Positive (or negative or no) interdependence among identities?

Scholars in the social identity tradition (e.g., [17, 18]) originally assumed that a negative relation existed between different aspects of identity within the wider self-concept [19, 20]. According to this perspective, as personal identity (i.e., the unique part of the self which concerns distinctiveness from others) becomes salient, individuals become less able to perceive the self in terms of its group membership(s) and similarities with others (social identity; i.e., the part of the self that is socially shared with others, including one’s group memberships; [17, 21–23]). Multiple social identities also ‘compete’ with one another, in order to become the “focal category” (e.g., metacontrast; [23], p.455). This means that, at least at times, multiple social identities can inhibit each other. For example, the more an individual acts in terms of their lawyer identity, the less they act in terms of their maternal female identity [9, 24, 25]. Thus, although it is generally recognised that personal and (multiple) social aspects of the self are not completely independent, they are often researched as if they were mutually exclusive [19, 26].

This theoretical background may be one reason why the question of inter-identity fit (i.e., degree of multiple identities overlap) has received little attention. However, more recent research implies that identities need not always be negatively related (see [27], for an overview). For instance, there are specific conditions (e.g., when group status or identity continuity is high) in which inter-identity fit becomes more strongly overlapping [28–30]. Moreover, studies that consider inter-identity fit (e.g., [31–33]) demonstrate that subordinate identities can be successfully embedded in a superordinate entity. Similarly, research on crossed categorizations suggests that two categories can be successfully integrated within a superordinate identity (“we”) as long as the two subcategories are not considered highly self-relevant [34]. Thus, there are several conditions under which hierarchically structured social identities are somewhat overlapping.

A growing body of research acknowledges that inter-identity fit can exist between potentially orthogonal identities as well (cf. the notion of “intersectionality” see [35, 36]). For example, Roccas and Brewer’s [37] work on social identity complexity presents four key types of inter-identity fit, ranging from compartmentalised (or independent) to more interdependent (e.g., dominance) relations. Research has, however, not extended substantially beyond these informative, yet predominantly descriptive studies (e.g., describing the management of fit between two social identities; [38, 39]). The cognitive-developmental model of social identity integration [40] is one notable exception. In this model the authors discuss how inter-identity fit can develop because of individual-level antecedents such as personal coping and perceived social support, which facilitate potentially effortful integrative processes that lead to stronger inter-identity fit (e.g., [11, 38]). However, in the current paper, we argue that this analysis overlooks the importance of the type of identity being managed (e.g., holisticness), which is a group characteristic.

Holistic groups and self-defining identities

Baray et al. [41] approached the question of how multiple identities fit together by focusing on a ‘special’ type of group with particular normative content (for an alternative identity fusion account, see [42]). Groups that prescribe a philosophy for life (e.g., a religious group) can paradoxically strengthen the personal identity of individuals who apply that philosophy (e.g., make an individual more certain of their personal attitudes or convictions). For example, an individual may use the group’s philosophy (e.g., Biblical teachings) to guide their individual behaviour in everyday life (e.g., motivating a personal decision about whether to prioritize self-achievement or self-enhancement goals), endowing them with a sense of certainty and self-assuredness about what is important to them personally. As a consequence, these social identities derived from highly holistic groups can promote a fit between that specific social identity and an individual’s personal identity [41, 43]. In contrast, groups that do not provide such a normative recipe for life only influence an individual in situ (e.g., TV show followers). Therefore, these groups low in holisticness only inform a specific social identity and this remains independent from other identities that an individual may possess.

Importantly however, Baray et al. [41, 43] did not study the inter-identity fit between different social identities. We move beyond this theoretical framework by exploring the inter-identity fit among multiple social identities. We focus on two specific social identities selected on the basis of a pilot study: religious (relatively high in holisticness) and gender identity (relatively low in holisticness). Specifically, we will investigate to what extent this holistic religious identity becomes more strongly fitting with an otherwise potentially non-overlapping or even conflicting social identity, in this case gender. In this way, we advance our understanding regarding the organisation of the self-concept by thoroughly exploring the inter-identity fit between two specific identities.

We further build on Baray et al.’s research [43] in two key ways. Firstly, we suggest that groups vary on a continuum from more holistic to less holistic (cf. Baray et al. [41], who relied on a dichotomy of self-defining vs. non self-defining). Secondly, we suggest that is it important differentiate holisticness—a group characteristic concerning the capacity for a group to prescribe a philosophy for life, which is socially shared and exists to some extent outside of the self—from the individual characteristic of the personal application of group philosophies to life in general. Notably, where we now refer to groups as more or less holistic, Baray et al. [41] used the term self-defining. We changed this terminology because we wanted to clearly differentiate the holisticness of groups—a group characteristic—from the subjectively applied self-definingness of social identities—a personal characteristic. Unlike Baray et al. [41], we propose that the extent to which an individual applies holistic group philosophies (i.e., self-definingness) will determine how strongly the social identity fits with other identities. Even though the group may have the capacity to be holistic (i.e., it prescribes a philosophy for life), individuals vary in the degree to which this identity is self-defining.

