Personalized neural networks underlie individual differences in ethnic identity exploration and resolution

Abstract

This study examined how ethnic identity relates to large-scale brain networks implicated in social interactions, social cognition, self-definition, and cognitive control. Group Iterative Multiple Model Estimation (GIMME) was used to create sparse, person-specific networks among the default mode and frontoparietal resting state networks in a diverse sample of 104 youths aged 17-21. Links between neural density (i.e., number of connections within and between these networks) and ethnic identity exploration and resolution were evaluated in the full sample. Ethnic identity resolution was positively related to frontoparietal network density, suggesting that having clarity about one’s ethnic group membership is associated with brain network organization reflecting cognitive control. These findings help fill a critical knowledge gap about the neural underpinnings of ethnic identity.

Introduction

Ethnic identity refers to one’s sense of self regarding one’s ethnic group connection and has been conceptualized as a cultural asset, providing promotive and protective benefits in youth development, largely evident amongst ethnic minority youth in the U.S (Neblett, Rivas-Drake, Umaña-Taylor, 2012; Perez-Brena et al., 2018). Social cognitive mechanisms underlying youths’ ethnic identity development and how it may contribute to adaptive adjustment have been offered (e.g., Neblett et al., 2012). Another kind of mechanism that may shed light on how ethnic identity functions as a promotive and protective factor concerns how this developmental process may be reflected in the brain. Considering neurodevelopmental frameworks in tandem with developmental frameworks centered on sociocultural contexts may offer novel insights as to how ethnic group connections translate into thoughts and behaviors that support adjustment (Qu & Telzer, 2018). Bridging our understanding of both social cognitive and neurodevelopmental mechanisms through which ethnic identity may function provides a more holistic and integrative developmental assessment of its properties in youth.

Currently, it is unclear how ethnic identity is reflected in the brain, let alone in novel investigations of the functional organization of the brain that underlies adaptive and maladaptive behaviors. Functional organization refers to synchronous activity among brain regions (or networks) implicated in basic and complex human cognitive, behavioral, social, and affective processes (Power et al., 2011; van den Heuvel & Pol, 2010). It has become increasingly clear that social interactions are similarly supported by synchronous activity among brain network regions (Feng et al., 2021). Revealing the ways in which these large-scale brain networks undergird psychological processes linked to ethnic identity will not only begin to define the role of the brain in adaptive behaviors evident when youth have positive connections to their ethnic group, but it will also help consider the role of ethnicity and identity in future neuroscience research, and vice-versa in developmental science research. Accordingly, the goal of the present study is to integrate social neuroscience and methodological advances in person-specific neural network mapping with the study of ethnic identity development. This study considered whether brain networks that support social cognition, self-definition, social interactions, and cognitive control are also relevant for ethnic identity development?

Ethnic Identity Development

Ethnic identity developmental processes include exploration, which refers to gaining knowledge about one’s ethnicity, and resolution, which refers to having clarity about what it means to belong to one’s ethnicity (Umaña-Taylor, 2011). The process of developing connections to an ethnic group is thought to occur universally across populations (Umaña-Taylor, 2011) and is supported by various socially salient contexts, such family, school, neighborhood, peers, and media (Rivas-Drake et al., 2017; Seaton et al., 2017). Thus, although developing a sense of connection to a salient social group centered around shared cultural traits may occur universally, the beliefs and attitudes regarding one’s ethnic group tend to be contingent on the group’s societal standing (i.e., minority status is often reflective of a country’s sociopolitical history). Compared to childhood, advanced cognitive abilities and heightened sensitivity to social-environmental cues make adolescence a significant period for ethnic identity development as youth develop even more complex perspectives of their ethnicity (Umaña-Taylor, Yazedjian, & Bàmaca-Gómez, 2004). During later adolescence, youth navigate new environments (e.g., college, employment), roles, and expectations, while also piecing together more complex perspectives regarding the role of race and ethnicity in their lives (Williams et al., 2020). Over the lifespan, ethnic identity development changes, and individuals continuously define and redefine their understanding about what it means to be a member of their ethnic group (Marks et al., 2020).

Research grounded in developmental frameworks has found that children, adolescents, and young adults who feel connected to their ethnic group enjoy better psychosocial adjustment, benefit in their academic outcomes, and engage in fewer health compromising behaviors such as substance use (Rivas-Drake et al., 2014; Smith & Silva, 2011; Umaña-Taylor et al., 2018). Youth with positive connections to their ethnic group have been shown to prioritize school, effectively engaging in academic tasks and school behaviors, and develop greater school engagement and academic efficacy that help navigate obstacles to academic success (Miller-Cotto & Brynes, 2016). Another example is reflected in youths’ involvement in health risk behaviors. Having a positive ethnic identity has been associated with less engagement with deviant peers as well as less involvement in substance use and other health compromising behaviors (Brook et al., 2010; Derlan & Umaña-Taylor, 2015; Grindal, & Nieri, 2016; Rivas-Drake et al., 2014; Zapolski et al., 2018). Despite these advances in knowledge, the role of the brain in how ethnic identity development translates into positive adjustment remains unknown. Integrating knowledge of how ethnic connections are embedded within the self with knowledge of brain functionality may reveal a more comprehensive understanding of how and for whom ethnic identity may function as a developmental and cultural asset.

Ethnic Identity Development and Social Reorientation

To help conceptualize ways that ethnic identity developmental processes are reflected in the brain, the current work draws on multiple frameworks that emphasize development in sociocultural context, social neuroscience, and identity. A social neuroscience perspective of adolescence posits that there is heightened neural sensitivity to social environmental cues during this period, which is thought to underlie social reorientation and support exploring and gaining experiences that inform identity development (Dahl, 2016; Galván & Tottenham, 2016; Pfeifer & Peake, 2012; Steinberg, 2008; Telzer, 2016; Telzer et al., 2018). This perspective is consistent with two different theories of identity development and identity-informed behaviors. For instance, the same social neuroscience perspective of social reorientation (sensitivity to social environmental cues and salient social others like peers and parents) may also explain the way interactions with ethnic group members support identity development. Current knowledge of ethnic identity developmental processes directly coheres with this social reorientation perspective (Umaña-Taylor et al., 2004; Umaña-Taylor et al., 2014). During adolescence in particular, youth interpret their ethnic group interactions with increased complexity and negotiate the views and perspectives of ethnic group members to develop their own sense of self (Quintana et al., 1999). A fundamental process of formulating one’s identity concerns integrating the perspective of others with one’s own to create a unique sense of self (see e.g., Rivas-Drake et al, 2017 for how this may occur with peers). The social and motivational saliency that youth attribute to ethnic group member interactions is, in turn, likely supported by neural processes implicated in self-reflections and social cognition (Pfeifer & Berkman, 2018; Welborn et al., 2018). Mentalizing, or thinking about the thoughts and perspectives of others, is an important aspect of identity development, consistent with social neuroscience perspectives on adolescent social reorientation (Pfeifer & Peake, 2012).

