Dual identity, bicultural identity integration and social identity complexity among Muslim minority adolescents

ABSTRACT

We examined cross-sectional and longitudinal associations between national and religious identification, Bicultural Identity Integration (BII), and Social Identity Complexity (SIC) among Muslim adolescents in the UK (Study 1, n = 773, M = 17.5 years) and the U.S. (Study 2, n = 190, MW1 = 14.1 years). Using person-oriented approaches, we identified four groups of adolescents. The two largest groups in both national contexts were “religiously-oriented strong dual identifiers” and “equally-strong dual identifiers”. The latter experienced less BII distance and more BII conflict, and perceived their identities as more similar and overlapping. These findings highlight that nuanced differences in strong dual identity patterns and trajectories have implications for how strong dual identities are experienced and perceived.

KEYWORDS:

• Dual identity
• religious identity
• national identity
• Muslims
• bicultural identity integration
• social identity complexity
• adolescents
• growth mixture model
• person-oriented

View correction statement:

CORRECTION NOTICE

Acculturation research has shown that ethnic minority adolescents use different strategies to combine their ethno-religious and national identities. These strategies include: a) “separation”, a strong sense of belonging to only the ethno-religious group; b) “assimilation”, a strong sense of belonging to only the national group; and c) “integration”, a strong sense of belonging to both groups (Phinney et al., 2006). The latter is also considered a dual identity (e.g., Wiley et al., 2019). Dual identifiers further vary in the degree to which they perceive their group memberships as harmonious versus conflicted, and blended versus compartmentalized (i.e., they show variations in Bicultural Identity Integration, BII; Benet‐Martínez & Haritatos, 2005) and the degree to which they perceive their group memberships as similar and overlapping (i.e., they show variations in Social Identity Complexity, SIC; Roccas & Brewer, 2002). Although these concepts all seek to understand how individuals make sense of and combine multiple identities, they have been largely studied in isolation. The present research examines, for the first time, the interrelationships (both cross-sectional and longitudinal) between dual identification, BII, and SIC.

We focused on Muslim adolescents in Western societies who are an at-risk population for identity-based threats due to widespread prejudice and discrimination (FRA, 2017; Pew Research Center, 2017) and populist public debates in which the religious identity of Muslims is speculated to conflict with Western values and culture (Ahmed & Matthes, 2017). We further focused on the intersection of religious and national identities as religious identity is a crucial form of ethnocultural identity for Muslim minorities (Phalet et al., 2018). Moreover, longitudinal research on BII and SIC is scarce (but see Schwartz et al., 2015 for BII; Knifsend et al., 2017, for SIC) and no research covering the concurrent development of dual identity, BII and SIC exists to date. Finally, we used a person-centered approach which allowed us to identify subgroups of adolescents with distinct religious and national identity patterns and developmental trajectories, which we subsequently linked to components of BII and SIC. Our research therefore contributes to a better understanding of: 1) how religious and national identities develop, 2) how these identities are experienced (e.g., are dually identified adolescents less conflicted than adolescents who prefer one identity over the other?), and 3) whether and how cognitive representations of identities (e.g., perceived group boundaries and similarities) are linked to identification processes and experiences of identity conflict or harmony.

Dual identification among Muslim minority adolescents

Muslim immigrants who strongly identify with both the ethno-religious ingroup and the superordinate national group report higher levels of psychological well-being (Dimitrova et al., 2015; Spiegler et al., 2019), greater involvement in collective action efforts (Klandermans et al., 2008; Simon & Ruhs, 2008), more trust in the national government (Saleem et al., 2019), and more positive intergroup attitudes and behaviors (Jugert et al., 2020; Saleem et al., 2018; Saleem & Ramasubramanian, 2019). However, some public and political discourses in Western societies constrain Muslims’ identity choices by questioning the compatibility of Islam and Western ways of life (Foner & Alba, 2008; Trittler, 2019). Such identity threats, experienced via interpersonal and group-level discrimination and/or exposure to unfavorable media depictions, are likely to explain why most Muslims in Europe (Fleischmann & Phalet, 2016; Kunst et al., 2012; Leszczensky et al., 2020; Martinovic & Verkuyten, 2012) and the U.S. (Balkaya et al., 2019; Hummel et al., 2020; Hutchison et al., 2015; Saleem et al., 2019) identify more strongly with their ethno-religious ingroup and less with the superordinate national group even when identification with both is high. Importantly, such an identification pattern still encompasses a sense of dual identity (Simon & Ruhs, 2008).

Prior research, using person-oriented analytical approaches, further indicates that Muslim minority adolescents’ dual identity (here: ethnic and national) is far from uniform and homogeneous. Spiegler et al. (2019), for example, identified four groups of Muslim adolescents: equally-medium and increasing dual identifiers (39%), ethnic and national identifiers who developed an equally-medium dual identity over time (39% and 8%, respectively), and stable ethnic identifiers (14%). Most adolescents thus developed equally-medium dual identities during adolescence. In addition, ethnic identifiers developing an equally-medium identity scored highest on well-being and health. Jugert et al. (2020), focusing on Turkish-Muslim adolescents in Germany, distinguished strong dual identifiers (56%), ethnic identifiers (29%), equally-medium dual identifiers (11%), and low ethnic identifiers (4%). In contrast to ethnic identifiers, strong-dual and equally-medium dual identifiers regarded themselves as both German and Turkish. Strong-dual and ethnic identifiers further self-categorized as primarily ethnic, while equally-medium dual identifiers indicated being primarily national. Equally-medium dual identifiers also perceived group boundaries as more permeable than strong dual identifiers. Subsequent latent transition analysis revealed that the number of strong dual identifiers decreased substantially over time, while the number of equally-medium dual identifiers increased, indicating that it becomes increasingly difficult for German-Muslims to uphold a strong dual identity during adolescence. Finally, S. Zhang et al. (2018), studying Turkish- and Moroccan-Dutch adults (who are overwhelmingly Muslim), identified four groups at a single point in time: equally-medium dual identifiers (47%), strong dual identifiers (25%), national identifiers (15%), and ethnic identifiers (13%). Strong dual identifiers were psychologically better adjusted than others. Together, these findings highlight distinct identity patterns and trajectories among Muslim minorities including, for example, patterns of integration, assimilation, and separation (Phinney et al., 2006). Moreover, there seem to be various forms of dual identity such as “equally-medium” and “strong”.

