Social categorization, such as forming distinct groups, leads to the development of social identities, where individuals identify with specific social groups (Turner et al., 1987; Abrams and Hogg, 1988). Social identity processes have far-reaching effects on various intergroup and psychological phenomena, such as leading to negative attitudes and discrimination between groups (Stephan and Stephan, 1985; Tajfel and Turner, 2004; Gordils et al., 2021, 2023). Social group identities are not fixed; they can be influenced by categorization processes, such as reducing identification with ingroups (e.g., Wilder and Allen, 1978; Gaertner et al., 1993; Dovidio et al., 1995; Van Bavel and Cunningham, 2010), or redefining group boundaries (Dovidio et al., 2000; Brewer, 2011). For instance, cooperation during a pandemic could signal shared goals, which may create a sense of community for groups that were previously considered outgroups (Bavel et al., 2020).

Similarly, the Common Ingroup Identity Model (CIIM) suggests that individuals from different groups can recategorize themselves into a superordinate identity category (Gaertner et al., 1993). In this scenario, members of subordinate groups view themselves as part of a broader, shared group identity. Embracing this superordinate identity, therefore, extends the favoritism that individuals show to members of their own group to those who were once perceived to belong to other groups, reducing intergroup tension (Dovidio et al., 2008).

Past work has documented the efficacy of adopting a common group identity on positive outcomes, including but not limited to more positive intergroup attitudes, a greater willingness for intergroup contact, helping outgroup members, and intergroup trust (Riek et al., 2010; Penner et al., 2013; Vezzali et al., 2015). Additionally, this model has been shown to be effective across myriad social contexts such as high schools, universities, and workplaces, and is applicable across different nationality groups (Hindriks et al., 2014; Dietz et al., 2015; Toprakkiran and Gordils, 2021; Cehajić-Clancy et al., 2023). Despite the robustness of this model, past work has also documented instances where the applications of the model may be less unequivocally clear. For instance, policies that highlight commonalities between groups, such as colorblindness, may hinder efforts for equality (Dovidio et al., 2016). Furthermore, increasing a common identity, such as a national identity, may promote negative attitudes toward those perceived as belonging to other national groups, such as immigrants (Wojcieszak and Garrett, 2018). Taken together, the common ingroup identity model sometimes fosters and sometimes hinders positive intergroup attitudes.

One explanation as to why increasing the salience of a common identity might not improve intergroup attitudes may be due to existing conceptualizations individuals hold of the given common identity. In other words, if common identities are construed differently across individuals from different groups, this might influence the outcomes of creating a common identity. However, the literature on common identity has mostly overlooked this potential issue. To better understand the conceptualizations of common identities, researchers must assess conceptual equivalence (i.e., whether a particular construct has the same meaning across social groups; Berry, 2002; Harachi et al., 2006; Trimble, 2007). For instance, if researchers plan to examine the outcomes of a common American identity between Democrats and Republicans, it might be necessary to understand the nuanced conceptualizations of what it means to be an American to members of each group.

One way of creating a common identity for Democrats and Republicans is by invoking a shared, superordinate national identity. Indeed, previous research shows that priming a common American identity decreases negative attitudes between Democrats and Republicans (Levendusky, 2018). However, as partisans may differ on their conceptualizations of what it means to be American, said differences may create contention between rival parties and attenuate positive effects of shared identities. As such, a common American identity might function as a double-edged sword.