The social neuroscience perspective of adolescence also aligns with the identity-value model of decision-making which emphasizes that the self is integral in how one interacts with one’s environment (Berkman et al., 2017; Pfeifer & Berkman, 2018). This framework acknowledges that different aspects of one’s identity, including social identities, may contribute to the values and motivations that underlie behaviors. Of the pieces of information that contribute to making decisions, engaging in a particular behavior may be valued more when it is relevant to one’s identity (Oyserman, 2007). Ethnic group connections may inform values and motivations over time and may contribute to behavior in numerous contexts. The identity-value model helps to link ethnic identity with thought and behavior and facilitates the consideration of aspects of the brain’s functional organization (i.e., networks) that may reflect this relationship.

Integrating social neuroscience and identity development frameworks reveals brain networks that underlie self-reflection, social cognition, and cognitive control: The default mode and frontoparietal networks are implicated in social reorientation processes and make potential candidate networks to consider for ethnic identity development. Furthermore, these frameworks do not diverge from, but actually converge with developmental models that emphasizes the role of ethnic group connection in adjustment (Coll et al., 1996; Neblett et al., 2012); however, weaving these various perspectives together illuminates potential mechanisms to help explain how ethnic group connection may translate to behavioral adjustment.

Resting State Networks and Ethnic Identity

Characterizing resting-state brain networks offers a way to understand the brain’s functional organization, how it supports complex human functionality, and the ways in which the environment (e.g., social support, stress, etc.) may influence functional organization (Barrett & Satpute, 2013). These networks are studied during the resting state because synchronous brain activity during an idle state can capture how network connections have been exercised, and hence, shaped and organized (Liégeois et al., 2017). Because there are no specific fMRI task-based paradigms to support hypotheses regarding neural correlates of ethnic identity processes, resting-state network approaches provide a useful starting point, as they offer a domain-general view of brain function and whether networks that support social reorientation processes (e.g., mentalizing, social interaction processing) are also implicated in ethnic identity development. Thus, insights gained from a resting state approach may provide the foundational basis for more explicit task-based examinations between the brain and ethnic identity.

The default mode network, and the mentalizing subnetwork in particular, is likely involved in social identity processes because this network is thought to reflect social affiliation and interactions (Spreng & Andrews-Hanna, 2015). The default mode network (which typically includes anterior medial prefrontal cortex and posterior cingulate cortex) is involved in internal mentation and is often implicated in self-generated or self-referential thinking, such as past-future autobiographical thinking, episodic memory, self-evaluation, as well as social cognition and mentalizing (Andrews-Hanna et al., 2010; Buckner & DiNicola, 2019). The mentalizing subnetwork (which consists of the dorsal medial prefrontal cortex, temporal parietal junction, lateral temporal cortex, and temporal pole) is thought to be involved in perspective-taking, social cognition, and theory of mind (Spreng & Andrews-Hanna, 2015), whereas the core default mode system is thought to be involved in self-evaluation and self-related processes (Barrett & Satpute, 2013). Graph theoretical approaches have shown greater connectivity strength within the default mode network with increase age from childhood to adolescence as well as greater global and local default mode network efficiency over time, demonstrating young adult like brain functional organization as adolescents age (Fan et al., 2021). A domain-general view of the default mode network suggests that remembering, thinking about the future, and taking another person’s perspective, all depend on the ability to draw on stored experiences to create meaningful mental moments in the present (Barret & Satpute, 2013). This domain-general view of the default mode network aligns well with current knowledge of ethnic identity development processes, for example, reflecting on what it means to be part of one’s ethnic group (i.e., resolution) based on past interactions with one’s ethnicity (i.e., exploration).

The frontoparietal network is also important for understanding the neural basis of ways that ethnic identity may relate to adjustment. This network is integral for facilitating communication across networks and helps support everyday functionality because it is a set of brain regions involved in actively (dis)engaging other networks to facilitate cognitive tasks. Brain regions that make up the frontoparietal network, such as the dorsolateral prefrontal cortex, frontal cortex, and inferior parietal lobe, are thought to underlie cognitive control and decision-making (Campbell et al., 2012; Scolari et al., 2015). The frontoparietal network can be viewed as a functional hub that balances activity between the default mode network (processing internal world) and the attentional network (processing external world) (Vincent et al., 2008). In addition, the frontoparietal network is structurally interspaced amid the default mode network (Spreng et al., 2013; Vincent et al., 2008), that is, within the prefrontal and parietal lobe. Throughout adolescence, the default mode and frontoparietal networks become increasingly distinct (e.g., negatively correlated) from an intrinsic connectivity standpoint (e.g., resting state), which have been linked with cognitive development over time (DeSerisy, et al., 2021). The decrease in functional connectivity of short-range (short-distance) connection and increase in functional connectivity of long-range connections reflects a developmental shift in brain organization comprised of more distributed and integrated large-scale networks (Fair, 2009). The frontoparietal network and the default mode network have been shown to coactivate during planning and goal-directed cognition (Gerlach et al., 2014; Spreng et al., 2010; Utevsky et al., 2016), such that strong functional interconnections between the networks result in better planning capability than do weak functional interconnections between the networks. Drawing from the integrated frameworks above (i.e., identity-value model), the frontoparietal network (with its role in cognitive control) in tandem with the default mode network (with its role in social cognition and self-referential thinking) may help explain how ethnic identity supports behavioral adjustment at the brain level. See the conceptual model in Figure 1.

Figure 1.

DMN¹ = Default Mode Network. FPN² = Frontoparietal Network. Other Networks³ = Brain networks unexamined in this study. Conceptualizing the role of resting state brain networks in how ethnic identity processes may translate into behavioral adjustment. The present study examines conceptual pathway A. The model proposes that exploration and resolution processes may implicate the default mode and frontoparietal networks, large-scale networks that support social cognition and self-referential thought as well as cognitive control, respectively. These networks may reflect two of the ways that ethnic identity relates to thought and behavior at the brain level.