For instance, some may act in ways consistent with their religious identity only when in church, others in more or less every situation (cf. [14]). For the former, religion is less self-defining than for the latter. The implication is that a holistic based identity that is only weakly self-defining remains relatively independent from other identities (i.e., Christian identity is active at church, administrator identity is active at work). In contrast, a strongly self-defining holistic based identity increases inter-identity fit with ancillary social identities, which might otherwise be in conflict with it (i.e., Christian identity being active at church, and at work). Notably, the influence of group philosophies should be visible not only in group members (e.g., devoted Christians) but also those who do not consider themselves group members but still share and apply some elements of the group’s values (e.g., use of Christian values by people who grew up in a Christian culture). Indeed, what is ultimately important in promoting inter-identity fit is the personal application of holistic group philosophies across situations (i.e., to one’s life).

Overview

We proposed that holistic groups have the capacity to provide a philosophy for life that can promote inter-identity fit within the self-concept, provided the individual considers it strongly self-defining (in the sense that they actively apply the group’s philosophies to guide their life). Thus, holisticness is defined as a group characteristic (i.e., how is the group seen in general) and self-definingness as an individual characteristic (i.e., how do you personally experience the identity). To examine this proposal, three studies investigate the inter-identity fit (i.e., an individual’s perception of the degree to which multiple identities within the self-concept overlap) between multiple (religious [more holistic] and gender [less holistic]) social identities as a function of their self-definingness. Our central hypothesis is that whether self-definingness promotes inter-identity fit depends on holisticness: When religion (as a holistic group) is self-defining, it will promote inter-identity fit and will therefore be strongly overlapping with gender identity. Conversely, when gender (as a less holistic group) is self-defining, it will not promote inter-identity fit and will therefore be non-overlapping with religious identity. (For a more detailed discussion about expectations concerning groups low in holisticness, see S2 Appendix.)

Importantly, we used different methodologies to tap into the notion of inter-identity fit between religion and gender. Study 1 focused on self-reported perceptions of inter-identity fit (e.g., whether the values of one social identity overlap with another). Study 2 more directly taps into the underlying cognitive representation of inter-identity fit by using relatively implicit measures of response times to traits associated with religion and gender to assess degree of cognitive overlap among multiple social identities (applying an experimental paradigm adapted from Aron, Aron, Tudor and Nelson [44], and Smith and Henry [45]). Finally, in Study 3 we developed a new measure of social identity content to assess inter-identity fit (i.e., a semantic network analysis of associated traits and characteristics which assesses the fit and overlap of actual identity content). Through the use of a multi-method approach assessing inter-identity fit, we seek convergent evidence for our core proposal that holistic groups promote inter-identity fit when individuals apply group philosophies to guide their life in general.

Method

Results

In order to select the most suitable identities for further investigation in Studies 1–3, holisticness and self-definingness means were inspected for each group (see Fig 1). A repeated measures analysis of variance (RM-ANOVA; with Greenhouse-Geisser degrees of freedom due to a sphericity violation; W (9) = .43, p < .001) showed that Gender and Nationality were perceived to be less holistic than Christianity, Healthy-living and Environmental groups (F (2.96, 334.79) = 27.03, p < .000001, η_p² = .19). Bonferroni corrected pairwise contrasts confirmed that Gender and Nationality were significantly less holistic than all three other groups. Furthermore, Nationality occupied an intermediate position, in that it also differed significantly from Gender. RM-ANOVAs (with Greenhouse-Geisser degrees of freedom due to a sphericity violation; W (9) = .46, p < .001) on self-definingness revealed some significant differences between groups (F (2.79, 315.66) = 3.68, p < .01, η_p² = .03), although effects were much less strong. Bonferroni corrected pairwise contrasts revealed that Gender and Healthy-living groups were experienced as significantly more self-defining than Christianity, but no further differences were found.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mean perceived holisticness and self-definingness of Gender, Nationality, Christianity, Healthy-Living and Environment.

https://doi.org/10.1371/journal.pone.0137879.g001

Thus, the pilot data largely confirmed expectations: Gender scored low on holisticness; Christianity, Healthy-living and Environmental groups were high in holisticness. Nationality emerged with higher holisticness than expected. In contrast, groups did not tend to differ so substantively in self-definingness (which emphasizes the utility of this variable in our planned moderation analysis). For this reason, we also tested which identities would have most potential to affect inter-identity fit. We thus tested our hypothesis that the impact of self-definingness for inter-identity fit would be moderated by holisticness: We expected an interaction between holisticness and self-definingness, whereby only highly holistic groups (Christianity, Healthy-living, Environment) would show positive relations between self-definingness and inter-identity fit (i.e., less holistic groups (Gender, Nationality) would not).

