychopathology (Nigg, 2017).

Despite robust, parallel lines of evidence from neuroscience and developmental psychology, few studies have integrated these perspectives by examining neural and temperament risk factors concurrently, especially during the preschool years. Notably, prefrontal cortex activation is associated with children’s effortful control and negative affect (J. G. Miller et al., 2023a; Quiñones-Camacho et al., 2019a). Recent work using a resting state functional connectivity naturalistic viewing paradigm established the feasibility of identifying associations between individual differences in executive function and prefrontal cortex activation (Eng et al., 2025). This work demonstrates that a naturalistic viewing fNIRS task can reliably capture meaningful individual differences in prefrontal connectivity during early childhood. Given that both DMN and FPN connectivity have been implicated in adult internalizing psychopathology, and that temperament traits such as negative affect and surgency are well-established behavioral risk markers, an integrated approach examining both neural and temperament correlates concurrently may better capture early vulnerability to internalizing problems. We build on this methodological foundation by using resting-state fNIRS to test whether DMN and FPN connectivity, alongside parent-reported temperament, are concurrently associated with internalizing symptoms in a community sample of typically developing 4- to 6-year-olds. This developmental period is marked by the rapid maturation of brain circuits and the consolidation of temperament traits, making it a critical window for understanding the origins of psychopathology (Brown and Jernigan, 2012, Campbell, 2017, Center on the Developing Child at Harvard University, 2024). Furthermore, most of the neuroimaging research on internalizing disorders has utilized functional magnetic resonance imaging (fMRI) in adolescent and adult populations, leaving a significant gap in our knowledge of the neural underpinnings of risk in early childhood (Kaiser et al., 2015; C. H. Miller et al., 2015; Solmi et al., 2022).

The remarkable consistency of this DMN/FPN pattern in adults raises a fundamental developmental question: When do these neural signatures of risk emerge? Are they present as a pre-existing vulnerability before the onset of a formal disorder, or do they develop as a "neural scar" resulting from chronic illness? Investigating these network dynamics in young, typically developing children provides a direct test of the hypothesis that these neural patterns are developmental precursors to psychopathology rather than its consequence. Though the Salience network and the Ventral Attention network are shown to be associated with internalizing symptoms (Karl et al., 2026, Pawlak et al., 2022, Sylvester et al., 2013), the present study focuses on the DMN and FPN because these networks are previously implicated in internalizing symptomology and include prefrontal cortical regions involved in self-referential processing, cognitive control, and emotion regulation (Friedman and Robbins, 2022, Grossmann, 2025). These prefrontal regions are also well suited for investigation using fNIRS, enabling the examination of neural systems implicated in internalizing symptoms in young children.

The present study focused on within-network functional connectivity because our primary aim was to determine whether previously established within-network DMN and FPN neural signatures of adult internalizing psychopathology are already evident during early childhood. Although between-network connectivity, particularly interactions involving the DMN and salience network, has also been implicated in internalizing psychopathology (Karl et al., 2026), large-scale functional network organization continues to mature throughout childhood with progressive refinement of interactions between intrinsic brain networks (Grayson & Fair, 2017). As networks become increasingly functionally segregated across development (He et al., 2019; Li et al., 2024; Rosenberg et al., 2020), it is important to characterize the integrity of individual networks to understand the emergence of neural risk for internalizing symptoms. Given our hypothesis-driven focus and the cortical coverage afforded by our fNIRS montage, we therefore restricted analyses to within-network connectivity. Future work with broader cortical coverage will be important for determining how between-network interactions contribute to emerging internalizing symptoms.

Functional near-infrared spectroscopy (fNIRS) is an optical neuroimaging modality particularly well-suited for bridging the gap in our knowledge of the neural underpinnings of internalizing risk in early childhood. As a non-invasive, portable, and relatively motion-tolerant technique, fNIRS makes it feasible to measure cortical brain function in young, active children who cannot easily tolerate the confines of an fMRI scanner (Pinti et al., 2020, Scholkmann et al., 2014, Yücel et al., 2021). The present study leverages this technology to investigate the concurrent associations of resting-state functional connectivity within the DMN and FPN, alongside core temperament traits, for internalizing symptoms in a community sample of 4- to 6-year-old children. Based on the adult and developmental literature, we formed the following hypotheses:

• 1.Higher internalizing symptoms will be associated with increased within-network DMN functional connectivity.
• 2.Higher internalizing symptoms will be associated with decreased within-network FPN functional connectivity.
• 3.Higher internalizing symptoms will be associated with higher temperament negative affect and lower surgency and effortful control.

As an exploratory analysis, we also investigated the relationship between externalizing symptoms and our predictors of interest, including DMN connectivity, FPN connectivity, negative affect, surgency, effortful control, household income and age. By considering externalizing symptoms, we aimed to examine the specificity of the DMN and FPN relationship to internalizing symptoms.