Bicultural identity integration

In addition to the strength of dual identification, acculturation research highlights the importance of examining the dynamics between these identities. Specifically, the Bicultural Identity Integration (BII) framework suggests that dual identifiers vary in their experiences (affectively and cognitively) of belonging to multiple social groups (Benet‐Martínez & Haritatos, 2005; see Benet-Martinez et al., 2021in press, for an extensive review of the accumulated literature on this construct). The BII model argues that individuals can experience their multiple identities as harmonious versus conflictual (which depends on the degree of tension between identities) and as blended versus compartmentalized (which depends on the perceived distance between identities). Feelings of conflict and blendedness are independent, such that, for example, blended individuals may experience their identities as either harmonious or conflicting (Van Der Werf et al., 2019).

Prior research on the link between dual identity and BII is scarce which is why it remains largely unclear how dual identifiers experience their duality (Wiley et al., 2019). While studies using composite scores of BII found positive associations between ethnic and national identity strength and identity integration (Manzi et al., 2014; Schwartz et al., 2015), research focusing on identity conflict and perceived incompatibility points toward a negative correlation between national identity and conflict, and a positive correlation between religious identity and conflict (Hutchison et al., 2015; Martinovic & Verkuyten, 2012). Notwithstanding this evidence, key limitations of much of this prior work are the use of composite scores for BII or an exclusive focus on BII conflict, the use of cross-sectional designs, and a variable-centered approach. The latter focuses on correlations between single identities and BII measures and does not account for dual identity. These limitations make it impossible to examine whether different types of dually identified individuals exist, with some experiencing their identities as more compartmentalized and conflicting than others. There is, however, the idea that individuals with equally strong identities are more likely to recognize discrepancies and conflicts between the norms and values associated with each identity (Tadmor & Tetlock, 2006), and that identity separation and compartmentalization might be effective ways to manage identity conflict (Amiot et al., 2007; Huynh et al., 2011, Figure 35.1).

Social identity complexity

A further concept that deals with multiple identities, which has rarely been studied in relation to dual identification or BII, is that of Social Identity Complexity (SIC) which captures how people subjectively combine their multiple ingroup identities into more or less complex cognitive representations thereof (Roccas & Brewer, 2002). SIC is subdivided into “overlap complexity” (recognizing that holding membership of one ingroup does not necessarily coincide with memberships of another ingroup) and “similarity complexity” (recognizing that the meaning and content underlying one ingroup may be distinct from another of one’s ingroups). A more (less) complex identity structure is thus present when individuals perceive relatively lower (higher) overlap and similarity between their multiple ingroups.

Low social identity complexity among majorities has been linked to more negative outgroup attitudes (Brewer & Pierce, 2005) and higher intergroup bias (Schmid et al., 2009) indicating that, for majorities, high identity overlap and similarity denotes nativism and privileging the dominant (white) cultural and religious group over other cultures and religions in the country. Research on SIC among minorities, however, is inconclusive as low SIC has been linked to less ethnically diverse friendships, weaker national identities and more negative attitudes toward integration (Knifsend et al., 2017; Verkuyten & Martinovic, 2012; A. Zhang et al., 2013). Others found low SIC to be unrelated to minorities’ national identity (Brewer et al., 2013) or even linked to more identity integration (Cárdenas et al., 2019). The latter findings suggest that SIC might operate differently among minorities for whom greater perceived overlap and similarity (which reflect lower complexity according to the definition of the construct) might denote a desire to bring two cultures into alignment within the self and within the dominant society, and thus a dual sense of belonging.

Very little is known about associations between BII and SIC because both concepts were developed in different literatures and subfields. BII was developed in the context of cultural psychology and aims to understand acculturation-related self and identity processes among immigrants. SIC, in contrast, is a social psychological concept rooted in social identity theory, and traditionally aimed at explaining intergroup attitudes, predominately of majority group members. However, given that both concepts deal with perceptions of multiple ingroups we consider it timely to, first, devote greater attention to the meaning and role of SIC among minority groups, and, second, to examine the interrelationship of SIC with BII, a key construct in the minority group literature. A critical distinction between both concepts is that BII focuses on self-perceptions (e.g., seeing oneself as a blend of Muslim and British cultures) while SIC focuses on group-perceptions (e.g., seeing Muslim and British as overlapping group categories). Thus, the relation between BII and SIC is not obvious. It is, for example, not clear if someone with a blended identity (feeling Muslim and British at the same time) would perceive the ingroups as more overlapping (all Muslims are British, and vice versa). In addition, there might be stronger associations between SIC and BII blendedness as both refer to perceptual aspects of how identities are organized, than between SIC and BII conflict as the latter taps into affective aspects (feeling conflicted and torn) of identity integration (Miramontez et al., 2008).

As becomes evident, the small but growing evidence base on SIC has primarily focused on potential consequences of complexity for intergroup attitudes, yet very little work has considered how the concept interrelates with processes of dual identification and BII – a critical consideration, especially in the context of understanding its relevance and meaning among ethnic minority members.

Overview and hypothesis

To examine how religious minorities make sense of their religious and national identities we conducted two studies that sought to explore the interrelationships between and co-development of dual identity, BII and SIC. In Study 1, a cross-sectional study in the UK, we used latent profile analysis to identify subgroups of adolescents with distinct identity patterns. In Study 2, a longitudinal study in the US, we used growth mixture modeling to identify subgroups of adolescents with distinct developmental trajectories. This approach allowed us to capture interindividual differences in (the development of) dual identification which we subsequently linked to BII and SIC. Due to the scarcity of prior work and its mixed findings, our research is exploratory. Yet, we expected to find heterogenous groups of Muslim adolescents such as strong or equally-medium dual identifiers. Moreover, we expected higher BII conflict among adolescents with equally-strong dual identities than among adolescents with more separated and compartmentalized identities. Finally, we expected a stronger association between SIC and BII blendedness than between SIC and BII conflict.