Further indication of the relations between default mode network, frontoparietal network, and ethnic identity come from a meta-analysis of the brain’s role in human social interactions (Feng et al., 2021). Studies that examined the neural correlates of social interactions, social norms, and social norm violation consistently point to a set of brain regions that reflect social cognition, motivation, and cognitive control. For example, studies on adolescent processing of social influences and social group evaluation find mentalizing regions like the dorsomedial prefrontal cortex, temporoparietal junction, and precuneus, as well as cognitive control regions (e.g., ventrolateral prefrontal cortex), are involved when youth adapt their perspective to align with salient social others, like peers and parents, and when identifying with in-group members (Moreria, Bavel, & Telzer, 2017; Welborn et al., 2016). Just as these neural processes support social reorientation, they may also be implicated in ethnic identity development. The intrinsic activity among these regions commonly observed in across studies on human social interaction processing mapped onto the default mode network as well as the salience and cognitive control networks (Feng et al., 2021). Indeed, despite the known role of large-scale networks in social interactions, the specific role of ethnic group social interactions is sorely under-studied. Ethnic identity reflects how social ties contribute to one’s sense of self, implicating brain networks involved in self-referential thinking and social cognition such as the default mode and frontoparietal network.

The Study of Ethnic Identity in the Brain

Ethnic identity development captures, through social ties and connections, identification with an ethnic group comprised of people who share cultural features. When considering the brain in this process, the aim is not to compare differences in brain function of people from different cultures (Qu & Telzer, 2018; Seligman et al., 2016). Ethnic identity can be informed by cultural experiences but does not necessarily reflect the embodiment of a cultural self (e.g., how behavioral norms are ingrained), and so, culture and ethnicity should not be equated. The key distinction is an individual’s identification and sense of understanding of belonging to their ethnic group, not endorsement of cultural norms. This distinction is particularly important for studying populations comprised of diverse ethnic groups, such as in the U.S, where each person may develop connections to their ethnic group in distinct and even unique ways. Therefore, studying ethnic identity processes and the brain requires methodological approaches that acknowledge intraindividual heterogeneity in such psychological processes (and social experiences that may shape these processes) – as well as the brain function that underlies them. The later adolescent years may provide a rich starting point for characterizing brain network organization as it relates to ethnic identity. This may help reveal the distinct and homogenous ways in which ethnic identity developmental experiences throughout adolescence (captured through exploration and resolution) relate to current profiles of brain network organization.

Group iterative multiple model estimation (GIMME; Gates & Molenaar, 2012) is a person-specific network mapping approach that assumes such heterogeneity, and thus, is well-suited to the examination of the neural underpinnings of identity-linked default mode and frontoparietal resting state networks. The modeling capabilities of GIMME allow it to capture the heterogeneity as well as potential homogeneity of the brain’s functional organization (e.g., Beltz et al., 2016). This is a critical advantage over traditional neuroimaging analytic approaches that average brain activity across individuals and potentially lose meaningful intraindividual heterogeneity (i.e., person-specific information) about the brain that may be relevant to the psychological processes of interest (Molenaar, 2004; e.g., Beltz et al., 2018). GIMME is data-driven and uses an iterative process to estimate connections among multiple brain regions based on individual and group-level patterns of brain activity (Gates & Molenaar, 2012). The connections can be contemporaneous (occurring at the same functional volume) or lagged (occurring from one functional volume to the next). In addition, GIMME has been shown to produce fewer spurious connections and to have greater specificity compared to other network mapping approaches (Gates & Molenaar, 2012). GIMME results can then be characterized in terms of network density, which reflects the proportion of connections that a particular subnetwork comprises in one’s overall network. Following others (e.g., Goetschius et al., 2020), this reflects how integrated the ROIs of a network are (within-network density) or how integrated different networks are with each other (e.g., between-network density); these densities can then be linked to psychosocial or environmental traits. In the present study, exploration and resolution were expected to be positively related to greater within and between default mode and frontoparietal network density.

The Current Study

Large-scale brain networks that support social cognition and self-referential thinking (i.e., the default mode network) as well as cognitive control (i.e., the frontoparietal network) are interactively implicated in processing social interactions, and thus, are also likely implicated in developing connections to salient social groups, such as one’s ethnic group. Similar processes and neural networks are also important according to a social reorientation lens of adolescence. The goal of the current study is to examine how ethnic identity is related to individual differences in connectivity within and between the default mode and frontoparietal networks at rest in individual youth, who are at a developmentally-relevant moment of identity exploration and resolution. We accomplished this using person-specific modeling framework that reflected potential heterogeneity and homogeneity in the functional organization of the brain that may be relevant for the complex, multi-dimensional, and unique ways youth develop their ethnic identities (e.g., Beltz et al., 2016; Beltz et al., 2018).

Brain network organization revealed by GIMME, such as network density (number of connections within- and between-networks; Beltz & Gates, 2017), was examined in relation to ethnic identity to reveal whether connectivity within a self-reflective network or a cognitive control network (or between them) was stronger or weaker for youth with greater exploration or resolution. It was hypothesized that having engaged in greater ethnic identity exploration and resolution would be associated with greater network density within the default mode network. It was also hypothesized that having engaged in greater ethnic identity exploration and resolution would be associated with greater density between the default mode and frontoparietal network, which may begin to provide insights into how ethnic identity is connected to thought and behavior at the brain level.

Exploratory analyses were considered given the literature on the saliency of ethnic identity development among ethnic and racial minority status youth (e.g., youth of color). Ethnic identity exploration and resolution have been observed in White youth and these measures have demonstrated similar psychometric properties among White youth compared to youth of color (Sladek et al., 2020). However, from a phenomenological perspective, these processes may look qualitatively different for White youth and youth of color (Martinez-Fuentes et al., 2020). Therefore, we explored whether relations among network densities and ethnic identity exploration and resolution differed between White youth and youth of color.

Method

A subsample of the participants from the Adolescent Health Risk Behavior (AHRB) study (N=2017) were invited to participate in neuroimaging. Of the 108 youth who participated in neuroimaging, 104 (M_age = 19.28, SD = 1.31, range = 17 – 21.45; 61 females) were included in these analyses. Three participants were excluded due to preprocessing issues (e.g., brain activity signal distorted), and one participant was excluded because of excessive motion (mean framewise displacement >.50). Seventy-four participants were White and 30 were youth of color, comprised of Black (n=15), Latinx (n=9), Asian (n=3), and mixed race-ethnicity youth (n=3). Parental education level was reported by youth and used as a proxy for socioeconomic status, ranging from 1 (Completed grade school or less) to 6 (Completed graduate or professional school after college) (M = 4.32, SD = 1.16). The neuroimaging sample did not differ from the parent AHRB sample with respect to demographic variables except for race-ethnicity, with the parent sample having a higher proportion of youth of color (45%) than the neuroimaging sample (30%) (see Supplemental Table S1).