2. Method

2.1. Participants

Seventy-eight children (44 female) between 4 and 6 years of age were retained after fNIRS quality control, of whom 66 had complete data across all study variables and constitute the analytic sample reported below (Mage = 65.8 months, SD = 11.6, range = 48–83 months). Consistent with a Research Domain Criteria (RDoC) framework, the present study employed a dimensional approach in a community sample rather than a case-control design. In line with RDoC principles, participants were not categorized into “healthy” versus “clinical” groups. Instead, the study was designed to capture continuous variability in neurobiological and behavioral functioning across early childhood (Pacheco et al., 2022, Sanislow et al., 2019). This approach allows for the examination of individual differences without imposing diagnostic boundaries during early development. Participants were recruited from the local community through flyers and online advertisements. An initial 100 children were enrolled, but data from 22 children were excluded from the final analysis due to excessive motion artifacts or poor signal quality in the fNIRS data (n = 14), a common challenge in pediatric neuroimaging (Scholkmann et al., 2014), the scan being cut short due to experimenter error (n = 1), being in physical contact with the parent during the fNIRS scan (n = 2), poor fNIRS cap placement (n = 2), refusal to wear the fNIRS cap (n = 1), issues with the video recording of the structured play session (n = 1), or having an autism spectrum disorder diagnosis (n = 1). The sample was predominantly White and from families with high socioeconomic status (see Table 1, Table 2). Parents provided written informed consent, and children provided verbal assent prior to participation. The study protocol was approved by the local Institutional Review Board.

Table 1.

Sociodemographic Characteristics of Total Participant Sample.

Sociodemographic Characteristic	Mean (M) or Count (n)	SD or %
Age in months, M	65.4	11.3
Sex, n
Male	34	44%
Female	44	56%
Race/Ethnicity, n
Black or African American	1	1%
Hispanic, Latino, or Spanish Origin	2	3%
Asian	1	1%
White	61	78%
Multiracial	13	17%
Annual household income, n
Less than $15,000	1	1%
$15,001 to $30,000	1	1%
$30,001 to $45,000	1	1%
$45,001 to $60,000	2	3%
$60,001 to $75,000	4	5%
$75,001 to $90,000	4	5%
$90,001 to $110,000	10	13%
$110,001 to $125,000	7	9%
$125,001 to $175,000	13	17%
$175,001 to $225,000	19	25%
$225,001 to $275,000	5	7%
More than $275,000	10	13%
Parent with a 4-year college degree, n
Neither parent	5	6%
One parent	14	18%
Both parents	59	76%

Note. SD, standard deviation. N = 78.

Table 2.

Sociodemographic Characteristics of Analytic Participant Sample.

Sociodemographic Characteristic	Mean (M) or Count (n)	SD or %
Age in months, M	65.8	11.5
Sex, n
Male	28	42%
Female	38	58%
Race/Ethnicity, n
Hispanic, Latino, or Spanish Origin	2	3%
Asian	1	1%
White	52	79%
Multiracial	11	17%
Annual household income, n
Less than $15,000	1	1%
$15,001 to $30,000	1	1%
$30,001 to $45,000	1	1%
$45,001 to $60,000	2	3%
$60,001 to $75,000	3	5%
$75,001 to $90,000	3	5%
$90,001 to $110,000	9	14%
$110,001 to $125,000	5	8%
$125,001 to $175,000	10	15%
$175,001 to $225,000	16	24%
$225,001 to $275,000	4	6%
More than $275,000	11	17%
Parent with a 4-year college degree, n
Neither parent	4	6%
One parent	12	18%
Both parents	50	76%

Note. SD, standard deviation. N = 66.

2.2. Procedure

Each child and a parent attended a single 1.5-hour laboratory visit. Upon arrival, parents completed a battery of online questionnaires while the child was familiarized with the lab environment and the fNIRS equipment. Participants were brought to a quiet, dimly lit testing room and seated at a child-sized table with the presentation computer. The participants' head circumference was measured before they were fitted with an appropriately sized fNIRS fabric cap (52-cm EasyCap, n = 78). The fNIRS caps were organized by 10–20 coordinates and placed on the head so that Fpz was centered above the nasion and equidistant between the nasion and AFz. The initial signal quality was checked for all channels, and any channels with poor quality were manually adjusted to ensure that all channels had acceptable signal quality before the recording. The child then completed a resting-state brain imaging session, during which they were instructed to sit still for a passive viewing task. Following the imaging session, the child participated in a structured play session (data not reported here). Families were compensated for their time.

2.3. Measures

Internalizing Symptoms. Parent-reported internalizing problems were measured using the Child Behavior Checklist (CBCL) for Ages 1.5–5 (Achenbach & Rescorla, 2000). The CBCL is a widely used, psychometrically robust measure of childhood emotional and behavioral problems. Questions are scored on a three-point Likert scale (0 = absent, 1 = occurs sometimes, 2 = occurs often). The Internalizing Problems scale raw score was used as the primary outcome variable. The Internalizing Problems scale included items such as is your child “too fearful or anxious” or “unhappy, sad, or depressed.” Of the 66 children in the analytic sample, 23 were older than 71 months and 43 were 71 months or younger (age range = 48–71 months among children ≤71 months). Because T-scores are not available for children older than 71 months on this CBCL form, raw scores were used to retain the full sample on a common metric. In a sensitivity analysis restricted to children aged 71 months or younger, the pattern of results was unchanged.