Bivariate correlations (see Table 1) provide support for this expectation: Self-definingness was generally more strongly related to inter-identity fit for the three groups which were expected to be high in holisticness and more weakly related in those low in holisticness. For example, Christianity shows a moderate to strong relation between its self-definingness (Table 1, row 10) and inter-identity fit (column 12), while gender shows no discernible relation between its self-definingness (row 2) and inter-identity fit (column 4). For a discussion of bivariate results concerning the relation between identification and self-definingness see the methodological section in S1 Appendix.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Bivariate correlations of identity variables for all five groups.

https://doi.org/10.1371/journal.pone.0137879.t001

In order to test our hypothesis more directly, we conducted multi-level regression, regressing a dummy variable of holisticness (0 = low holisticness (Gender, Nationality), 1 = high holisticness (Christianity, Healthy-living, Environment); step 1), mean centred self-definingness (step 2) and the interaction between holisticness and self-definingness (step 3) on our dependent variable—inter-identity fit—while controlling for gender and age (both of which were non-significant). A multi-level (random-intercept) regression was conducted in MLwiN (version 2.23) to control for the repeated measures design (i.e., participants answered the questions for all five groups). The intraclass correlation indicated that 50.37% of variance occurred at the participant-level, reinforcing the importance of this multi-level control.

Regressions confirm the hypothesized interaction: Holisticness (B = 0.09, SE = 0.08, t (554) = 1.17, p = .12) was only related to overall group fit through its interaction with self-definingness (B = 0.10, SE = 0.50, t (554) = 1.92, p < .03). There was also a main effect of self-deiningness (B = 0.10, SE = 0.50, t (108) = 2.18, p < .02), indicating that—somewhat unexpectedly -self-definingness does still have some positive association with inter-identity fit for groups low in holisticness (which again seems primarily attributable to the more holistic nature of Nationality than initially expected). Notably however, this association is significantly higher for groups high in holisticness. The overall fit of the model was good, in comparison to a model with only controls entered (χ²_difference (3) = 42.88, p < .001). Thus, as expected, the extent holistic philosophies are applied to life (i.e., self-definingness) was central to the ability of holistic groups to promote inter-identity fit.

Discussion

Results confirmed our expectation that there is substantial variation between groups in the extent to which they (a) holistically prescribe a philosophy for life, and (b) are actually used across all life contexts in this self-defining way. Moreover, the fit between multiple social identities was influenced by a combined effect of self-definingness and holisticness: When an identity was both holistic and strongly self-defining, it fitted more strongly with multiple (otherwise unrelated) social identities. In other cases, when either the identity was less holistic and/or it was more weakly self-defining, different aspects of the self fitted together more weakly. Interestingly, holisticness appears to be a quality of group philosophies which can be either self-selected (e.g., Christianity) or ascribed (e.g., Nationality). Moreover, we provide important preliminary evidence that Environmental and Healthy-living groups (in addition to religion) provide philosophies for life and are also associated with stronger inter-identity fit (although less strongly than religion). This suggests it is the group’s ideologies and the way they are personally applied which is key to promoting inter-identity fit.

Findings strongly and consistently identified two groups that best characterize more and less holistic identities: Christianity and Gender respectively. Results indicate that although gender carries great personal significance and meaning, in modern society (at least in this culturally Western sample) its social identity content does not generally prescribe its members with a particular philosophy to follow that truly guides their life (e.g., determines personal goals), in the same way that Christianity does (e.g., promoting inter-identity fit). Importantly for this research, gender and Christianity are orthogonal identities (i.e., being a woman says nothing about one’s religion), so their fit can be attributed to the individual’s representation of the identity. Moreover, because gender is naturally occurring (i.e., not personally chosen) variation in fit between religious and gender identities should be induced by the use of holistic group philosophies, not pre-existing similarities or differences between identity content. Consequently, gender and Christianity are expected to be a good starting point for our analysis, and will be utilized in the proceeding studies. Notably, these studies focus on the female gender identity only. There were methodological and theoretical reasons for this. Methodologically, if more than one gender were used, the differing contents of female and male identities means that they could not be aggregated nor even directly compared in studies 2 and 3. Theoretically, it has often been found that women tend to consider gender a more central aspect of their self-concept (e.g., [50]) and thus we thought it likely that cleaner and clearer effects may be achieved in female populations. Thus, holding these group factors constant allowed us to focus on the application of different measurement methods in studies 1–3 in order to demonstrate convergent evidence for our hypothesis.