Specifically, how people define the boundaries of a broader American identity might determine who they exclude from this common identity, and thus persist as categorizing compatriots as outgroup members. For example, for those who believe that the American identity necessitates being Christian, increasing an American identity may be linked to higher negative attitudes toward those who are not (or perceived to not be) Christian. Christian nationalism implies the exclusion of non-Christian religious groups from a national identity (Delehanty et al., 2019). In the United States, the importance of religion is linked to anti-Muslim prejudice in Christian respondents (Ogan et al., 2014). Additionally, for individuals who believe that being American encompasses being born in the U.S., a higher American identity may be related to higher negative attitudes toward those who are foreign-born. Higher national identity may increase hostility toward immigrants by enhancing the salience of group boundaries (Wojcieszak and Garrett, 2018). While 14% of the United States population is made up of immigrants (Budiman, 2020), 41% of U.S. Americans believe that immigration should be decreased (Saad, 2023). Further, for individuals who believe that speaking English is necessary to be a true American, a higher American identity may be linked to higher negative attitudes toward those who do not speak English. In a 2017 Pew Research survey, 70% of respondents reported that it is very important to be able to speak English to be considered “American” (Stokes, 2017). The literature on linguistic prejudice shows that accented speech (i.e., accents spoken by minoritized groups and/or viewed as foreign) is related to interpersonal evaluations and the differentiation between ingroup and outgroup members (Fuertes et al., 2012).

Together, those that endorse a more restrictive and exclusionary definition of “being American” may be less likely to view outgroup members as part of the superordinate ingroup, and thus are likely to view outgroups more negatively (or less positively), and by extension, less likely to benefit from the advantages posited by the common ingroup identity model. In the present work, we specifically focus on the aforementioned factors (i.e., being a Christian, being born in the U.S., and being able to speak English) and the degree to which partisans find these factors important for what it means to be “truly American.”

Previous work found that Democrats and Republicans define what it means to be American differently. More specifically, when it comes to being “truly American,” 89% of Republicans compared to 65% of Democrats believe that speaking English is important, 86% of Republicans compared to 59% of Democrats believe that sharing U.S. customs and traditions is important, 48% of Republicans compared to 25% of Democrats believe that being Christian is important, and 46% of Republicans compared to 25% of Democrats believe that having been born in the U.S is important (Connaughton, 2021). Given these differences, the aims of the present research are 2-fold. First, we sought to conceptually replicate whether American identity (i.e., superordinate identity) is related to positive intergroup attitudes between Democrats and Republicans and whether what it means to be an American differs between Democrats and Republicans. Second, we examined whether the extent to which people define what it means to be an American restrictively moderates the relationship between American Identity and intergroup attitudes.

All survey materials, data, and R code have been made publicly available at the Open Science Framework and can be accessed at: https://osf.io/n3r8y/?view_only=c7332bfa772741db82c50c29029612a8.

Data was collected from 1,322 United States citizens over the age of 18, between 10/27/2020 and 11/11/2020¹. All participants were recruited from MTurk, and they received $0.50 for their participation. Thirty-seven (3%) individuals failed the attention checks, 3 (0.2%) individuals completed the study outside of the U.S., and 10 (0.8%) individuals had duplicate IP addresses. These exclusions resulted in a sample of N = 1,272 participants (56% Female, 44% Male, 0.2% Other; 59% Democrat, 41% Republican; 74% White, 9% Black or African American, 9% Asian, 2% Hispanic or Latino, 4% multiracial, and 1% other).

We found no differences between Democrats and Republicans in our sample, in terms of race/ethnicity, X(6,1,266)2 = 9.77, p = 0.13, sex, X(3,1,271)2 = 3.67, p = 0.30, education, X(7,1,272)2 = 13.26, p = 0.066, and age, t_(537.17) = −1.88, p = 0.061. However, we found that on average, Republicans in our sample had higher income compared to Democrats t_{(1, 125.6)} = −3.34, p < 0.001, which is representative of the general population (Pew Research Center, 2023).

First, participants completed demographic information, which included a question asking, “Which of the following do you primarily identify as?” with the choices “Democrat” or “Republican.” As the U.S. representative democracy governmental system is dominated by two political parties, we only offered these choices to our participants (e.g., out of 100 senators, only three are not affiliated with either party; US House of Representatives Press Gallery, 2023). Second, they completed survey questions². This survey was designed to be conditional so that each participant received questions based on their party identity. For instance, if the participant identified as a Democrat, they were asked to report their American identity, restrictive Americanism, how much they like Republicans, and how warm they perceive Republicans to be (and vice versa for participants who identified as Republicans). We included three attention check items across the study. Participants were debriefed after completing the study.