Temperament. Child temperament was assessed using the parent-report Child Behavior Questionnaire (CBQ) – Very Short Form (Putnam & Rothbart, 2006). This measure provides scores on three broad, well-validated temperament dimensions: Negative Affect, Surgency/Extraversion, and Effortful Control. These scales demonstrate acceptable to marginal internal consistency in our sample, with Cronbach’s α = .73 for Surgency, α = .69 for Negative Affect, and α = .68 for Effortful Control (see Table 3). Questions are scored on a 7-point Likert scale, with an option for “not applicable.” A score of 1 is considered “extremely untrue of your child” and a score of 7 is considered “extremely true of your child.” Negative affect was measured through questions such as “Becomes sad when told to do something s/he does not want to do.” Surgency/Extroversion was measured through questions such as “Is full of energy, even in the evening.” Effortful control was measured through questions such as “Can easily stop an activity when s/he is told no."

Table 3.

Descriptive Statistics for Study Variables.

Variable	N	Min	Max	Mean	SD	Cronbach’s Alpha
Age (months)	66	48.000	83.000	65.833	11.555
DMN HbO connectivity (Fisher-z)	66	-0.220	0.445	0.057	0.146
FPN HbO connectivity (Fisher-z)	66	-0.102	0.589	0.199	0.144
CBQ Surgency	66	2.185	6.107	4.400	0.823	0.729
CBQ Negative Affect	66	2.490	5.424	3.980	0.678	0.691
CBQ Effortful Control	66	3.667	6.521	5.334	0.617	0.676
CBCL Internalizing Problems	66	0.000	25.000	7.530	5.892
CBCL Externalizing Problems	66	0.000	30.000	9.697	6.961
Household Income	66	1.000	12.000	8.727	2.605
Sex	66
Female	38
Male	28

Socio-Economic Status (SES). SES was operationalized using parent-reported total annual household income, a standard proxy in developmental research (Singh & Rajendra, 2024).

Demographic Questions. Parents were asked to report their child’s birthdate, gender,

race and ethnicity, and parent level of education (see Table 1).

2.4. fNIRS data acquisition and processing

Resting-state fNIRS data were acquired using a continuous-wave fNIRS system at a sampling rate of 4.17 Hz, using a NIRx NIRScoutX system and NIRStar acquisition software. A standardized optode cap was placed on the child's head, with emitters and detectors arranged to provide coverage over cortical regions associated with the DMN and FPN (see Fig. 1). NIRS data were recorded for up to 12.5 min while the child sat at rest while watching Disney’s Fantasia 2000 (Goldberg, 1999). Children watched the 12.5 min Rhapsody in Blue segment, an animated clip with no dialogue, set to classical music composed by George Gershwin. This clip was selected as a way to keep the participant’s attention and minimize motion for at least 10 min of data collection (Wang et al., 2017) without being overly stimulating or already familiar to the child. Previous work assessing functional connectivity measured during naturalistic viewing has found that movie watching may amplify individual differences in behaviorally relevant brain networks (Finn & Bandettini, 2021). Further work suggests that functional connectivity is highly consistent across different movies, and conclusions drawn based on movie-watching may be generalizable (Tian et al., 2021). In a developmental sample, playing a non-stimulating animated clip, without dialogue allowed for higher subject compliance. Triggers were used to indicate when the child began and ended their resting state task. Raw light intensity data were converted to changes in oxygenated (Δ[HbO]) and deoxygenated (Δ[HbR]) hemoglobin concentration using the modified Beer-Lambert law (Scholkmann et al., 2014). The processing pipeline, conducted using FC-NIRS software (Xu et al., 2015), included several critical steps to ensure high data quality. Age-appropriate differential pathlength factor values of 5.5 for 760 nm and 4.5 for 850 nm were used (Scholkmann & Wolf, 2013). Data were band-pass filtered (0.01–0.08 Hz) to isolate the low-frequency fluctuations characteristic of resting-state neural activity. Motion artifacts were detected by calculating the moving standard deviation within a 2-second sliding time window and flagging any values more than five standard deviations above the mean of the moving standard deviation; flagged artifacts were then corrected using PCA. Finally, extracerebral and systemic physiological signals (e.g., scalp blood flow and low-frequency vascular oscillations) were removed using the 'shortchannel' method of the Global Signal Correction module in FC-NIRS (version 3.0; Xu et al., 2015), with the eight short-separation channels (see Fig. 1) specified as reference channels, so that signal from these channels was regressed out of all long-separation channels. Time-series data were trimmed as needed to realign the start and stop times of the data with the start and stop times of the resting-state task. Channels with a signal-to-noise ratio (calculated by the mean signal intensity divided by the standard deviation of the signal intensity over time) below a threshold of 12.5 for both wavelengths were marked as poor-quality channels and excluded from further analyses. This SNR threshold corresponds to a coefficient of variation (CV) of 8%, a signal quality criterion that has been commonly used to identify variability of the signal and noisy fNIRS channels (Collins-Jones et al., 2024, Frijia et al., 2021). Additionally, participants were excluded from further analyses if 50% or more of their channels were excluded at this stage.