Study 1

Method

Sample and Procedure

The analysis is based on data from adolescents in their last two years of secondary education in a small town in the North of England. The sample contained 773 adolescents (M = 17.07 years, SD = 0.59 years, 46.8% male) who self-categorized as Muslim. Adolescents’ ethnic background was either Bangladeshi or Pakistani (45.8%, and 54.2%, respectively). The majority had British citizenship (96.5%). Respondents completed an online survey during regular college hours either at the beginning of the academic year 2015/2016 (n = 426; 55.1%), or the beginning of the academic year 2016/2017 (n = 347; 44.9%). Participants gave informed consent to participate in the study. Ethical approval was obtained from the University of Oxford, Medical Sciences Interdivisional Research Ethics Committee.

Measures

Muslim and British identification

Muslim and British identification were measured separately with five items each (e.g., “Being [British/Muslim] is an important part of who I am”). The items loaded on a single factor explaining 71.5% and 69.4% of the variance in religious and national identity, respectively.¹ Response options ranged from 1‚ “strongly disagree” to 5, “strongly agree”. The five items were combined to create composite scores (α _British = .88 and α _Muslim = .88).

Bicultural identity integration (BII)

BII was assessed with eight items adapted from Benet‐Martínez and Haritatos (2005). The items loaded on two factors explaining 64.6% of the variance. The first factor, BII distance vs. blendedness (α = .73) was measured with three items (e.g., “I keep Muslim and British cultures separate”). The second factor, BII conflict vs. harmony (α = .85) was measured with five items (e.g., “I am conflicted between the British and Muslim ways of doing things”). Respondents reported their agreement with each item using a 5-point scale from 1‚ “strongly disagree” to 5, “strongly agree”.

Social identity complexity (SIC)

SIC overlap was assessed with 4 items: 1) “What proportion of British people do you think are Muslim?”, 2) “What proportion of Muslim people do you think are British?” (0, “none of them”; 100 “all of them”), 3) “Imagine you meet a British person for the first time; what is the likelihood that he/she would also be Muslim?”, and 4) “Imagine you meet a Muslim person for the first time; what is the likelihood that he/she would also be British?” (1,“extremely likely”; 11, “extremely unlikely”). The likelihood items were recoded, so that higher values indicated higher perceived overlap (i.e., lower overlap complexity). We transformed both scales into 5-point scales and computed the mean (α = .65). SIC similarity was assessed with two items: 1) “To what extent do you think being British and being Muslim means something similar or different?” (1, “means something very similar”; 5 “means something very different”), and 2) “How similar or different do you think the typical British person and the typical Muslim person are?” (1, “very similar”; 5, “very different”). Both items were recoded, so that higher values indicated greater perceived similarity (i.e., lower similarity complexity; r = .30, p < .001).

Analytical procedure

To identify subgroups of adolescents, we conducted a Latent Profile Analysis (LPA) in MPlus using the national and religious identification scales. Both scales were allowed to correlate, and the correlation was fixed to be equal across classes. We used a stepwise procedure in which one additional class (k) was added to the model at a time and the fit of the more parsimonious model was compared with the model with one additional class. To decide on the number of classes, the Bayesian Information Criterion (BIC) was used which should be lower when compared to the k − 1 class solution. We further used the Lo–Mendell–Rubin Likelihood Ratio Test (LMR–LRT) and Bootstrapped Likelihood Ratio Test (BLRT) that indicate, when significant, that a k-class solution fits the data better than a k − 1 class solution. Parsimony and theoretical meaning of the classes were also considered. Following class enumeration, we examined whether religious and national identity strength differed across classes. To do so, we used the MODEL TEST function of MPlus which yields a Wald test of parameter constraints. A significant Wald test indicates a significant difference between two classes. In a final step, we examined associations between class membership, demographic characteristics, BII and SIC. We used a 3-step maximum likelihood (ML) procedure that adjusts for classification errors (Vermunt, 2010), and the AUXILIARY, R3STEP and BCH commands of Mplus (Asparouhov & Muthén, 2014).

Results

Descriptive statistics are shown in Table 1. On average, both religious and national identities were strong (above the midpoint of the scale; ps < .001, ds ≥ 1.41). Religious identities were stronger than national identities, t(713) = 19.53, p < .001, d = 0.87. Both identities were positively correlated (r = .30, p < .001), experienced as more compartmentalized than blended (p < .001, d = 0.79), and more harmonious than conflicted (p < .001, d= 0.81), and perceived as more distant than overlapping (p < .001, d= 0.56), and more different than similar (p < .001, d= 0.77).

Table 1. Descriptive statistics Study 1.

	M(SD)	1	2	3	4	5
1. National identity	3.95(0.68)
2. Religious identity	4.52(0.63)	.30***
3. BII distance	3.15(0.79)	−.03	.14***
4. BII conflict	2.60(0.81)	−.09*	−.08*	.31***
5. SIC overlap	2.69(0.56)	.06	.11**	.08*	.01
6. SIC similarity	2.68(0.77)	.11**	−.08*	−.18***	−.03	.00

Note. BII = Bicultural Identity Integration; SIC = Social Identity Complexity.

* p< 0.05, ** p < 0.01, *** p < 0.001.

Classes of adolescents with distinct identity patterns

We used Latent Profile Analysis (LPA) to identify subgroups of adolescents with distinct identity patterns. The model fit statistics of the class solutions are presented in Table 2. The BIC and BLRT pointed toward a 5-class solution, the LMR-LRT toward a 4-class solution which we favored due to its parsimony.²

Table 2. Model fit statistics of the Latent Profile Analysis (LPA) and class sizes.