Procedures

Consistent with the aims of the parent AHRB study, participants were recruited into the neuroimaging subsample based on their health risk behaviors, aiming to have a distribution of youth who displayed typical versus more frequent levels of health risks (see Demidenko et al., 2020 for a detailed description). In addition, participants were recruited on the basis of sex (58% female) and race-ethnicity (70% White, 30% youth of color), aiming to have equal proportions of males and females and significant racial-ethnic diversity.

Recruited participants were contacted and screened for MRI eligibility (e.g., hazardous metals in bodies). Eligible participants were scheduled for a scan session approximately a week after they provided in-person informed consent or assent with parental consent (if under age 18). While in the scanner, structural brain images, resting state brain activity, diffusion tensor brain images, and brain activity during two behavioral tasks were acquired. Neuroimaging sessions were one hour long, and participants were compensated up to $140 for the entire visit. Procedures of the AHRB study were approved by the University of Michigan Institutional Review Board (IRB) (IRBMED Approval #HUM00097653; PI: Keating, D).

Either the day before or the day of the scan, participants also completed a brief survey that contained the ethnic identity measures reported here. For 20 of the 22 participants who did not provide information about their ethnic identity exploration and resolution during the neuroimaging portion of the study, this information was imputed from the same measures administered during the parent study; 2 participants were missing these data. There was an average of 7.4 months (SD = 2.7) between the larger study survey and MRI scan for these 20 participants. Ethnic identity exploration and resolution scores from these 20 youth did not differ from the scores of youth who completed the measures at the time of the scan; exploration, t(100) = 1.06, p = .29; resolution, t(100) = 1.67, p = .09, nor did the youth differ on main demographic variables (see Supplemental Table S2).

Resting State fMRI Data Acquisition.

Resting state fMRI data was acquired using a GE Discovery MR750 3.0 Tesla scanner with a standard adult-sized coil (Milwaukee, WI). During the resting state scan, participants laid in the scanner for 8 minutes with their eyes open and fixated on a cross hair ‘+’ displayed on a presentation screen. Eye activity was monitored with an eye tracker by a research staff to make sure participants kept eyes open. Functional T2*-weighted BOLD images were acquired using a multiband EPI sequence (MB factor=6) of 60 contiguous axial 2.4 x 2.4 x 2.4 mm slices (TR = 800ms, TE = 30ms, flip angle = 52°, FOV = 21.6 cm, 90x90 matrix). For preprocessing, a high resolution T1-weighted anatomical (SPGR PROMO) was also acquired (TR = 7.0s, TE = 2.9s, flip angle = 8°, Field of View (FOV) = 25.6 cm, slice thickness = 1 x 1 x 1mm, 208 sagittal slices; matrix = 256 x 256). Slices in the functional and structural sequences were prescribed in the same locations.

Preprocessing.

As described in Circ et al. (2017), fMRI data were reconstructed, realignment and fieldmap correction was applied in SPM12 to each T2* run to recover inhomogeneity of signal in the B0 field, and physiological noise was removed using RETROICOR (Glover, Li, & Ress, 2000). FMRI data processing was carried out using FEAT (FMRI Expert Analysis Tool) Version 6.00, part of FSL (FMRIB’s Software Library, www.fmrib.ox.ac.uk/fsl). The first eight volumes were deleted to get a more stable neural signal. Resting state functional images were then aligned onto each participant’s anatomical structural image and registered onto standard MNI space using FLIRT (Jenkinson & Smith, 2001; Jenkinson, Bannister, Brady, & Smith, 2002). Data were realigned for head movement using 6 parameter frame-wise displacement (MCFLIRT; Jenkinson, Bannister, Brady, & Smith, 2002) and were spatially smoothed using a Gaussian kernel of FWHM 5mm. FSL brain extraction tool was used to extract the brain from the skull using BET (Smith 2002). Grand mean intensity normalization of the entire 4D dataset was done by single multiplicative factor.

Additional preprocessing steps were applied in order to remove motion, white matter, and cerebral spinal fluid tissue related artifacts, which are known to introduce noise and potentially bias resting state fMRI analysis (Circ et al., 2017; Satterthwaite et al., 2013). Steps were similar to Goetschius et al. (2020) for preparing resting state data for GIMME (Beltz et al., 2019). Independent components analysis-automatic removal of motion artifact (ICA-AROMA) was used to detect and regress out motion-related artifacts using FSL (Pruim et al., 2015). White matter (WM) and cerebral spinal fluid (CSF) masks were created and used to regress out tissue related artifacts from the ICA-AROMA denoised functional subject specific data through nuisance regression (Caballero-Goudes & Reynolds, 2017). A high pass-temporal filter of 100 Hz was applied to the resting state fMRI data. Mean frame-wise displacement (FD) values for each participant were calculated using fsl_motion outliers before (M_sample=.12, SD=.06) and after (M_sample=.03, SD = .02) motion artifact denoising (see supplemental Figure S1 for mean FD, pre- and post-preprocessing).

Measures

ROI Time Series Extraction.

Functional BOLD timeseries were extracted from 17 ROIs comprising the default mode and frontoparietal networks that were defined a priori. The ROIs and their central coordinates are listed in Table 1. ROIs in the default mode, and its sub mentalizing network, were taken from previous work (Andrews-Hanna et al., 2010). ROIs in the frontoparietal network were determined using the Power Atlas due to its reliable and widely-used detection of this network at rest (Power et al., 2011). Spheres with 10mm diameters were created around the central coordinate of each ROI, and then the mean BOLD signal of each ROI sphere (at each of 592 functional volumes) was extracted for each participant. There were no BOLD signal time-series data missing. These ROI timeseries were down-sampled to every other volume (296 total time points) to reduce the high temporal resolution of the multiband data, and thus, to increase power to detect cross-ROI relations characterizing connectivity within- and between-networks (e.g., Beltz & Gates, 2017; Sate et al., 2010).

Table 1.