Method

Results

We expected that only self-definingness of religion (as more holistic) would be associated with stronger fit among social identities. In order to test this hypothesis separately, simple regressions were carried out with the outcome variables of inter-identity fit predicted by, (1) self-definingness of religious identity, and (2) self-definingness of gender identity. Age was entered as a control.

First, regressions using self-defining religiosity as a predictor revealed self-definingness of religion was strongly, positively associated with inter-identity fit: The more self-defining one’s religious identity, the higher the perceived fit of religion in gender (B = 0.21, SE = 0.04, β = .39, 95% CI: 0.23, 0.55; t (129) = 4.91, p < .0001) and gender in religion (B = 0.19, SE = 0.04, β = .32, 95% CI: 0.18, 0.51; t (129) = 4.23, p < .001).

Second, regressions using self-defining gender as a predictor showed that it was not, across the board, strongly associated with inter-identity fit: Indeed, it was not as strongly related to religion’s fit in gender (B = 0.16, SE = 0.09, β = .16, 95% CI: -0.01, 0.33; t (129) = 1.86, p > .06) or the fit of gender into one’s religious identity (B = 0.21, SE = 0.09, β = .20, 95% CI: 0.03, 0.37; t (129) = 2.36, p < .03), although effect sizes were notable.

Reviewing findings on the basis of effect sizes (95% confidence intervals calculated around standardized betas using the “Methods for the Behavioural, Educational and Social Sciences” Package in R [51]), we can see that self-definingness of gender has relatively little influence on inter-identity fit, but self-definingness of religiosity had relatively strong influence. Specifically, self-defining religiosity is not only associated with religious aspects of an individual’s identity, but also has a strong positive association with inter-identity fit, bidirectionally informing the fit of religion in gender.

Discussion

We found that the self-definingness of religion and gender influenced inter-identity fit largely as expected. The self-definingness of religious identity was strongly associated with its fit with the social identity of gender: Crucially, the more self-defining religious identity was, the greater the perceived fit between it and female identity. Although, the self-definingness of gender identity was more strongly associated with perceptions of inter-identity fit than initially predicted, in line with the expectations, the effects of religion were much stronger. Thus, self-definingness of religious identity was reliably associated with inter-identity fit.

Method

Results

Self-trait associations: Response latencies.

The data was analyzed using multi-level analysis conducted in MLwiN (version 2.18). This analysis takes into account that RTs are nested within the individual.

A random intercept model (allowing individuals to differ randomly in their average RT) was built. In the first step, the control variable of age was entered. In the second step, three categorical independent variables–‘Self’, ‘Christian’ and ‘Women’- were entered. They indicated whether specific traits were descriptive of Christians and women, (dichotomized so that responses from 1 to 3 were coded as ‘no’, those ranging from 5 to 7 were coded ‘yes’, with 4 coded as missing: see [45], [52]) and of the self (using the dichotomous yes/no response to the response latency task). Each independent variable was dummy coded 0 (no; reference category) and 1 (yes). In the third step, the Match-Types were entered (with Self-Identities Mismatch as the reference group). Step four included individual-level continuous variables of self-defining religion and gender (centered). Finally, step five comprised the interaction between self-defining religion (centered) and Match-Types, and similarly step six included the interaction between self-defining gender (centered) and Match-Types.

An initial inspection of independence and normality of residuals suggested that our final model was an acceptable fit to the data. Intraclass correlation indicated that a substantial 30% of the variance was accounted for by between-individual differences in RTs [57]. This confirms the need for a multi-level analysis. The final model is presented in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Summary of Multi-level regression model predicating reaction times for trait classifications.

https://doi.org/10.1371/journal.pone.0137879.t002

Our focal prediction was that there would be a significant interaction between Match-Type and self-defining religiosity. Naturally, we expect that traits associated with self and Christianity are responded to more quickly among individuals with stronger self-defining religiosity. However, the critical test of the hypothesis is if stronger self-definingness of religiosity predicts faster responses to traits with a full-match between the self and both one’s ingroups (i.e., overlap between all three identities). This interaction would reflect the stronger cognitive overlap of multiple social identities within the self. In contrast, self-definingness of gender was expected to be unrelated to all RTs.