C	BIC	LMR–LRT	BLRT	E	Class sizes
2	2635.61	−1410.39 **	−1410.39 ***	0.990	14,726
3	2487.75	−1291.38 **	−1291.38 ***	0.896	10,141,589
4	2373.31	−1207.54 **	−1207.54 ***	0.912	38,10,507,185
5	2340.77	−1140.41	−1140.41 ***	0.878	10,30,109,169,423

Notes. C = Classes, E = Entropy. Class sizes are reported based on the estimated posterior probabilities. Higher class solutions inadmissible.

** p< 0.01, *** p < 0.001.

Figure 1 depicts the 4-class solution and Table 3 shows how the classes differed in terms of BII and SIC. In general, respondents scored high on both identity measures. Beyond this, there were slight but important differentiations in identity strength which, albeit small, did correlate in meaningful ways with BII and SIC. The first class, “religiously-oriented strong dual identifiers” (69%), had very strong religious, M = 4.84, SD = 0.02, and strong but significantly weaker national identities, M = 4.05, SD = 0.03, W(1) = 651.60, p < .001, d = 1.20. Respondents in this class reported comparatively high BII distance and SIC overlap, and low BII conflict and SIC similarity. The second class, “equally-strong dual identifiers” (25%), had strong religious, M = 4.08, SD = 0.06, and slightly weaker national identities, M = 3.83, SD = 0.04, W(1) = 18.73, p < .001, d = 0.33. We label this second class “equally-strong” as the difference between their identities was much smaller than the difference in the “religiously-oriented” class, as indicated by the effect sizes and evident in Figure 1. Respondents in this class reported lower BII distance than “religiously-oriented strong dual identifiers”, and comparatively high BII conflict, SIC similarity and SIC overlap. The third class, “equally-medium dual identifiers” (5%), had equally strong, medium religious, M = 3.20, SD = 0.14, and national identities, M = 3.44, SD = 0.09, W(1) = 2.86, p = .091. They were characterized by high SIC similarity and low BII distance. The fourth class, national identifiers (1%), had medium national, M = 3.27, SD = 0.49, and significantly weaker, extremely low religious identities, M = 1.28, SD = 0.09, W(1) = 17.73, p < .001. We did not compare this class with the other classes as it included only ten respondents.

Table 3. Estimated means per latent class.

	Religiously-oriented strong dual identity (C1)	Equally-strong dual identity (C2)	Equally-medium dual identity (C3)	National identity (C4)
BII conflict	2.53 (0.04)	2.80 (0.06)	2.70 (0.11)	2.46 (0.33)	C1< C2
BII distance	3.22 (0.04)	3.06 (0.06)	2.78 (0.10)	2.83 (0.37)	C1> C2> C3
SIC similarity	2.63 (0.04)	2.73 (0.06)	2.95 (0.08)	2.75 (0.11)	C1< C2,C3
SIC overlap	2.63 (0.03)	2.57 (0.05)	2.40 (0.13)	2.19 (0.10)

Note. BII = Bicultural Identity Integration; SIC = Social Identity Complexity. Class comparisons based on Chi-square tests (df = 1). Significant differences based on p < .05.

Further comparisons showed that the three larger classes differed in the strength of religious (ps < .001) and national (ps ≤ .038) identity. There were no differences between the classes in terms of adolescents’ age, gender and citizenship. However, Pakistani (in contrast to Bangladeshi) respondents, and those recruited in 2016 (in contrast to 2015), were under-represented in the equally-medium dual identity class.

Figure 1. Four classes of adolescents with distinct identity patterns.

Discussion

All but 1% of adolescents in our sample showed a pattern of dual identification. Yet, three distinct forms of such a dual identity emerged: “religiously-oriented strong” (69%), “equally-strong” (25%), and “equally-medium” (5%). Given that the majority of respondents’ identity patterns were defined by the first two classes, as well as the very small cell sizes of the other two classes, we focus our discussion on the two larger classes. Adolescents with “religiously-oriented strong dual identities” identified strongly with the religious and national group but relatively more so with the religious group. This identification pattern was related to experiences of comparatively greater distance between both identities and less identity conflict. This goes against the popular view that more religious Muslims must struggle to combine religious and national identities. Arguably, these adolescents experienced less identity conflict precisely by focusing on their religious commitment. Affirming one identity more strongly than the other (in this case the religious identity) might thus help adolescents to manage conflicting normative expectations. Adolescents with “equally-strong dual identities” also felt strongly attached to both groups, but their religious identity was weaker, so that both identities were more balanced. Relative to the first group, they reported lower distance but higher internal conflict. This suggests that “equally-strong dual identifiers” are more likely to perceive discrepancies between their identities as they gravitate toward both groups and perceive the values of both groups as equally important (Tadmor & Tetlock, 2006). They may be engaged in resolving tensions and work their way through these conflicts by forming complex cognitive links between their identities – as indicated, for example, by their high levels of perceived identity similarity and overlap.

Our findings further showed that “equally-strong dual identifiers” had less complex social identities in terms of SIC similarity than “religiously-oriented strong dual identifiers”. This supports the idea that SIC might operate differently among minorities for whom high overlap and similarity might signify an attempt to reconcile and combine their dual identities. Moreover, BII distance and SIC similarity appeared to be two sides of the same coin, at least when considering national and religious identity categories among religious minority members: because, whereas “religiously-oriented strong dual identifiers” reported greater BII distance and less SIC similarity, “equally-strong dual identifiers” reported lower BII distance and greater SIC similarity. This is in line with Miramontez et al. (2008) and suggests that BII distance and SIC similarity both refer to perceptual aspects of identity integration.

The findings of Study 1 provide an important initial step to understanding how dually identified individuals make sense of and experience their dual identity. Yet, this study can only provide a snapshot of what is actually a highly complex, dynamic developmental process. Thus, to understand better the multiple options of developing a dual identity, BII and SIC during adolescence, we conducted a second longitudinal study among Muslim minority adolescents in the US.