Region of Interest MNI Coordinates

Region of Interest	Abbreviation	MNI Coordinate (x)	MNI Coordinate (y)	MNI Coordinate (z)
Default Mode Network
Anterior Medial Prefrontal Cortex	aMPFC	−6	52	−2
Posterior Cingulate Cortex	PCC	−8	−56	26
Dorsal Medial Prefrontal Cortex	dmPFC	0	52	26
L Temporal Parietal Junction	L TPJ	−54	−54	28
R Temporal Parietal Junction	R TPJ	54	−54	28
L Lateral Temporal Cortex	L LTC	−60	−24	−18
R Lateral Temporal Cortex	R LTC	60	−24	−18
L Temporal Pole	L TP	−50	14	−40
R Temporal Pole	R TP	50	14	−40
Frontoparietal Network
L Dorsolateral Prefrontal Cortex	L dlPFC	−44	27	33
R Dorsolateral Prefrontal Cortex	R dlPFC	46	28	31
L Frontal Cortex	L Frontal	−42	7	36
R Frontal Cortex	R Frontal	44	8	34
L Inferior Parietal Lobe	L IPL	−53	−50	39
R Inferior Parietal Lobe	R IPL	54	−44	43
L Intraparietal Sulcus	L IPS	−32	−58	46
R Intraparietal Sulcus	R IPS	32	−59	41

Ethnic identity.

The Ethnic Identity Scale (Umaña-Taylor et al., 2004) includes two subscales: exploration (e.g., “I have experienced things that reflect my ethnicity, such as eating food, listening to music, and watching movies;” 7 items; α = .85) and resolution (e.g., “I have a clear sense of what my ethnicity means to me”; 4 items; α = .85). Participants indicated their endorsement of the 11 statements on a scale from 1 (Does not describe me at all) to 4 (Describes me very well). Items of each subscale were averaged to calculate mean scores, such that higher values indicated more exploration and resolution, respectively. There were no missing item-level data. Among the whole sample, ethnic identity exploration (M = 2.28, SD = .85) and resolution (M = 2.64, SD = .87) were positively correlated with each other, r(100) = .62, p <.001).

Analysis Plan

ROI timeseries for each individual were entered into GIMME analyses (Gates & Molenaar, 2012) using the gimme package in R studio (v. 1.2.5033).

$${\eta }_i\left(t\right)=\left(A_i+A^g\right){\eta }_i\left(t\right)+\left({\mathrm{\Phi }}_{1,i}+{\mathrm{\Phi }}_1^g\right){\eta }_i\left(t-1\right)+{\zeta }_i\left(t\right)$$ Equation 1

GIMME is defined as a person-specific unified structural equal model in equation 1 above. ${\eta }_i\left(t\right)$ is the p-variate ROI time series at time t = 1, 2, 3, …T. p is the number of ROIs (17 ROIs) and T is the length of the timeseries (296). A is the p-p dimensional matrix of contemporaneous relations among ROIs. $\mathrm{\Phi }$ is the p-p dimensional matrix of lagged relations among ROIs. Superscript g indicates group-level estimates and subscript i indicates person specific estimates. ${\zeta }_i\left(t\right)$ is the p-variate error time series assumed to contain no temporal dependencies. The equation is a form of structural vector autoregressive model, and it is embedded within a search algorithm that uses Lagrange Multiplier Equivalences to determine the order of parameter, or connection, estimation (prioritizing parameters that are significant for the majority of the sample and then progressing to parameters each individual needs to ensure a good fitting network. The goal is not to maximize individual differences, but to fit a descriptive model of contemporaneous and lagged ROI associations to each individual in a way that incorporates connections that are significant and present in the majority (i.e., 75%) of the sample. If no connections are significant for 75% of the sample, then there would be no group-level structure reflecting homogeneity. If there are many connections significant for 75% of the sample that models fit well before progressing to the individual-level, then we would conclude the sample is relatively homogeneous.

In order to estimate personalized networks, GIMME begins with a “null” model that only includes autoregressive ROI connections (i.e., estimates of how much each ROI predicts itself at the next functional volume) because this is shown to improve the recovery of connections in temporally dense data (Lane et al., 2019), such as resting state fMRI. Group-level contemporaneous (i.e., same volume) or lagged (i.e., next volume) directed connections that would significantly improve network fit for the majority of participants in the sample (>75%) according to Lagrange Multiplier equivalents tests (i.e., modification indices; Sörbom, 1989) are then iteratively added to each individual’s network. When no more connections would improve network fit for the majority of the sample, individual-level contemporaneous or lagged directed connections that would significantly improve network fit for an individual participant according to Lagrange Multiplier equivalents tests are then iteratively added until the network fits each participants data well. Excellent model fit is determined by meeting two of four criteria: RMSEA ≤ .05, SRMR ≤ .05, CFI ≥ .95, and NNFI ≥ .95. Thus, final person-specific models are sparse – containing autoregressive connections as well as statistically meaningful connections that were estimated for everyone in the sample and that were estimated just for an individual; all connections have a person-specific direction (positive or negative) and magnitude. Details of model fitting and identification are described in Beltz and Gates (2017), Gates and Molenaar (2012), and Lane and Gates (2017). See Figure 2 for overview.

Figure 2.

Summary of unified structural equation (uSEM) model fitting with the GIMME algorithm. This process estimates a unique network for every participant in the sample, but the network contains some group-level connections that were fit for everyone and some individual-level connections that were potentially fit only for that person. Thus, networks are sparse (i.e., only contain connections meaningful for the group and individual), and all connections have a beta-weight estimated for each participant’s unique BOLD time series.

Network densities were calculated from the final person-specific GIMME models in order to characterize heterogenous resting state brain function to be linked to ethnic identity. Specifically, density was calculated for each participant by dividing the number of connections within and between their default mode and frontoparietal networks by the total number of connections in their GIMME model, reflecting the extent to which connections among ROIs that constitute the default mode, for instance, contribute to the overall network function (Beltz & Gates, 2017). For example, two individuals may have the same number of default mode network connections in their personalized models (e.g., 5), but these connections might represent a different proportion of their overall networks if one person has a denser network (e.g., 10) than the other person (e.g., 5).

These network densities were used in hierarchical linear regressions to examine associations with ethnic identity exploration and resolution (see supplemental table S3 for zero-order correlations amongst key study variables). For all regressions, post preprocessing mean FD motion estimates (calculated after motion denoising) and age were covariates entered in Step 1, and exploration and resolution were entered in Step 2. Network densities (either within or between default mode and frontoparietal) were outcomes in separate models. Type I error was .05. In the exploratory analyses, youth of color status (0 = White and 1 = youth of color) and its interaction with exploration and resolution were entered as Step 3 into the primary study models.