Results show a main effect of Match-Type. Crucially, this was qualified by the predicted interaction between Match-Type and self-defining religion (see Fig 2). The increase in model fit is highly significant (χ²_difference = 41.66, p < .001), which indicates significant differences between slopes. Moreover, the direction of the interaction confirms predictions. In line with expectations, the Self-Identities Match slope is steepest and significantly negative (B = -0.029, SE = 0.005, t (8781) = 5.89, p < .0001). So, the more religion is self-defining for an individual, the quicker they responded to traits that overlap across all three identities. Notably, the Self—Religious Identity Match slope was also significantly quicker (B = -0.024, SE = 0.008, t (8781) = 3.00, p < .05). However, no significant effect of Self—Gender Identity Match was found.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Log transformed reaction times representing the interaction between match-type and self-defining Christianity (centred).

https://doi.org/10.1371/journal.pone.0137879.g002

The significant interaction was followed up with tests of differences between slope estimates (see Table 3). Results indicate that overlap between self and Christian traits is strongly associated with the facilitation of fast RTs. Although adding an additional overlap with women to this is not associated with significantly faster RTs (i.e., Self—Religious Identity Match does not differ significantly from a Self—Identities Match), important substantive differences exist between these two Match-Types. Self—Identities Match differs from both Self—Identities Mismatch and Self—Gender Identity Match, while Self—Religious Identity Match only differs significantly from the former. So, Self—Identities Match differs strongly and consistently from other Match-Types, while Self—Religious Identity Match occupies an intermediary position. Thus, although overlap between the self and one’s religious identity appears to be (unsurprisingly) important for strongly self-defining Christians, the additional overlap with the ancillary identity of gender may be substantively important for the self.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Contrasts between beta estimates of the interaction between different Match-Types and self-defining religion.

https://doi.org/10.1371/journal.pone.0137879.t003

As expected, the interaction between Match-Type and self-defining gender was not significant. Notably, a marginally significant effect of self-defining gender with Self—Gender Identity Match emerged (B = -0.013, SE = 0.009, t (8781) = 1.44, p = .07), indicating that stronger self-definingness of gender is associated with faster responses to traits overlapping with the self and the female identity (in comparison to a Self-Identities Mismatch). No other significant results were found.

In sum, inspection of between-slope differences confirms that self-definingness of religion is associated with—not only for the inclusion of one’s religious identity into the self—but also the cognitive overlap in the representation of the self and religious identity with the ancillary identity of gender. In contrast, although self-definingness of gender may be associated with the fit between one’s gender identity and the self, it shows no substantive association with its fit with the ancillary religious identity.

Discussion

In Study 2, we replicated support for our hypothesis using more implicit measures. An RT analysis revealed that religiosity had consequences for cognitive representations of inter-identity fit and judgment processes, influencing both the speed and type of self-descriptiveness judgments made. In line with predictions, cognitive overlap between the self and one’s religious and gender identities was stronger at higher levels of self-definingness; evidenced by the stronger facilitation of responses to traits overlapping between these three identities, and inhibition of non-overlapping traits. Building upon previous findings [45, 52], Study 2 therefore shows that overlap can exist in the representation of the self and multiple social identities. Moreover, self-definingness of the more holistic religious identity moderates this cognitive fit.

In contrast, self-definingness of the less holistic gender only had consequences for that specific social identity. Although subjectively people report levels of self-definingness for gender and for religion that are (on average) equally high (see Pilot), gender’s self-definingness does not appear to regulate the fit among different aspects of self in the same way that religion does. Indeed, we have found only limited evidence of an association between inter-identity fit and self-definingness of gender using both explicit self-reports (Study 1) and more implicit (RT) measures tapping into cognitive representations (Study 2).

Method

Results

This analysis investigated the association between the self-definingness of gender and Christianity and the fit between the content of these two social identities. The content and structure of (strongly and weakly self-defining) Christian and gender identities were compared and contrasted by conducting a semantic network analysis on the ART lists. All networks and relevant descriptives are presented in Fig 3 and Table 4, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Aggregated networks of traits associated with female and Christian identities for (a) strongly self-defining Christians; (b) weakly self-defining Christians; and (c) strongly self-defining women; and (d) weakly self-defining women.

https://doi.org/10.1371/journal.pone.0137879.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Network descriptives of both the strongly and weakly self-defining networks for (left) religion and gender (right).

https://doi.org/10.1371/journal.pone.0137879.t004

Network shape and content.

As can be seen in panel (a) of Fig 3, the network for strongly self-defining Christians was highly interconnected (see “Figure notes” in S7 Appendix). The density suggests better connectedness than in any other network (see Table 4); a finding which is reinforced by the small network diameter (presenting the shortest possible number of links that must be crossed when moving between the two most distant nodes; see Table 4, left half).