Fig. 1.

Schematic of fNIRS Channels (numbered 1–52) with Short-Separation Channels on 10–20 Landmark Grid. Short-Separation Channels (numbers 4, 11, 16, 26, 30, 41, 45, and 49) shown in purple. Blue channels indicate channels selected to create regions of interest (ROI) for the Default Mode Network (DMN). Green channels indicate channels selected to create regions of interest (ROI) for the Frontoparietal Network (FPN).

2.5. Functional connectivity (FC) Analysis

A seed-based correlation analysis was performed on the preprocessed Δ[HbO] time-series data. Seeds were defined based on the channel locations corresponding to key nodes of the DMN and FPN. The default mode network (DMN) was created by averaging the left and right hemisphere correlations between three channels in the medial prefrontal cortex [corresponding with the Fpz electrode (10–20 system), or the superior and middle frontal gyri in the LPBA40 atlas] and three channels in the lateral temporal cortex [corresponding with T7 and T8 electrodes (10–20 system), or the superior and middle temporal gyri (LPBA)]. The frontoparietal network (FPN) was created by averaging the left and right hemisphere correlations between two channels in the dorsolateral prefrontal cortex [corresponding with the F3 and F4 electrodes (10–20 system), or the middle frontal gyrus (LPBA)] and two channels in the posterior parietal cortex [corresponding with CP3 and CP4 electrodes (10–20 system), or the supramarginal, postcentral, and angular gyri (LPBA)]. For each network, the time-series from all channels within that network were averaged to create a representative network time-series. The functional connectivity for each network was then calculated as the average Pearson correlation coefficient between each individual channel in the network and the representative network time-series. The Pearson correlation coefficient values were then Fisher-z transformed for normalization. This resulted in a single value for each participant representing the average within-network FC for the DMN and the FPN.

2.6. Statistical analysis

The primary analysis was a multiple linear regression using the entry method to test the concurrent associations of these variables with internalizing symptoms. The CBCL Internalizing raw score served as the outcome variable. The predictor variables entered into the model were: DMN within-network FC, FPN within-network FC, CBQ Negative Affect score, CBQ Surgency score, CBQ Effortful Control score, household income (SES), and the child's age in months. A Shapiro–Wilk test of the model residuals indicated that they did not significantly deviate from normality (W = 0.97, p = .072) and variance inflation factors indicated no problematic multicollinearity among predictors. As an exploratory analysis, we also conducted a multiple linear regression considering the CBCL Externalizing raw score, with the same predictor variables as the internalizing symptoms model. Standardized regression coefficients are reported for both models. Throughout, the term predictor refers only to the role a variable plays in the regression model and is not intended to imply temporal or causal precedence. Child sex was included as a covariate in both models in response to reviewer comment. Sex was coded as female = 1, male = 2. The pattern of results was unchanged with the inclusion of sex in the model.

3. Results

Of the 78 children in the final sample, complete data across all predictor variables were available for 66 participants. The remaining 12 were excluded from the regression analyses via listwise deletion due to missing data on one or more predictors. Exploratory Pearson correlation analyses were conducted to examine associations among all variables in the model (see Table 4). DMN connectivity was positively associated with FPN connectivity (r = .34, p = .005) and with CBCL Internalizing Problems (r = .30, p = .01). Surgency was negatively associated with effortful control (r = −.39, p = .001) and CBCL Internalizing Problems (r = −.39, p = .001). Negative affect was positively associated with CBCL Internalizing Problems (r = .49, p < .001) and Externalizing Problems (r = .42, p < .001) and negatively associated with household income (r = −.28, p = .022). Effortful control was negatively associated with Externalizing Problems (r = −.42, p < .001). Internalizing and Externalizing Problems were strongly positively correlated (r = .60, p < .001), while Externalizing Problems were negatively associated with household income (r = −.32, p = .009). No significant associations were observed between age or gender and the other study variables, and all remaining correlations were nonsignificant (p > .05).

Table 4.

Correlations with confidence intervals.