Study 2

Methods

Sample and procedure

Respondents were recruited from three schools and community sites in the metro Detroit, Michigan and Columbus, Ohio areas to participate in three annual waves of data collection starting in 2015. We specifically chose to work with these charter schools and community sites as they have programs (e.g., English as a second language), services (multilingual translators), and staff that are knowledgeable about and sensitive to Muslim students’ concerns. Self–identified Muslim students were invited to participate in the study in exchange for 10 USD payment. Parental and student consent were obtained consistent with Institutional Review Board guidelines. Respondents completed paper surveys in English in group settings. The total sample comprised 211 adolescents, but we focused on those able to identify their branch of Islam (n= 190, M = 14.14 years, SD = 0.92, range 12-17, 45.2% male). Adolescents were mostly Sunni (73.2%) or Shia (22.1%). The three largest, self–identified ethnic groups in the sample were Arabs (66.8%), Somali (15.3%) and African (6.8%). The majority were born in the US (73.0%), those who were born abroad had lived in the US for 8 years (M = 8.05, SD = 4.09, range 1-16 years).

Of the total sample, 77.9% participated again at W2 (2016) and 67.4% continued to participate at W3 (2017). Respondents who dropped out after W1 had weaker religious identities (M= 4.29, SD = 0.83) than respondents who completed at least the first two waves (M= 4.59, SD = 0.52, Welch (1, 50.47) = 4.94, p = 0.031, d = 0.50), and stronger national identities (M= 3.95, SD = 0.58) than respondents who completed the first two waves (M= 3.67, SD = 0.85, Welch (1, 95.43) = 5.80, p = 0.018, d = 0.35). Respondents who dropped out after either W1 or W2 reported higher SIC similarity (M= 2.61, SD = 0.89) than respondents who participated in all three waves (M= 2.29, SD = 0.97, F(1,188) = 4.92, p = 0.028, d = 0.34).

Measures

Muslim and American identification

We used the same measures (items and response scales) as in Study 1 to assess Muslim (αs: W1 = .86, W2 = .90, W3 = .91) and American (αs: W1-W3 = .88) identification. The items loaded on a single factor explaining 67.9%, 71.7% and 74.9% of the variance of religious identity, and 67.6%, 68.6% and 67.8% of the variance of national identity at W1-W3, respectively.

Bicultural identity integration (BII)

BII was assessed with the same eight items used in Study 1. At each wave of measurement, the items loaded on two factors explaining 56.5%, 61.0% and 0.9% of the variance at W1-W3, respectively (αs distance: W1 = .62, W2 = .70, W3 = .76, αs conflict: W1 = .77, W2 = .81, W3 = .90).

Social identity complexity (SIC)

SIC overlap was assessed with one item: “What percentage of Americans do you think are Muslims?”. The response options ranged from 1 (0% – None), 2 (10%) to 11 (100% – ALL). We recoded and transformed the scale into a 5-point scale with higher values indicating greater overlap. SIC similarity was assessed with the same two items used in Study 1. Both were recoded, so that higher values indicated greater similarity (rs: W1 = .33, W2 = .38, W3 = .34, ps < .001).

Analytical procedure

We used a multi-step analytical procedure. First, we estimated a parallel process latent growth curve model (LGCM) based on the longitudinal measures of religious and national identity, BII and SIC to gain an overall impression of developmental changes. Second, we estimated a parallel process growth mixture model (GMM) based on the longitudinal measures of religious and national identity. This person-oriented approach uncovered classes of adolescents who followed different developmental trajectories. Third, we examined whether the classes of dual identity development differed on W1 demographic characteristics and in their initial level and rate of change in BII and SIC (the procedure is explained in Study 1). We used full information maximum likelihood (FIML) to handle missing data and a robust estimator (MLR) to account for potential non-normality.

Results

Descriptive statistics are shown in Table 4. Both identities were strong (above the scale’s midpoint), and religious identity was stronger than national identity at each assessment (ps < 0.001). Although not related in W1, religious and national identities were positively correlated in W2 and W3.

Table 4. Descriptive statistics Study 2.

	M(SD)	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
W1 National ID	3.73 (0.80)
W1 Religious ID	4.52 (0.61)	−.06
W1 BII distance	3.28 (0.89)	−.22**	.05
W1 BII conflict	2.63 (0.85)	−.13	−.17*	.26***
W1 SIC overlap	2.50 (0.77)	−.09	−.10	.13	.12
W1 SIC similarity	2.40 (0.95)	.31***	−.15*	−.21**	.02	.18*
W2 National ID	3.75 (0.81)	.35***	.12	−.33***	−.04	−.01	.10
W2 Religious ID	4.50 (0.69)	−.01	.35***	.15	−.24**	−.13	−.13	.21**
W2 BII distance	3.16 (0.94)	−.10	−.03	.62***	.09	.09	−.06	−.21**	.03
W2 BII conflict	2.55 (0.88)	−.11	−.15	.30***	.41***	.22**	.16	.03	−.24**	.20*
W2 SIC overlap	2.43 (0.78)	−.05	−.20*	.10	.19*	.55***	.04	.02	−.06	.03	.05
W2 SIC similarity	2.51 (1.03)	.08	.00	−.22**	.12	.13	.32***	.18*	−.05	−.20*	.06	.17*
W3 National ID	3.88 (0.78)	.39***	.00	−.16	−.08	−.16	.13	.49***	.13	−.25*	−.09	−.00	.12
W3 Religious ID	4.40 (0.74)	.08	.28**	−.03	−.28**	.06	−.03	.18	.34***	.05	−.29**	−.11	−.29**	.25**
W3 BII distance	2.87 (0.92)	−.25**	−.02	.30**	.06	.16	−.15	−.31**	.00	.54**	.07	.01	−.29**	−.36***	.11
W3 BII conflict	2.62 (0.96)	.01	−.14	.07	.30**	.13	.06	.05	−.05	.20*	.49***	.13	.04	−.13	−.16	.33***
W3 SIC overlap	2.48 (0.86)	.03	−.14	−.00	.21*	.48***	.25**	−.03	−.20*	.06	.05	.58***	.07	−.13	−.25**	.11	.12
W3 SIC similarity	2.63 (0.98)	−.01	−.13	−.05	.15	.06	.14	.11	−.16	−.14	−.02	.24*	.40***	.04	−.33***	−.12	−.05	.23**

Note. ID = identity; BII = Bicultural Identity Integration; SIC = Social Identity Complexity. W1, W2, W3 denotes waves 1, 2 and 3.