Results

All 104 personalized GIMME models converged. Average model fit indices across all participants suggest that the networks fit the data well: RMSEA = .053 (SD = .005), SRMR = .045 (SD =.004), CFI = .949 (SD = .004), and NNFI = .923 (SD = .006). See Supplemental Table S4 for model fit indices for each participant.

Figure 3 is a visual summary of all personalized networks. Solid lines depict contemporaneous connections, and dashed lines depict lagged connections. Bold lines represent connections that were estimated for everyone in the sample (i.e., group-level connections), and gray lines depict connections that were uniquely estimated for individuals (i.e., individual-level connections). Beyond the autoregressive connections, only one data-driven group-level connection between the left Inferior Parietal Lobe and the left Temporal Parietal Junction was detected. The presence of this connection is not surprising because the two brain regions are spatially proximal. The small number of group-level connections indicates that there is substantial heterogeneity (reflected by variation in the presence and magnitude of individual-level connections) in resting state brain network connectivity across participants.

Figure 3.

Summary of GIMME default mode and frontoparietal resting state network connectivity across participants (N = 104) showcasing substantial heterogeneity. See Table 1 for ROI acronyms. Blue ROIs comprise the default mode network, and green ROIs comprise the frontoparietal network. Black lines represent group-level connections that were estimated for all participants. Gray lines represent individual-level connections that were estimated for some participants, with the proportion of participants reflected by line thickness. Solid lines depict contemporaneous connections, and dashed lines depict lagged connections.

To illustrate the heterogeneity in brain connectivity across individuals, examples of two participants’ networks are depicted in Figure 4a and Figure 4b. Again, solid lines depict contemporaneous connections, and dashed lines depict lagged connections. Each connection has a corresponding beta weight reflected by line thickness, and red lines indicate positive connections and blue lines indicate negative connections. These participants clearly have diverse networks. The personalized network depicted on the left (Figure 4a) is relatively dense, with more negative contemporaneous connections compared to the personalized network depicted on the right (Figure 4b). To help characterize how personalized networks are organized, network densities were calculated. For example, the personalized network depicted on the left has about 32% of their brain network comprised of within default mode connections and 32% comprised of within frontoparietal connections. The personalized network depicted on the right has about 26% of their brain network comprised of within default mode connections and 56% comprised of within frontoparietal connections.

Figure 4.

GIMME default mode and frontoparietal resting state network connectivity for two illustrative participants. See Table 1 for ROI acronyms. Blue ROIs comprise the default mode network, and green ROIs comprise the frontoparietal network. Solid lines depict contemporaneous connections, and dashed lines depict lagged connections. Each connection has a corresponding beta weight reflected by line thickness, with red lines indicating positive weights and blue lines indicating negative weights. The network depicted on the left, Figure 4a: default mode density = 32% and frontoparietal density = 32% for a participant who had an exploration composite = 2.83 and a resolution composite = 2.66. The network depicted on the right, Figure 4b: default mode density = 26% and frontoparietal density = 55% for a participant who had an exploration composite = 1.66 and a resolution composite = 3.66.

Across all participants, within default mode connections (M = 10.73, SD = 3.72; range = 3 – 26), within frontoparietal connections (M = 13.48, SD = 3.37; range = 6 – 23), and between network connections (M = 18.58, SD = 8.07; range = 8 – 69) were present. Across all estimated network densities (i.e., within or between connections divided by total connections), within default mode density accounted for 8% to 47% of network connections and within frontoparietal density accounted for 2% to 57% of network connections. Between-network connections ranged from 27% to 65%. This variable distribution of network densities is indicative of the substantial heterogeneity in brain functional organization during the resting state.

Network Density and Ethnic Identity

A summary of hierarchical linear regression results for ethnic identity exploration and resolution as statistical predictors of within- and between-network densities, controlling for motion and age are presented in Table 2.

Table 2.

Summary of regressions of ethnic identity exploration and resolution and network densities

	Default Mode Density				Frontoparietal Density				Between-Network Density
	B	b	SE	t	B	b	SE	t	B	b	SE	t
Step 1
Intercept		.29***	.01	46.79		.37***	.01	45.82		.48***	.01	58.89
Motion	−.14	−.34	.25	−1.38	−.21	−.75*	.33	−2.32	.18	.59	.33	1.79
Age	.04	.00	.00	.31	.08	.01	.01	1.04	−.06	−.00	.01	−.57
R2	.02				.06				.04
Model	F = 1.08 (2,99), p =.34				F = 3.09 (2,99), p =.05				F =1.97 (2,99), p = .15
Step 2
Intercept		.29***	.01	46.79		.37***	.01	48.82		.48	.01	58.89
Motion	−.14	−.34	.25	−1.38	−.23	−.75*	.36	−2.38	.18	.59	.33	1.79
Age	.03	.00	.00	.31	.10	.01	.01	1.04	−.06	−.00	.01	−.57
Exploration	.26	.02*	.01	2.03	−.20	−.02	.01	−1.60	−.04	−.00	.01	−.35
Resolution	−.18	−.01	.01	−1.46	.30	.03*	.01	2.43	−.00	−.00	.01	−.02
Model	F (4, 97) = 1.60, p = .18				F (4, 97) = 3.08, p = .02				F (4, 97) = 1.02, p = .40
ΔR2	.04				.05				> .01
	F (2,97) = 2.10, p = .13				F (2,97) = 2.95, p = .06				F (2,97) = .11, p = .89

Note.

^*p < .05.

^**p < .01.

^***p < .001

For density within the default mode network, the step 1 model of covariates was not significant. In the step 2 model that included exploration and resolution, ethnic identity exploration was positively associated with higher default mode network density (B =.26, b = .02, SE = .01, p = .04), but the model did not significantly explain the variance in network density. This finding, however, was consistent with our hypotheses.

For density within the frontoparietal network, the step 1 model of covariates was significant. Motion was negatively associated with density (B = −.22, b = −.75, SE = .36, p = .02). The step 2 model that included exploration and resolution was also significant. Ethnic identity resolution was positively associated with higher frontoparietal network density (B = .30, b = .03, SE = .01, p = .02) and the model significantly explained the variance in network density. This finding is also consistent with our hypotheses, and inferences were similar in a model excluding the motion covariate.

For density between the default mode and frontoparietal network, the step 1 model of covariates was not significant. The step 2 model with exploration and resolution was also not significant. Neither exploration (B = −.04, b = −.00, SE = .01, p = .72) nor resolution (B = −.00, b = −.00, SE = .01, p = .98) were associated with between-network connectivity.