This observational study examined whether the definition of what it means to be American moderates the relationship between identifying as American and intergroup attitudes (liking and perceived warmth) between Democrats and Republicans. Based on preliminary exploratory correlational analyses, we planned to examine the associations between American identity and intergroup attitudes: liking and warmth, separately controlling for political party (Republican = 1, and Democrat = 0) and controlling for political orientation. Then, we examined whether restrictive Americanism moderated these relationships. We took composite scores of all variables that were assessed with multiple items by taking the average of the items, and all continuous variables were standardized. Since the restrictive Americanism items had high inter-item correlations (rs ≥ 0.5, ps < 0.001; see Table 1) and internal consistency (α = 0.81), we took the average of these items to create a composite score (higher values corresponding to those who believe that being Christian, born in the U.S., and speaking English were integral to being American). Because we wanted to test if restrictive Americanism moderates the relationship between American identity and intergroup attitudes regardless of participants' political party affiliation (i.e., Democrat or Republican) and political orientation, we also included political party and political orientation as covariates in separate regression analyses⁴.

On average, participants found speaking English most important to being American (see Table 1; M = 3.63, SD = 1.40), followed by being born in the U.S. (M = 3.29, SD = 1.50), and being Christian (M = 2.32, SD = 1.59). Compared to Democrats, Republicans viewed speaking English [t_{(1, 259.9)} = −14.58, p < 0.001], being born in the U.S. [t_{(1, 216.9)} = −13.12, p < 0.001], and being Christian [t_{(1, 007.8)} = −14.50, p < 0.001] more important to being American.

Our main outcome variables (i.e., outgroup liking and warmth) were correlated with each other (r = 0.58, p < 0.001; see Table 2). American identity was positively related to liking (r = 0.20, p < 0.001) and warmth (r = 0.22, p < 0.001). Compared to Democrats, Republicans were more likely to score higher in restrictive Americanism [i.e., a more restrictive definition of what being “truly” American means; t_{(1, 200)} = −17.22, p < 0.001] and American identity [t_{(1, 235.5)} = −14.01, p < 0.001]. Furthermore, those who had a more restrictive view of what it means to be “truly” American were higher in American identity (r = 0.53, p < 0.001).

First, we found that American identity was positively related to outgroup liking (β = 0.083, p = 0.018) and perceived outgroup warmth (β = 0.091, p = 0.0076) when we controlled for restrictive Americanism, the interaction between American identity and restrictive Americanism, and party⁵ (i.e., Democrat or Republican; see Table 2). This relationship was moderated by restrictive Americanism for liking (β = −0.14, p < 0.001) and warmth (β = −0.082, p = 0.0033). More specifically, higher restrictive Americanism (in terms of being a Christian, being born in the U.S., and speaking English), had a negative effect on how American identity was related to liking members of the opposing party and perceiving them as warmer, when controlling for party. Moreover, follow up regression analyses with three-way interactions showed that the size of this interaction effect did not differ across Democrat and Republican participants. Three-way interaction terms between American identity, restrictive Americanism, and party in separately predicting outgroup liking and warmth were not significant (β_liking = −0.091, p = 0.22, β_warmth = −0.058, p = 0.40)⁶.

We ran the same analyses, controlling for political orientation instead of party, and we found similar results⁷. American identity was positively related to outgroup liking (β = 0.079, p = 0.023) and perceived outgroup warmth (β = 0.096, p = 0.0046) when we controlled for restrictive Americanism, the interaction between American identity and restrictive Americanism, and political orientation (see Table 3). This relationship was moderated by restrictive Americanism for liking (β = −0.14, p < 0.001) and warmth (β = −0.075, p = 0.0073). Higher restrictive Americanism had a negative effect on how American identity was related to liking members of the opposing party and perceiving them as warmer, controlling for participants' political orientation. Moreover, follow up regression analyses with three-way interactions showed that the interaction between American identity and restrictive Americanism did not depend on political orientation when it comes to liking (β_liking = −0.055, p = 0.065) but it did for warmth (β_warmth = −0.095, p = 0.0014)⁸.