⁺p< 0.10, * p< 0.05, ** p < 0.01, *** p < 0.001.

Average changes in religious and national identities, BII and SIC

To get an overall impression of stability and change in our study variables, we specified a parallel process LGCM. This model estimated intercepts and slopes, which can be interpreted as an adolescent’s initial level and rate of change over time, respectively. We specified a model in which the intercepts of the manifest variables were fixed at 0, the time scores of the slope factors were fixed at 0, 1, and 2, and all latent intercepts and slopes correlated.

The model fit the data well: χ²(df) = 90.54 (81), p = .220, CFI = .982, TLI = .967, RMSEA = .025, 90% CI [.000, .049], SRMR = .045. The results indicated that, on average, adolescents had very strong and stable religious identities, b(SE) = 4.52 (0.04), p < 0.001, m(SE) = −0.05 (0.03), p = 0.116, and strong and increasing national identities, b(SE) = 3.72 (0.06), p < 0.001, m(SE) = 0.10 (0.04), p = 0.009. BII distance was medium to high at W1 and decreased over time, b(SE) = 3.29 (0.07), p < 0.001, m(SE) = −0.22 (0.04), p < 0.001. BII conflict was medium to low and stable, b(SE) = 2.62 (0.06), p < 0.001, m(SE) = −0.03 (0.05), p = 0.535. SIC similarity was medium to low and increased over time, b(SE) = 2.40 (0.07), p < 0.001, m(SE) = 0.13 (0.05), p = 0.013, and SIC overlap was medium to low and stable, b(SE) = 2.48 (0.05), p < 0.001, m(SE) = −0.01 (0.04), p = 0.801. Inspection of the correlations between the intercepts and slopes showed that stronger initial national identities were linked to an increase in religious identity (r = 0.06, p = 0.020) and a decrease in SIC similarity (r = −0.11, p = 0.006). Lower initial levels of SIC similarity were linked to an increase in national identity (r = −0.06, p = 0.058), and an increase in national identity was linked to an increase in SIC similarity (r = 0.05, p = 0.060). Lower levels of BII conflict were linked to an increase in religious identity (r = −0.05, p = 0.093). An increase in religious identity was linked to a decrease in SIC similarity (r = −0.07, p = 0.003) and a decrease in SIC overlap (r = −0.04, p = 0.009). Higher initial BII distance was linked to a decrease in BII conflict (r = −0.07, p = 0.061), and an increase in BII distance was linked to an increase in BII conflict (r = 0.09, p < 0.001).³

Classes of adolescents with distinct religious and national identity trajectories

To examine if there was variability in the development of religious and national identities, we specified a parallel process GMM. The model fit statistics are shown in Table 5. To decide on the number of classes, we used the criteria explained in Study 1. The BLRT pointed toward a six-class solution which we rejected on grounds of parsimony and because BIC and LMR-LRT clearly pointed toward a four-class solution. The four classes are shown in Figure 2. Following class enumeration, we examined associations between class membership and BII and SIC. The results are shown in Table 6.

Table 5. Model fit statistics of the GMM and class sizes.

C	BIC	LMR–LRT	BLRT	E	Class sizes
2	1995.92	−978.04 *	−978.04 ***	0.994	186,4
3	1959.66	−937.62	−937.62 ***	0.965	4,11,175
4	1935.29	−906.37 *	−906.58 ***	0.933	148,3,28,11
5	1936.35	−881.07	−881.07 ***	0.937	3,27,5,147,8
6	1937.16	−868.48	−868.48 *	0.923	8,3,5,129,8,37

Notes. C = Classes, E = Entropy. Class sizes are reported based on the estimated posterior probabilities. Higher-class solutions inadmissible.

* p< 0.05, *** p < 0.001.

Table 6. Estimated means per latent class.

		Religiously-oriented strong dual identity (C1)	Equally-strong dual identity (C2)	Religious to national (C3)	Incorporating religious (C4)	Results Wald tests
BII conflict	Intercept	2.50 (0.07)***	2.79 (0.15)***	3.38 (0.44)***	4.21 (0.57)***	C1< C2^+
	Slope	−0.01 (0.05)	0.11 (0.14)	−0.34 (0.23)	−0.41 (0.27)
BII distance	Intercept	3.28 (0.07)***	3.06 (0.15)***	3.68 (0.35)***	3.67 (0.06)***
	Slope	−0.17 (0.05)***	−0.26 (0.10)**	−0.46 (0.19)*	−0.22 (0.39)
SIC similarity	Intercept	2.38 (0.08)***	2.66 (0.19)***	1.90 (0.36)***	2.53 (0.71)***
	Slope	0.08 (0.06)	0.08 (0.18)	0.77 (0.25) **	0.25 (0.31)
SIC overlap	Intercept	2.44 (0.06)***	2.75 (0.16)***	2.30 (0.26)***	2.75 (0.58)	C1< C2^+
	Slope	−0.02 (0.04)	−0.10 (0.09)	0.15 (0.24)	0.27 (0.52)

Note. BII = Bicultural Identity Integration; SIC = Social Identity Complexity. Unstandardized effects (standard errors in parentheses). Class comparisons based on Wald tests (df = 1). The two smaller classes were not included in class comparisons due to small sample size.⁺ p< 0.10, * p< 0.05, ** p < 0.01, *** p < 0.001.