Exploratory Analysis.

Exploratory analyses considered whether associations between ethnic identity and resting state network densities varied between White youth (n = 74) and youth of color (n = 30). In general, White youth had lower levels of ethnic identity exploration (M = 2.02, SD = .69) compared to youth of color (M = 2.95, SD = .85), t(100) = −5.79, p < .001. White youth also had lower levels of ethnic identity resolution (M = 2.35, SD = .78) compared to youth of color (M =3.36, SD = .66), t(100) = −6.20, p < .001. In the hierarchical regression analyses reported above, an indicator of youth of color status (0 = White and 1 = youth of color) and its interaction with exploration and resolution were entered in the models (see Supplemental Table S5). Regarding the default mode network, neither youth of color status nor its interactions statistically predicted density, and the association between exploration and default mode density was no longer significant. Regarding the frontoparietal network, neither youth of color status nor its interactions statistically predicted density, but the association between resolution and frontoparietal density remained significant (B = .33, b = .03, SE = .01, p = .02). Regarding between network density, neither youth of color status nor its interactions were statistically significant predictors, and inferences about other predictors in the model were unchanged.

Discussion

Ethnic identity development refers to the process of exploring one’s ethnicity and developing a sense of clarity about what it means to belong to one’s ethnic group. Despite the burgeoning knowledge on the relations between ethnic identity and youth adjustment, there has remained a key conceptual gap about ways ethnic identity development processes are reflected in the brain. This work aimed to fill that knowledge gap, proposing that the default mode and frontoparietal networks that underlie self-referential thinking, social cognition, and cognitive control are relevant for ethnic identity developmental processes. GIMME, a person-specific analysis approach well-suited to the study of heterogeneity, was used to estimate personalized brain connectivity networks. Indeed, there were more person-specific than group-level connections within participants’ neural networks; put another way, each youth evinced different brain connectivity patterns amongst the ROIs that are potentially implicated in ethnic identity development. It was hypothesized that ethnic identity exploration and resolution would relate to neural network organization, such that more exploration and resolution would be associated with greater network densities (e.g., proportional connectivity within and between these networks). Findings showed that youth who reported greater clarity about their ethnic group membership demonstrated more connectivity within their frontoparietal network and suggested that youth who reported greater ethnic identity exploration evinced more connectivity within their default mode network.

Frontoparietal Network Density.

Having clarity about what it means to belong to one’s ethnic group was associated with having more connectivity within the frontoparietal network during the resting state. This finding supports our hypothesis and is novel, as no previous research has examined relations between ethnic identity developmental processes and the brain. Although directionality cannot be asserted, there are plausible explanations for how more resolution development is linked to denser networks. For instance, brain networks captured during the resting state are a reflection of how the brain has been shaped and organized by lived experiences, and the frontoparietal network represents a set of brain regions that work together to support cognitive control. It has been previously noted that youth with more clarity about their ethnic group membership demonstrate better behavioral adjustment, such as engaging in less substance use and circumnavigating deviant peers (Brook et al., 2010; Derlan & Umaña-Taylor, 2015; Zapolski et al., 2018). Thus, increased frontoparietal network density could be a reflection of the known links between resolution and behavioral adjustment. That is, youth with more resolution may engage cognitive control processes when engaging in behaviors that are promotive to one’s adjustment, such as resisting peer influence and engaging in behaviors that support academic achievement. The everyday ways in which a youth’s sense of ethnic group connection supports behavioral adjustment may involve cognitive control processes and reflect denser frontoparietal network over time. This does not mean that achieving some sort of positive ethnic identity resolution state is promotive for behavioral adjustment, but rather, that resolution as a developmental process (or the exercise of negotiating what one’s ethnic identity means to oneself) is promotive. Moreover, the process of negotiating the perspective of salient social others, such as ethnic group members, to develop clarity about one’s sense of self could exercise the frontoparietal network. This is consistent with research that has highlighted the involvement of the central executive network in a myriad of social interactions (Feng et al., 2021), especially during social influence and mentalizing processes (Moreira et al., 2017; Welborn et al 2016).

Motion also explained frontoparietal network density. Participants that moved more in the MRI scanner had less density in their frontoparietal network connections. Although this may seem to challenge the interpretation of network links with resolution, the opposite association would have been most concerning (i.e., with more motion tied to more frontoparietal network connectivity), as previous research has shown that specific characteristics of measured brain networks, such as long-distance connection strength, are sensitive to motion (Circ et al., 2017; Power et al., 2015). Moreover, significant motion would have made it harder to detect links with behavior. In fact, when the motion covariate was removed from the model, the association of resolution remained significant, making it unlikely that resolution effects were related to statistical noise. Additionally, the association of resolution remained present in the exploratory analysis (discussed below) despite adding three additional predictors, suggesting that the effect is robust.

Default Mode Network Density.

Default mode network organization was expected to relate to ethnic identity exploration and resolution. Self-referential thinking and social cognition are defining features of the default mode network (Buckner, & DiNicola, 2019). This network is relevant for ethnic identity development which involves thinking about the perspectives of ethnic group members to make sense of one’s ethnic group membership and one’s self-concept. Exploration of one’s ethnicity, participating in culturally salient events, gaining knowledge about one’s heritage, and learning from other ethnic group members involve self-reflective processes and social cognition. Default mode network density was positively associated with exploration in our study, but the finding was not as robust as the finding for resolution and frontoparietal network density because the overall model was not significant, and the effect was not present in exploratory analyses that included interactions. It is nonetheless a promising finding for future research consideration.

A possible explanation for the limited scope of this finding may concern the heterogeneity in the exploration processes itself. For example, exploration reflects the process of gaining ethnic-relevant knowledge and experiences but does not necessarily equate to achieving a “positive” state of one’s sense of ethnic identity. (This is actually what precipitated the conceptual disaggregation of exploration and resolution as related and interdependent, but distinct processes; Umaña-Taylor et al., 2004.) Perhaps a better approach is to examine the interaction between exploration and resolution. For example, those with more exploration and a greater sense of clarity may have more resonating implications for the default mode network functional organization.

Between-Network Density.

Connections between the default mode and frontoparietal networks were proposed to examine whether the brain supports the known relations between ethnic identity development and behavioral adjustment. However, ethnic identity exploration and resolution were not associated with between-network density. There was a wider distribution in the presence of this type of connection across personalized models compared to within-network connections. Resting state networks are best characterized (and defined) by their within-network synchronized activity (van den Heuvel & Pol, 2010; Liégeois et al., 2017). Therefore, within-network density as indicator of the brains’ functional organization may be optimal for assessing during the resting state than between-network density.