Looking beyond network shape to content, female and Christian concepts are also highly integrated in the networks of strongly self-defining Christians. Not only do a substantial proportion of links bind uniquely female and Christian concepts to overlapping concepts (i.e., concepts which occurred on both Christian and female lists), but Table 5 indicates that the central nodes within this network are overlapping identity-concepts. Measures of degree, closeness, and betweenness converge to reveal ‘strong’, ‘caring’, ‘hard-working’ and ‘loving’ as key hubs. These nodes are adjacent to over half the concepts in the network, have the closest average distance to other concepts in the network, and help bridging concepts that would otherwise be unconnected. In this way, even uniquely female and Christian concepts are bought into close proximity, suggesting strong positive association. Thus, the content of female and religious identities of strongly self-defining Christians appears to be very closely intertwined.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Strongly self-defining (left) and weakly self-defining (right) Religion network descriptives for important identity concepts in each network.

https://doi.org/10.1371/journal.pone.0137879.t005

In contrast, the network of participants who are weakly self-defining Christians is more diffuse, as indicated by the deflated density and inflated diameter. Moreover, the weakly self-defining network shows a fairly clean segregation between uniquely female and Christian traits; few of which are connected via overlapping concepts. Indeed, some of the more central traits hubs within this network tend to be uniquely Christian and female, (e.g., ‘believers’, ‘religious’ and ‘strong’, ‘nurturing’, respectively; see Table 5). This indicates that these nodes are important bridges between other uniquely religious or female concepts and the rest of the network, which correspondingly emphasizes the segregation between these groups of identity concepts. Notably, the higher betweenness scores of the weakly self-defining network (in comparison to the strongly self-defining) suggests that overall connectedness of the network is lower, in that the network would be more sensitive to the breakdown in connectedness of a few key concepts. So, although some joint concepts are shared between women and Christians, the links within the lists are much stronger and more robust than those overlapping between them.

Interestingly, the node ‘strong’ is a uniquely female attribute amongst weakly self-defined Christians, but an overlapping concept in the strongly self-defined Christians (see “Overlap Coding” in S7 Appendix). It seems poignant that strongly self-defining individuals bridge these two identities using a concept that is most central to the female identity. This is an interesting indication that the high degree of interconnection among strongly self-defined Christians is not merely a matter of Christian values dictating gender identity: there may also be a recursive influence of gender characteristics informing Christian identity in a bi-directional interaction.

Table 4 (right half) shows that the networks of strong and weakly self-defining gender participants are highly comparable to each other. Similarly, Table 6 indicates that the key concepts and their descriptives are not substantively different for both gender networks. The characteristics of these two networks fall somewhere in between the religion networks (which is unsurprising as they are based on the same data). The lack of difference between the two networks shows that self-definingness of gender has little consequence for the fit between gender and religious identity-concepts.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Strongly self-defining (left) and weakly self-defining (right) gender network descriptives for important identity concepts in each network.

https://doi.org/10.1371/journal.pone.0137879.t006

Discussion

Study 3 offered a novel insight into the actual content of people’s Christian and gender identities through our application of a new measurement method. Specifically, we found that traits associated with these social identities were related to holisticness and self-definingness in the predicted way: Self-definingness of religiosity was associated with greater overlap between religious and gender identities and an increase in actual cognitive overlap. Thus, self-definingness of religion was associated with increased overlap between the actual social identity content of religion and gender. In line with our previous findings, self-definingness of gender was not associated with the integration of religious and gender identities. Identity content of highly self-defining religious participants, by contrast, comprised a web weaving together otherwise orthogonal facets of the self by increasing inter-identity fit. Thus, the Study 3 findings provide convergent evidence for our hypothesis and simultaneously validates our novel network method of tapping into inter-identity fit.

Theoretical implications

Our findings imply that it is important to consider both the type of group (i.e., group characteristic holisticness) and how they are personally applied (i.e., individual characteristic self-definingness). Results across all studies confirmed that only when a group was highly holistic, was stronger self-definingness related to stronger fit between Christian and gender identities. So, although holistic groups are “special” in the sense that they provide a philosophy for life and have the potential to promote inter-identity fit, this potential can only be realized by the individual’s application of group philosophies for life. It is ultimately individuals' self-definingness that allows these “special” group philosophies to strengthen inter-identity fit and come out, transcending different situations.

Of course, our results are not the first to emphasize the special role that social identities such as religion [14, 15, 63, 64] or political activism (e.g., ‘identity extension’, [65]; ‘personalism’, [66]) can play: These identities have a tendency to be all-defining in life. However, we move beyond this earlier work by concretely exploring what that means for the way such identities fit within the wider self-concept. Moreover, as far as we are aware, this research presents the first attempt to unify such varying identities under one rubric: Socially shared identity content (i.e., holisticness) provides the ideological basis for a life-philosophy, but only the individual’s use of that content (i.e., the self-definingness) ultimately actualises this potential (i.e., allows multiple social identities to fit together or overlap). Thus, our line of thought moves beyond previous conceptualizations of religious identity by viewing it as one (prototypical) member of a larger family of holistic groups.