The first class, “religiously-oriented strong dual identifiers” (78%), had very strong and stable religious identities, b(SE) = 4.71 (0.03), m(SE) = −0.03 (0.03), p = 0.320, and strong and slightly increasing national identities, b(SE) = 3.80 (0.06), m(SE) = 0.08 (0.04), p = 0.054. They reported the lowest level of BII conflict and, as the slopes did not differ between classes, continued to experience less conflict than others across time. They also perceived comparatively less overlap between their identities. The second class, “equally-strong dual identifiers” (14%) had equally stable and strong religious, b(SE) = 3.73 (0.12), m(SE) = 0.04 (0.07), p = 0.605, and national, b(SE) = 3.73 (0.10), m(SE) = 0.05 (0.11), p = 0.621, identities. Their religious identities were weaker than the religious identities of adolescents in the “religiously-oriented strong dual identity” class, W(1) = 89.32, p < 0.001. In all other aspects (i.e., change in religious identity, level and change in national identity) “religiously-oriented” and “equally” strong dual identifiers were similar (ps > 0.416). “Equally-strong dual identifiers” reported the highest levels of SIC similarity and overlap and significantly more BII conflict than “religiously-oriented strong dual identifiers” at the onset of the study and across time. The third class, “Religious to national” (6%), had very strong, sharply decreasing religious identities, b(SE) = 4.79 (0.06), m(SE) = −1.04 (0.19), p < 0.001, and relatively low, sharply increasing national identities, b(SE) = 2.62 (0.35), m(SE) = 0.45 (0.16), p = 0.004. This was accompanied by sharp increases in SIC similarity. The final class, “Incorporating religious” (2%), had weak and sharply increasing religious identities, b(SE) = 1.37 (0.26), m(SE) = 1.05 (0.33), p = 0.002, and stable and strong national identities, b(SE) = 3.87 (0.40), m(SE) = 0.01 (0.09), p = 0.935. Given the small sizes of the latter two classes (i.e., eleven and three respondents) we did not compare them to the other classes.

A comparison of the larger classes on W1 demographic characteristics indicated that adolescents following the Sunni (versus Shia) denomination of Islam and those with higher (versus lower) GPA were more likely in the “religiously-oriented strong dual identity” class. The classes did not differ in terms of gender or place of birth (US vs not).

Figure 2. Four classes of dual identity development.

Discussion

We identified four distinct groups of adolescents that differed in the combination, development and experiences of their dual identities. We focus our discussion on the two larger groups, “religiously-oriented” and “equally” strong dual identifiers, which together accounted for 92% of the sample. “Religiously-oriented strong dual identifiers” identified strongly (above the scale midpoint) with both of their identities, but relatively more so with their religious identity. They reported relatively high BII distance (above the scale midpoint) and the lowest level of BII conflict, which is in line with the results of Study 1 and corroborates the idea that compartmentalization (at a very high level) can be an effective way to manage multiple and potentially conflicting identities. “Equally-strong dual identifiers”, in contrast, identified equally strongly with both identities and experienced more internal conflict. The consistently higher levels of SIC similarity and overlap among “equally-strong dual identifiers” could indicate that they are in the process of actively integrating their identities.

General discussion

The aim of our research was to explore interrelationships between dual identity, Bicultural Identity Integration (BII) and Social Identity Complexity (SIC). To this end, we conducted two studies: a cross-sectional one in the UK and a longitudinal one in the US. We focused on Muslim minority adolescents, an at-risk population for identity-based threats, and used person-oriented analytical approaches to identify subgroups of adolescents with different dual identity patterns and developmental trajectories. In both national contexts, the majority of adolescents held two distinct patterns of identification: one group reflecting a “religiously-oriented strong dual identity” (69% and 78% in the UK and US, respectively) and another reflecting an “equally-strong dual identity” (25% and 14% in the UK and US, respectively). “Equally-strong dual identifiers” experienced their identities as more conflicting than “religiously-oriented strong dual identifiers”. Additional smaller subgroups of adolescents emerged in both contexts, but, given their extremely small size, we focus our discussion on the two larger groups.

“Religiously-oriented” and “equally-strong” dual identifiers identified strongly with both the religious and national group, thus reflecting a pattern of strong dual identity, contrary to claims that Muslim religious identity is incompatible with Western values and culture. As the group of “religiously-oriented strong dual identifiers” considerably outnumbered the group of “equally-strong dual identifiers”, most adolescents identified more strongly with the religious relative to the national group. This indicates that religious identity is a vital and essential identity source (Phalet et al., 2018). This is not, however, unique to Muslims as other religious groups, not studied in our research, such as Christians in the US, also see themselves as Christian, first, and American, second (Pew Research Center, 2011).

Regarding developmental changes in dual identity, our longitudinal findings from the US indicate that religious identities might develop earlier than national identities as religious identities were of great importance at the onset of the study, with little to no variation over time, while national identities became even stronger during adolescence. This suggests that a religious sense of self develops and manifests early in life (Phalet et al., 2018) while national identities still gain in importance during adolescence (Spiegler et al., 2019). Religious and national identity development appeared to be coupled processes, as initially stronger national identities were linked to increases in religious identity. The initial level of religious identity was not, however, related to changes in national identity. This suggests that religious identity neither interferes with nor fosters the development of a national identity among Muslim adolescents in the U.S., and that national identity development is affected by other variables not included in our research such as experiences of discrimination and intergroup contact (Fleischmann et al., 2019; Leszczensky et al., 2020; Saleem et al., 2018; Sirin et al., 2008), or exposure to negative media stereotypes (Saleem & Ramasubramanian, 2019; Saleem et al., 2019). The positive association between initial national identity and change in religious identity further indicated that religious and national identities became more integrated (here, equal in strength) with time.

Our longitudinal findings also contribute to a better understanding of developmental changes in BII and SIC. On average, adolescents experienced their religious and national identities as more harmonious than conflicted, with no apparent changes over time. Further, they experienced their identities as more distanced than blended at the age of 14, but over time the two identities became more integrated. These findings are in line with prior research among Muslims in the UK and US who largely prefer to describe themselves as hyphenated (British-Muslim or Muslim-American) and experience their identities as more complementary than conflicting (Hutchison et al., 2015; Sirin et al., 2008). Our longitudinal findings did not reveal any developmental changes in SIC overlap, but an increase in SIC similarity which seemed to be driven by a small group of adolescents (i.e., “religious to national” who comprised 6% of the US sample). On average, adolescents thus consistently perceived their religious and national identities as slightly more distant than overlapping and more different than similar from 14 to 16 years. Our findings further showed that an increase in religious identity was linked to a decrease in perceived identity overlap and similarity, while an increase in national identity was associated with an increase in identity similarity. The strength of multiple ingroup identities is thus related to how they are organized within the self. We did not find longitudinal associations between BII and SIC, indicating that the development of self-perceptions (e.g., perceiving identities as blended) and group-perceptions (e.g., perceiving groups as similar) might be independent processes.