The role of youth of color status.

The interactions between group status (i.e., being a youth of color versus not) and ethnic identity exploration and resolution did not help explain network densities. Ethnic identity processes of exploration and resolution can be thought as universally applicable but may be qualitatively distinct for ethnic-racial minority youth in the U.S. where race- and ethnicity-salient experiences are tied to social stratification (Coll et al., 1996). Although the psychometric similarities in the exploration and resolution measures supports pooling youth of color together into a single group, a multi-group approach would be better and an individual-level approach through intensive longitudinal measurement would be best. Any observed effects from a pooled group could potentially mask whether any subgroup is driving the effect, or could obscure individual differences.

However, this should not discard the role of race in the way that race-salient experiences are reflected in the functional organization of the brain. This analysis, as an early step for studying ethnic identity in the brain, was important, but only a first step. Although, there was no particular hypothesis for whether and how race would impact the main findings, this exploratory analysis was still merited given the differential saliency of ethnicity and race among youth of color vis-à-vis White youth in the U.S. Additional considerations are needed in conceptualizing how ethnic identity content, that is one’s beliefs and attitudes about their group membership, relates to the brain as the nature of identity content does vary substantially by ethnic-group membership based on their unique sociohistorical contexts. Thus, an entirely different set of integrated theoretical formulations, including neural networks, would be important to consider for conceptualizing relations between ethnic identity content and the brain.

Study Considerations

This study is novel in its conceptual framing of the research question that integrates developmental and neuroscience as well as advanced quantitative methods. A fundamental aspect of social identity development is social cognition and, particularly for adolescent identity development, social reorientation (Telzer et al., 2018). We leveraged this idea to draw conceptual overlaps between knowledge about brain networks and ethnic identity development. The proposed conceptual framing should not be viewed as a mechanistic explanation of how ethnic identity development is reflected in the brain, but rather, as one potential way to link a set of ideas to conceptualize how ethnic identity development may be implicated in the brain (Jabareen, 2009). Thus, continued research relying on similar and other plausible conceptual framing is needed to further explore ways that ethnic identity processes are interrelated with brain and behavior.

In the present study, more individual-level than group-level connections were observed within participants’ person-specific neural networks; in fact, there was only one group-level connection between ROIs (i.e., between the proximal left IPL and TPJ). In other words, there was substantial heterogeneity in brain connectivity patterns captured during the resting-state in the sample (see Figure 3 and Figure 4). This person-specific modeling approach was vital because meaningful information about the unique ways that ethnic identity defining experiences are internalized and reflected in the brain would be lost in approaches that averaged brain activity (based on the assumption that group-level connections would apply to all youth in the sample). From a methodological perspective, the heterogenous and multidimensional aspects of ethnic identity development and the brain were accurately estimated through this person-specific approach which does not average across participants (Molenaar, 2004). Likewise, a person-specific approach to brain functional organization most clearly matches research questions about individual differences in psychosocial or environmental factors.

The age range of the present sample was limited to 17 through 21 years, though no significant effects of age were observed in predicting network densities. From a resting state perspective, age is important to account for in predicting brain network connectivity, which has characteristics dependent on age, such as strength and detection of long and short distance connections (Fair et al., 2008; Fair et al., 2009). With respect to ethnic identity development, our limited age range of youth in later adolescence could have narrowed viable variability in exploration and resolution given that these processes are known to manifest in a developmental age-stage manner across adolescence (Umaña-Taylor et al., 2014). However, it may have also been advantageous because it allowed us to capture resting state brain function after youth had gone through the majority of adolescence during which ethnic identity developmental processes are known to be salient, meaning that we captured the adolescent developmental “endpoint” (Rivas-Drake & Umaña-Taylor, 2019); having such a point is vital for grounding and encouraging future research. Ethnic identity development does not end in early-mid adolescence, though; it continuously formulates even among young adults when new perspectives may emerge linked to one’s changing environments (Marks et al., 2020). A younger sample with a wider age range examined longitudinally would provide unique insight into how network level neural organization coincides with exploration and resolution development. For instance, it might provide some indications of directionality between ethnic identity and brain network organization; it is plausible that ethnic identity developmental experiences shape brain functional organization and vice versa.

Although the study sample helps address calls for more diverse samples in neuroimaging research (Falk et al., 2013), a much more diverse and larger sample is important for continued research on brain network organization as it relates to ethnic identity development, specifically among youth of color. A larger sample means more personalized networks that can help to distinguish homogenous and heterogeneous ways that ethnic identity experiences relate to brain network organization.

To further understand links between ethnic identity development and the brain, it must be acknowledged that ethnic identity development does not happen in a social vacuum of positive social interaction. Ethnic identity development is also formulated from negative social interactions as well, such as ethnic-racial discrimination, that should be considered in future research (Yip, 2018). Therefore, future studies should consider other brain networks, like the salience network (see Feng et al., 2021), that may help account for the motivational and affective components involved in identity development. Alternative neuroimaging methodologies, such as exploratory whole-brain network approaches, could identify additional relevant brain networks for ethnic identity (or could help determine if the link between ethnic identity exploration and brain connectivity detected here is indeed unique to the default mode network, for instance). However, this would assume a homogenous relation between ethnic identity and the brain across adolescents and young adults, so it also has drawbacks. Novel fMRI experimental tasks, designed to emulate specific ethnic identity processes, coupled with person-specific frameworks, would help gain greater specificity in the links between neural processes and ethnic identity development, and are also a promising area for future work.

Conclusion

Although the consequential effects of ethnic identity development on thought, cognition, and behavior are known, a foundational conceptual gap concerns the ways in which ethnic identity development, if at all, is reflected in the brain. This study begins to fill that gap by considering the role of functional brain networks in how ethnic identity developmental processes relate to thoughts and behaviors using an integrative approach that relied on multiple theories. This present study revealed that youth who had more clarity about their ethnic identity had more connectivity within their frontoparietal network, consistent with what is known about ethnic identity resolution and adjustment from self-report studies. It also suggested that youth who were exploring and gaining more knowledge about their heritage may have more connectivity within their default mode network. Importantly, the procedures used in this study did not involve average group comparisons, and instead, focused on continuous, culturally salient processes and their likely heterogeneity through person-specific neural network modeling. Researchers can take this approach and apply it to other social groups and identity processes.

Data Availability:

The datasets generated and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.