To disentangle the interactions between American identity and restrictive Americanism (controlling for party), we used simple slopes analysis, using the interactions package in R (v1. 1.5; Long, 2019; see Figure 1). We specifically looked at restrictive Americanism at three levels: low (mean – 1 SD), average, and high (mean + 1 SD). We found that for participants who were low in terms of restrictive Americanism, American identity positively predicted liking (β = 0.22, p < 0.001; see Figure 1) and warmth (β = 0.17, p < 0.001). Similarly, at mean level restrictive Americanism, American identity positively predicted liking (β = 0.082, p = 0.019) and warmth (β = 0.091, p = 0.0076). However, for participants who were high in restrictive Americanism, these effects did not hold for either liking (β = −0.059, p = 0.24) or warmth (β = 0.0093, p = 0.85). In other words, for participants who had a less restrictive conceptualization of “American,” their American identity was positively related to their liking and warmth toward their political outgroup. For participants who had a more restrictive conceptualization of “American,” their American identity was not related to liking and warmth, regardless of which party they identified with. To disentangle the interactions between American identity and restrictive Americanism when controlling for political orientation, we repeated the same analyses and found similar results (see Supplementary material).

This work is the first to document the moderating role of conceptualizations of national American identity on relationships between American identity and intergroup attitudes. It is important to note that we only examined a subset of facets related to being American (i.e., Christian, born in U.S., speaks English). While we chose these characteristics based on past work (Li and Brewer, 2004), conceptualizations of American identity extend to many other facets. For instance, people might associate being an American with a commitment to certain values, such as democracy, equality, liberalism, multiculturalism, liberty, and freedom (Gleason, 1980; Gallup, 2003; Theiss-Morse, 2009). Future research would benefit from understanding what these characteristics are, especially for attributes that may be endorsed by both Democrats and Republicans, as they may help maximize the utility of the Common Ingroup Identity Model in improving intergroup relations.

Besides expanding on the multi-faceted conceptualization of what an American identity means to Democrats and Republicans, future work would also benefit from taking an intersectional approach to understanding the role of concept equivalence in common identity processes. That is, certain aspects of restrictive Americanism, or an exclusive conceptualization of what it means to be an American may resonate more with certain groups, especially subgroups of political parties such as populists. Being born in the U.S. might not be as important to being American for immigrants. Additionally, these views might differ across generations. For instance, younger Americans (18–29 years old) viewed openness to people from all over the world as more ‘essential to who we are as a nation' compared to older Americans (65+ years; Hartig, 2018). Moreover, speaking English might not be as important to being American for Hispanic or Latino Americans, given 68% of Hispanic or Latino Americans speak Spanish at home (Krogstad et al., 2015). Additionally, it is necessary to examine how a dual identity approach, where partisans endorse both their party identities and national identities, would influence relations between partisans. Since the focus on a common identity without dual identity can increase identity threat (Crisp et al., 2006), a common national identity that may not align with one's party identity may further increase polarization.

In considering the conceptualization of what it means to be American, it is also crucial to understand how important these characteristics are to individuals. Since our study did not ask participants about their attitudes toward these identity facets, we cannot distinguish those who value Christianity, nativism, and speaking English from those who do not. In other words, someone might view Christianity to be important to being American, without viewing Christianity to be important to their own identity. Furthermore, the Ingroup Projection Model suggests that group members may project characteristics of their ingroup to a broader identity, viewing their group as prototypical of this identity (Wenzel et al., 2008). Asking respondents to rate how important a characteristic is to their own identity, alongside assessing how important it is to being American is essential in understanding this phenomenon. Future directions include exploring this additional dimension to parse concept equivalence from subjective perceptions of the components underlying American identity.