Variable	1	2	3	4	5	6	7	8	9
1. Sex
2. Age (months)	-.04
	[-.28,.20]
3. DMN	-.11	.16
	[-.34,.14]	[-.08,.39]
4. FPN	-.10	-.18	.34**
	[-.33,.15]	[-.41,.06]	[.11,.54]
5. Surgency	.02	-.22	.01	.02
	[-.23,.26]	[-.44,.03]	[-.23,.26]	[-.22,.26]
6. Negative Affect	-.07	-.13	.04	.15	-.23
	[-.31,.17]	[-.36,.11]	[-.20,.28]	[-.10,.38]	[-.45,.01]
7. Effortful Control	-.21	.09	-.15	.06	-.39**	-.07
	[-.43,.03]	[-.16,.32]	[-.38,.09]	[-.19,.30]	[-.58, −.17]	[-.31,.17]
8. Internalizing Problems	-.01	-.09	.30*	.12	-.39**	.49**	-.04
	[-.25,.23]	[-.33,.15]	[.06,.51]	[-.12,.36]	[-.58, −.16]	[.28,.65]	[-.28,.20]
9. Externalizing Problems	.14	-.22	.22	.11	.20	.42**	-.42**	.60**
	[-.10,.37]	[-.44,.02]	[-.02,.44]	[-.13,.35]	[-.04,.42]	[.19,.60]	[-.60, −.20]	[.42,.74]
10. Household Income	.16	.17	-.02	-.11	-.10	-.28*	.18	-.13	-.32**
	[-.08,.39]	[-.07,.40]	[-.26,.22]	[-.35,.13]	[-.33,.15]	[-.49, −.04]	[-.06,.41]	[-.36,.11]	[-.52, −.08]

Note. Values in square brackets indicate the 95% confidence interval for each correlation. The confidence interval is a plausible range of population correlations that could have caused the sample correlation (Cumming, 2014). * Indicates p < .05. ** indicates p < .01. This table represents an exploratory analysis; p values are unadjusted for multiple comparisons.

A multiple linear regression was conducted to evaluate how well DMN and FPN functional connectivity, temperament, SES, and age were associated with internalizing symptoms (see Table 5). The overall regression model was statistically significant, F(8, 57) = 5.66, p < .001, and accounted for approximately 44.3% of the variance in internalizing problems (R² =.443, Adjusted R² =.364). Examination of the individual predictors revealed that three variables made unique, statistically significant contributions to the model. DMN connectivity was significantly and positively associated with internalizing symptoms (β =.33, t = 2.94, p = .005) (see Fig. 2). Negative Affect was also significantly and positively associated with internalizing symptoms (β =.36, t = 3.24, p = .002) (see Fig. 3), while Surgency was significantly and negatively associated with internalizing symptoms (β = −.39, t = -3.32, p = .002) (see Fig. 4). These results indicate that, when controlling for all other variables in the model, children with higher DMN connectivity, higher negative affect, and lower surgency tended to have more parent-reported internalizing problems. Contrary to our hypothesis, FPN connectivity was not significantly associated with internalizing symptoms (β = −.06, t = -0.55, p = .585). Similarly, Effortful Control (β = −.09, t = -0.76, p = .452), household income (β = −.03, t = -.28, p = .780), age (β = −.18, t = -1.63, p = .108), and sex (β =.03, t = .30, p = .768) were not significantly associated with internalizing symptoms in the final model. In the internalizing and externalizing models, sex was coded as 1 = female, 2 = male, and age was controlled but not significant.

Table 5.

Multiple Linear Regression Models Results.

nternalizing
Predictor	B	SE	β	t	p	Fit
(Intercept)	17.73	11.73		1.51	.136
DMN connectivity	13.34	4.53	0.33	2.94	.005*
FPN connectivity	-2.50	4.55	-0.06	-0.55	.585
Surgency	-2.79	0.84	-0.39	-3.32	.002**
Negative Affect	3.11	0.96	0.36	3.24	.002**
Effortful Control	-0.85	1.13	-0.09	-0.76	.452
SES	-0.07	0.24	-0.03	-0.28	.780
Age	-0.09	0.06	-0.18	-1.63	.108
Sex	0.37	1.25	0.03	0.30	.768
						R2 = .443**
						95% CI [.17,.53]
Externalizing
(Intercept)	7.17	14.03		0.51	.611
DMN connectivity	10.09	5.42	0.21	1.86	.068
FPN connectivity	-1.44	5.44	-0.03	-0.27	.792
Surgency	1.30	1.00	0.15	1.29	.201
Negative Affect	3.90	1.15	0.38	3.39	.001**
Effortful Control	-2.56	1.35	-0.23	-1.90	.063
SES	-0.43	0.29	-0.16	-1.49	.143
Age	-0.07	0.07	-0.12	-1.10	.276
Sex	2.27	1.49	0.16	1.52	.134
						R2 = .430** 95% CI [.15,.52]

Note.

* indicates p < .05. ** indicates p < .01.

Fig. 2.

Partial Regression plot showing the association between Default Mode Network (DMN) connectivity and internalizing symptoms, with both axes residualized on the remaining terms in the model. DMN connectivity was measured using fNIRS as changes in oxygenated hemoglobin (ΔHbO) during rest. Internalizing symptoms were assessed using the Child Behavior Checklist (CBCL). The solid line represents the fitted regression line, and the shaded region indicates the 95% confidence interval. The scatter plot represents individual participant data points (n = 66). A significant positive association was observed, suggesting higher DMN connectivity is linked to greater internalizing symptoms (β =.33, t = 2.94, p = .005). DMN connectivity was not significantly associated with externalizing symptoms (β =.21, t = 1.86, p = .068).

Fig. 3.