The most significant and striking finding of our research was that adolescents with “equally-strong dual identities” experienced their identities as more blended, but also more conflicting, than adolescents with “religiously-oriented strong dual identities”. This supports the notion that individuals who perceive both identities as equally important are more likely to experience identity conflict than those who identify more strongly with one of the groups (Tadmor & Tetlock, 2006) and that more separated and compartmentalized identities (at a very high level) might be an effective identity management strategy for adolescents facing identity threats (Amiot et al., 2007; Huynh et al., 2011). It also shows that apparently small differences in dual identity patterns and trajectories have an impact on how these identities are perceived and experienced. At first sight, “religiously-oriented” and “equally” strong dual identifiers may appear very similar due to their strong dual identities. However, the crucial difference between both groups was the strength of their religious identity and the resulting gap between both identities which was associated in meaningful ways with BII and SIC. The comparatively higher levels of identity conflict among “equally-strong dual identifiers” further indicate that a positive correlation between religious and national identities is not necessarily indicative of identity compatibility. Prior research has equated a positive correlation with compatibility and a negative correlation with conflict. From our point of view, this might have been misleading. A positive correlation simply shows that individuals who identify strongly with one group are likely to identify strongly with the other group. A correlation does not reveal whether the identities are experienced as conflicting or not. It is thus important to measure how identities are actually experienced.⁴

Another key aspect that our work has highlighted concerns the meaning of the social identity complexity concept. Our findings suggest that the underlying meaning of SIC may differ, in the case of minority groups, when examining relationships between identities that are held only by the minority group (e.g., Muslim religious identity), compared with categories shared by both the minority and the majority group (e.g., British or American national identity). In this case, for minorities, greater perceived overlap and similarity between these identities may not be indicative of lower complexity (as theoretically defined) but may in fact reflect a different form of complexity in which minority individuals manage to cognitively align two very different identities. In contrast, minority individuals who show less perceived overlap and similarity between their religious and national identities (i.e., individuals who would be characterized as high in SIC, as per the construct definition) may in fact be drawing a clearer distinction between the two categories which allows them to retain strong identification with their minority identities (for example, the increases in religious identification in Study 2 being related with less perceived overlap and similarity between the religious and national identities). It remains, however, for future research to examine such potential differences in the conceptual meaning of social identity complexity by conducting a more focused comparison of minority and majority group members.

An important question for future research is how adolescents cope with comparatively high levels of identity conflict. Our correlational findings do not suggest that identity conflicts are resolved by a focus on religious identity as higher initial levels of identity conflict did not predict an increase in religious identity. Instead, higher initial levels of identity conflict predicted a decrease in religious identity, indicating that, at least among Muslim American adolescents, identity conflicts are resolved by identifying less with the religious group (or vice versa: lower initial levels of identity conflict allowed adolescents to develop stronger religious identities). Future research is also needed to examine whether and how differences in identity integration and complexity have implications for psychological and socio-cultural adaptation. BII conflict, for example, seems to be positively associated with identity exploration (Huynh et al., 2018), active engagement, personal growth (Jones & Hynie, 2017), and more creative and complex ways of thinking (Tadmor & Tetlock, 2006). Moreover, higher levels of blendedness are associated with more adaptive conflict management strategies and well-being (Arias‐Valenzuela et al., 2019), and a higher motivation to integrate (Ward et al., 2018). Yet, whether there are differences in adjustment between “religiously-oriented” and “equally” strong dual identifiers remains a question for future research.

Despite the various strengths of our research (e.g., the use of person-centered approaches, the longitudinal design in Study 2, and the replication of findings across two national contexts), there are limitations we acknowledge. First, we focused on Muslim minority adolescents, which makes it imperative to examine whether our findings generalize to other ethno-religious minority groups. Second, although our analytical approach is well-suited to detect heterogeneity in a population it is limited in its analysis of causal relationships. Moreover, our relatively small sample sizes prevent us drawing firm conclusions about adolescents with less normative dual identity patterns and trajectories. Additionally, Study 2 is longitudinal, but it still covers only a relatively short time period during adolescence. Finally, there was selective panel attrition in Study 2, which is why the findings on dual identity and SIC similarity require replication, ideally in larger and more representative samples.

Conclusion

Despite widespread speculation that the religious identity of Muslim immigrants is incompatible with or an obstacle to their successful integration in Western societies, the majority of Muslim adolescents in our British and North American samples strongly identified with both their religious and national identities. The largest group, “religiously-oriented strong dual identifiers”, strongly identified with both groups, albeit having a stronger religious compared to national identification. The second largest group, dual identifiers, equally and strongly identified with both identities and perceived them as more similar, balanced, and blended, yet experienced relatively more conflict between these identities compared to the first group. These findings highlight individual differences in dual identity and the complex interplay of how dual identities are expressed and experienced among Muslim immigrants in Western societies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here

Notes

1. Our measures and items can be found in the online supplementary materials. To examine the factor structure of the dual identity and BII in both studies, we conducted principal component factor analyses with oblimin rotation.

2. Results of the five-class solution can be found in the online supplementary material.

3. Given the relatively small sample and exploratory nature of our research, we included findings significant at p < .10. We do not report here correlations between intercepts (e.g., intercept national and intercept religious identity), and between intercepts and slopes of the same process (e.g., intercept national and slope national identity). These can be found in the online supplementary materials.

4. To further support this, we have conducted post-hoc analyses which can be found in the online supplementary material.