The work presented herein shows that there was a moderate positive relationship between American identity and restrictive Americanism (r = 0.53, p < 0.001), indicating that those who tend to identify more strongly as American were also more likely to define this identity restrictively. Moreover, participants who were lower in restrictive Americanism and American identity also reported more negative attitudes toward members of the other party (see Figure 1). A combination of low American identity and low restrictive Americanism might represent a group of people who not only score low on American identity, but intentionally oppose or dislike American identity, since it entails values that they do not support. This may be linked to more negative attitudes toward the opposing party, if restrictive Americanism is linked with perceptions of the other parties' ideologies. In our sample, Democrats scored lower than Republicans in American identity, restrictive Americanism, and expressed more negative attitudes toward outgroup members (see Table 4). Since American identity is less clearly defined among Democrats, and Republican leaders often use restrictive definitions of American identity to gain support (Mellow, 2020; Hanson et al., 2023), Democrats may perceive American identity to be akin to Republican identity. Besides further understanding the sociodemographic and psychological characteristics of individuals who are low in American identity and restrictive Americanism, a notable advance would be to differentiate between how people personally define what it means to be American and how they think society and members of the opposing party define what it means to be American.

This data was collected as part of a larger project related to timing effects around presidential elections, and additional analyses confirmed that results hold when controlling for timing effects (i.e., pre-, during, and post-election). Even though the broader data collection period of this study was still around the 2020 election, our findings replicated previous work that was conducted during other time points (Levendusky, 2018; Connaughton, 2021). Regardless, it is still possible that our data collection around the Presidential election coincided with increased political tensions, and that the salience of restrictive Americanism, American identity, and party identity were higher at this time. Additionally, it is important to note that while differences between the composition of Republicans and Democrats in our sample were not significant in terms of race/ethnicity, sex, education, and age, the composition of Republicans at the national level tends to be more White, less formally educated, and older compared to Democrats (Pew Research Center, 2023). As such, future work would benefit from examining these associations during non-election years, as well as in more representative samples.

In terms of measurement, even though we used multiple items to assess warmth, we used a single-item measure to assess liking. While multiple-item measures tend to exhibit higher predictive validity compared to single-item measures (Diamantopoulos et al., 2012), intergroup relations researchers often use single-item feeling thermometers as main outcome measures, including those assessing outgroup liking (e.g., Crisp et al., 2009; Seger et al., 2014; Jacoby-Senghor et al., 2019). Indeed, single-item measures across research domains exhibit comparable validity and reliability to their multi-item equivalents (Bergkvist and Rossiter, 2007; Ahmad et al., 2014; Ang and Eisend, 2018; Allen et al., 2022). Feeling thermometers have also been successful in measuring sentiments toward members of other parties, being related to preferences for social distance and discrimination in economic games (Iyengar et al., 2019; Gidron et al., 2023). Additionally, while many studies focusing on intergroup relations use liking and warmth interchangeably (e.g., Fiske et al., 1999; Dasgupta and Greenwald, 2001; Inbar et al., 2012), we focused on liking and warmth separately to get a more nuanced understanding of inter-partisan relations, as attitudes can be complex and must be measured accordingly (Krosnick et al., 2005).

When it comes to generalizability, our findings are limited to a single, correlational study in the U.S. political context. Researchers must examine whether these relationships are causal by experimentally manipulating the strength and conceptualizations of a common national identity. In doing so, researchers must explore how priming different common national identities might affect participants (e.g., based on party affiliation, immigration status, and race/ethnicity). Moreover, while we specifically focused on the U.S. political context, political polarization is a global issue (Carothers and O'Donahue, 2019). To assess the generalizability of our findings, examining whether the issue of conceptual equivalence emerges in other political contexts (e.g., other countries and other parties) is paramount, especially in coalition-based democratic governments around the world.

Finally, while the effects we found are consistent across both measures of intergroup relations we used, interaction effects were quite small (Funder and Ozer, 2019). Small effects can be beneficial to advancing theory and application (Prentice and Miller, 1992; Rosenthal et al., 2000). However, they may also indicate limited applicability of the studied phenomenon to real-life outcomes. This limitation highlights the importance of conducting future work to understand the possible consequences of a common identity in improving inter-partisan attitudes.