The current work further offers a broad and nuanced perspective on the fit between multiple social identities. For instance, some have suggested that the fit between different facets of identity is neutral (i.e., independence) or negative [9], as are antagonistic inter-identity relations [19]. The present research suggests, in line with other work [2, 12, 35, 38, 40], that there is little support for universal antagonism and only qualified support for independence. But, perhaps more importantly, we also show that strong fit can emerge between multiple social identities in particular cases, such as those involving religion (as highly holistic and strongly self-defining) and gender. Notably, although this fit was strong at times, overlap between identities was never absolute (cf. mismatching ‘Match-Types’ in Study 2, and network idiosyncrasies in Study 3); identities did not become synonymous or merged into a unitary identity, but always retained some distinctiveness. In this way, the relation presented was between two unique identities, which shared important similarities and differences, but were shaped to be increasingly congruent with one-another. Thus, our line of thought suggests that the fit between multiple social identities can be thought of as reflecting a continuum ranging from non-overlapping to highly overlapping, varying as a function of both identity self-definingness and holisticness. Importantly, because gender is naturally occurring (i.e., not personally chosen) this observed increase in cognitive fit in strongly self-defining Christians should have been induced more organically through an internal process of increased inter-identity fit. Studies 2 and 3 indicate that holistic identities transform pre-existing social identity content, by mobilizing specific aspects of ancillary social identities that are consistent (or at least not inconsistent; e.g., loving) with its ideology, whilst diminishing the prominence of inconsistent traits (e.g., shopper). Holistic groups may therefore promote a more unified self-concept through the alignment of multiple, otherwise disparate, social identities. Together, our findings therefore suggest that the multifaceted self (e.g., [17, 21]) need not be inconsistent with individuals’ ability to achieve a more stable and unitary sense of self: Holisticness and self-definingness offer a pathway to achieving greater consistency over multiple and distinct self-aspects.

Practical implications

Our findings also have more practical implications, particularly when considering the specific religious and gender identities involved in our studies. There is growing evidence in the literature to suggest the important role that a religious identity can play in the self-concept [63, 64, 67, 15]. The present findings provide support for this stance. Our results align with Allport and Ross’s [14] distinction between intrinsic and extrinsic religiosity, and provide the first empirical evidence concerning the intrapsychological consequences of intrinsic religiosity. Additionally, results converge with Ysseldyk et al. [63, 64], suggesting that the guiding, socially shared, belief-system offered by religions may be an important factor amplifying the cognitive and emotional importance of these identities, relative to other (less holistic) social identities. Finally, our results are also consistent with Koole et al. [15], who argue that religion can non-invasively influence all areas of an individual’s experience—including action, attention, affect, and perceived meaning—over diverse contexts, via implicit self-regulation. Our findings shed new light on what factors may underlie those arguments: not just individual-level self-definition, but also socially shared identity content.

Regarding gender, our research clearly shows it to be a highly influential identity for many in our sample. But despite this, gender does not define inter-identity fit in the same way that religion does. We attribute this to the lack of a uniform ideology attached to gender identity content: Although there is nothing to stop gender from regulating relations among different aspects of self, in the general population based samples we recruited, it was not clearly related to inter-identity fit. Importantly however, we expect that in particular contexts (e.g., politicization), gender identity can transform, taking on a new personal meaning (e.g., feminism) and thereby come to define the whole self—and influence inter-identity fit—in ways that are more consistent across the whole category [68].

Limitations and directions for future research

Although Study 3 employed a semantic network analysis to test our hypothesis, one may wonder about its validity. Indeed, although the methods of social network analysis used in our approach are tried and trusted [69, 61], the application to self-reported identity-associations is entirely novel, as far as we know. As with all new methods, the validity of this method needs thorough assessment. Nevertheless, the results of this study converged with the other measurement methods we used. More generally, we are quite optimistic about what this method has to offer. For instance, it is very easy to administer, and can be administered to measure identity content of any social group or to any aspect of identity. Furthermore, this method is relatively implicit and non-reactive, and allows for both description and quantification. So, whereas the RT task in Study 2 might be criticised for the pre-selection of traits (which may suffer from confounded meaning). ART does not suffer from such limitations because choice and meaning were obtained from the participants themselves. Future research should embrace this new method to test its potential to measure identity content.