Partial Regression plot showing the association between negative affect and internalizing symptoms in children, with both axes residualized on the remaining terms in the model. Negative affect was measured using the Child Behavior Questionnaire (CBQ) and represents a composite temperament score reflecting a predisposition to experience negative emotions such as fear, sadness, and frustration. The solid line represents the fitted regression line, and the shaded region shows the 95% confidence interval. The scatter plot represents individual participant data points (n = 66). Higher negative emotionality was associated with more internalizing symptoms.

Fig. 4.

Partial regression plot showing the association between surgency and internalizing symptoms in children, with both axes residualized on the remaining terms in the model. Surgency was measured using the Child Behavior Questionnaire (CBQ) and represents a composite temperament score reflecting positive emotionality, activity level, and approach behaviors. The solid line represents the fitted regression line, and the shaded region shows the 95% confidence interval. The scatter plot represents individual participant data points (n = 66). Higher Surgency was associated with fewer internalizing symptoms.

The second multiple linear regression model evaluated how well DMN and FPN functional connectivity, temperament, SES, age, and sex were associated with externalizing symptoms. Neither DMN (β =.21, t = 1.86, p = .07) nor FPN (β = −.03, t = -.27, p = .79) connectivity was significantly associated with externalizing symptoms. Notably, the direction and magnitude of the DMN–externalizing association were similar to those observed for internalizing symptoms. Surgency (β =.15, t = 1.29, p = .20), effortful control (β = −.23, t = -1.90, p = .06), household income (β = −.16, t = -1.49, p = .14), age (β = −.12, t = -1.10, p = .28) and sex (β =.16, t = 1.52, p = .134) were also not significantly associated with externalizing symptoms in the model. However, negative affect was significantly and positively associated with externalizing symptoms (β =.38, t = 3.91, p < .001) (see Fig. 5). When controlling for all other variables in the model, children with higher negative affect and lower effortful control exhibited more externalizing symptoms.

Fig. 5.

Partial regression plot showing the association between negative affect and externalizing symptoms in children, with both axes residualized on the remaining terms in the model. Negative affect was measured using the Child Behavior Questionnaire (CBQ) and represents a composite temperament score reflecting a predisposition to experience negative emotions such as fear, sadness, and frustration. The solid line represents the fitted regression line, and the shaded region shows the 95% confidence interval. The scatter plot represents individual participant data points (n = 66). Higher negative affect was significantly associated with more externalizing symptoms.

4. Discussion

This study examined concurrent neural and temperament correlates of internalizing symptoms in early childhood. In a community sample of 4- to 6-year-old children, we found that greater within-network DMN functional connectivity, higher negative affect, and lower surgency were each independently associated with higher parent-reported internalizing problems. Contrary to our hypotheses, effortful control was not significantly associated with internalizing symptoms. Additionally, contrary to hypotheses derived from the adult literature, FPN connectivity was not associated with internalizing symptoms in this young age group. These findings suggest that the neural and temperament foundations of vulnerability to internalizing disorders are established by the preschool years, but that the specific neural signatures of risk may follow distinct developmental timetables.

The positive association between DMN connectivity and internalizing symptoms represents a downward extension of findings from adult and adolescent psychopathology (Kaiser et al., 2015; C. H. Miller et al., 2015; Tooley et al., 2022). Because our montage captured only the superficial anterior medial prefrontal and lateral temporal nodes of the DMN, this association should be understood as pertaining to those nodes rather than to the network as a whole (see Limitations). This result suggests that hyperconnectivity of the DMN, a neural pattern consistently linked to rumination and maladaptive self-referential thought in adult depression (Chou et al., 2023, Fang et al., 2025, Hamilton et al., 2015, Yan et al., 2019), is not merely a consequence of chronic illness. Instead, its association with subclinical symptoms in very young children supports the view that it may represent a primary, early emerging biological vulnerability. In early childhood, this increased DMN integration may not manifest as the complex, abstract rumination seen in adults, but rather as a more foundational processing bias toward internal states at the expense of engagement with the external world. This could be a neural mechanism underlying behaviors like social withdrawal and excessive shyness, which are hallmarks of the internalizing spectrum in young children (Rubin & Mills, 1991).

Although DMN connectivity was not significantly associated with externalizing symptoms, we note that the direction and magnitude of this association were broadly similar to those observed for internalizing symptoms. Accordingly, the present data do not provide strong evidence for specificity of the DMN–internalizing link, and future studies with larger samples will be needed to determine whether the relationship between DMN hyperconnectivity and behavior problems is domain-specific or reflects a more general vulnerability. That said, the observed pattern is broadly consistent with theoretical accounts linking the DMN to internally oriented thought (Son et al., 2023, Yu et al., 2019), which may be more directly implicated in internalizing than externalizing symptomatology, though this remains to be confirmed.

Externalizing symptoms were significantly associated with higher negative affect and lower effortful control, aligning with previous literature that suggests poor executive function as related to greater levels of broad externalizing problems (Yang et al., 2022). Together, these findings suggest dissociable pathways in early childhood, where internalizing symptoms appear more closely related to variation in internally oriented neural networks, though we note that this dissociation was not definitive given the similar direction of the DMN–externalizing association. Nonetheless, the pattern underscores the importance of considering domain-specific mechanisms when examining early neural correlates of psychopathology, and different developmental pathways in the maturation of neural networks.