One valuable application of the ART method could be in addressing the juncture between holisticness/self-definingness and the intersectionality literature [35, 36]. Both research areas converge on their aim to explore the fit between multiple identities, but holisticness/self-definingness focuses more strongly on why this occurs, while intersectionality tends to focus on the how meaning of these identities depends on each other [35]. Thus, in many ways, these theories are complementary. Moreover, considering the aspect of meaning creation between multiple identities, we agree that intersections between identities can be important cognitive spaces for transformation of identity content (e.g., using a group philosophy to increase overlap between identities). Although the extent to which the fit highlighted in the present paper are additive or transformative remains an empirical question, ART makes it possible to distinguish these types of relations by comparing (1) the intersectional female-Christian identity, with (2) the intersection between the two social categories of female and Christian. This method may therefore be useful for empirically supplementing and corroborating claims in the intersectionality literature [70].

We further note that our choice of samples and groups necessarily limits the generalizability of our findings. Firstly, although we recruited diverse samp7les (student and community samples from the Netherlands and U.S., respectively), they were all taken from individualist societies. Conclusions may therefore be limited to independent cultures. Nevertheless, the findings by Baray et al. [41] among Turkish national action party members suggest comparable processes may also be found in holistic groups in non-Western cultures too. Secondly, the inclusion of non-religious individuals in our sample means that results should be interpreted with caution. Importantly however, we had theoretical reasons for including these individuals, and no substantive statistical differences were found between religious and non-religious individuals (see S5 Appendix). Thirdly, it may be the case that self-selection (into one’s religious group, but not one’s gender group) may have biased our findings. Although including non-religious people in our sample arguably helps to alleviate this critique, possible alternative explanations concerning self-selection of high identifiers of course remain; for example it could be the case that those who are devotedly religious have a need for cognitive structure and simplicity (cf. [37]). Importantly however, we believe that self-selection does not necessarily question the presence of the relations reported, because inter-identity fit (i.e., between religion and gender) were measured within the same participants: Inter-identity fit is not an individual difference characteristic based on personal predispositions, but concerns the way that individuals apply holistic ideologies.

A potentially more pressing limitation of the present research is that it only focuses on Christianity and gender. We made this deliberate choice on in order to ‘cleanly’ compare three studies employing different methods to tap into inter-identity fit. Thus, we present strong evidence supporting the stronger fit between gender and religion when religion is strongly self-defining. However conclusions drawn from these analyses are, for now necessarily limited to these groups. Nevertheless, our findings equally provide some preliminary support for the utility of these constructs and their potential to apply more widely: The Pilot study hinted towards wider applications to different groups (including environmental, health and nationality) whereas similar processes to those highlighted in the present research also emerged in Extremist Turkish political organisations [41], as in Muslim religious identity and martial arts groups who prescribed a particular philosophy [43]. Finally, we are also currently in the process of gathering and analysing data for a promising replication of the present effects using the environmental identity.

One more specific concern regarding generalizability is that Christianity may simply have exerted a direct influence on gender identity content, and that this confounded our results. Indeed, the Bible explicitly informs gender roles (e.g., 1 Timothy 2: 9–15, New International Version). Although further tests in other religions with less clear doctrines (e.g. humanism) may help to address this limitation in the future, we also have some evidence that alleviates this concern. Specifically, we find variance in the extent that religious identities are found to define gender identities—which varies in association with self-definingness. Moreover, there was some suggestion that stronger inter-identity fit was facilitated through a bidirectional process (i.e., gender and religion mutually informing each other; e.g., Study 3 female words such as ‘strong’ becoming associated with religion in strongly self-defining women only, and in Study 1 self-definingness of religion was associated with the fit of religion in gender and vice versa), rather than a simple relation of gender being defined by religion. Indeed, Study 3 showed that overlap between religion and gender was achieved on content which may not be archetypally religiously defined (i.e., strong, focused).

By and large, we are optimistic that our findings do not reflect a quirky exception due to Christianity alone. Future research can build on the present work by corroborating its external validity and addressing its internal validity. As for external validity, future research should focus on and compare different religions and aspects of fit between Christianity and gender, for example, under conditions where fit may come more strongly under question (e.g., for homosexual Christians’ gender or sexual identity; [11]. Furthermore, we think it is important to explore beyond religious identities to other holistic identities, such as those of politicized activists or professional athletes. This could give us insight into how fit between multiple identities within activists or athletes can aid their pursuit of a philosophy for life in the face of potentially competing goals (e.g., goal of lowing carbon footprint v. goal to visit foreign and distant country). As for internal validity, future research could manipulate group holisticness (i.e., by manipulating the extent a group is perceived to prescribe a philosophy for life). However, it is unclear how viable it would be to test the consequences of holisticness experimentally: Given that inter-identity fit is an outcome of both holisticness and self-definingness it is unclear whether intrapersonal consequences (i.e., varying inter-identity fit) will be observable in short experimentally controllable periods. Despite this, gender may be an ideal identity to utilise for such a procedure, because results demonstrate that (a) it does not generally function as highly holistic groups do, and (b) individuals accept it can be strongly self-defining.