The presence of this DMN-symptom relationship in a non-clinical sample of young children, where symptoms exist on a continuum, adds to a growing body of evidence for DMN hyperconnectivity as a candidate marker for internalizing disorders. This interpretation is consistent with prior work in early childhood showing that early brain connectivity patterns predict risk for later psychopathology (Xia et al., 2018, Xie et al., 2026). Extending this literature, the present findings demonstrate that individual differences in DMN connectivity are associated with subclinical internalizing symptoms in preschool-aged children. The logic is as follows: (1) adult clinical studies establish DMN hyperconnectivity as a correlate of MDD (Kaiser et al., 2015); (2) the current study demonstrates a linear relationship between DMN connectivity and symptoms in young, non-diagnosed children; (3) this is consistent with the view that the neural pattern exists on a continuum and is present before the typical age of onset for a formal disorder. Therefore, this finding provides evidence for a model in which DMN hyperconnectivity represents an early-emerging neural correlate of vulnerability to internalizing symptoms.

The lack of an association between FPN connectivity and internalizing symptoms is a critical finding that must be interpreted through a developmental lens. Rather than a failure to replicate adult findings, this null result may provide an accurate snapshot of the FPN's maturational state in early childhood. The FPN is the brain's hub for "top-down" cognitive control and deliberate, effortful emotion regulation strategies like reappraisal (Menon, 2011, Ochsner and Gross, 2005). This network undergoes a particularly protracted developmental course, with its structural and functional architecture continuing to mature throughout childhood and into adolescence (Chen et al., 2023, Tooley et al., 2022, Wendelken et al., 2017). Studies consistently show that children exhibit weaker FPN activation during cognitive control tasks compared to adults, and their ability to leverage FPN-mediated strategies to modulate emotional responses improves with age (McRae et al., 2012, Silvers et al., 2012).

The combination of a significant DMN finding and a null FPN finding in our sample is consistent with, though it cannot establish, a developmental lag between the maturation and pathological involvement of these two key networks. Because connectivity was measured at a single timepoint, this account is offered as a hypothesis rather than a conclusion. The DMN matures relatively early in development, whereas the FPN matures much later (Chen et al., 2023, Fair et al., 2008, Gao et al., 2015). One possible explanation for our results, therefore, is that vulnerability to internalizing problems first manifests in the brain systems that are functionally mature enough to display dysregulation. In early childhood, this is the DMN, via hyperconnectivity that may bias the child toward an internal focus. The role of FPN hypoconnectivity or decreased within network connectivity, reflecting deficits in cognitive control, may emerge as a significant risk factor only later in development—perhaps during adolescence—when these circuits are expected to be fully engaged for self-regulation and are now failing to do so. This suggests a dynamic, two-stage model of evolving neural risk. However, confirming this developmental account would require future work tracking the same children across multiple developmental stages. Furthermore, some evidence suggests the DMN may act as a "scaffold" for the immature FPN during typical development, with stronger DMN-FPN coupling observed in children than in adults (Chen et al., 2023). It is plausible that at this young age, the FPN is not yet sufficiently independent or mature to be a distinct source of variance contributing to psychopathology.

Our statistical model indicates that DMN connectivity, negative affect, and surgency are each independently associated with internalizing symptoms. This finding supports a multi-pathway model of risk, where biological predispositions in brain function and early emerging behavioral dispositions both make unique contributions to vulnerability. The results showing that high negative affect and low surgency are associated with internalizing problems are highly consistent with decades of developmental theory and research (Rubin and Mills, 1991, Muris and Ollendick, 2005, Northerner et al., 2016, Gartstein et al., 2012). High negative affect reflects a child's intrinsic reactivity to stress, while low surgency indicates a tendency toward shyness, low positive emotionality, and behavioral inhibition—a classic internalizing profile. The current findings are in favor of an etiological model where multiple, distinct risk factors can converge on the same clinical phenotype. In other words, a child may be at risk due to their brain connectivity, their temperament, or an additive combination of both.

This study has several notable strengths. It is among the first to examine concurrent neural and temperament correlates of internalizing symptoms within the preschool period. The successful use of fNIRS allowed for the collection of functional neuroimaging data from this challenging-to-scan population, providing a unique window into early brain-behavior associations (Pinti et al., 2020). The multimodal approach, integrating brain and behavior, provides a more holistic understanding of risk than either modality alone. However, the findings must be considered in light of several limitations. First, the cross-sectional design precludes any causal inferences; we cannot determine the direction of these associations over time. Second, we must acknowledge the technical limitations of fNIRS, which include lower spatial resolution compared to fMRI and a sensitivity that is primarily limited to the cortical surface (Pinti et al., 2020, Yücel et al., 2021). Though previous work establishes fNIRS as able to produce resting-state functional connectivity networks comparable to those produced by fMRI (Duan et al., 2012), due to the limited penetration depth of fNIRS we are unable to measure the posterior cingulate cortex implicated in the DMN or amygdala implicated in internalizing symptoms (Burghy et al., 2012, Foster and Koslov, 2025, Gerin et al., 2019, Merz et al., 2018). Additionally, much of the medial prefrontal component of the DMN extends dorsally and ventrally along the medial wall and cannot be captured by our fNIRS montage (Andrews-Hanna, 2012). Our DMN index therefore reflects connectivity between superficial anterior mPFC and lateral temporal cortex, a cortical-surface subset of the network, and our findings should be interpreted as pertaining to these nodes rather than to the DMN as a whole. Although DMN–salience network connectivity has been associated with internalizing symptoms (Karl et al., 2026), our fNIRS montage did not provide sufficient coverage of key salience network regions (e.g., anterior insula) to reliably characterize whole-network interactions. Consequently, analyses were limited to within-network connectivity of the DMN and FPN, which were well represented by the recording montage. Third, the sample was relatively homogenous, consisting of primarily White children from high-SES families, which may limit the generalizability of the findings to more diverse populations. Fourth, results should be interpreted considering the study sample size (Marek et al., 2022). Additionally, we examine internalizing symptoms as a general construct, based on the CBCL. The CBCL/1½-5 is created for children up to 5 years old, and our sample includes some 6 year olds. However, in our sensitivity analysis including only children 71 months or younger, the pattern of results was unchanged for both internalizing and externalizing symptoms. Prior work has identified differential neural associations between internalizing symptom subtypes including withdrawn/depressed behavior, somatic complaints, and anxious/depressed behaviors (Thomas et al., 2024). Future work should examine DMN and FPN connectivity and internalizing symptom subscales as well as other contextual factors that may influence the development of internalizing symptoms, such as early life environment and adversity (Albertina et al., 2024, Buss et al., 2025). Finally, 22 of the 100 children enrolled were excluded prior to analysis, 14 of them (14%) for fNIRS data quality reasons specifically, a persistent challenge in pediatric fNIRS that highlights the ongoing need for methodological refinement (Scholkmann et al., 2014). Notably, a recent meta-analysis suggests an average attrition rate of 34.23% in fNIRS work with younger, infant samples (Baek et al., 2023). Future research should prioritize longitudinal designs. Following this cohort into middle childhood and adolescence would allow for a direct test of the "developmental lag" hypothesis: does FPN connectivity emerge as a significant predictor of internalizing trajectories as children age? Integrating additional levels of analysis, such as genetic data and measures of environmental adversity (e.g., adverse childhood experiences), would provide an even more comprehensive and ecologically valid model of risk and resilience (Albertina et al., 2024, Tooley et al., 2022).

The findings of this study hold significant translational potential. The identification of DMN connectivity and specific temperament traits as early, independent risk markers could inform the development of novel, targeted, and timely prevention strategies. For instance, children identified with a temperament profile of high negative affect and low surgency could benefit from established behavioral interventions focused on building emotion regulation skills and fostering social engagement. In parallel, children identified with DMN hyperconnectivity might one day benefit from interventions aimed at modulating self-focused attention, such as mindfulness-based practices that have been adapted for young children.

Prior work has linked individual differences in prefrontal resting-state functional connectivity to temperament dimensions and executive function, suggesting that these neural and behavioral risk markers may share overlapping neurodevelopmental origins (Eng et al., 2025, Li et al., 2017; J. G. Miller et al., 2023b; Quiñones-Camacho et al., 2019b). However, DMN connectivity was not significantly correlated with any of the measured temperament dimensions, indicating that variation in DMN connectivity is largely independent of individual differences in temperament in our sample. Likewise, internalizing symptoms were not significantly associated with effortful control, although negative affect and surgency remained significant behavioral factors. Critically, the association between DMN connectivity and internalizing symptoms remained significant after adjusting for temperament and other covariates. This suggests that DMN hyperconnectivity is associated with internalizing symptoms independent of, rather than merely as a proxy for, temperamental vulnerability. Together, these findings suggest that neural and behavioral factors contribute unique variance to early internalizing symptoms, supporting the possibility that they represent complementary, rather than redundant, pathways of vulnerability.

In conclusion, this study demonstrates that the roots of internalizing psychopathology are evident in both brain function and behavior by the preschool years. The early emergence of DMN dysregulation, coupled with well-established temperament vulnerabilities, appears to set the stage for later difficulties. The absence of an FPN association at this age raises the possibility that the neurobiology of risk is not static but shifts as the brain matures, a possibility that longitudinal designs will be needed to evaluate. Understanding this developmental timing is a crucial step toward identifying the right children at the right time for the right intervention, ultimately improving long-term mental health outcomes.

CRediT authorship contribution statement

Halle Miller: Writing – review & editing, Writing – original draft, Visualization, Validation, Project administration, Investigation, Formal analysis, Data curation, Conceptualization. Johanna Chajes: Writing – review & editing, Resources, Project administration, Funding acquisition, Conceptualization. Tobias Grossmann: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Formal analysis, Data curation, Conceptualization.

Declaration of Competing Interest

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

Data availability

Data will be made available on request.